Merge tag 'for-linus-v3.5-rc7' of git://oss.sgi.com/xfs/xfs
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / net / sctp / socket.c
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 <linux/export.h>
82 #include <net/sock.h>
83 #include <net/sctp/sctp.h>
84 #include <net/sctp/sm.h>
85
86 /* WARNING: Please do not remove the SCTP_STATIC attribute to
87 * any of the functions below as they are used to export functions
88 * used by a project regression testsuite.
89 */
90
91 /* Forward declarations for internal helper functions. */
92 static int sctp_writeable(struct sock *sk);
93 static void sctp_wfree(struct sk_buff *skb);
94 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
95 size_t msg_len);
96 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
97 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
98 static int sctp_wait_for_accept(struct sock *sk, long timeo);
99 static void sctp_wait_for_close(struct sock *sk, long timeo);
100 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
101 union sctp_addr *addr, int len);
102 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
103 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
104 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
105 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
106 static int sctp_send_asconf(struct sctp_association *asoc,
107 struct sctp_chunk *chunk);
108 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
109 static int sctp_autobind(struct sock *sk);
110 static void sctp_sock_migrate(struct sock *, struct sock *,
111 struct sctp_association *, sctp_socket_type_t);
112 static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
113
114 extern struct kmem_cache *sctp_bucket_cachep;
115 extern long sysctl_sctp_mem[3];
116 extern int sysctl_sctp_rmem[3];
117 extern int sysctl_sctp_wmem[3];
118
119 static int sctp_memory_pressure;
120 static atomic_long_t sctp_memory_allocated;
121 struct percpu_counter sctp_sockets_allocated;
122
123 static void sctp_enter_memory_pressure(struct sock *sk)
124 {
125 sctp_memory_pressure = 1;
126 }
127
128
129 /* Get the sndbuf space available at the time on the association. */
130 static inline int sctp_wspace(struct sctp_association *asoc)
131 {
132 int amt;
133
134 if (asoc->ep->sndbuf_policy)
135 amt = asoc->sndbuf_used;
136 else
137 amt = sk_wmem_alloc_get(asoc->base.sk);
138
139 if (amt >= asoc->base.sk->sk_sndbuf) {
140 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
141 amt = 0;
142 else {
143 amt = sk_stream_wspace(asoc->base.sk);
144 if (amt < 0)
145 amt = 0;
146 }
147 } else {
148 amt = asoc->base.sk->sk_sndbuf - amt;
149 }
150 return amt;
151 }
152
153 /* Increment the used sndbuf space count of the corresponding association by
154 * the size of the outgoing data chunk.
155 * Also, set the skb destructor for sndbuf accounting later.
156 *
157 * Since it is always 1-1 between chunk and skb, and also a new skb is always
158 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
159 * destructor in the data chunk skb for the purpose of the sndbuf space
160 * tracking.
161 */
162 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
163 {
164 struct sctp_association *asoc = chunk->asoc;
165 struct sock *sk = asoc->base.sk;
166
167 /* The sndbuf space is tracked per association. */
168 sctp_association_hold(asoc);
169
170 skb_set_owner_w(chunk->skb, sk);
171
172 chunk->skb->destructor = sctp_wfree;
173 /* Save the chunk pointer in skb for sctp_wfree to use later. */
174 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
175
176 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
177 sizeof(struct sk_buff) +
178 sizeof(struct sctp_chunk);
179
180 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
181 sk->sk_wmem_queued += chunk->skb->truesize;
182 sk_mem_charge(sk, chunk->skb->truesize);
183 }
184
185 /* Verify that this is a valid address. */
186 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
187 int len)
188 {
189 struct sctp_af *af;
190
191 /* Verify basic sockaddr. */
192 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
193 if (!af)
194 return -EINVAL;
195
196 /* Is this a valid SCTP address? */
197 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
198 return -EINVAL;
199
200 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
201 return -EINVAL;
202
203 return 0;
204 }
205
206 /* Look up the association by its id. If this is not a UDP-style
207 * socket, the ID field is always ignored.
208 */
209 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
210 {
211 struct sctp_association *asoc = NULL;
212
213 /* If this is not a UDP-style socket, assoc id should be ignored. */
214 if (!sctp_style(sk, UDP)) {
215 /* Return NULL if the socket state is not ESTABLISHED. It
216 * could be a TCP-style listening socket or a socket which
217 * hasn't yet called connect() to establish an association.
218 */
219 if (!sctp_sstate(sk, ESTABLISHED))
220 return NULL;
221
222 /* Get the first and the only association from the list. */
223 if (!list_empty(&sctp_sk(sk)->ep->asocs))
224 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
225 struct sctp_association, asocs);
226 return asoc;
227 }
228
229 /* Otherwise this is a UDP-style socket. */
230 if (!id || (id == (sctp_assoc_t)-1))
231 return NULL;
232
233 spin_lock_bh(&sctp_assocs_id_lock);
234 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
235 spin_unlock_bh(&sctp_assocs_id_lock);
236
237 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
238 return NULL;
239
240 return asoc;
241 }
242
243 /* Look up the transport from an address and an assoc id. If both address and
244 * id are specified, the associations matching the address and the id should be
245 * the same.
246 */
247 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
248 struct sockaddr_storage *addr,
249 sctp_assoc_t id)
250 {
251 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
252 struct sctp_transport *transport;
253 union sctp_addr *laddr = (union sctp_addr *)addr;
254
255 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
256 laddr,
257 &transport);
258
259 if (!addr_asoc)
260 return NULL;
261
262 id_asoc = sctp_id2assoc(sk, id);
263 if (id_asoc && (id_asoc != addr_asoc))
264 return NULL;
265
266 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
267 (union sctp_addr *)addr);
268
269 return transport;
270 }
271
272 /* API 3.1.2 bind() - UDP Style Syntax
273 * The syntax of bind() is,
274 *
275 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
276 *
277 * sd - the socket descriptor returned by socket().
278 * addr - the address structure (struct sockaddr_in or struct
279 * sockaddr_in6 [RFC 2553]),
280 * addr_len - the size of the address structure.
281 */
282 SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
283 {
284 int retval = 0;
285
286 sctp_lock_sock(sk);
287
288 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
289 sk, addr, addr_len);
290
291 /* Disallow binding twice. */
292 if (!sctp_sk(sk)->ep->base.bind_addr.port)
293 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
294 addr_len);
295 else
296 retval = -EINVAL;
297
298 sctp_release_sock(sk);
299
300 return retval;
301 }
302
303 static long sctp_get_port_local(struct sock *, union sctp_addr *);
304
305 /* Verify this is a valid sockaddr. */
306 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
307 union sctp_addr *addr, int len)
308 {
309 struct sctp_af *af;
310
311 /* Check minimum size. */
312 if (len < sizeof (struct sockaddr))
313 return NULL;
314
315 /* V4 mapped address are really of AF_INET family */
316 if (addr->sa.sa_family == AF_INET6 &&
317 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
318 if (!opt->pf->af_supported(AF_INET, opt))
319 return NULL;
320 } else {
321 /* Does this PF support this AF? */
322 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
323 return NULL;
324 }
325
326 /* If we get this far, af is valid. */
327 af = sctp_get_af_specific(addr->sa.sa_family);
328
329 if (len < af->sockaddr_len)
330 return NULL;
331
332 return af;
333 }
334
335 /* Bind a local address either to an endpoint or to an association. */
336 SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
337 {
338 struct sctp_sock *sp = sctp_sk(sk);
339 struct sctp_endpoint *ep = sp->ep;
340 struct sctp_bind_addr *bp = &ep->base.bind_addr;
341 struct sctp_af *af;
342 unsigned short snum;
343 int ret = 0;
344
345 /* Common sockaddr verification. */
346 af = sctp_sockaddr_af(sp, addr, len);
347 if (!af) {
348 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
349 sk, addr, len);
350 return -EINVAL;
351 }
352
353 snum = ntohs(addr->v4.sin_port);
354
355 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
356 ", port: %d, new port: %d, len: %d)\n",
357 sk,
358 addr,
359 bp->port, snum,
360 len);
361
362 /* PF specific bind() address verification. */
363 if (!sp->pf->bind_verify(sp, addr))
364 return -EADDRNOTAVAIL;
365
366 /* We must either be unbound, or bind to the same port.
367 * It's OK to allow 0 ports if we are already bound.
368 * We'll just inhert an already bound port in this case
369 */
370 if (bp->port) {
371 if (!snum)
372 snum = bp->port;
373 else if (snum != bp->port) {
374 SCTP_DEBUG_PRINTK("sctp_do_bind:"
375 " New port %d does not match existing port "
376 "%d.\n", snum, bp->port);
377 return -EINVAL;
378 }
379 }
380
381 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
382 return -EACCES;
383
384 /* See if the address matches any of the addresses we may have
385 * already bound before checking against other endpoints.
386 */
387 if (sctp_bind_addr_match(bp, addr, sp))
388 return -EINVAL;
389
390 /* Make sure we are allowed to bind here.
391 * The function sctp_get_port_local() does duplicate address
392 * detection.
393 */
394 addr->v4.sin_port = htons(snum);
395 if ((ret = sctp_get_port_local(sk, addr))) {
396 return -EADDRINUSE;
397 }
398
399 /* Refresh ephemeral port. */
400 if (!bp->port)
401 bp->port = inet_sk(sk)->inet_num;
402
403 /* Add the address to the bind address list.
404 * Use GFP_ATOMIC since BHs will be disabled.
405 */
406 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
407
408 /* Copy back into socket for getsockname() use. */
409 if (!ret) {
410 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
411 af->to_sk_saddr(addr, sk);
412 }
413
414 return ret;
415 }
416
417 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
418 *
419 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
420 * at any one time. If a sender, after sending an ASCONF chunk, decides
421 * it needs to transfer another ASCONF Chunk, it MUST wait until the
422 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
423 * subsequent ASCONF. Note this restriction binds each side, so at any
424 * time two ASCONF may be in-transit on any given association (one sent
425 * from each endpoint).
426 */
427 static int sctp_send_asconf(struct sctp_association *asoc,
428 struct sctp_chunk *chunk)
429 {
430 int retval = 0;
431
432 /* If there is an outstanding ASCONF chunk, queue it for later
433 * transmission.
434 */
435 if (asoc->addip_last_asconf) {
436 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
437 goto out;
438 }
439
440 /* Hold the chunk until an ASCONF_ACK is received. */
441 sctp_chunk_hold(chunk);
442 retval = sctp_primitive_ASCONF(asoc, chunk);
443 if (retval)
444 sctp_chunk_free(chunk);
445 else
446 asoc->addip_last_asconf = chunk;
447
448 out:
449 return retval;
450 }
451
452 /* Add a list of addresses as bind addresses to local endpoint or
453 * association.
454 *
455 * Basically run through each address specified in the addrs/addrcnt
456 * array/length pair, determine if it is IPv6 or IPv4 and call
457 * sctp_do_bind() on it.
458 *
459 * If any of them fails, then the operation will be reversed and the
460 * ones that were added will be removed.
461 *
462 * Only sctp_setsockopt_bindx() is supposed to call this function.
463 */
464 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
465 {
466 int cnt;
467 int retval = 0;
468 void *addr_buf;
469 struct sockaddr *sa_addr;
470 struct sctp_af *af;
471
472 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
473 sk, addrs, addrcnt);
474
475 addr_buf = addrs;
476 for (cnt = 0; cnt < addrcnt; cnt++) {
477 /* The list may contain either IPv4 or IPv6 address;
478 * determine the address length for walking thru the list.
479 */
480 sa_addr = addr_buf;
481 af = sctp_get_af_specific(sa_addr->sa_family);
482 if (!af) {
483 retval = -EINVAL;
484 goto err_bindx_add;
485 }
486
487 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
488 af->sockaddr_len);
489
490 addr_buf += af->sockaddr_len;
491
492 err_bindx_add:
493 if (retval < 0) {
494 /* Failed. Cleanup the ones that have been added */
495 if (cnt > 0)
496 sctp_bindx_rem(sk, addrs, cnt);
497 return retval;
498 }
499 }
500
501 return retval;
502 }
503
504 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
505 * associations that are part of the endpoint indicating that a list of local
506 * addresses are added to the endpoint.
507 *
508 * If any of the addresses is already in the bind address list of the
509 * association, we do not send the chunk for that association. But it will not
510 * affect other associations.
511 *
512 * Only sctp_setsockopt_bindx() is supposed to call this function.
513 */
514 static int sctp_send_asconf_add_ip(struct sock *sk,
515 struct sockaddr *addrs,
516 int addrcnt)
517 {
518 struct sctp_sock *sp;
519 struct sctp_endpoint *ep;
520 struct sctp_association *asoc;
521 struct sctp_bind_addr *bp;
522 struct sctp_chunk *chunk;
523 struct sctp_sockaddr_entry *laddr;
524 union sctp_addr *addr;
525 union sctp_addr saveaddr;
526 void *addr_buf;
527 struct sctp_af *af;
528 struct list_head *p;
529 int i;
530 int retval = 0;
531
532 if (!sctp_addip_enable)
533 return retval;
534
535 sp = sctp_sk(sk);
536 ep = sp->ep;
537
538 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
539 __func__, sk, addrs, addrcnt);
540
541 list_for_each_entry(asoc, &ep->asocs, asocs) {
542
543 if (!asoc->peer.asconf_capable)
544 continue;
545
546 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
547 continue;
548
549 if (!sctp_state(asoc, ESTABLISHED))
550 continue;
551
552 /* Check if any address in the packed array of addresses is
553 * in the bind address list of the association. If so,
554 * do not send the asconf chunk to its peer, but continue with
555 * other associations.
556 */
557 addr_buf = addrs;
558 for (i = 0; i < addrcnt; i++) {
559 addr = addr_buf;
560 af = sctp_get_af_specific(addr->v4.sin_family);
561 if (!af) {
562 retval = -EINVAL;
563 goto out;
564 }
565
566 if (sctp_assoc_lookup_laddr(asoc, addr))
567 break;
568
569 addr_buf += af->sockaddr_len;
570 }
571 if (i < addrcnt)
572 continue;
573
574 /* Use the first valid address in bind addr list of
575 * association as Address Parameter of ASCONF CHUNK.
576 */
577 bp = &asoc->base.bind_addr;
578 p = bp->address_list.next;
579 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
580 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
581 addrcnt, SCTP_PARAM_ADD_IP);
582 if (!chunk) {
583 retval = -ENOMEM;
584 goto out;
585 }
586
587 /* Add the new addresses to the bind address list with
588 * use_as_src set to 0.
589 */
590 addr_buf = addrs;
591 for (i = 0; i < addrcnt; i++) {
592 addr = addr_buf;
593 af = sctp_get_af_specific(addr->v4.sin_family);
594 memcpy(&saveaddr, addr, af->sockaddr_len);
595 retval = sctp_add_bind_addr(bp, &saveaddr,
596 SCTP_ADDR_NEW, GFP_ATOMIC);
597 addr_buf += af->sockaddr_len;
598 }
599 if (asoc->src_out_of_asoc_ok) {
600 struct sctp_transport *trans;
601
602 list_for_each_entry(trans,
603 &asoc->peer.transport_addr_list, transports) {
604 /* Clear the source and route cache */
605 dst_release(trans->dst);
606 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
607 2*asoc->pathmtu, 4380));
608 trans->ssthresh = asoc->peer.i.a_rwnd;
609 trans->rto = asoc->rto_initial;
610 trans->rtt = trans->srtt = trans->rttvar = 0;
611 sctp_transport_route(trans, NULL,
612 sctp_sk(asoc->base.sk));
613 }
614 }
615 retval = sctp_send_asconf(asoc, chunk);
616 }
617
618 out:
619 return retval;
620 }
621
622 /* Remove a list of addresses from bind addresses list. Do not remove the
623 * last address.
624 *
625 * Basically run through each address specified in the addrs/addrcnt
626 * array/length pair, determine if it is IPv6 or IPv4 and call
627 * sctp_del_bind() on it.
628 *
629 * If any of them fails, then the operation will be reversed and the
630 * ones that were removed will be added back.
631 *
632 * At least one address has to be left; if only one address is
633 * available, the operation will return -EBUSY.
634 *
635 * Only sctp_setsockopt_bindx() is supposed to call this function.
636 */
637 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
638 {
639 struct sctp_sock *sp = sctp_sk(sk);
640 struct sctp_endpoint *ep = sp->ep;
641 int cnt;
642 struct sctp_bind_addr *bp = &ep->base.bind_addr;
643 int retval = 0;
644 void *addr_buf;
645 union sctp_addr *sa_addr;
646 struct sctp_af *af;
647
648 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
649 sk, addrs, addrcnt);
650
651 addr_buf = addrs;
652 for (cnt = 0; cnt < addrcnt; cnt++) {
653 /* If the bind address list is empty or if there is only one
654 * bind address, there is nothing more to be removed (we need
655 * at least one address here).
656 */
657 if (list_empty(&bp->address_list) ||
658 (sctp_list_single_entry(&bp->address_list))) {
659 retval = -EBUSY;
660 goto err_bindx_rem;
661 }
662
663 sa_addr = addr_buf;
664 af = sctp_get_af_specific(sa_addr->sa.sa_family);
665 if (!af) {
666 retval = -EINVAL;
667 goto err_bindx_rem;
668 }
669
670 if (!af->addr_valid(sa_addr, sp, NULL)) {
671 retval = -EADDRNOTAVAIL;
672 goto err_bindx_rem;
673 }
674
675 if (sa_addr->v4.sin_port &&
676 sa_addr->v4.sin_port != htons(bp->port)) {
677 retval = -EINVAL;
678 goto err_bindx_rem;
679 }
680
681 if (!sa_addr->v4.sin_port)
682 sa_addr->v4.sin_port = htons(bp->port);
683
684 /* FIXME - There is probably a need to check if sk->sk_saddr and
685 * sk->sk_rcv_addr are currently set to one of the addresses to
686 * be removed. This is something which needs to be looked into
687 * when we are fixing the outstanding issues with multi-homing
688 * socket routing and failover schemes. Refer to comments in
689 * sctp_do_bind(). -daisy
690 */
691 retval = sctp_del_bind_addr(bp, sa_addr);
692
693 addr_buf += af->sockaddr_len;
694 err_bindx_rem:
695 if (retval < 0) {
696 /* Failed. Add the ones that has been removed back */
697 if (cnt > 0)
698 sctp_bindx_add(sk, addrs, cnt);
699 return retval;
700 }
701 }
702
703 return retval;
704 }
705
706 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
707 * the associations that are part of the endpoint indicating that a list of
708 * local addresses are removed from the endpoint.
709 *
710 * If any of the addresses is already in the bind address list of the
711 * association, we do not send the chunk for that association. But it will not
712 * affect other associations.
713 *
714 * Only sctp_setsockopt_bindx() is supposed to call this function.
715 */
716 static int sctp_send_asconf_del_ip(struct sock *sk,
717 struct sockaddr *addrs,
718 int addrcnt)
719 {
720 struct sctp_sock *sp;
721 struct sctp_endpoint *ep;
722 struct sctp_association *asoc;
723 struct sctp_transport *transport;
724 struct sctp_bind_addr *bp;
725 struct sctp_chunk *chunk;
726 union sctp_addr *laddr;
727 void *addr_buf;
728 struct sctp_af *af;
729 struct sctp_sockaddr_entry *saddr;
730 int i;
731 int retval = 0;
732 int stored = 0;
733
734 chunk = NULL;
735 if (!sctp_addip_enable)
736 return retval;
737
738 sp = sctp_sk(sk);
739 ep = sp->ep;
740
741 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
742 __func__, sk, addrs, addrcnt);
743
744 list_for_each_entry(asoc, &ep->asocs, asocs) {
745
746 if (!asoc->peer.asconf_capable)
747 continue;
748
749 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
750 continue;
751
752 if (!sctp_state(asoc, ESTABLISHED))
753 continue;
754
755 /* Check if any address in the packed array of addresses is
756 * not present in the bind address list of the association.
757 * If so, do not send the asconf chunk to its peer, but
758 * continue with other associations.
759 */
760 addr_buf = addrs;
761 for (i = 0; i < addrcnt; i++) {
762 laddr = addr_buf;
763 af = sctp_get_af_specific(laddr->v4.sin_family);
764 if (!af) {
765 retval = -EINVAL;
766 goto out;
767 }
768
769 if (!sctp_assoc_lookup_laddr(asoc, laddr))
770 break;
771
772 addr_buf += af->sockaddr_len;
773 }
774 if (i < addrcnt)
775 continue;
776
777 /* Find one address in the association's bind address list
778 * that is not in the packed array of addresses. This is to
779 * make sure that we do not delete all the addresses in the
780 * association.
781 */
782 bp = &asoc->base.bind_addr;
783 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
784 addrcnt, sp);
785 if ((laddr == NULL) && (addrcnt == 1)) {
786 if (asoc->asconf_addr_del_pending)
787 continue;
788 asoc->asconf_addr_del_pending =
789 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
790 if (asoc->asconf_addr_del_pending == NULL) {
791 retval = -ENOMEM;
792 goto out;
793 }
794 asoc->asconf_addr_del_pending->sa.sa_family =
795 addrs->sa_family;
796 asoc->asconf_addr_del_pending->v4.sin_port =
797 htons(bp->port);
798 if (addrs->sa_family == AF_INET) {
799 struct sockaddr_in *sin;
800
801 sin = (struct sockaddr_in *)addrs;
802 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
803 } else if (addrs->sa_family == AF_INET6) {
804 struct sockaddr_in6 *sin6;
805
806 sin6 = (struct sockaddr_in6 *)addrs;
807 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
808 }
809 SCTP_DEBUG_PRINTK_IPADDR("send_asconf_del_ip: keep the last address asoc: %p ",
810 " at %p\n", asoc, asoc->asconf_addr_del_pending,
811 asoc->asconf_addr_del_pending);
812 asoc->src_out_of_asoc_ok = 1;
813 stored = 1;
814 goto skip_mkasconf;
815 }
816
817 /* We do not need RCU protection throughout this loop
818 * because this is done under a socket lock from the
819 * setsockopt call.
820 */
821 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
822 SCTP_PARAM_DEL_IP);
823 if (!chunk) {
824 retval = -ENOMEM;
825 goto out;
826 }
827
828 skip_mkasconf:
829 /* Reset use_as_src flag for the addresses in the bind address
830 * list that are to be deleted.
831 */
832 addr_buf = addrs;
833 for (i = 0; i < addrcnt; i++) {
834 laddr = addr_buf;
835 af = sctp_get_af_specific(laddr->v4.sin_family);
836 list_for_each_entry(saddr, &bp->address_list, list) {
837 if (sctp_cmp_addr_exact(&saddr->a, laddr))
838 saddr->state = SCTP_ADDR_DEL;
839 }
840 addr_buf += af->sockaddr_len;
841 }
842
843 /* Update the route and saddr entries for all the transports
844 * as some of the addresses in the bind address list are
845 * about to be deleted and cannot be used as source addresses.
846 */
847 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
848 transports) {
849 dst_release(transport->dst);
850 sctp_transport_route(transport, NULL,
851 sctp_sk(asoc->base.sk));
852 }
853
854 if (stored)
855 /* We don't need to transmit ASCONF */
856 continue;
857 retval = sctp_send_asconf(asoc, chunk);
858 }
859 out:
860 return retval;
861 }
862
863 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
864 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
865 {
866 struct sock *sk = sctp_opt2sk(sp);
867 union sctp_addr *addr;
868 struct sctp_af *af;
869
870 /* It is safe to write port space in caller. */
871 addr = &addrw->a;
872 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
873 af = sctp_get_af_specific(addr->sa.sa_family);
874 if (!af)
875 return -EINVAL;
876 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
877 return -EINVAL;
878
879 if (addrw->state == SCTP_ADDR_NEW)
880 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
881 else
882 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
883 }
884
885 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
886 *
887 * API 8.1
888 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
889 * int flags);
890 *
891 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
892 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
893 * or IPv6 addresses.
894 *
895 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
896 * Section 3.1.2 for this usage.
897 *
898 * addrs is a pointer to an array of one or more socket addresses. Each
899 * address is contained in its appropriate structure (i.e. struct
900 * sockaddr_in or struct sockaddr_in6) the family of the address type
901 * must be used to distinguish the address length (note that this
902 * representation is termed a "packed array" of addresses). The caller
903 * specifies the number of addresses in the array with addrcnt.
904 *
905 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
906 * -1, and sets errno to the appropriate error code.
907 *
908 * For SCTP, the port given in each socket address must be the same, or
909 * sctp_bindx() will fail, setting errno to EINVAL.
910 *
911 * The flags parameter is formed from the bitwise OR of zero or more of
912 * the following currently defined flags:
913 *
914 * SCTP_BINDX_ADD_ADDR
915 *
916 * SCTP_BINDX_REM_ADDR
917 *
918 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
919 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
920 * addresses from the association. The two flags are mutually exclusive;
921 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
922 * not remove all addresses from an association; sctp_bindx() will
923 * reject such an attempt with EINVAL.
924 *
925 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
926 * additional addresses with an endpoint after calling bind(). Or use
927 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
928 * socket is associated with so that no new association accepted will be
929 * associated with those addresses. If the endpoint supports dynamic
930 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
931 * endpoint to send the appropriate message to the peer to change the
932 * peers address lists.
933 *
934 * Adding and removing addresses from a connected association is
935 * optional functionality. Implementations that do not support this
936 * functionality should return EOPNOTSUPP.
937 *
938 * Basically do nothing but copying the addresses from user to kernel
939 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
940 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
941 * from userspace.
942 *
943 * We don't use copy_from_user() for optimization: we first do the
944 * sanity checks (buffer size -fast- and access check-healthy
945 * pointer); if all of those succeed, then we can alloc the memory
946 * (expensive operation) needed to copy the data to kernel. Then we do
947 * the copying without checking the user space area
948 * (__copy_from_user()).
949 *
950 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
951 * it.
952 *
953 * sk The sk of the socket
954 * addrs The pointer to the addresses in user land
955 * addrssize Size of the addrs buffer
956 * op Operation to perform (add or remove, see the flags of
957 * sctp_bindx)
958 *
959 * Returns 0 if ok, <0 errno code on error.
960 */
961 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
962 struct sockaddr __user *addrs,
963 int addrs_size, int op)
964 {
965 struct sockaddr *kaddrs;
966 int err;
967 int addrcnt = 0;
968 int walk_size = 0;
969 struct sockaddr *sa_addr;
970 void *addr_buf;
971 struct sctp_af *af;
972
973 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
974 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
975
976 if (unlikely(addrs_size <= 0))
977 return -EINVAL;
978
979 /* Check the user passed a healthy pointer. */
980 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
981 return -EFAULT;
982
983 /* Alloc space for the address array in kernel memory. */
984 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
985 if (unlikely(!kaddrs))
986 return -ENOMEM;
987
988 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
989 kfree(kaddrs);
990 return -EFAULT;
991 }
992
993 /* Walk through the addrs buffer and count the number of addresses. */
994 addr_buf = kaddrs;
995 while (walk_size < addrs_size) {
996 if (walk_size + sizeof(sa_family_t) > addrs_size) {
997 kfree(kaddrs);
998 return -EINVAL;
999 }
1000
1001 sa_addr = addr_buf;
1002 af = sctp_get_af_specific(sa_addr->sa_family);
1003
1004 /* If the address family is not supported or if this address
1005 * causes the address buffer to overflow return EINVAL.
1006 */
1007 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1008 kfree(kaddrs);
1009 return -EINVAL;
1010 }
1011 addrcnt++;
1012 addr_buf += af->sockaddr_len;
1013 walk_size += af->sockaddr_len;
1014 }
1015
1016 /* Do the work. */
1017 switch (op) {
1018 case SCTP_BINDX_ADD_ADDR:
1019 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1020 if (err)
1021 goto out;
1022 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1023 break;
1024
1025 case SCTP_BINDX_REM_ADDR:
1026 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1027 if (err)
1028 goto out;
1029 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1030 break;
1031
1032 default:
1033 err = -EINVAL;
1034 break;
1035 }
1036
1037 out:
1038 kfree(kaddrs);
1039
1040 return err;
1041 }
1042
1043 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1044 *
1045 * Common routine for handling connect() and sctp_connectx().
1046 * Connect will come in with just a single address.
1047 */
1048 static int __sctp_connect(struct sock* sk,
1049 struct sockaddr *kaddrs,
1050 int addrs_size,
1051 sctp_assoc_t *assoc_id)
1052 {
1053 struct sctp_sock *sp;
1054 struct sctp_endpoint *ep;
1055 struct sctp_association *asoc = NULL;
1056 struct sctp_association *asoc2;
1057 struct sctp_transport *transport;
1058 union sctp_addr to;
1059 struct sctp_af *af;
1060 sctp_scope_t scope;
1061 long timeo;
1062 int err = 0;
1063 int addrcnt = 0;
1064 int walk_size = 0;
1065 union sctp_addr *sa_addr = NULL;
1066 void *addr_buf;
1067 unsigned short port;
1068 unsigned int f_flags = 0;
1069
1070 sp = sctp_sk(sk);
1071 ep = sp->ep;
1072
1073 /* connect() cannot be done on a socket that is already in ESTABLISHED
1074 * state - UDP-style peeled off socket or a TCP-style socket that
1075 * is already connected.
1076 * It cannot be done even on a TCP-style listening socket.
1077 */
1078 if (sctp_sstate(sk, ESTABLISHED) ||
1079 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1080 err = -EISCONN;
1081 goto out_free;
1082 }
1083
1084 /* Walk through the addrs buffer and count the number of addresses. */
1085 addr_buf = kaddrs;
1086 while (walk_size < addrs_size) {
1087 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1088 err = -EINVAL;
1089 goto out_free;
1090 }
1091
1092 sa_addr = addr_buf;
1093 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1094
1095 /* If the address family is not supported or if this address
1096 * causes the address buffer to overflow return EINVAL.
1097 */
1098 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1099 err = -EINVAL;
1100 goto out_free;
1101 }
1102
1103 port = ntohs(sa_addr->v4.sin_port);
1104
1105 /* Save current address so we can work with it */
1106 memcpy(&to, sa_addr, af->sockaddr_len);
1107
1108 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1109 if (err)
1110 goto out_free;
1111
1112 /* Make sure the destination port is correctly set
1113 * in all addresses.
1114 */
1115 if (asoc && asoc->peer.port && asoc->peer.port != port)
1116 goto out_free;
1117
1118
1119 /* Check if there already is a matching association on the
1120 * endpoint (other than the one created here).
1121 */
1122 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1123 if (asoc2 && asoc2 != asoc) {
1124 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1125 err = -EISCONN;
1126 else
1127 err = -EALREADY;
1128 goto out_free;
1129 }
1130
1131 /* If we could not find a matching association on the endpoint,
1132 * make sure that there is no peeled-off association matching
1133 * the peer address even on another socket.
1134 */
1135 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1136 err = -EADDRNOTAVAIL;
1137 goto out_free;
1138 }
1139
1140 if (!asoc) {
1141 /* If a bind() or sctp_bindx() is not called prior to
1142 * an sctp_connectx() call, the system picks an
1143 * ephemeral port and will choose an address set
1144 * equivalent to binding with a wildcard address.
1145 */
1146 if (!ep->base.bind_addr.port) {
1147 if (sctp_autobind(sk)) {
1148 err = -EAGAIN;
1149 goto out_free;
1150 }
1151 } else {
1152 /*
1153 * If an unprivileged user inherits a 1-many
1154 * style socket with open associations on a
1155 * privileged port, it MAY be permitted to
1156 * accept new associations, but it SHOULD NOT
1157 * be permitted to open new associations.
1158 */
1159 if (ep->base.bind_addr.port < PROT_SOCK &&
1160 !capable(CAP_NET_BIND_SERVICE)) {
1161 err = -EACCES;
1162 goto out_free;
1163 }
1164 }
1165
1166 scope = sctp_scope(&to);
1167 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1168 if (!asoc) {
1169 err = -ENOMEM;
1170 goto out_free;
1171 }
1172
1173 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1174 GFP_KERNEL);
1175 if (err < 0) {
1176 goto out_free;
1177 }
1178
1179 }
1180
1181 /* Prime the peer's transport structures. */
1182 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1183 SCTP_UNKNOWN);
1184 if (!transport) {
1185 err = -ENOMEM;
1186 goto out_free;
1187 }
1188
1189 addrcnt++;
1190 addr_buf += af->sockaddr_len;
1191 walk_size += af->sockaddr_len;
1192 }
1193
1194 /* In case the user of sctp_connectx() wants an association
1195 * id back, assign one now.
1196 */
1197 if (assoc_id) {
1198 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1199 if (err < 0)
1200 goto out_free;
1201 }
1202
1203 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1204 if (err < 0) {
1205 goto out_free;
1206 }
1207
1208 /* Initialize sk's dport and daddr for getpeername() */
1209 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1210 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1211 af->to_sk_daddr(sa_addr, sk);
1212 sk->sk_err = 0;
1213
1214 /* in-kernel sockets don't generally have a file allocated to them
1215 * if all they do is call sock_create_kern().
1216 */
1217 if (sk->sk_socket->file)
1218 f_flags = sk->sk_socket->file->f_flags;
1219
1220 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1221
1222 err = sctp_wait_for_connect(asoc, &timeo);
1223 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1224 *assoc_id = asoc->assoc_id;
1225
1226 /* Don't free association on exit. */
1227 asoc = NULL;
1228
1229 out_free:
1230
1231 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1232 " kaddrs: %p err: %d\n",
1233 asoc, kaddrs, err);
1234 if (asoc)
1235 sctp_association_free(asoc);
1236 return err;
1237 }
1238
1239 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1240 *
1241 * API 8.9
1242 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1243 * sctp_assoc_t *asoc);
1244 *
1245 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1246 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1247 * or IPv6 addresses.
1248 *
1249 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1250 * Section 3.1.2 for this usage.
1251 *
1252 * addrs is a pointer to an array of one or more socket addresses. Each
1253 * address is contained in its appropriate structure (i.e. struct
1254 * sockaddr_in or struct sockaddr_in6) the family of the address type
1255 * must be used to distengish the address length (note that this
1256 * representation is termed a "packed array" of addresses). The caller
1257 * specifies the number of addresses in the array with addrcnt.
1258 *
1259 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1260 * the association id of the new association. On failure, sctp_connectx()
1261 * returns -1, and sets errno to the appropriate error code. The assoc_id
1262 * is not touched by the kernel.
1263 *
1264 * For SCTP, the port given in each socket address must be the same, or
1265 * sctp_connectx() will fail, setting errno to EINVAL.
1266 *
1267 * An application can use sctp_connectx to initiate an association with
1268 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1269 * allows a caller to specify multiple addresses at which a peer can be
1270 * reached. The way the SCTP stack uses the list of addresses to set up
1271 * the association is implementation dependent. This function only
1272 * specifies that the stack will try to make use of all the addresses in
1273 * the list when needed.
1274 *
1275 * Note that the list of addresses passed in is only used for setting up
1276 * the association. It does not necessarily equal the set of addresses
1277 * the peer uses for the resulting association. If the caller wants to
1278 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1279 * retrieve them after the association has been set up.
1280 *
1281 * Basically do nothing but copying the addresses from user to kernel
1282 * land and invoking either sctp_connectx(). This is used for tunneling
1283 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1284 *
1285 * We don't use copy_from_user() for optimization: we first do the
1286 * sanity checks (buffer size -fast- and access check-healthy
1287 * pointer); if all of those succeed, then we can alloc the memory
1288 * (expensive operation) needed to copy the data to kernel. Then we do
1289 * the copying without checking the user space area
1290 * (__copy_from_user()).
1291 *
1292 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1293 * it.
1294 *
1295 * sk The sk of the socket
1296 * addrs The pointer to the addresses in user land
1297 * addrssize Size of the addrs buffer
1298 *
1299 * Returns >=0 if ok, <0 errno code on error.
1300 */
1301 SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
1302 struct sockaddr __user *addrs,
1303 int addrs_size,
1304 sctp_assoc_t *assoc_id)
1305 {
1306 int err = 0;
1307 struct sockaddr *kaddrs;
1308
1309 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1310 __func__, sk, addrs, addrs_size);
1311
1312 if (unlikely(addrs_size <= 0))
1313 return -EINVAL;
1314
1315 /* Check the user passed a healthy pointer. */
1316 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1317 return -EFAULT;
1318
1319 /* Alloc space for the address array in kernel memory. */
1320 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1321 if (unlikely(!kaddrs))
1322 return -ENOMEM;
1323
1324 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1325 err = -EFAULT;
1326 } else {
1327 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1328 }
1329
1330 kfree(kaddrs);
1331
1332 return err;
1333 }
1334
1335 /*
1336 * This is an older interface. It's kept for backward compatibility
1337 * to the option that doesn't provide association id.
1338 */
1339 SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1340 struct sockaddr __user *addrs,
1341 int addrs_size)
1342 {
1343 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1344 }
1345
1346 /*
1347 * New interface for the API. The since the API is done with a socket
1348 * option, to make it simple we feed back the association id is as a return
1349 * indication to the call. Error is always negative and association id is
1350 * always positive.
1351 */
1352 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1353 struct sockaddr __user *addrs,
1354 int addrs_size)
1355 {
1356 sctp_assoc_t assoc_id = 0;
1357 int err = 0;
1358
1359 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1360
1361 if (err)
1362 return err;
1363 else
1364 return assoc_id;
1365 }
1366
1367 /*
1368 * New (hopefully final) interface for the API.
1369 * We use the sctp_getaddrs_old structure so that use-space library
1370 * can avoid any unnecessary allocations. The only defferent part
1371 * is that we store the actual length of the address buffer into the
1372 * addrs_num structure member. That way we can re-use the existing
1373 * code.
1374 */
1375 SCTP_STATIC int sctp_getsockopt_connectx3(struct sock* sk, int len,
1376 char __user *optval,
1377 int __user *optlen)
1378 {
1379 struct sctp_getaddrs_old param;
1380 sctp_assoc_t assoc_id = 0;
1381 int err = 0;
1382
1383 if (len < sizeof(param))
1384 return -EINVAL;
1385
1386 if (copy_from_user(&param, optval, sizeof(param)))
1387 return -EFAULT;
1388
1389 err = __sctp_setsockopt_connectx(sk,
1390 (struct sockaddr __user *)param.addrs,
1391 param.addr_num, &assoc_id);
1392
1393 if (err == 0 || err == -EINPROGRESS) {
1394 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1395 return -EFAULT;
1396 if (put_user(sizeof(assoc_id), optlen))
1397 return -EFAULT;
1398 }
1399
1400 return err;
1401 }
1402
1403 /* API 3.1.4 close() - UDP Style Syntax
1404 * Applications use close() to perform graceful shutdown (as described in
1405 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1406 * by a UDP-style socket.
1407 *
1408 * The syntax is
1409 *
1410 * ret = close(int sd);
1411 *
1412 * sd - the socket descriptor of the associations to be closed.
1413 *
1414 * To gracefully shutdown a specific association represented by the
1415 * UDP-style socket, an application should use the sendmsg() call,
1416 * passing no user data, but including the appropriate flag in the
1417 * ancillary data (see Section xxxx).
1418 *
1419 * If sd in the close() call is a branched-off socket representing only
1420 * one association, the shutdown is performed on that association only.
1421 *
1422 * 4.1.6 close() - TCP Style Syntax
1423 *
1424 * Applications use close() to gracefully close down an association.
1425 *
1426 * The syntax is:
1427 *
1428 * int close(int sd);
1429 *
1430 * sd - the socket descriptor of the association to be closed.
1431 *
1432 * After an application calls close() on a socket descriptor, no further
1433 * socket operations will succeed on that descriptor.
1434 *
1435 * API 7.1.4 SO_LINGER
1436 *
1437 * An application using the TCP-style socket can use this option to
1438 * perform the SCTP ABORT primitive. The linger option structure is:
1439 *
1440 * struct linger {
1441 * int l_onoff; // option on/off
1442 * int l_linger; // linger time
1443 * };
1444 *
1445 * To enable the option, set l_onoff to 1. If the l_linger value is set
1446 * to 0, calling close() is the same as the ABORT primitive. If the
1447 * value is set to a negative value, the setsockopt() call will return
1448 * an error. If the value is set to a positive value linger_time, the
1449 * close() can be blocked for at most linger_time ms. If the graceful
1450 * shutdown phase does not finish during this period, close() will
1451 * return but the graceful shutdown phase continues in the system.
1452 */
1453 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1454 {
1455 struct sctp_endpoint *ep;
1456 struct sctp_association *asoc;
1457 struct list_head *pos, *temp;
1458 unsigned int data_was_unread;
1459
1460 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1461
1462 sctp_lock_sock(sk);
1463 sk->sk_shutdown = SHUTDOWN_MASK;
1464 sk->sk_state = SCTP_SS_CLOSING;
1465
1466 ep = sctp_sk(sk)->ep;
1467
1468 /* Clean up any skbs sitting on the receive queue. */
1469 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1470 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1471
1472 /* Walk all associations on an endpoint. */
1473 list_for_each_safe(pos, temp, &ep->asocs) {
1474 asoc = list_entry(pos, struct sctp_association, asocs);
1475
1476 if (sctp_style(sk, TCP)) {
1477 /* A closed association can still be in the list if
1478 * it belongs to a TCP-style listening socket that is
1479 * not yet accepted. If so, free it. If not, send an
1480 * ABORT or SHUTDOWN based on the linger options.
1481 */
1482 if (sctp_state(asoc, CLOSED)) {
1483 sctp_unhash_established(asoc);
1484 sctp_association_free(asoc);
1485 continue;
1486 }
1487 }
1488
1489 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1490 !skb_queue_empty(&asoc->ulpq.reasm) ||
1491 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1492 struct sctp_chunk *chunk;
1493
1494 chunk = sctp_make_abort_user(asoc, NULL, 0);
1495 if (chunk)
1496 sctp_primitive_ABORT(asoc, chunk);
1497 } else
1498 sctp_primitive_SHUTDOWN(asoc, NULL);
1499 }
1500
1501 /* On a TCP-style socket, block for at most linger_time if set. */
1502 if (sctp_style(sk, TCP) && timeout)
1503 sctp_wait_for_close(sk, timeout);
1504
1505 /* This will run the backlog queue. */
1506 sctp_release_sock(sk);
1507
1508 /* Supposedly, no process has access to the socket, but
1509 * the net layers still may.
1510 */
1511 sctp_local_bh_disable();
1512 sctp_bh_lock_sock(sk);
1513
1514 /* Hold the sock, since sk_common_release() will put sock_put()
1515 * and we have just a little more cleanup.
1516 */
1517 sock_hold(sk);
1518 sk_common_release(sk);
1519
1520 sctp_bh_unlock_sock(sk);
1521 sctp_local_bh_enable();
1522
1523 sock_put(sk);
1524
1525 SCTP_DBG_OBJCNT_DEC(sock);
1526 }
1527
1528 /* Handle EPIPE error. */
1529 static int sctp_error(struct sock *sk, int flags, int err)
1530 {
1531 if (err == -EPIPE)
1532 err = sock_error(sk) ? : -EPIPE;
1533 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1534 send_sig(SIGPIPE, current, 0);
1535 return err;
1536 }
1537
1538 /* API 3.1.3 sendmsg() - UDP Style Syntax
1539 *
1540 * An application uses sendmsg() and recvmsg() calls to transmit data to
1541 * and receive data from its peer.
1542 *
1543 * ssize_t sendmsg(int socket, const struct msghdr *message,
1544 * int flags);
1545 *
1546 * socket - the socket descriptor of the endpoint.
1547 * message - pointer to the msghdr structure which contains a single
1548 * user message and possibly some ancillary data.
1549 *
1550 * See Section 5 for complete description of the data
1551 * structures.
1552 *
1553 * flags - flags sent or received with the user message, see Section
1554 * 5 for complete description of the flags.
1555 *
1556 * Note: This function could use a rewrite especially when explicit
1557 * connect support comes in.
1558 */
1559 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1560
1561 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1562
1563 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1564 struct msghdr *msg, size_t msg_len)
1565 {
1566 struct sctp_sock *sp;
1567 struct sctp_endpoint *ep;
1568 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1569 struct sctp_transport *transport, *chunk_tp;
1570 struct sctp_chunk *chunk;
1571 union sctp_addr to;
1572 struct sockaddr *msg_name = NULL;
1573 struct sctp_sndrcvinfo default_sinfo;
1574 struct sctp_sndrcvinfo *sinfo;
1575 struct sctp_initmsg *sinit;
1576 sctp_assoc_t associd = 0;
1577 sctp_cmsgs_t cmsgs = { NULL };
1578 int err;
1579 sctp_scope_t scope;
1580 long timeo;
1581 __u16 sinfo_flags = 0;
1582 struct sctp_datamsg *datamsg;
1583 int msg_flags = msg->msg_flags;
1584
1585 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1586 sk, msg, msg_len);
1587
1588 err = 0;
1589 sp = sctp_sk(sk);
1590 ep = sp->ep;
1591
1592 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1593
1594 /* We cannot send a message over a TCP-style listening socket. */
1595 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1596 err = -EPIPE;
1597 goto out_nounlock;
1598 }
1599
1600 /* Parse out the SCTP CMSGs. */
1601 err = sctp_msghdr_parse(msg, &cmsgs);
1602
1603 if (err) {
1604 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1605 goto out_nounlock;
1606 }
1607
1608 /* Fetch the destination address for this packet. This
1609 * address only selects the association--it is not necessarily
1610 * the address we will send to.
1611 * For a peeled-off socket, msg_name is ignored.
1612 */
1613 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1614 int msg_namelen = msg->msg_namelen;
1615
1616 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1617 msg_namelen);
1618 if (err)
1619 return err;
1620
1621 if (msg_namelen > sizeof(to))
1622 msg_namelen = sizeof(to);
1623 memcpy(&to, msg->msg_name, msg_namelen);
1624 msg_name = msg->msg_name;
1625 }
1626
1627 sinfo = cmsgs.info;
1628 sinit = cmsgs.init;
1629
1630 /* Did the user specify SNDRCVINFO? */
1631 if (sinfo) {
1632 sinfo_flags = sinfo->sinfo_flags;
1633 associd = sinfo->sinfo_assoc_id;
1634 }
1635
1636 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1637 msg_len, sinfo_flags);
1638
1639 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1640 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1641 err = -EINVAL;
1642 goto out_nounlock;
1643 }
1644
1645 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1646 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1647 * If SCTP_ABORT is set, the message length could be non zero with
1648 * the msg_iov set to the user abort reason.
1649 */
1650 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1651 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1652 err = -EINVAL;
1653 goto out_nounlock;
1654 }
1655
1656 /* If SCTP_ADDR_OVER is set, there must be an address
1657 * specified in msg_name.
1658 */
1659 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1660 err = -EINVAL;
1661 goto out_nounlock;
1662 }
1663
1664 transport = NULL;
1665
1666 SCTP_DEBUG_PRINTK("About to look up association.\n");
1667
1668 sctp_lock_sock(sk);
1669
1670 /* If a msg_name has been specified, assume this is to be used. */
1671 if (msg_name) {
1672 /* Look for a matching association on the endpoint. */
1673 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1674 if (!asoc) {
1675 /* If we could not find a matching association on the
1676 * endpoint, make sure that it is not a TCP-style
1677 * socket that already has an association or there is
1678 * no peeled-off association on another socket.
1679 */
1680 if ((sctp_style(sk, TCP) &&
1681 sctp_sstate(sk, ESTABLISHED)) ||
1682 sctp_endpoint_is_peeled_off(ep, &to)) {
1683 err = -EADDRNOTAVAIL;
1684 goto out_unlock;
1685 }
1686 }
1687 } else {
1688 asoc = sctp_id2assoc(sk, associd);
1689 if (!asoc) {
1690 err = -EPIPE;
1691 goto out_unlock;
1692 }
1693 }
1694
1695 if (asoc) {
1696 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1697
1698 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1699 * socket that has an association in CLOSED state. This can
1700 * happen when an accepted socket has an association that is
1701 * already CLOSED.
1702 */
1703 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1704 err = -EPIPE;
1705 goto out_unlock;
1706 }
1707
1708 if (sinfo_flags & SCTP_EOF) {
1709 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1710 asoc);
1711 sctp_primitive_SHUTDOWN(asoc, NULL);
1712 err = 0;
1713 goto out_unlock;
1714 }
1715 if (sinfo_flags & SCTP_ABORT) {
1716
1717 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1718 if (!chunk) {
1719 err = -ENOMEM;
1720 goto out_unlock;
1721 }
1722
1723 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1724 sctp_primitive_ABORT(asoc, chunk);
1725 err = 0;
1726 goto out_unlock;
1727 }
1728 }
1729
1730 /* Do we need to create the association? */
1731 if (!asoc) {
1732 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1733
1734 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1735 err = -EINVAL;
1736 goto out_unlock;
1737 }
1738
1739 /* Check for invalid stream against the stream counts,
1740 * either the default or the user specified stream counts.
1741 */
1742 if (sinfo) {
1743 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1744 /* Check against the defaults. */
1745 if (sinfo->sinfo_stream >=
1746 sp->initmsg.sinit_num_ostreams) {
1747 err = -EINVAL;
1748 goto out_unlock;
1749 }
1750 } else {
1751 /* Check against the requested. */
1752 if (sinfo->sinfo_stream >=
1753 sinit->sinit_num_ostreams) {
1754 err = -EINVAL;
1755 goto out_unlock;
1756 }
1757 }
1758 }
1759
1760 /*
1761 * API 3.1.2 bind() - UDP Style Syntax
1762 * If a bind() or sctp_bindx() is not called prior to a
1763 * sendmsg() call that initiates a new association, the
1764 * system picks an ephemeral port and will choose an address
1765 * set equivalent to binding with a wildcard address.
1766 */
1767 if (!ep->base.bind_addr.port) {
1768 if (sctp_autobind(sk)) {
1769 err = -EAGAIN;
1770 goto out_unlock;
1771 }
1772 } else {
1773 /*
1774 * If an unprivileged user inherits a one-to-many
1775 * style socket with open associations on a privileged
1776 * port, it MAY be permitted to accept new associations,
1777 * but it SHOULD NOT be permitted to open new
1778 * associations.
1779 */
1780 if (ep->base.bind_addr.port < PROT_SOCK &&
1781 !capable(CAP_NET_BIND_SERVICE)) {
1782 err = -EACCES;
1783 goto out_unlock;
1784 }
1785 }
1786
1787 scope = sctp_scope(&to);
1788 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1789 if (!new_asoc) {
1790 err = -ENOMEM;
1791 goto out_unlock;
1792 }
1793 asoc = new_asoc;
1794 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1795 if (err < 0) {
1796 err = -ENOMEM;
1797 goto out_free;
1798 }
1799
1800 /* If the SCTP_INIT ancillary data is specified, set all
1801 * the association init values accordingly.
1802 */
1803 if (sinit) {
1804 if (sinit->sinit_num_ostreams) {
1805 asoc->c.sinit_num_ostreams =
1806 sinit->sinit_num_ostreams;
1807 }
1808 if (sinit->sinit_max_instreams) {
1809 asoc->c.sinit_max_instreams =
1810 sinit->sinit_max_instreams;
1811 }
1812 if (sinit->sinit_max_attempts) {
1813 asoc->max_init_attempts
1814 = sinit->sinit_max_attempts;
1815 }
1816 if (sinit->sinit_max_init_timeo) {
1817 asoc->max_init_timeo =
1818 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1819 }
1820 }
1821
1822 /* Prime the peer's transport structures. */
1823 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1824 if (!transport) {
1825 err = -ENOMEM;
1826 goto out_free;
1827 }
1828 }
1829
1830 /* ASSERT: we have a valid association at this point. */
1831 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1832
1833 if (!sinfo) {
1834 /* If the user didn't specify SNDRCVINFO, make up one with
1835 * some defaults.
1836 */
1837 memset(&default_sinfo, 0, sizeof(default_sinfo));
1838 default_sinfo.sinfo_stream = asoc->default_stream;
1839 default_sinfo.sinfo_flags = asoc->default_flags;
1840 default_sinfo.sinfo_ppid = asoc->default_ppid;
1841 default_sinfo.sinfo_context = asoc->default_context;
1842 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1843 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1844 sinfo = &default_sinfo;
1845 }
1846
1847 /* API 7.1.7, the sndbuf size per association bounds the
1848 * maximum size of data that can be sent in a single send call.
1849 */
1850 if (msg_len > sk->sk_sndbuf) {
1851 err = -EMSGSIZE;
1852 goto out_free;
1853 }
1854
1855 if (asoc->pmtu_pending)
1856 sctp_assoc_pending_pmtu(asoc);
1857
1858 /* If fragmentation is disabled and the message length exceeds the
1859 * association fragmentation point, return EMSGSIZE. The I-D
1860 * does not specify what this error is, but this looks like
1861 * a great fit.
1862 */
1863 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1864 err = -EMSGSIZE;
1865 goto out_free;
1866 }
1867
1868 /* Check for invalid stream. */
1869 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1870 err = -EINVAL;
1871 goto out_free;
1872 }
1873
1874 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1875 if (!sctp_wspace(asoc)) {
1876 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1877 if (err)
1878 goto out_free;
1879 }
1880
1881 /* If an address is passed with the sendto/sendmsg call, it is used
1882 * to override the primary destination address in the TCP model, or
1883 * when SCTP_ADDR_OVER flag is set in the UDP model.
1884 */
1885 if ((sctp_style(sk, TCP) && msg_name) ||
1886 (sinfo_flags & SCTP_ADDR_OVER)) {
1887 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1888 if (!chunk_tp) {
1889 err = -EINVAL;
1890 goto out_free;
1891 }
1892 } else
1893 chunk_tp = NULL;
1894
1895 /* Auto-connect, if we aren't connected already. */
1896 if (sctp_state(asoc, CLOSED)) {
1897 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1898 if (err < 0)
1899 goto out_free;
1900 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1901 }
1902
1903 /* Break the message into multiple chunks of maximum size. */
1904 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1905 if (!datamsg) {
1906 err = -ENOMEM;
1907 goto out_free;
1908 }
1909
1910 /* Now send the (possibly) fragmented message. */
1911 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1912 sctp_chunk_hold(chunk);
1913
1914 /* Do accounting for the write space. */
1915 sctp_set_owner_w(chunk);
1916
1917 chunk->transport = chunk_tp;
1918 }
1919
1920 /* Send it to the lower layers. Note: all chunks
1921 * must either fail or succeed. The lower layer
1922 * works that way today. Keep it that way or this
1923 * breaks.
1924 */
1925 err = sctp_primitive_SEND(asoc, datamsg);
1926 /* Did the lower layer accept the chunk? */
1927 if (err)
1928 sctp_datamsg_free(datamsg);
1929 else
1930 sctp_datamsg_put(datamsg);
1931
1932 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1933
1934 if (err)
1935 goto out_free;
1936 else
1937 err = msg_len;
1938
1939 /* If we are already past ASSOCIATE, the lower
1940 * layers are responsible for association cleanup.
1941 */
1942 goto out_unlock;
1943
1944 out_free:
1945 if (new_asoc)
1946 sctp_association_free(asoc);
1947 out_unlock:
1948 sctp_release_sock(sk);
1949
1950 out_nounlock:
1951 return sctp_error(sk, msg_flags, err);
1952
1953 #if 0
1954 do_sock_err:
1955 if (msg_len)
1956 err = msg_len;
1957 else
1958 err = sock_error(sk);
1959 goto out;
1960
1961 do_interrupted:
1962 if (msg_len)
1963 err = msg_len;
1964 goto out;
1965 #endif /* 0 */
1966 }
1967
1968 /* This is an extended version of skb_pull() that removes the data from the
1969 * start of a skb even when data is spread across the list of skb's in the
1970 * frag_list. len specifies the total amount of data that needs to be removed.
1971 * when 'len' bytes could be removed from the skb, it returns 0.
1972 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1973 * could not be removed.
1974 */
1975 static int sctp_skb_pull(struct sk_buff *skb, int len)
1976 {
1977 struct sk_buff *list;
1978 int skb_len = skb_headlen(skb);
1979 int rlen;
1980
1981 if (len <= skb_len) {
1982 __skb_pull(skb, len);
1983 return 0;
1984 }
1985 len -= skb_len;
1986 __skb_pull(skb, skb_len);
1987
1988 skb_walk_frags(skb, list) {
1989 rlen = sctp_skb_pull(list, len);
1990 skb->len -= (len-rlen);
1991 skb->data_len -= (len-rlen);
1992
1993 if (!rlen)
1994 return 0;
1995
1996 len = rlen;
1997 }
1998
1999 return len;
2000 }
2001
2002 /* API 3.1.3 recvmsg() - UDP Style Syntax
2003 *
2004 * ssize_t recvmsg(int socket, struct msghdr *message,
2005 * int flags);
2006 *
2007 * socket - the socket descriptor of the endpoint.
2008 * message - pointer to the msghdr structure which contains a single
2009 * user message and possibly some ancillary data.
2010 *
2011 * See Section 5 for complete description of the data
2012 * structures.
2013 *
2014 * flags - flags sent or received with the user message, see Section
2015 * 5 for complete description of the flags.
2016 */
2017 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2018
2019 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2020 struct msghdr *msg, size_t len, int noblock,
2021 int flags, int *addr_len)
2022 {
2023 struct sctp_ulpevent *event = NULL;
2024 struct sctp_sock *sp = sctp_sk(sk);
2025 struct sk_buff *skb;
2026 int copied;
2027 int err = 0;
2028 int skb_len;
2029
2030 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
2031 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
2032 "len", len, "knoblauch", noblock,
2033 "flags", flags, "addr_len", addr_len);
2034
2035 sctp_lock_sock(sk);
2036
2037 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2038 err = -ENOTCONN;
2039 goto out;
2040 }
2041
2042 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2043 if (!skb)
2044 goto out;
2045
2046 /* Get the total length of the skb including any skb's in the
2047 * frag_list.
2048 */
2049 skb_len = skb->len;
2050
2051 copied = skb_len;
2052 if (copied > len)
2053 copied = len;
2054
2055 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2056
2057 event = sctp_skb2event(skb);
2058
2059 if (err)
2060 goto out_free;
2061
2062 sock_recv_ts_and_drops(msg, sk, skb);
2063 if (sctp_ulpevent_is_notification(event)) {
2064 msg->msg_flags |= MSG_NOTIFICATION;
2065 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2066 } else {
2067 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2068 }
2069
2070 /* Check if we allow SCTP_SNDRCVINFO. */
2071 if (sp->subscribe.sctp_data_io_event)
2072 sctp_ulpevent_read_sndrcvinfo(event, msg);
2073 #if 0
2074 /* FIXME: we should be calling IP/IPv6 layers. */
2075 if (sk->sk_protinfo.af_inet.cmsg_flags)
2076 ip_cmsg_recv(msg, skb);
2077 #endif
2078
2079 err = copied;
2080
2081 /* If skb's length exceeds the user's buffer, update the skb and
2082 * push it back to the receive_queue so that the next call to
2083 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2084 */
2085 if (skb_len > copied) {
2086 msg->msg_flags &= ~MSG_EOR;
2087 if (flags & MSG_PEEK)
2088 goto out_free;
2089 sctp_skb_pull(skb, copied);
2090 skb_queue_head(&sk->sk_receive_queue, skb);
2091
2092 /* When only partial message is copied to the user, increase
2093 * rwnd by that amount. If all the data in the skb is read,
2094 * rwnd is updated when the event is freed.
2095 */
2096 if (!sctp_ulpevent_is_notification(event))
2097 sctp_assoc_rwnd_increase(event->asoc, copied);
2098 goto out;
2099 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2100 (event->msg_flags & MSG_EOR))
2101 msg->msg_flags |= MSG_EOR;
2102 else
2103 msg->msg_flags &= ~MSG_EOR;
2104
2105 out_free:
2106 if (flags & MSG_PEEK) {
2107 /* Release the skb reference acquired after peeking the skb in
2108 * sctp_skb_recv_datagram().
2109 */
2110 kfree_skb(skb);
2111 } else {
2112 /* Free the event which includes releasing the reference to
2113 * the owner of the skb, freeing the skb and updating the
2114 * rwnd.
2115 */
2116 sctp_ulpevent_free(event);
2117 }
2118 out:
2119 sctp_release_sock(sk);
2120 return err;
2121 }
2122
2123 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2124 *
2125 * This option is a on/off flag. If enabled no SCTP message
2126 * fragmentation will be performed. Instead if a message being sent
2127 * exceeds the current PMTU size, the message will NOT be sent and
2128 * instead a error will be indicated to the user.
2129 */
2130 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2131 char __user *optval,
2132 unsigned int optlen)
2133 {
2134 int val;
2135
2136 if (optlen < sizeof(int))
2137 return -EINVAL;
2138
2139 if (get_user(val, (int __user *)optval))
2140 return -EFAULT;
2141
2142 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2143
2144 return 0;
2145 }
2146
2147 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2148 unsigned int optlen)
2149 {
2150 struct sctp_association *asoc;
2151 struct sctp_ulpevent *event;
2152
2153 if (optlen > sizeof(struct sctp_event_subscribe))
2154 return -EINVAL;
2155 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2156 return -EFAULT;
2157
2158 /*
2159 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2160 * if there is no data to be sent or retransmit, the stack will
2161 * immediately send up this notification.
2162 */
2163 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2164 &sctp_sk(sk)->subscribe)) {
2165 asoc = sctp_id2assoc(sk, 0);
2166
2167 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2168 event = sctp_ulpevent_make_sender_dry_event(asoc,
2169 GFP_ATOMIC);
2170 if (!event)
2171 return -ENOMEM;
2172
2173 sctp_ulpq_tail_event(&asoc->ulpq, event);
2174 }
2175 }
2176
2177 return 0;
2178 }
2179
2180 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2181 *
2182 * This socket option is applicable to the UDP-style socket only. When
2183 * set it will cause associations that are idle for more than the
2184 * specified number of seconds to automatically close. An association
2185 * being idle is defined an association that has NOT sent or received
2186 * user data. The special value of '0' indicates that no automatic
2187 * close of any associations should be performed. The option expects an
2188 * integer defining the number of seconds of idle time before an
2189 * association is closed.
2190 */
2191 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2192 unsigned int optlen)
2193 {
2194 struct sctp_sock *sp = sctp_sk(sk);
2195
2196 /* Applicable to UDP-style socket only */
2197 if (sctp_style(sk, TCP))
2198 return -EOPNOTSUPP;
2199 if (optlen != sizeof(int))
2200 return -EINVAL;
2201 if (copy_from_user(&sp->autoclose, optval, optlen))
2202 return -EFAULT;
2203
2204 return 0;
2205 }
2206
2207 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2208 *
2209 * Applications can enable or disable heartbeats for any peer address of
2210 * an association, modify an address's heartbeat interval, force a
2211 * heartbeat to be sent immediately, and adjust the address's maximum
2212 * number of retransmissions sent before an address is considered
2213 * unreachable. The following structure is used to access and modify an
2214 * address's parameters:
2215 *
2216 * struct sctp_paddrparams {
2217 * sctp_assoc_t spp_assoc_id;
2218 * struct sockaddr_storage spp_address;
2219 * uint32_t spp_hbinterval;
2220 * uint16_t spp_pathmaxrxt;
2221 * uint32_t spp_pathmtu;
2222 * uint32_t spp_sackdelay;
2223 * uint32_t spp_flags;
2224 * };
2225 *
2226 * spp_assoc_id - (one-to-many style socket) This is filled in the
2227 * application, and identifies the association for
2228 * this query.
2229 * spp_address - This specifies which address is of interest.
2230 * spp_hbinterval - This contains the value of the heartbeat interval,
2231 * in milliseconds. If a value of zero
2232 * is present in this field then no changes are to
2233 * be made to this parameter.
2234 * spp_pathmaxrxt - This contains the maximum number of
2235 * retransmissions before this address shall be
2236 * considered unreachable. If a value of zero
2237 * is present in this field then no changes are to
2238 * be made to this parameter.
2239 * spp_pathmtu - When Path MTU discovery is disabled the value
2240 * specified here will be the "fixed" path mtu.
2241 * Note that if the spp_address field is empty
2242 * then all associations on this address will
2243 * have this fixed path mtu set upon them.
2244 *
2245 * spp_sackdelay - When delayed sack is enabled, this value specifies
2246 * the number of milliseconds that sacks will be delayed
2247 * for. This value will apply to all addresses of an
2248 * association if the spp_address field is empty. Note
2249 * also, that if delayed sack is enabled and this
2250 * value is set to 0, no change is made to the last
2251 * recorded delayed sack timer value.
2252 *
2253 * spp_flags - These flags are used to control various features
2254 * on an association. The flag field may contain
2255 * zero or more of the following options.
2256 *
2257 * SPP_HB_ENABLE - Enable heartbeats on the
2258 * specified address. Note that if the address
2259 * field is empty all addresses for the association
2260 * have heartbeats enabled upon them.
2261 *
2262 * SPP_HB_DISABLE - Disable heartbeats on the
2263 * speicifed address. Note that if the address
2264 * field is empty all addresses for the association
2265 * will have their heartbeats disabled. Note also
2266 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2267 * mutually exclusive, only one of these two should
2268 * be specified. Enabling both fields will have
2269 * undetermined results.
2270 *
2271 * SPP_HB_DEMAND - Request a user initiated heartbeat
2272 * to be made immediately.
2273 *
2274 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2275 * heartbeat delayis to be set to the value of 0
2276 * milliseconds.
2277 *
2278 * SPP_PMTUD_ENABLE - This field will enable PMTU
2279 * discovery upon the specified address. Note that
2280 * if the address feild is empty then all addresses
2281 * on the association are effected.
2282 *
2283 * SPP_PMTUD_DISABLE - This field will disable PMTU
2284 * discovery upon the specified address. Note that
2285 * if the address feild is empty then all addresses
2286 * on the association are effected. Not also that
2287 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2288 * exclusive. Enabling both will have undetermined
2289 * results.
2290 *
2291 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2292 * on delayed sack. The time specified in spp_sackdelay
2293 * is used to specify the sack delay for this address. Note
2294 * that if spp_address is empty then all addresses will
2295 * enable delayed sack and take on the sack delay
2296 * value specified in spp_sackdelay.
2297 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2298 * off delayed sack. If the spp_address field is blank then
2299 * delayed sack is disabled for the entire association. Note
2300 * also that this field is mutually exclusive to
2301 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2302 * results.
2303 */
2304 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2305 struct sctp_transport *trans,
2306 struct sctp_association *asoc,
2307 struct sctp_sock *sp,
2308 int hb_change,
2309 int pmtud_change,
2310 int sackdelay_change)
2311 {
2312 int error;
2313
2314 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2315 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2316 if (error)
2317 return error;
2318 }
2319
2320 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2321 * this field is ignored. Note also that a value of zero indicates
2322 * the current setting should be left unchanged.
2323 */
2324 if (params->spp_flags & SPP_HB_ENABLE) {
2325
2326 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2327 * set. This lets us use 0 value when this flag
2328 * is set.
2329 */
2330 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2331 params->spp_hbinterval = 0;
2332
2333 if (params->spp_hbinterval ||
2334 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2335 if (trans) {
2336 trans->hbinterval =
2337 msecs_to_jiffies(params->spp_hbinterval);
2338 } else if (asoc) {
2339 asoc->hbinterval =
2340 msecs_to_jiffies(params->spp_hbinterval);
2341 } else {
2342 sp->hbinterval = params->spp_hbinterval;
2343 }
2344 }
2345 }
2346
2347 if (hb_change) {
2348 if (trans) {
2349 trans->param_flags =
2350 (trans->param_flags & ~SPP_HB) | hb_change;
2351 } else if (asoc) {
2352 asoc->param_flags =
2353 (asoc->param_flags & ~SPP_HB) | hb_change;
2354 } else {
2355 sp->param_flags =
2356 (sp->param_flags & ~SPP_HB) | hb_change;
2357 }
2358 }
2359
2360 /* When Path MTU discovery is disabled the value specified here will
2361 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2362 * include the flag SPP_PMTUD_DISABLE for this field to have any
2363 * effect).
2364 */
2365 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2366 if (trans) {
2367 trans->pathmtu = params->spp_pathmtu;
2368 sctp_assoc_sync_pmtu(asoc);
2369 } else if (asoc) {
2370 asoc->pathmtu = params->spp_pathmtu;
2371 sctp_frag_point(asoc, params->spp_pathmtu);
2372 } else {
2373 sp->pathmtu = params->spp_pathmtu;
2374 }
2375 }
2376
2377 if (pmtud_change) {
2378 if (trans) {
2379 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2380 (params->spp_flags & SPP_PMTUD_ENABLE);
2381 trans->param_flags =
2382 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2383 if (update) {
2384 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2385 sctp_assoc_sync_pmtu(asoc);
2386 }
2387 } else if (asoc) {
2388 asoc->param_flags =
2389 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2390 } else {
2391 sp->param_flags =
2392 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2393 }
2394 }
2395
2396 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2397 * value of this field is ignored. Note also that a value of zero
2398 * indicates the current setting should be left unchanged.
2399 */
2400 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2401 if (trans) {
2402 trans->sackdelay =
2403 msecs_to_jiffies(params->spp_sackdelay);
2404 } else if (asoc) {
2405 asoc->sackdelay =
2406 msecs_to_jiffies(params->spp_sackdelay);
2407 } else {
2408 sp->sackdelay = params->spp_sackdelay;
2409 }
2410 }
2411
2412 if (sackdelay_change) {
2413 if (trans) {
2414 trans->param_flags =
2415 (trans->param_flags & ~SPP_SACKDELAY) |
2416 sackdelay_change;
2417 } else if (asoc) {
2418 asoc->param_flags =
2419 (asoc->param_flags & ~SPP_SACKDELAY) |
2420 sackdelay_change;
2421 } else {
2422 sp->param_flags =
2423 (sp->param_flags & ~SPP_SACKDELAY) |
2424 sackdelay_change;
2425 }
2426 }
2427
2428 /* Note that a value of zero indicates the current setting should be
2429 left unchanged.
2430 */
2431 if (params->spp_pathmaxrxt) {
2432 if (trans) {
2433 trans->pathmaxrxt = params->spp_pathmaxrxt;
2434 } else if (asoc) {
2435 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2436 } else {
2437 sp->pathmaxrxt = params->spp_pathmaxrxt;
2438 }
2439 }
2440
2441 return 0;
2442 }
2443
2444 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2445 char __user *optval,
2446 unsigned int optlen)
2447 {
2448 struct sctp_paddrparams params;
2449 struct sctp_transport *trans = NULL;
2450 struct sctp_association *asoc = NULL;
2451 struct sctp_sock *sp = sctp_sk(sk);
2452 int error;
2453 int hb_change, pmtud_change, sackdelay_change;
2454
2455 if (optlen != sizeof(struct sctp_paddrparams))
2456 return - EINVAL;
2457
2458 if (copy_from_user(&params, optval, optlen))
2459 return -EFAULT;
2460
2461 /* Validate flags and value parameters. */
2462 hb_change = params.spp_flags & SPP_HB;
2463 pmtud_change = params.spp_flags & SPP_PMTUD;
2464 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2465
2466 if (hb_change == SPP_HB ||
2467 pmtud_change == SPP_PMTUD ||
2468 sackdelay_change == SPP_SACKDELAY ||
2469 params.spp_sackdelay > 500 ||
2470 (params.spp_pathmtu &&
2471 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2472 return -EINVAL;
2473
2474 /* If an address other than INADDR_ANY is specified, and
2475 * no transport is found, then the request is invalid.
2476 */
2477 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
2478 trans = sctp_addr_id2transport(sk, &params.spp_address,
2479 params.spp_assoc_id);
2480 if (!trans)
2481 return -EINVAL;
2482 }
2483
2484 /* Get association, if assoc_id != 0 and the socket is a one
2485 * to many style socket, and an association was not found, then
2486 * the id was invalid.
2487 */
2488 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2489 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2490 return -EINVAL;
2491
2492 /* Heartbeat demand can only be sent on a transport or
2493 * association, but not a socket.
2494 */
2495 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2496 return -EINVAL;
2497
2498 /* Process parameters. */
2499 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2500 hb_change, pmtud_change,
2501 sackdelay_change);
2502
2503 if (error)
2504 return error;
2505
2506 /* If changes are for association, also apply parameters to each
2507 * transport.
2508 */
2509 if (!trans && asoc) {
2510 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2511 transports) {
2512 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2513 hb_change, pmtud_change,
2514 sackdelay_change);
2515 }
2516 }
2517
2518 return 0;
2519 }
2520
2521 /*
2522 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2523 *
2524 * This option will effect the way delayed acks are performed. This
2525 * option allows you to get or set the delayed ack time, in
2526 * milliseconds. It also allows changing the delayed ack frequency.
2527 * Changing the frequency to 1 disables the delayed sack algorithm. If
2528 * the assoc_id is 0, then this sets or gets the endpoints default
2529 * values. If the assoc_id field is non-zero, then the set or get
2530 * effects the specified association for the one to many model (the
2531 * assoc_id field is ignored by the one to one model). Note that if
2532 * sack_delay or sack_freq are 0 when setting this option, then the
2533 * current values will remain unchanged.
2534 *
2535 * struct sctp_sack_info {
2536 * sctp_assoc_t sack_assoc_id;
2537 * uint32_t sack_delay;
2538 * uint32_t sack_freq;
2539 * };
2540 *
2541 * sack_assoc_id - This parameter, indicates which association the user
2542 * is performing an action upon. Note that if this field's value is
2543 * zero then the endpoints default value is changed (effecting future
2544 * associations only).
2545 *
2546 * sack_delay - This parameter contains the number of milliseconds that
2547 * the user is requesting the delayed ACK timer be set to. Note that
2548 * this value is defined in the standard to be between 200 and 500
2549 * milliseconds.
2550 *
2551 * sack_freq - This parameter contains the number of packets that must
2552 * be received before a sack is sent without waiting for the delay
2553 * timer to expire. The default value for this is 2, setting this
2554 * value to 1 will disable the delayed sack algorithm.
2555 */
2556
2557 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2558 char __user *optval, unsigned int optlen)
2559 {
2560 struct sctp_sack_info params;
2561 struct sctp_transport *trans = NULL;
2562 struct sctp_association *asoc = NULL;
2563 struct sctp_sock *sp = sctp_sk(sk);
2564
2565 if (optlen == sizeof(struct sctp_sack_info)) {
2566 if (copy_from_user(&params, optval, optlen))
2567 return -EFAULT;
2568
2569 if (params.sack_delay == 0 && params.sack_freq == 0)
2570 return 0;
2571 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2572 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2573 pr_warn("Use struct sctp_sack_info instead\n");
2574 if (copy_from_user(&params, optval, optlen))
2575 return -EFAULT;
2576
2577 if (params.sack_delay == 0)
2578 params.sack_freq = 1;
2579 else
2580 params.sack_freq = 0;
2581 } else
2582 return - EINVAL;
2583
2584 /* Validate value parameter. */
2585 if (params.sack_delay > 500)
2586 return -EINVAL;
2587
2588 /* Get association, if sack_assoc_id != 0 and the socket is a one
2589 * to many style socket, and an association was not found, then
2590 * the id was invalid.
2591 */
2592 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2593 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2594 return -EINVAL;
2595
2596 if (params.sack_delay) {
2597 if (asoc) {
2598 asoc->sackdelay =
2599 msecs_to_jiffies(params.sack_delay);
2600 asoc->param_flags =
2601 (asoc->param_flags & ~SPP_SACKDELAY) |
2602 SPP_SACKDELAY_ENABLE;
2603 } else {
2604 sp->sackdelay = params.sack_delay;
2605 sp->param_flags =
2606 (sp->param_flags & ~SPP_SACKDELAY) |
2607 SPP_SACKDELAY_ENABLE;
2608 }
2609 }
2610
2611 if (params.sack_freq == 1) {
2612 if (asoc) {
2613 asoc->param_flags =
2614 (asoc->param_flags & ~SPP_SACKDELAY) |
2615 SPP_SACKDELAY_DISABLE;
2616 } else {
2617 sp->param_flags =
2618 (sp->param_flags & ~SPP_SACKDELAY) |
2619 SPP_SACKDELAY_DISABLE;
2620 }
2621 } else if (params.sack_freq > 1) {
2622 if (asoc) {
2623 asoc->sackfreq = params.sack_freq;
2624 asoc->param_flags =
2625 (asoc->param_flags & ~SPP_SACKDELAY) |
2626 SPP_SACKDELAY_ENABLE;
2627 } else {
2628 sp->sackfreq = params.sack_freq;
2629 sp->param_flags =
2630 (sp->param_flags & ~SPP_SACKDELAY) |
2631 SPP_SACKDELAY_ENABLE;
2632 }
2633 }
2634
2635 /* If change is for association, also apply to each transport. */
2636 if (asoc) {
2637 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2638 transports) {
2639 if (params.sack_delay) {
2640 trans->sackdelay =
2641 msecs_to_jiffies(params.sack_delay);
2642 trans->param_flags =
2643 (trans->param_flags & ~SPP_SACKDELAY) |
2644 SPP_SACKDELAY_ENABLE;
2645 }
2646 if (params.sack_freq == 1) {
2647 trans->param_flags =
2648 (trans->param_flags & ~SPP_SACKDELAY) |
2649 SPP_SACKDELAY_DISABLE;
2650 } else if (params.sack_freq > 1) {
2651 trans->sackfreq = params.sack_freq;
2652 trans->param_flags =
2653 (trans->param_flags & ~SPP_SACKDELAY) |
2654 SPP_SACKDELAY_ENABLE;
2655 }
2656 }
2657 }
2658
2659 return 0;
2660 }
2661
2662 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2663 *
2664 * Applications can specify protocol parameters for the default association
2665 * initialization. The option name argument to setsockopt() and getsockopt()
2666 * is SCTP_INITMSG.
2667 *
2668 * Setting initialization parameters is effective only on an unconnected
2669 * socket (for UDP-style sockets only future associations are effected
2670 * by the change). With TCP-style sockets, this option is inherited by
2671 * sockets derived from a listener socket.
2672 */
2673 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2674 {
2675 struct sctp_initmsg sinit;
2676 struct sctp_sock *sp = sctp_sk(sk);
2677
2678 if (optlen != sizeof(struct sctp_initmsg))
2679 return -EINVAL;
2680 if (copy_from_user(&sinit, optval, optlen))
2681 return -EFAULT;
2682
2683 if (sinit.sinit_num_ostreams)
2684 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2685 if (sinit.sinit_max_instreams)
2686 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2687 if (sinit.sinit_max_attempts)
2688 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2689 if (sinit.sinit_max_init_timeo)
2690 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2691
2692 return 0;
2693 }
2694
2695 /*
2696 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2697 *
2698 * Applications that wish to use the sendto() system call may wish to
2699 * specify a default set of parameters that would normally be supplied
2700 * through the inclusion of ancillary data. This socket option allows
2701 * such an application to set the default sctp_sndrcvinfo structure.
2702 * The application that wishes to use this socket option simply passes
2703 * in to this call the sctp_sndrcvinfo structure defined in Section
2704 * 5.2.2) The input parameters accepted by this call include
2705 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2706 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2707 * to this call if the caller is using the UDP model.
2708 */
2709 static int sctp_setsockopt_default_send_param(struct sock *sk,
2710 char __user *optval,
2711 unsigned int optlen)
2712 {
2713 struct sctp_sndrcvinfo info;
2714 struct sctp_association *asoc;
2715 struct sctp_sock *sp = sctp_sk(sk);
2716
2717 if (optlen != sizeof(struct sctp_sndrcvinfo))
2718 return -EINVAL;
2719 if (copy_from_user(&info, optval, optlen))
2720 return -EFAULT;
2721
2722 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2723 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2724 return -EINVAL;
2725
2726 if (asoc) {
2727 asoc->default_stream = info.sinfo_stream;
2728 asoc->default_flags = info.sinfo_flags;
2729 asoc->default_ppid = info.sinfo_ppid;
2730 asoc->default_context = info.sinfo_context;
2731 asoc->default_timetolive = info.sinfo_timetolive;
2732 } else {
2733 sp->default_stream = info.sinfo_stream;
2734 sp->default_flags = info.sinfo_flags;
2735 sp->default_ppid = info.sinfo_ppid;
2736 sp->default_context = info.sinfo_context;
2737 sp->default_timetolive = info.sinfo_timetolive;
2738 }
2739
2740 return 0;
2741 }
2742
2743 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2744 *
2745 * Requests that the local SCTP stack use the enclosed peer address as
2746 * the association primary. The enclosed address must be one of the
2747 * association peer's addresses.
2748 */
2749 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2750 unsigned int optlen)
2751 {
2752 struct sctp_prim prim;
2753 struct sctp_transport *trans;
2754
2755 if (optlen != sizeof(struct sctp_prim))
2756 return -EINVAL;
2757
2758 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2759 return -EFAULT;
2760
2761 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2762 if (!trans)
2763 return -EINVAL;
2764
2765 sctp_assoc_set_primary(trans->asoc, trans);
2766
2767 return 0;
2768 }
2769
2770 /*
2771 * 7.1.5 SCTP_NODELAY
2772 *
2773 * Turn on/off any Nagle-like algorithm. This means that packets are
2774 * generally sent as soon as possible and no unnecessary delays are
2775 * introduced, at the cost of more packets in the network. Expects an
2776 * integer boolean flag.
2777 */
2778 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2779 unsigned int optlen)
2780 {
2781 int val;
2782
2783 if (optlen < sizeof(int))
2784 return -EINVAL;
2785 if (get_user(val, (int __user *)optval))
2786 return -EFAULT;
2787
2788 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2789 return 0;
2790 }
2791
2792 /*
2793 *
2794 * 7.1.1 SCTP_RTOINFO
2795 *
2796 * The protocol parameters used to initialize and bound retransmission
2797 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2798 * and modify these parameters.
2799 * All parameters are time values, in milliseconds. A value of 0, when
2800 * modifying the parameters, indicates that the current value should not
2801 * be changed.
2802 *
2803 */
2804 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2805 {
2806 struct sctp_rtoinfo rtoinfo;
2807 struct sctp_association *asoc;
2808
2809 if (optlen != sizeof (struct sctp_rtoinfo))
2810 return -EINVAL;
2811
2812 if (copy_from_user(&rtoinfo, optval, optlen))
2813 return -EFAULT;
2814
2815 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2816
2817 /* Set the values to the specific association */
2818 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2819 return -EINVAL;
2820
2821 if (asoc) {
2822 if (rtoinfo.srto_initial != 0)
2823 asoc->rto_initial =
2824 msecs_to_jiffies(rtoinfo.srto_initial);
2825 if (rtoinfo.srto_max != 0)
2826 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2827 if (rtoinfo.srto_min != 0)
2828 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2829 } else {
2830 /* If there is no association or the association-id = 0
2831 * set the values to the endpoint.
2832 */
2833 struct sctp_sock *sp = sctp_sk(sk);
2834
2835 if (rtoinfo.srto_initial != 0)
2836 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2837 if (rtoinfo.srto_max != 0)
2838 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2839 if (rtoinfo.srto_min != 0)
2840 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2841 }
2842
2843 return 0;
2844 }
2845
2846 /*
2847 *
2848 * 7.1.2 SCTP_ASSOCINFO
2849 *
2850 * This option is used to tune the maximum retransmission attempts
2851 * of the association.
2852 * Returns an error if the new association retransmission value is
2853 * greater than the sum of the retransmission value of the peer.
2854 * See [SCTP] for more information.
2855 *
2856 */
2857 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2858 {
2859
2860 struct sctp_assocparams assocparams;
2861 struct sctp_association *asoc;
2862
2863 if (optlen != sizeof(struct sctp_assocparams))
2864 return -EINVAL;
2865 if (copy_from_user(&assocparams, optval, optlen))
2866 return -EFAULT;
2867
2868 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2869
2870 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2871 return -EINVAL;
2872
2873 /* Set the values to the specific association */
2874 if (asoc) {
2875 if (assocparams.sasoc_asocmaxrxt != 0) {
2876 __u32 path_sum = 0;
2877 int paths = 0;
2878 struct sctp_transport *peer_addr;
2879
2880 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2881 transports) {
2882 path_sum += peer_addr->pathmaxrxt;
2883 paths++;
2884 }
2885
2886 /* Only validate asocmaxrxt if we have more than
2887 * one path/transport. We do this because path
2888 * retransmissions are only counted when we have more
2889 * then one path.
2890 */
2891 if (paths > 1 &&
2892 assocparams.sasoc_asocmaxrxt > path_sum)
2893 return -EINVAL;
2894
2895 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2896 }
2897
2898 if (assocparams.sasoc_cookie_life != 0) {
2899 asoc->cookie_life.tv_sec =
2900 assocparams.sasoc_cookie_life / 1000;
2901 asoc->cookie_life.tv_usec =
2902 (assocparams.sasoc_cookie_life % 1000)
2903 * 1000;
2904 }
2905 } else {
2906 /* Set the values to the endpoint */
2907 struct sctp_sock *sp = sctp_sk(sk);
2908
2909 if (assocparams.sasoc_asocmaxrxt != 0)
2910 sp->assocparams.sasoc_asocmaxrxt =
2911 assocparams.sasoc_asocmaxrxt;
2912 if (assocparams.sasoc_cookie_life != 0)
2913 sp->assocparams.sasoc_cookie_life =
2914 assocparams.sasoc_cookie_life;
2915 }
2916 return 0;
2917 }
2918
2919 /*
2920 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2921 *
2922 * This socket option is a boolean flag which turns on or off mapped V4
2923 * addresses. If this option is turned on and the socket is type
2924 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2925 * If this option is turned off, then no mapping will be done of V4
2926 * addresses and a user will receive both PF_INET6 and PF_INET type
2927 * addresses on the socket.
2928 */
2929 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2930 {
2931 int val;
2932 struct sctp_sock *sp = sctp_sk(sk);
2933
2934 if (optlen < sizeof(int))
2935 return -EINVAL;
2936 if (get_user(val, (int __user *)optval))
2937 return -EFAULT;
2938 if (val)
2939 sp->v4mapped = 1;
2940 else
2941 sp->v4mapped = 0;
2942
2943 return 0;
2944 }
2945
2946 /*
2947 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2948 * This option will get or set the maximum size to put in any outgoing
2949 * SCTP DATA chunk. If a message is larger than this size it will be
2950 * fragmented by SCTP into the specified size. Note that the underlying
2951 * SCTP implementation may fragment into smaller sized chunks when the
2952 * PMTU of the underlying association is smaller than the value set by
2953 * the user. The default value for this option is '0' which indicates
2954 * the user is NOT limiting fragmentation and only the PMTU will effect
2955 * SCTP's choice of DATA chunk size. Note also that values set larger
2956 * than the maximum size of an IP datagram will effectively let SCTP
2957 * control fragmentation (i.e. the same as setting this option to 0).
2958 *
2959 * The following structure is used to access and modify this parameter:
2960 *
2961 * struct sctp_assoc_value {
2962 * sctp_assoc_t assoc_id;
2963 * uint32_t assoc_value;
2964 * };
2965 *
2966 * assoc_id: This parameter is ignored for one-to-one style sockets.
2967 * For one-to-many style sockets this parameter indicates which
2968 * association the user is performing an action upon. Note that if
2969 * this field's value is zero then the endpoints default value is
2970 * changed (effecting future associations only).
2971 * assoc_value: This parameter specifies the maximum size in bytes.
2972 */
2973 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2974 {
2975 struct sctp_assoc_value params;
2976 struct sctp_association *asoc;
2977 struct sctp_sock *sp = sctp_sk(sk);
2978 int val;
2979
2980 if (optlen == sizeof(int)) {
2981 pr_warn("Use of int in maxseg socket option deprecated\n");
2982 pr_warn("Use struct sctp_assoc_value instead\n");
2983 if (copy_from_user(&val, optval, optlen))
2984 return -EFAULT;
2985 params.assoc_id = 0;
2986 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2987 if (copy_from_user(&params, optval, optlen))
2988 return -EFAULT;
2989 val = params.assoc_value;
2990 } else
2991 return -EINVAL;
2992
2993 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2994 return -EINVAL;
2995
2996 asoc = sctp_id2assoc(sk, params.assoc_id);
2997 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2998 return -EINVAL;
2999
3000 if (asoc) {
3001 if (val == 0) {
3002 val = asoc->pathmtu;
3003 val -= sp->pf->af->net_header_len;
3004 val -= sizeof(struct sctphdr) +
3005 sizeof(struct sctp_data_chunk);
3006 }
3007 asoc->user_frag = val;
3008 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3009 } else {
3010 sp->user_frag = val;
3011 }
3012
3013 return 0;
3014 }
3015
3016
3017 /*
3018 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3019 *
3020 * Requests that the peer mark the enclosed address as the association
3021 * primary. The enclosed address must be one of the association's
3022 * locally bound addresses. The following structure is used to make a
3023 * set primary request:
3024 */
3025 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3026 unsigned int optlen)
3027 {
3028 struct sctp_sock *sp;
3029 struct sctp_association *asoc = NULL;
3030 struct sctp_setpeerprim prim;
3031 struct sctp_chunk *chunk;
3032 struct sctp_af *af;
3033 int err;
3034
3035 sp = sctp_sk(sk);
3036
3037 if (!sctp_addip_enable)
3038 return -EPERM;
3039
3040 if (optlen != sizeof(struct sctp_setpeerprim))
3041 return -EINVAL;
3042
3043 if (copy_from_user(&prim, optval, optlen))
3044 return -EFAULT;
3045
3046 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3047 if (!asoc)
3048 return -EINVAL;
3049
3050 if (!asoc->peer.asconf_capable)
3051 return -EPERM;
3052
3053 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3054 return -EPERM;
3055
3056 if (!sctp_state(asoc, ESTABLISHED))
3057 return -ENOTCONN;
3058
3059 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3060 if (!af)
3061 return -EINVAL;
3062
3063 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3064 return -EADDRNOTAVAIL;
3065
3066 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3067 return -EADDRNOTAVAIL;
3068
3069 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3070 chunk = sctp_make_asconf_set_prim(asoc,
3071 (union sctp_addr *)&prim.sspp_addr);
3072 if (!chunk)
3073 return -ENOMEM;
3074
3075 err = sctp_send_asconf(asoc, chunk);
3076
3077 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
3078
3079 return err;
3080 }
3081
3082 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3083 unsigned int optlen)
3084 {
3085 struct sctp_setadaptation adaptation;
3086
3087 if (optlen != sizeof(struct sctp_setadaptation))
3088 return -EINVAL;
3089 if (copy_from_user(&adaptation, optval, optlen))
3090 return -EFAULT;
3091
3092 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3093
3094 return 0;
3095 }
3096
3097 /*
3098 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3099 *
3100 * The context field in the sctp_sndrcvinfo structure is normally only
3101 * used when a failed message is retrieved holding the value that was
3102 * sent down on the actual send call. This option allows the setting of
3103 * a default context on an association basis that will be received on
3104 * reading messages from the peer. This is especially helpful in the
3105 * one-2-many model for an application to keep some reference to an
3106 * internal state machine that is processing messages on the
3107 * association. Note that the setting of this value only effects
3108 * received messages from the peer and does not effect the value that is
3109 * saved with outbound messages.
3110 */
3111 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3112 unsigned int optlen)
3113 {
3114 struct sctp_assoc_value params;
3115 struct sctp_sock *sp;
3116 struct sctp_association *asoc;
3117
3118 if (optlen != sizeof(struct sctp_assoc_value))
3119 return -EINVAL;
3120 if (copy_from_user(&params, optval, optlen))
3121 return -EFAULT;
3122
3123 sp = sctp_sk(sk);
3124
3125 if (params.assoc_id != 0) {
3126 asoc = sctp_id2assoc(sk, params.assoc_id);
3127 if (!asoc)
3128 return -EINVAL;
3129 asoc->default_rcv_context = params.assoc_value;
3130 } else {
3131 sp->default_rcv_context = params.assoc_value;
3132 }
3133
3134 return 0;
3135 }
3136
3137 /*
3138 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3139 *
3140 * This options will at a minimum specify if the implementation is doing
3141 * fragmented interleave. Fragmented interleave, for a one to many
3142 * socket, is when subsequent calls to receive a message may return
3143 * parts of messages from different associations. Some implementations
3144 * may allow you to turn this value on or off. If so, when turned off,
3145 * no fragment interleave will occur (which will cause a head of line
3146 * blocking amongst multiple associations sharing the same one to many
3147 * socket). When this option is turned on, then each receive call may
3148 * come from a different association (thus the user must receive data
3149 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3150 * association each receive belongs to.
3151 *
3152 * This option takes a boolean value. A non-zero value indicates that
3153 * fragmented interleave is on. A value of zero indicates that
3154 * fragmented interleave is off.
3155 *
3156 * Note that it is important that an implementation that allows this
3157 * option to be turned on, have it off by default. Otherwise an unaware
3158 * application using the one to many model may become confused and act
3159 * incorrectly.
3160 */
3161 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3162 char __user *optval,
3163 unsigned int optlen)
3164 {
3165 int val;
3166
3167 if (optlen != sizeof(int))
3168 return -EINVAL;
3169 if (get_user(val, (int __user *)optval))
3170 return -EFAULT;
3171
3172 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3173
3174 return 0;
3175 }
3176
3177 /*
3178 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3179 * (SCTP_PARTIAL_DELIVERY_POINT)
3180 *
3181 * This option will set or get the SCTP partial delivery point. This
3182 * point is the size of a message where the partial delivery API will be
3183 * invoked to help free up rwnd space for the peer. Setting this to a
3184 * lower value will cause partial deliveries to happen more often. The
3185 * calls argument is an integer that sets or gets the partial delivery
3186 * point. Note also that the call will fail if the user attempts to set
3187 * this value larger than the socket receive buffer size.
3188 *
3189 * Note that any single message having a length smaller than or equal to
3190 * the SCTP partial delivery point will be delivered in one single read
3191 * call as long as the user provided buffer is large enough to hold the
3192 * message.
3193 */
3194 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3195 char __user *optval,
3196 unsigned int optlen)
3197 {
3198 u32 val;
3199
3200 if (optlen != sizeof(u32))
3201 return -EINVAL;
3202 if (get_user(val, (int __user *)optval))
3203 return -EFAULT;
3204
3205 /* Note: We double the receive buffer from what the user sets
3206 * it to be, also initial rwnd is based on rcvbuf/2.
3207 */
3208 if (val > (sk->sk_rcvbuf >> 1))
3209 return -EINVAL;
3210
3211 sctp_sk(sk)->pd_point = val;
3212
3213 return 0; /* is this the right error code? */
3214 }
3215
3216 /*
3217 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3218 *
3219 * This option will allow a user to change the maximum burst of packets
3220 * that can be emitted by this association. Note that the default value
3221 * is 4, and some implementations may restrict this setting so that it
3222 * can only be lowered.
3223 *
3224 * NOTE: This text doesn't seem right. Do this on a socket basis with
3225 * future associations inheriting the socket value.
3226 */
3227 static int sctp_setsockopt_maxburst(struct sock *sk,
3228 char __user *optval,
3229 unsigned int optlen)
3230 {
3231 struct sctp_assoc_value params;
3232 struct sctp_sock *sp;
3233 struct sctp_association *asoc;
3234 int val;
3235 int assoc_id = 0;
3236
3237 if (optlen == sizeof(int)) {
3238 pr_warn("Use of int in max_burst socket option deprecated\n");
3239 pr_warn("Use struct sctp_assoc_value instead\n");
3240 if (copy_from_user(&val, optval, optlen))
3241 return -EFAULT;
3242 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3243 if (copy_from_user(&params, optval, optlen))
3244 return -EFAULT;
3245 val = params.assoc_value;
3246 assoc_id = params.assoc_id;
3247 } else
3248 return -EINVAL;
3249
3250 sp = sctp_sk(sk);
3251
3252 if (assoc_id != 0) {
3253 asoc = sctp_id2assoc(sk, assoc_id);
3254 if (!asoc)
3255 return -EINVAL;
3256 asoc->max_burst = val;
3257 } else
3258 sp->max_burst = val;
3259
3260 return 0;
3261 }
3262
3263 /*
3264 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3265 *
3266 * This set option adds a chunk type that the user is requesting to be
3267 * received only in an authenticated way. Changes to the list of chunks
3268 * will only effect future associations on the socket.
3269 */
3270 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3271 char __user *optval,
3272 unsigned int optlen)
3273 {
3274 struct sctp_authchunk val;
3275
3276 if (!sctp_auth_enable)
3277 return -EACCES;
3278
3279 if (optlen != sizeof(struct sctp_authchunk))
3280 return -EINVAL;
3281 if (copy_from_user(&val, optval, optlen))
3282 return -EFAULT;
3283
3284 switch (val.sauth_chunk) {
3285 case SCTP_CID_INIT:
3286 case SCTP_CID_INIT_ACK:
3287 case SCTP_CID_SHUTDOWN_COMPLETE:
3288 case SCTP_CID_AUTH:
3289 return -EINVAL;
3290 }
3291
3292 /* add this chunk id to the endpoint */
3293 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3294 }
3295
3296 /*
3297 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3298 *
3299 * This option gets or sets the list of HMAC algorithms that the local
3300 * endpoint requires the peer to use.
3301 */
3302 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3303 char __user *optval,
3304 unsigned int optlen)
3305 {
3306 struct sctp_hmacalgo *hmacs;
3307 u32 idents;
3308 int err;
3309
3310 if (!sctp_auth_enable)
3311 return -EACCES;
3312
3313 if (optlen < sizeof(struct sctp_hmacalgo))
3314 return -EINVAL;
3315
3316 hmacs= memdup_user(optval, optlen);
3317 if (IS_ERR(hmacs))
3318 return PTR_ERR(hmacs);
3319
3320 idents = hmacs->shmac_num_idents;
3321 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3322 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3323 err = -EINVAL;
3324 goto out;
3325 }
3326
3327 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3328 out:
3329 kfree(hmacs);
3330 return err;
3331 }
3332
3333 /*
3334 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3335 *
3336 * This option will set a shared secret key which is used to build an
3337 * association shared key.
3338 */
3339 static int sctp_setsockopt_auth_key(struct sock *sk,
3340 char __user *optval,
3341 unsigned int optlen)
3342 {
3343 struct sctp_authkey *authkey;
3344 struct sctp_association *asoc;
3345 int ret;
3346
3347 if (!sctp_auth_enable)
3348 return -EACCES;
3349
3350 if (optlen <= sizeof(struct sctp_authkey))
3351 return -EINVAL;
3352
3353 authkey= memdup_user(optval, optlen);
3354 if (IS_ERR(authkey))
3355 return PTR_ERR(authkey);
3356
3357 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3358 ret = -EINVAL;
3359 goto out;
3360 }
3361
3362 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3363 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3364 ret = -EINVAL;
3365 goto out;
3366 }
3367
3368 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3369 out:
3370 kfree(authkey);
3371 return ret;
3372 }
3373
3374 /*
3375 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3376 *
3377 * This option will get or set the active shared key to be used to build
3378 * the association shared key.
3379 */
3380 static int sctp_setsockopt_active_key(struct sock *sk,
3381 char __user *optval,
3382 unsigned int optlen)
3383 {
3384 struct sctp_authkeyid val;
3385 struct sctp_association *asoc;
3386
3387 if (!sctp_auth_enable)
3388 return -EACCES;
3389
3390 if (optlen != sizeof(struct sctp_authkeyid))
3391 return -EINVAL;
3392 if (copy_from_user(&val, optval, optlen))
3393 return -EFAULT;
3394
3395 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3396 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3397 return -EINVAL;
3398
3399 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3400 val.scact_keynumber);
3401 }
3402
3403 /*
3404 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3405 *
3406 * This set option will delete a shared secret key from use.
3407 */
3408 static int sctp_setsockopt_del_key(struct sock *sk,
3409 char __user *optval,
3410 unsigned int optlen)
3411 {
3412 struct sctp_authkeyid val;
3413 struct sctp_association *asoc;
3414
3415 if (!sctp_auth_enable)
3416 return -EACCES;
3417
3418 if (optlen != sizeof(struct sctp_authkeyid))
3419 return -EINVAL;
3420 if (copy_from_user(&val, optval, optlen))
3421 return -EFAULT;
3422
3423 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3424 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3425 return -EINVAL;
3426
3427 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3428 val.scact_keynumber);
3429
3430 }
3431
3432 /*
3433 * 8.1.23 SCTP_AUTO_ASCONF
3434 *
3435 * This option will enable or disable the use of the automatic generation of
3436 * ASCONF chunks to add and delete addresses to an existing association. Note
3437 * that this option has two caveats namely: a) it only affects sockets that
3438 * are bound to all addresses available to the SCTP stack, and b) the system
3439 * administrator may have an overriding control that turns the ASCONF feature
3440 * off no matter what setting the socket option may have.
3441 * This option expects an integer boolean flag, where a non-zero value turns on
3442 * the option, and a zero value turns off the option.
3443 * Note. In this implementation, socket operation overrides default parameter
3444 * being set by sysctl as well as FreeBSD implementation
3445 */
3446 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3447 unsigned int optlen)
3448 {
3449 int val;
3450 struct sctp_sock *sp = sctp_sk(sk);
3451
3452 if (optlen < sizeof(int))
3453 return -EINVAL;
3454 if (get_user(val, (int __user *)optval))
3455 return -EFAULT;
3456 if (!sctp_is_ep_boundall(sk) && val)
3457 return -EINVAL;
3458 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3459 return 0;
3460
3461 if (val == 0 && sp->do_auto_asconf) {
3462 list_del(&sp->auto_asconf_list);
3463 sp->do_auto_asconf = 0;
3464 } else if (val && !sp->do_auto_asconf) {
3465 list_add_tail(&sp->auto_asconf_list,
3466 &sctp_auto_asconf_splist);
3467 sp->do_auto_asconf = 1;
3468 }
3469 return 0;
3470 }
3471
3472
3473 /* API 6.2 setsockopt(), getsockopt()
3474 *
3475 * Applications use setsockopt() and getsockopt() to set or retrieve
3476 * socket options. Socket options are used to change the default
3477 * behavior of sockets calls. They are described in Section 7.
3478 *
3479 * The syntax is:
3480 *
3481 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3482 * int __user *optlen);
3483 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3484 * int optlen);
3485 *
3486 * sd - the socket descript.
3487 * level - set to IPPROTO_SCTP for all SCTP options.
3488 * optname - the option name.
3489 * optval - the buffer to store the value of the option.
3490 * optlen - the size of the buffer.
3491 */
3492 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3493 char __user *optval, unsigned int optlen)
3494 {
3495 int retval = 0;
3496
3497 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3498 sk, optname);
3499
3500 /* I can hardly begin to describe how wrong this is. This is
3501 * so broken as to be worse than useless. The API draft
3502 * REALLY is NOT helpful here... I am not convinced that the
3503 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3504 * are at all well-founded.
3505 */
3506 if (level != SOL_SCTP) {
3507 struct sctp_af *af = sctp_sk(sk)->pf->af;
3508 retval = af->setsockopt(sk, level, optname, optval, optlen);
3509 goto out_nounlock;
3510 }
3511
3512 sctp_lock_sock(sk);
3513
3514 switch (optname) {
3515 case SCTP_SOCKOPT_BINDX_ADD:
3516 /* 'optlen' is the size of the addresses buffer. */
3517 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3518 optlen, SCTP_BINDX_ADD_ADDR);
3519 break;
3520
3521 case SCTP_SOCKOPT_BINDX_REM:
3522 /* 'optlen' is the size of the addresses buffer. */
3523 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3524 optlen, SCTP_BINDX_REM_ADDR);
3525 break;
3526
3527 case SCTP_SOCKOPT_CONNECTX_OLD:
3528 /* 'optlen' is the size of the addresses buffer. */
3529 retval = sctp_setsockopt_connectx_old(sk,
3530 (struct sockaddr __user *)optval,
3531 optlen);
3532 break;
3533
3534 case SCTP_SOCKOPT_CONNECTX:
3535 /* 'optlen' is the size of the addresses buffer. */
3536 retval = sctp_setsockopt_connectx(sk,
3537 (struct sockaddr __user *)optval,
3538 optlen);
3539 break;
3540
3541 case SCTP_DISABLE_FRAGMENTS:
3542 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3543 break;
3544
3545 case SCTP_EVENTS:
3546 retval = sctp_setsockopt_events(sk, optval, optlen);
3547 break;
3548
3549 case SCTP_AUTOCLOSE:
3550 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3551 break;
3552
3553 case SCTP_PEER_ADDR_PARAMS:
3554 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3555 break;
3556
3557 case SCTP_DELAYED_SACK:
3558 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3559 break;
3560 case SCTP_PARTIAL_DELIVERY_POINT:
3561 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3562 break;
3563
3564 case SCTP_INITMSG:
3565 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3566 break;
3567 case SCTP_DEFAULT_SEND_PARAM:
3568 retval = sctp_setsockopt_default_send_param(sk, optval,
3569 optlen);
3570 break;
3571 case SCTP_PRIMARY_ADDR:
3572 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3573 break;
3574 case SCTP_SET_PEER_PRIMARY_ADDR:
3575 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3576 break;
3577 case SCTP_NODELAY:
3578 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3579 break;
3580 case SCTP_RTOINFO:
3581 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3582 break;
3583 case SCTP_ASSOCINFO:
3584 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3585 break;
3586 case SCTP_I_WANT_MAPPED_V4_ADDR:
3587 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3588 break;
3589 case SCTP_MAXSEG:
3590 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3591 break;
3592 case SCTP_ADAPTATION_LAYER:
3593 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3594 break;
3595 case SCTP_CONTEXT:
3596 retval = sctp_setsockopt_context(sk, optval, optlen);
3597 break;
3598 case SCTP_FRAGMENT_INTERLEAVE:
3599 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3600 break;
3601 case SCTP_MAX_BURST:
3602 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3603 break;
3604 case SCTP_AUTH_CHUNK:
3605 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3606 break;
3607 case SCTP_HMAC_IDENT:
3608 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3609 break;
3610 case SCTP_AUTH_KEY:
3611 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3612 break;
3613 case SCTP_AUTH_ACTIVE_KEY:
3614 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3615 break;
3616 case SCTP_AUTH_DELETE_KEY:
3617 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3618 break;
3619 case SCTP_AUTO_ASCONF:
3620 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3621 break;
3622 default:
3623 retval = -ENOPROTOOPT;
3624 break;
3625 }
3626
3627 sctp_release_sock(sk);
3628
3629 out_nounlock:
3630 return retval;
3631 }
3632
3633 /* API 3.1.6 connect() - UDP Style Syntax
3634 *
3635 * An application may use the connect() call in the UDP model to initiate an
3636 * association without sending data.
3637 *
3638 * The syntax is:
3639 *
3640 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3641 *
3642 * sd: the socket descriptor to have a new association added to.
3643 *
3644 * nam: the address structure (either struct sockaddr_in or struct
3645 * sockaddr_in6 defined in RFC2553 [7]).
3646 *
3647 * len: the size of the address.
3648 */
3649 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3650 int addr_len)
3651 {
3652 int err = 0;
3653 struct sctp_af *af;
3654
3655 sctp_lock_sock(sk);
3656
3657 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3658 __func__, sk, addr, addr_len);
3659
3660 /* Validate addr_len before calling common connect/connectx routine. */
3661 af = sctp_get_af_specific(addr->sa_family);
3662 if (!af || addr_len < af->sockaddr_len) {
3663 err = -EINVAL;
3664 } else {
3665 /* Pass correct addr len to common routine (so it knows there
3666 * is only one address being passed.
3667 */
3668 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3669 }
3670
3671 sctp_release_sock(sk);
3672 return err;
3673 }
3674
3675 /* FIXME: Write comments. */
3676 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3677 {
3678 return -EOPNOTSUPP; /* STUB */
3679 }
3680
3681 /* 4.1.4 accept() - TCP Style Syntax
3682 *
3683 * Applications use accept() call to remove an established SCTP
3684 * association from the accept queue of the endpoint. A new socket
3685 * descriptor will be returned from accept() to represent the newly
3686 * formed association.
3687 */
3688 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3689 {
3690 struct sctp_sock *sp;
3691 struct sctp_endpoint *ep;
3692 struct sock *newsk = NULL;
3693 struct sctp_association *asoc;
3694 long timeo;
3695 int error = 0;
3696
3697 sctp_lock_sock(sk);
3698
3699 sp = sctp_sk(sk);
3700 ep = sp->ep;
3701
3702 if (!sctp_style(sk, TCP)) {
3703 error = -EOPNOTSUPP;
3704 goto out;
3705 }
3706
3707 if (!sctp_sstate(sk, LISTENING)) {
3708 error = -EINVAL;
3709 goto out;
3710 }
3711
3712 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3713
3714 error = sctp_wait_for_accept(sk, timeo);
3715 if (error)
3716 goto out;
3717
3718 /* We treat the list of associations on the endpoint as the accept
3719 * queue and pick the first association on the list.
3720 */
3721 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3722
3723 newsk = sp->pf->create_accept_sk(sk, asoc);
3724 if (!newsk) {
3725 error = -ENOMEM;
3726 goto out;
3727 }
3728
3729 /* Populate the fields of the newsk from the oldsk and migrate the
3730 * asoc to the newsk.
3731 */
3732 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3733
3734 out:
3735 sctp_release_sock(sk);
3736 *err = error;
3737 return newsk;
3738 }
3739
3740 /* The SCTP ioctl handler. */
3741 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3742 {
3743 int rc = -ENOTCONN;
3744
3745 sctp_lock_sock(sk);
3746
3747 /*
3748 * SEQPACKET-style sockets in LISTENING state are valid, for
3749 * SCTP, so only discard TCP-style sockets in LISTENING state.
3750 */
3751 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3752 goto out;
3753
3754 switch (cmd) {
3755 case SIOCINQ: {
3756 struct sk_buff *skb;
3757 unsigned int amount = 0;
3758
3759 skb = skb_peek(&sk->sk_receive_queue);
3760 if (skb != NULL) {
3761 /*
3762 * We will only return the amount of this packet since
3763 * that is all that will be read.
3764 */
3765 amount = skb->len;
3766 }
3767 rc = put_user(amount, (int __user *)arg);
3768 break;
3769 }
3770 default:
3771 rc = -ENOIOCTLCMD;
3772 break;
3773 }
3774 out:
3775 sctp_release_sock(sk);
3776 return rc;
3777 }
3778
3779 /* This is the function which gets called during socket creation to
3780 * initialized the SCTP-specific portion of the sock.
3781 * The sock structure should already be zero-filled memory.
3782 */
3783 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3784 {
3785 struct sctp_endpoint *ep;
3786 struct sctp_sock *sp;
3787
3788 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3789
3790 sp = sctp_sk(sk);
3791
3792 /* Initialize the SCTP per socket area. */
3793 switch (sk->sk_type) {
3794 case SOCK_SEQPACKET:
3795 sp->type = SCTP_SOCKET_UDP;
3796 break;
3797 case SOCK_STREAM:
3798 sp->type = SCTP_SOCKET_TCP;
3799 break;
3800 default:
3801 return -ESOCKTNOSUPPORT;
3802 }
3803
3804 /* Initialize default send parameters. These parameters can be
3805 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3806 */
3807 sp->default_stream = 0;
3808 sp->default_ppid = 0;
3809 sp->default_flags = 0;
3810 sp->default_context = 0;
3811 sp->default_timetolive = 0;
3812
3813 sp->default_rcv_context = 0;
3814 sp->max_burst = sctp_max_burst;
3815
3816 /* Initialize default setup parameters. These parameters
3817 * can be modified with the SCTP_INITMSG socket option or
3818 * overridden by the SCTP_INIT CMSG.
3819 */
3820 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3821 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3822 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3823 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3824
3825 /* Initialize default RTO related parameters. These parameters can
3826 * be modified for with the SCTP_RTOINFO socket option.
3827 */
3828 sp->rtoinfo.srto_initial = sctp_rto_initial;
3829 sp->rtoinfo.srto_max = sctp_rto_max;
3830 sp->rtoinfo.srto_min = sctp_rto_min;
3831
3832 /* Initialize default association related parameters. These parameters
3833 * can be modified with the SCTP_ASSOCINFO socket option.
3834 */
3835 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3836 sp->assocparams.sasoc_number_peer_destinations = 0;
3837 sp->assocparams.sasoc_peer_rwnd = 0;
3838 sp->assocparams.sasoc_local_rwnd = 0;
3839 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3840
3841 /* Initialize default event subscriptions. By default, all the
3842 * options are off.
3843 */
3844 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3845
3846 /* Default Peer Address Parameters. These defaults can
3847 * be modified via SCTP_PEER_ADDR_PARAMS
3848 */
3849 sp->hbinterval = sctp_hb_interval;
3850 sp->pathmaxrxt = sctp_max_retrans_path;
3851 sp->pathmtu = 0; // allow default discovery
3852 sp->sackdelay = sctp_sack_timeout;
3853 sp->sackfreq = 2;
3854 sp->param_flags = SPP_HB_ENABLE |
3855 SPP_PMTUD_ENABLE |
3856 SPP_SACKDELAY_ENABLE;
3857
3858 /* If enabled no SCTP message fragmentation will be performed.
3859 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3860 */
3861 sp->disable_fragments = 0;
3862
3863 /* Enable Nagle algorithm by default. */
3864 sp->nodelay = 0;
3865
3866 /* Enable by default. */
3867 sp->v4mapped = 1;
3868
3869 /* Auto-close idle associations after the configured
3870 * number of seconds. A value of 0 disables this
3871 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3872 * for UDP-style sockets only.
3873 */
3874 sp->autoclose = 0;
3875
3876 /* User specified fragmentation limit. */
3877 sp->user_frag = 0;
3878
3879 sp->adaptation_ind = 0;
3880
3881 sp->pf = sctp_get_pf_specific(sk->sk_family);
3882
3883 /* Control variables for partial data delivery. */
3884 atomic_set(&sp->pd_mode, 0);
3885 skb_queue_head_init(&sp->pd_lobby);
3886 sp->frag_interleave = 0;
3887
3888 /* Create a per socket endpoint structure. Even if we
3889 * change the data structure relationships, this may still
3890 * be useful for storing pre-connect address information.
3891 */
3892 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3893 if (!ep)
3894 return -ENOMEM;
3895
3896 sp->ep = ep;
3897 sp->hmac = NULL;
3898
3899 SCTP_DBG_OBJCNT_INC(sock);
3900
3901 local_bh_disable();
3902 percpu_counter_inc(&sctp_sockets_allocated);
3903 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
3904 if (sctp_default_auto_asconf) {
3905 list_add_tail(&sp->auto_asconf_list,
3906 &sctp_auto_asconf_splist);
3907 sp->do_auto_asconf = 1;
3908 } else
3909 sp->do_auto_asconf = 0;
3910 local_bh_enable();
3911
3912 return 0;
3913 }
3914
3915 /* Cleanup any SCTP per socket resources. */
3916 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
3917 {
3918 struct sctp_sock *sp;
3919
3920 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3921
3922 /* Release our hold on the endpoint. */
3923 sp = sctp_sk(sk);
3924 if (sp->do_auto_asconf) {
3925 sp->do_auto_asconf = 0;
3926 list_del(&sp->auto_asconf_list);
3927 }
3928 sctp_endpoint_free(sp->ep);
3929 local_bh_disable();
3930 percpu_counter_dec(&sctp_sockets_allocated);
3931 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
3932 local_bh_enable();
3933 }
3934
3935 /* API 4.1.7 shutdown() - TCP Style Syntax
3936 * int shutdown(int socket, int how);
3937 *
3938 * sd - the socket descriptor of the association to be closed.
3939 * how - Specifies the type of shutdown. The values are
3940 * as follows:
3941 * SHUT_RD
3942 * Disables further receive operations. No SCTP
3943 * protocol action is taken.
3944 * SHUT_WR
3945 * Disables further send operations, and initiates
3946 * the SCTP shutdown sequence.
3947 * SHUT_RDWR
3948 * Disables further send and receive operations
3949 * and initiates the SCTP shutdown sequence.
3950 */
3951 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3952 {
3953 struct sctp_endpoint *ep;
3954 struct sctp_association *asoc;
3955
3956 if (!sctp_style(sk, TCP))
3957 return;
3958
3959 if (how & SEND_SHUTDOWN) {
3960 ep = sctp_sk(sk)->ep;
3961 if (!list_empty(&ep->asocs)) {
3962 asoc = list_entry(ep->asocs.next,
3963 struct sctp_association, asocs);
3964 sctp_primitive_SHUTDOWN(asoc, NULL);
3965 }
3966 }
3967 }
3968
3969 /* 7.2.1 Association Status (SCTP_STATUS)
3970
3971 * Applications can retrieve current status information about an
3972 * association, including association state, peer receiver window size,
3973 * number of unacked data chunks, and number of data chunks pending
3974 * receipt. This information is read-only.
3975 */
3976 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3977 char __user *optval,
3978 int __user *optlen)
3979 {
3980 struct sctp_status status;
3981 struct sctp_association *asoc = NULL;
3982 struct sctp_transport *transport;
3983 sctp_assoc_t associd;
3984 int retval = 0;
3985
3986 if (len < sizeof(status)) {
3987 retval = -EINVAL;
3988 goto out;
3989 }
3990
3991 len = sizeof(status);
3992 if (copy_from_user(&status, optval, len)) {
3993 retval = -EFAULT;
3994 goto out;
3995 }
3996
3997 associd = status.sstat_assoc_id;
3998 asoc = sctp_id2assoc(sk, associd);
3999 if (!asoc) {
4000 retval = -EINVAL;
4001 goto out;
4002 }
4003
4004 transport = asoc->peer.primary_path;
4005
4006 status.sstat_assoc_id = sctp_assoc2id(asoc);
4007 status.sstat_state = asoc->state;
4008 status.sstat_rwnd = asoc->peer.rwnd;
4009 status.sstat_unackdata = asoc->unack_data;
4010
4011 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4012 status.sstat_instrms = asoc->c.sinit_max_instreams;
4013 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4014 status.sstat_fragmentation_point = asoc->frag_point;
4015 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4016 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4017 transport->af_specific->sockaddr_len);
4018 /* Map ipv4 address into v4-mapped-on-v6 address. */
4019 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4020 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4021 status.sstat_primary.spinfo_state = transport->state;
4022 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4023 status.sstat_primary.spinfo_srtt = transport->srtt;
4024 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4025 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4026
4027 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4028 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4029
4030 if (put_user(len, optlen)) {
4031 retval = -EFAULT;
4032 goto out;
4033 }
4034
4035 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
4036 len, status.sstat_state, status.sstat_rwnd,
4037 status.sstat_assoc_id);
4038
4039 if (copy_to_user(optval, &status, len)) {
4040 retval = -EFAULT;
4041 goto out;
4042 }
4043
4044 out:
4045 return retval;
4046 }
4047
4048
4049 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4050 *
4051 * Applications can retrieve information about a specific peer address
4052 * of an association, including its reachability state, congestion
4053 * window, and retransmission timer values. This information is
4054 * read-only.
4055 */
4056 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4057 char __user *optval,
4058 int __user *optlen)
4059 {
4060 struct sctp_paddrinfo pinfo;
4061 struct sctp_transport *transport;
4062 int retval = 0;
4063
4064 if (len < sizeof(pinfo)) {
4065 retval = -EINVAL;
4066 goto out;
4067 }
4068
4069 len = sizeof(pinfo);
4070 if (copy_from_user(&pinfo, optval, len)) {
4071 retval = -EFAULT;
4072 goto out;
4073 }
4074
4075 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4076 pinfo.spinfo_assoc_id);
4077 if (!transport)
4078 return -EINVAL;
4079
4080 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4081 pinfo.spinfo_state = transport->state;
4082 pinfo.spinfo_cwnd = transport->cwnd;
4083 pinfo.spinfo_srtt = transport->srtt;
4084 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4085 pinfo.spinfo_mtu = transport->pathmtu;
4086
4087 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4088 pinfo.spinfo_state = SCTP_ACTIVE;
4089
4090 if (put_user(len, optlen)) {
4091 retval = -EFAULT;
4092 goto out;
4093 }
4094
4095 if (copy_to_user(optval, &pinfo, len)) {
4096 retval = -EFAULT;
4097 goto out;
4098 }
4099
4100 out:
4101 return retval;
4102 }
4103
4104 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4105 *
4106 * This option is a on/off flag. If enabled no SCTP message
4107 * fragmentation will be performed. Instead if a message being sent
4108 * exceeds the current PMTU size, the message will NOT be sent and
4109 * instead a error will be indicated to the user.
4110 */
4111 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4112 char __user *optval, int __user *optlen)
4113 {
4114 int val;
4115
4116 if (len < sizeof(int))
4117 return -EINVAL;
4118
4119 len = sizeof(int);
4120 val = (sctp_sk(sk)->disable_fragments == 1);
4121 if (put_user(len, optlen))
4122 return -EFAULT;
4123 if (copy_to_user(optval, &val, len))
4124 return -EFAULT;
4125 return 0;
4126 }
4127
4128 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4129 *
4130 * This socket option is used to specify various notifications and
4131 * ancillary data the user wishes to receive.
4132 */
4133 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4134 int __user *optlen)
4135 {
4136 if (len <= 0)
4137 return -EINVAL;
4138 if (len > sizeof(struct sctp_event_subscribe))
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 */
4158 static 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. */
4174 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4175 {
4176 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4177 struct socket *sock;
4178 struct sctp_af *af;
4179 int err = 0;
4180
4181 if (!asoc)
4182 return -EINVAL;
4183
4184 /* An association cannot be branched off from an already peeled-off
4185 * socket, nor is this supported for tcp style sockets.
4186 */
4187 if (!sctp_style(sk, UDP))
4188 return -EINVAL;
4189
4190 /* Create a new socket. */
4191 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4192 if (err < 0)
4193 return err;
4194
4195 sctp_copy_sock(sock->sk, sk, asoc);
4196
4197 /* Make peeled-off sockets more like 1-1 accepted sockets.
4198 * Set the daddr and initialize id to something more random
4199 */
4200 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4201 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4202
4203 /* Populate the fields of the newsk from the oldsk and migrate the
4204 * asoc to the newsk.
4205 */
4206 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4207
4208 *sockp = sock;
4209
4210 return err;
4211 }
4212 EXPORT_SYMBOL(sctp_do_peeloff);
4213
4214 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4215 {
4216 sctp_peeloff_arg_t peeloff;
4217 struct socket *newsock;
4218 int retval = 0;
4219
4220 if (len < sizeof(sctp_peeloff_arg_t))
4221 return -EINVAL;
4222 len = sizeof(sctp_peeloff_arg_t);
4223 if (copy_from_user(&peeloff, optval, len))
4224 return -EFAULT;
4225
4226 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4227 if (retval < 0)
4228 goto out;
4229
4230 /* Map the socket to an unused fd that can be returned to the user. */
4231 retval = sock_map_fd(newsock, 0);
4232 if (retval < 0) {
4233 sock_release(newsock);
4234 goto out;
4235 }
4236
4237 SCTP_DEBUG_PRINTK("%s: sk: %p newsk: %p sd: %d\n",
4238 __func__, sk, newsock->sk, retval);
4239
4240 /* Return the fd mapped to the new socket. */
4241 peeloff.sd = retval;
4242 if (put_user(len, optlen))
4243 return -EFAULT;
4244 if (copy_to_user(optval, &peeloff, len))
4245 retval = -EFAULT;
4246
4247 out:
4248 return retval;
4249 }
4250
4251 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4252 *
4253 * Applications can enable or disable heartbeats for any peer address of
4254 * an association, modify an address's heartbeat interval, force a
4255 * heartbeat to be sent immediately, and adjust the address's maximum
4256 * number of retransmissions sent before an address is considered
4257 * unreachable. The following structure is used to access and modify an
4258 * address's parameters:
4259 *
4260 * struct sctp_paddrparams {
4261 * sctp_assoc_t spp_assoc_id;
4262 * struct sockaddr_storage spp_address;
4263 * uint32_t spp_hbinterval;
4264 * uint16_t spp_pathmaxrxt;
4265 * uint32_t spp_pathmtu;
4266 * uint32_t spp_sackdelay;
4267 * uint32_t spp_flags;
4268 * };
4269 *
4270 * spp_assoc_id - (one-to-many style socket) This is filled in the
4271 * application, and identifies the association for
4272 * this query.
4273 * spp_address - This specifies which address is of interest.
4274 * spp_hbinterval - This contains the value of the heartbeat interval,
4275 * in milliseconds. If a value of zero
4276 * is present in this field then no changes are to
4277 * be made to this parameter.
4278 * spp_pathmaxrxt - This contains the maximum number of
4279 * retransmissions before this address shall be
4280 * considered unreachable. If a value of zero
4281 * is present in this field then no changes are to
4282 * be made to this parameter.
4283 * spp_pathmtu - When Path MTU discovery is disabled the value
4284 * specified here will be the "fixed" path mtu.
4285 * Note that if the spp_address field is empty
4286 * then all associations on this address will
4287 * have this fixed path mtu set upon them.
4288 *
4289 * spp_sackdelay - When delayed sack is enabled, this value specifies
4290 * the number of milliseconds that sacks will be delayed
4291 * for. This value will apply to all addresses of an
4292 * association if the spp_address field is empty. Note
4293 * also, that if delayed sack is enabled and this
4294 * value is set to 0, no change is made to the last
4295 * recorded delayed sack timer value.
4296 *
4297 * spp_flags - These flags are used to control various features
4298 * on an association. The flag field may contain
4299 * zero or more of the following options.
4300 *
4301 * SPP_HB_ENABLE - Enable heartbeats on the
4302 * specified address. Note that if the address
4303 * field is empty all addresses for the association
4304 * have heartbeats enabled upon them.
4305 *
4306 * SPP_HB_DISABLE - Disable heartbeats on the
4307 * speicifed address. Note that if the address
4308 * field is empty all addresses for the association
4309 * will have their heartbeats disabled. Note also
4310 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4311 * mutually exclusive, only one of these two should
4312 * be specified. Enabling both fields will have
4313 * undetermined results.
4314 *
4315 * SPP_HB_DEMAND - Request a user initiated heartbeat
4316 * to be made immediately.
4317 *
4318 * SPP_PMTUD_ENABLE - This field will enable PMTU
4319 * discovery upon the specified address. Note that
4320 * if the address feild is empty then all addresses
4321 * on the association are effected.
4322 *
4323 * SPP_PMTUD_DISABLE - This field will disable PMTU
4324 * discovery upon the specified address. Note that
4325 * if the address feild is empty then all addresses
4326 * on the association are effected. Not also that
4327 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4328 * exclusive. Enabling both will have undetermined
4329 * results.
4330 *
4331 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4332 * on delayed sack. The time specified in spp_sackdelay
4333 * is used to specify the sack delay for this address. Note
4334 * that if spp_address is empty then all addresses will
4335 * enable delayed sack and take on the sack delay
4336 * value specified in spp_sackdelay.
4337 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4338 * off delayed sack. If the spp_address field is blank then
4339 * delayed sack is disabled for the entire association. Note
4340 * also that this field is mutually exclusive to
4341 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4342 * results.
4343 */
4344 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4345 char __user *optval, int __user *optlen)
4346 {
4347 struct sctp_paddrparams params;
4348 struct sctp_transport *trans = NULL;
4349 struct sctp_association *asoc = NULL;
4350 struct sctp_sock *sp = sctp_sk(sk);
4351
4352 if (len < sizeof(struct sctp_paddrparams))
4353 return -EINVAL;
4354 len = sizeof(struct sctp_paddrparams);
4355 if (copy_from_user(&params, optval, len))
4356 return -EFAULT;
4357
4358 /* If an address other than INADDR_ANY is specified, and
4359 * no transport is found, then the request is invalid.
4360 */
4361 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
4362 trans = sctp_addr_id2transport(sk, &params.spp_address,
4363 params.spp_assoc_id);
4364 if (!trans) {
4365 SCTP_DEBUG_PRINTK("Failed no transport\n");
4366 return -EINVAL;
4367 }
4368 }
4369
4370 /* Get association, if assoc_id != 0 and the socket is a one
4371 * to many style socket, and an association was not found, then
4372 * the id was invalid.
4373 */
4374 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4375 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4376 SCTP_DEBUG_PRINTK("Failed no association\n");
4377 return -EINVAL;
4378 }
4379
4380 if (trans) {
4381 /* Fetch transport values. */
4382 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4383 params.spp_pathmtu = trans->pathmtu;
4384 params.spp_pathmaxrxt = trans->pathmaxrxt;
4385 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4386
4387 /*draft-11 doesn't say what to return in spp_flags*/
4388 params.spp_flags = trans->param_flags;
4389 } else if (asoc) {
4390 /* Fetch association values. */
4391 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4392 params.spp_pathmtu = asoc->pathmtu;
4393 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4394 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4395
4396 /*draft-11 doesn't say what to return in spp_flags*/
4397 params.spp_flags = asoc->param_flags;
4398 } else {
4399 /* Fetch socket values. */
4400 params.spp_hbinterval = sp->hbinterval;
4401 params.spp_pathmtu = sp->pathmtu;
4402 params.spp_sackdelay = sp->sackdelay;
4403 params.spp_pathmaxrxt = sp->pathmaxrxt;
4404
4405 /*draft-11 doesn't say what to return in spp_flags*/
4406 params.spp_flags = sp->param_flags;
4407 }
4408
4409 if (copy_to_user(optval, &params, len))
4410 return -EFAULT;
4411
4412 if (put_user(len, optlen))
4413 return -EFAULT;
4414
4415 return 0;
4416 }
4417
4418 /*
4419 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4420 *
4421 * This option will effect the way delayed acks are performed. This
4422 * option allows you to get or set the delayed ack time, in
4423 * milliseconds. It also allows changing the delayed ack frequency.
4424 * Changing the frequency to 1 disables the delayed sack algorithm. If
4425 * the assoc_id is 0, then this sets or gets the endpoints default
4426 * values. If the assoc_id field is non-zero, then the set or get
4427 * effects the specified association for the one to many model (the
4428 * assoc_id field is ignored by the one to one model). Note that if
4429 * sack_delay or sack_freq are 0 when setting this option, then the
4430 * current values will remain unchanged.
4431 *
4432 * struct sctp_sack_info {
4433 * sctp_assoc_t sack_assoc_id;
4434 * uint32_t sack_delay;
4435 * uint32_t sack_freq;
4436 * };
4437 *
4438 * sack_assoc_id - This parameter, indicates which association the user
4439 * is performing an action upon. Note that if this field's value is
4440 * zero then the endpoints default value is changed (effecting future
4441 * associations only).
4442 *
4443 * sack_delay - This parameter contains the number of milliseconds that
4444 * the user is requesting the delayed ACK timer be set to. Note that
4445 * this value is defined in the standard to be between 200 and 500
4446 * milliseconds.
4447 *
4448 * sack_freq - This parameter contains the number of packets that must
4449 * be received before a sack is sent without waiting for the delay
4450 * timer to expire. The default value for this is 2, setting this
4451 * value to 1 will disable the delayed sack algorithm.
4452 */
4453 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4454 char __user *optval,
4455 int __user *optlen)
4456 {
4457 struct sctp_sack_info params;
4458 struct sctp_association *asoc = NULL;
4459 struct sctp_sock *sp = sctp_sk(sk);
4460
4461 if (len >= sizeof(struct sctp_sack_info)) {
4462 len = sizeof(struct sctp_sack_info);
4463
4464 if (copy_from_user(&params, optval, len))
4465 return -EFAULT;
4466 } else if (len == sizeof(struct sctp_assoc_value)) {
4467 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4468 pr_warn("Use struct sctp_sack_info instead\n");
4469 if (copy_from_user(&params, optval, len))
4470 return -EFAULT;
4471 } else
4472 return - EINVAL;
4473
4474 /* Get association, if sack_assoc_id != 0 and the socket is a one
4475 * to many style socket, and an association was not found, then
4476 * the id was invalid.
4477 */
4478 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4479 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4480 return -EINVAL;
4481
4482 if (asoc) {
4483 /* Fetch association values. */
4484 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4485 params.sack_delay = jiffies_to_msecs(
4486 asoc->sackdelay);
4487 params.sack_freq = asoc->sackfreq;
4488
4489 } else {
4490 params.sack_delay = 0;
4491 params.sack_freq = 1;
4492 }
4493 } else {
4494 /* Fetch socket values. */
4495 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4496 params.sack_delay = sp->sackdelay;
4497 params.sack_freq = sp->sackfreq;
4498 } else {
4499 params.sack_delay = 0;
4500 params.sack_freq = 1;
4501 }
4502 }
4503
4504 if (copy_to_user(optval, &params, len))
4505 return -EFAULT;
4506
4507 if (put_user(len, optlen))
4508 return -EFAULT;
4509
4510 return 0;
4511 }
4512
4513 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4514 *
4515 * Applications can specify protocol parameters for the default association
4516 * initialization. The option name argument to setsockopt() and getsockopt()
4517 * is SCTP_INITMSG.
4518 *
4519 * Setting initialization parameters is effective only on an unconnected
4520 * socket (for UDP-style sockets only future associations are effected
4521 * by the change). With TCP-style sockets, this option is inherited by
4522 * sockets derived from a listener socket.
4523 */
4524 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4525 {
4526 if (len < sizeof(struct sctp_initmsg))
4527 return -EINVAL;
4528 len = sizeof(struct sctp_initmsg);
4529 if (put_user(len, optlen))
4530 return -EFAULT;
4531 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4532 return -EFAULT;
4533 return 0;
4534 }
4535
4536
4537 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4538 char __user *optval, int __user *optlen)
4539 {
4540 struct sctp_association *asoc;
4541 int cnt = 0;
4542 struct sctp_getaddrs getaddrs;
4543 struct sctp_transport *from;
4544 void __user *to;
4545 union sctp_addr temp;
4546 struct sctp_sock *sp = sctp_sk(sk);
4547 int addrlen;
4548 size_t space_left;
4549 int bytes_copied;
4550
4551 if (len < sizeof(struct sctp_getaddrs))
4552 return -EINVAL;
4553
4554 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4555 return -EFAULT;
4556
4557 /* For UDP-style sockets, id specifies the association to query. */
4558 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4559 if (!asoc)
4560 return -EINVAL;
4561
4562 to = optval + offsetof(struct sctp_getaddrs,addrs);
4563 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4564
4565 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4566 transports) {
4567 memcpy(&temp, &from->ipaddr, sizeof(temp));
4568 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4569 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4570 if (space_left < addrlen)
4571 return -ENOMEM;
4572 if (copy_to_user(to, &temp, addrlen))
4573 return -EFAULT;
4574 to += addrlen;
4575 cnt++;
4576 space_left -= addrlen;
4577 }
4578
4579 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4580 return -EFAULT;
4581 bytes_copied = ((char __user *)to) - optval;
4582 if (put_user(bytes_copied, optlen))
4583 return -EFAULT;
4584
4585 return 0;
4586 }
4587
4588 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4589 size_t space_left, int *bytes_copied)
4590 {
4591 struct sctp_sockaddr_entry *addr;
4592 union sctp_addr temp;
4593 int cnt = 0;
4594 int addrlen;
4595
4596 rcu_read_lock();
4597 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4598 if (!addr->valid)
4599 continue;
4600
4601 if ((PF_INET == sk->sk_family) &&
4602 (AF_INET6 == addr->a.sa.sa_family))
4603 continue;
4604 if ((PF_INET6 == sk->sk_family) &&
4605 inet_v6_ipv6only(sk) &&
4606 (AF_INET == addr->a.sa.sa_family))
4607 continue;
4608 memcpy(&temp, &addr->a, sizeof(temp));
4609 if (!temp.v4.sin_port)
4610 temp.v4.sin_port = htons(port);
4611
4612 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4613 &temp);
4614 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4615 if (space_left < addrlen) {
4616 cnt = -ENOMEM;
4617 break;
4618 }
4619 memcpy(to, &temp, addrlen);
4620
4621 to += addrlen;
4622 cnt ++;
4623 space_left -= addrlen;
4624 *bytes_copied += addrlen;
4625 }
4626 rcu_read_unlock();
4627
4628 return cnt;
4629 }
4630
4631
4632 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4633 char __user *optval, int __user *optlen)
4634 {
4635 struct sctp_bind_addr *bp;
4636 struct sctp_association *asoc;
4637 int cnt = 0;
4638 struct sctp_getaddrs getaddrs;
4639 struct sctp_sockaddr_entry *addr;
4640 void __user *to;
4641 union sctp_addr temp;
4642 struct sctp_sock *sp = sctp_sk(sk);
4643 int addrlen;
4644 int err = 0;
4645 size_t space_left;
4646 int bytes_copied = 0;
4647 void *addrs;
4648 void *buf;
4649
4650 if (len < sizeof(struct sctp_getaddrs))
4651 return -EINVAL;
4652
4653 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4654 return -EFAULT;
4655
4656 /*
4657 * For UDP-style sockets, id specifies the association to query.
4658 * If the id field is set to the value '0' then the locally bound
4659 * addresses are returned without regard to any particular
4660 * association.
4661 */
4662 if (0 == getaddrs.assoc_id) {
4663 bp = &sctp_sk(sk)->ep->base.bind_addr;
4664 } else {
4665 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4666 if (!asoc)
4667 return -EINVAL;
4668 bp = &asoc->base.bind_addr;
4669 }
4670
4671 to = optval + offsetof(struct sctp_getaddrs,addrs);
4672 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4673
4674 addrs = kmalloc(space_left, GFP_KERNEL);
4675 if (!addrs)
4676 return -ENOMEM;
4677
4678 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4679 * addresses from the global local address list.
4680 */
4681 if (sctp_list_single_entry(&bp->address_list)) {
4682 addr = list_entry(bp->address_list.next,
4683 struct sctp_sockaddr_entry, list);
4684 if (sctp_is_any(sk, &addr->a)) {
4685 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4686 space_left, &bytes_copied);
4687 if (cnt < 0) {
4688 err = cnt;
4689 goto out;
4690 }
4691 goto copy_getaddrs;
4692 }
4693 }
4694
4695 buf = addrs;
4696 /* Protection on the bound address list is not needed since
4697 * in the socket option context we hold a socket lock and
4698 * thus the bound address list can't change.
4699 */
4700 list_for_each_entry(addr, &bp->address_list, list) {
4701 memcpy(&temp, &addr->a, sizeof(temp));
4702 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4703 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4704 if (space_left < addrlen) {
4705 err = -ENOMEM; /*fixme: right error?*/
4706 goto out;
4707 }
4708 memcpy(buf, &temp, addrlen);
4709 buf += addrlen;
4710 bytes_copied += addrlen;
4711 cnt ++;
4712 space_left -= addrlen;
4713 }
4714
4715 copy_getaddrs:
4716 if (copy_to_user(to, addrs, bytes_copied)) {
4717 err = -EFAULT;
4718 goto out;
4719 }
4720 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4721 err = -EFAULT;
4722 goto out;
4723 }
4724 if (put_user(bytes_copied, optlen))
4725 err = -EFAULT;
4726 out:
4727 kfree(addrs);
4728 return err;
4729 }
4730
4731 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4732 *
4733 * Requests that the local SCTP stack use the enclosed peer address as
4734 * the association primary. The enclosed address must be one of the
4735 * association peer's addresses.
4736 */
4737 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4738 char __user *optval, int __user *optlen)
4739 {
4740 struct sctp_prim prim;
4741 struct sctp_association *asoc;
4742 struct sctp_sock *sp = sctp_sk(sk);
4743
4744 if (len < sizeof(struct sctp_prim))
4745 return -EINVAL;
4746
4747 len = sizeof(struct sctp_prim);
4748
4749 if (copy_from_user(&prim, optval, len))
4750 return -EFAULT;
4751
4752 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4753 if (!asoc)
4754 return -EINVAL;
4755
4756 if (!asoc->peer.primary_path)
4757 return -ENOTCONN;
4758
4759 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4760 asoc->peer.primary_path->af_specific->sockaddr_len);
4761
4762 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4763 (union sctp_addr *)&prim.ssp_addr);
4764
4765 if (put_user(len, optlen))
4766 return -EFAULT;
4767 if (copy_to_user(optval, &prim, len))
4768 return -EFAULT;
4769
4770 return 0;
4771 }
4772
4773 /*
4774 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4775 *
4776 * Requests that the local endpoint set the specified Adaptation Layer
4777 * Indication parameter for all future INIT and INIT-ACK exchanges.
4778 */
4779 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4780 char __user *optval, int __user *optlen)
4781 {
4782 struct sctp_setadaptation adaptation;
4783
4784 if (len < sizeof(struct sctp_setadaptation))
4785 return -EINVAL;
4786
4787 len = sizeof(struct sctp_setadaptation);
4788
4789 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4790
4791 if (put_user(len, optlen))
4792 return -EFAULT;
4793 if (copy_to_user(optval, &adaptation, len))
4794 return -EFAULT;
4795
4796 return 0;
4797 }
4798
4799 /*
4800 *
4801 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4802 *
4803 * Applications that wish to use the sendto() system call may wish to
4804 * specify a default set of parameters that would normally be supplied
4805 * through the inclusion of ancillary data. This socket option allows
4806 * such an application to set the default sctp_sndrcvinfo structure.
4807
4808
4809 * The application that wishes to use this socket option simply passes
4810 * in to this call the sctp_sndrcvinfo structure defined in Section
4811 * 5.2.2) The input parameters accepted by this call include
4812 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4813 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4814 * to this call if the caller is using the UDP model.
4815 *
4816 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4817 */
4818 static int sctp_getsockopt_default_send_param(struct sock *sk,
4819 int len, char __user *optval,
4820 int __user *optlen)
4821 {
4822 struct sctp_sndrcvinfo info;
4823 struct sctp_association *asoc;
4824 struct sctp_sock *sp = sctp_sk(sk);
4825
4826 if (len < sizeof(struct sctp_sndrcvinfo))
4827 return -EINVAL;
4828
4829 len = sizeof(struct sctp_sndrcvinfo);
4830
4831 if (copy_from_user(&info, optval, len))
4832 return -EFAULT;
4833
4834 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4835 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4836 return -EINVAL;
4837
4838 if (asoc) {
4839 info.sinfo_stream = asoc->default_stream;
4840 info.sinfo_flags = asoc->default_flags;
4841 info.sinfo_ppid = asoc->default_ppid;
4842 info.sinfo_context = asoc->default_context;
4843 info.sinfo_timetolive = asoc->default_timetolive;
4844 } else {
4845 info.sinfo_stream = sp->default_stream;
4846 info.sinfo_flags = sp->default_flags;
4847 info.sinfo_ppid = sp->default_ppid;
4848 info.sinfo_context = sp->default_context;
4849 info.sinfo_timetolive = sp->default_timetolive;
4850 }
4851
4852 if (put_user(len, optlen))
4853 return -EFAULT;
4854 if (copy_to_user(optval, &info, len))
4855 return -EFAULT;
4856
4857 return 0;
4858 }
4859
4860 /*
4861 *
4862 * 7.1.5 SCTP_NODELAY
4863 *
4864 * Turn on/off any Nagle-like algorithm. This means that packets are
4865 * generally sent as soon as possible and no unnecessary delays are
4866 * introduced, at the cost of more packets in the network. Expects an
4867 * integer boolean flag.
4868 */
4869
4870 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4871 char __user *optval, int __user *optlen)
4872 {
4873 int val;
4874
4875 if (len < sizeof(int))
4876 return -EINVAL;
4877
4878 len = sizeof(int);
4879 val = (sctp_sk(sk)->nodelay == 1);
4880 if (put_user(len, optlen))
4881 return -EFAULT;
4882 if (copy_to_user(optval, &val, len))
4883 return -EFAULT;
4884 return 0;
4885 }
4886
4887 /*
4888 *
4889 * 7.1.1 SCTP_RTOINFO
4890 *
4891 * The protocol parameters used to initialize and bound retransmission
4892 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4893 * and modify these parameters.
4894 * All parameters are time values, in milliseconds. A value of 0, when
4895 * modifying the parameters, indicates that the current value should not
4896 * be changed.
4897 *
4898 */
4899 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4900 char __user *optval,
4901 int __user *optlen) {
4902 struct sctp_rtoinfo rtoinfo;
4903 struct sctp_association *asoc;
4904
4905 if (len < sizeof (struct sctp_rtoinfo))
4906 return -EINVAL;
4907
4908 len = sizeof(struct sctp_rtoinfo);
4909
4910 if (copy_from_user(&rtoinfo, optval, len))
4911 return -EFAULT;
4912
4913 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4914
4915 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4916 return -EINVAL;
4917
4918 /* Values corresponding to the specific association. */
4919 if (asoc) {
4920 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4921 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4922 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4923 } else {
4924 /* Values corresponding to the endpoint. */
4925 struct sctp_sock *sp = sctp_sk(sk);
4926
4927 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4928 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4929 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4930 }
4931
4932 if (put_user(len, optlen))
4933 return -EFAULT;
4934
4935 if (copy_to_user(optval, &rtoinfo, len))
4936 return -EFAULT;
4937
4938 return 0;
4939 }
4940
4941 /*
4942 *
4943 * 7.1.2 SCTP_ASSOCINFO
4944 *
4945 * This option is used to tune the maximum retransmission attempts
4946 * of the association.
4947 * Returns an error if the new association retransmission value is
4948 * greater than the sum of the retransmission value of the peer.
4949 * See [SCTP] for more information.
4950 *
4951 */
4952 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4953 char __user *optval,
4954 int __user *optlen)
4955 {
4956
4957 struct sctp_assocparams assocparams;
4958 struct sctp_association *asoc;
4959 struct list_head *pos;
4960 int cnt = 0;
4961
4962 if (len < sizeof (struct sctp_assocparams))
4963 return -EINVAL;
4964
4965 len = sizeof(struct sctp_assocparams);
4966
4967 if (copy_from_user(&assocparams, optval, len))
4968 return -EFAULT;
4969
4970 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4971
4972 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4973 return -EINVAL;
4974
4975 /* Values correspoinding to the specific association */
4976 if (asoc) {
4977 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4978 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4979 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4980 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4981 * 1000) +
4982 (asoc->cookie_life.tv_usec
4983 / 1000);
4984
4985 list_for_each(pos, &asoc->peer.transport_addr_list) {
4986 cnt ++;
4987 }
4988
4989 assocparams.sasoc_number_peer_destinations = cnt;
4990 } else {
4991 /* Values corresponding to the endpoint */
4992 struct sctp_sock *sp = sctp_sk(sk);
4993
4994 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
4995 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
4996 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
4997 assocparams.sasoc_cookie_life =
4998 sp->assocparams.sasoc_cookie_life;
4999 assocparams.sasoc_number_peer_destinations =
5000 sp->assocparams.
5001 sasoc_number_peer_destinations;
5002 }
5003
5004 if (put_user(len, optlen))
5005 return -EFAULT;
5006
5007 if (copy_to_user(optval, &assocparams, len))
5008 return -EFAULT;
5009
5010 return 0;
5011 }
5012
5013 /*
5014 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5015 *
5016 * This socket option is a boolean flag which turns on or off mapped V4
5017 * addresses. If this option is turned on and the socket is type
5018 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5019 * If this option is turned off, then no mapping will be done of V4
5020 * addresses and a user will receive both PF_INET6 and PF_INET type
5021 * addresses on the socket.
5022 */
5023 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5024 char __user *optval, int __user *optlen)
5025 {
5026 int val;
5027 struct sctp_sock *sp = sctp_sk(sk);
5028
5029 if (len < sizeof(int))
5030 return -EINVAL;
5031
5032 len = sizeof(int);
5033 val = sp->v4mapped;
5034 if (put_user(len, optlen))
5035 return -EFAULT;
5036 if (copy_to_user(optval, &val, len))
5037 return -EFAULT;
5038
5039 return 0;
5040 }
5041
5042 /*
5043 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5044 * (chapter and verse is quoted at sctp_setsockopt_context())
5045 */
5046 static int sctp_getsockopt_context(struct sock *sk, int len,
5047 char __user *optval, int __user *optlen)
5048 {
5049 struct sctp_assoc_value params;
5050 struct sctp_sock *sp;
5051 struct sctp_association *asoc;
5052
5053 if (len < sizeof(struct sctp_assoc_value))
5054 return -EINVAL;
5055
5056 len = sizeof(struct sctp_assoc_value);
5057
5058 if (copy_from_user(&params, optval, len))
5059 return -EFAULT;
5060
5061 sp = sctp_sk(sk);
5062
5063 if (params.assoc_id != 0) {
5064 asoc = sctp_id2assoc(sk, params.assoc_id);
5065 if (!asoc)
5066 return -EINVAL;
5067 params.assoc_value = asoc->default_rcv_context;
5068 } else {
5069 params.assoc_value = sp->default_rcv_context;
5070 }
5071
5072 if (put_user(len, optlen))
5073 return -EFAULT;
5074 if (copy_to_user(optval, &params, len))
5075 return -EFAULT;
5076
5077 return 0;
5078 }
5079
5080 /*
5081 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5082 * This option will get or set the maximum size to put in any outgoing
5083 * SCTP DATA chunk. If a message is larger than this size it will be
5084 * fragmented by SCTP into the specified size. Note that the underlying
5085 * SCTP implementation may fragment into smaller sized chunks when the
5086 * PMTU of the underlying association is smaller than the value set by
5087 * the user. The default value for this option is '0' which indicates
5088 * the user is NOT limiting fragmentation and only the PMTU will effect
5089 * SCTP's choice of DATA chunk size. Note also that values set larger
5090 * than the maximum size of an IP datagram will effectively let SCTP
5091 * control fragmentation (i.e. the same as setting this option to 0).
5092 *
5093 * The following structure is used to access and modify this parameter:
5094 *
5095 * struct sctp_assoc_value {
5096 * sctp_assoc_t assoc_id;
5097 * uint32_t assoc_value;
5098 * };
5099 *
5100 * assoc_id: This parameter is ignored for one-to-one style sockets.
5101 * For one-to-many style sockets this parameter indicates which
5102 * association the user is performing an action upon. Note that if
5103 * this field's value is zero then the endpoints default value is
5104 * changed (effecting future associations only).
5105 * assoc_value: This parameter specifies the maximum size in bytes.
5106 */
5107 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5108 char __user *optval, int __user *optlen)
5109 {
5110 struct sctp_assoc_value params;
5111 struct sctp_association *asoc;
5112
5113 if (len == sizeof(int)) {
5114 pr_warn("Use of int in maxseg socket option deprecated\n");
5115 pr_warn("Use struct sctp_assoc_value instead\n");
5116 params.assoc_id = 0;
5117 } else if (len >= sizeof(struct sctp_assoc_value)) {
5118 len = sizeof(struct sctp_assoc_value);
5119 if (copy_from_user(&params, optval, sizeof(params)))
5120 return -EFAULT;
5121 } else
5122 return -EINVAL;
5123
5124 asoc = sctp_id2assoc(sk, params.assoc_id);
5125 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5126 return -EINVAL;
5127
5128 if (asoc)
5129 params.assoc_value = asoc->frag_point;
5130 else
5131 params.assoc_value = sctp_sk(sk)->user_frag;
5132
5133 if (put_user(len, optlen))
5134 return -EFAULT;
5135 if (len == sizeof(int)) {
5136 if (copy_to_user(optval, &params.assoc_value, len))
5137 return -EFAULT;
5138 } else {
5139 if (copy_to_user(optval, &params, len))
5140 return -EFAULT;
5141 }
5142
5143 return 0;
5144 }
5145
5146 /*
5147 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5148 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5149 */
5150 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5151 char __user *optval, int __user *optlen)
5152 {
5153 int val;
5154
5155 if (len < sizeof(int))
5156 return -EINVAL;
5157
5158 len = sizeof(int);
5159
5160 val = sctp_sk(sk)->frag_interleave;
5161 if (put_user(len, optlen))
5162 return -EFAULT;
5163 if (copy_to_user(optval, &val, len))
5164 return -EFAULT;
5165
5166 return 0;
5167 }
5168
5169 /*
5170 * 7.1.25. Set or Get the sctp partial delivery point
5171 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5172 */
5173 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5174 char __user *optval,
5175 int __user *optlen)
5176 {
5177 u32 val;
5178
5179 if (len < sizeof(u32))
5180 return -EINVAL;
5181
5182 len = sizeof(u32);
5183
5184 val = sctp_sk(sk)->pd_point;
5185 if (put_user(len, optlen))
5186 return -EFAULT;
5187 if (copy_to_user(optval, &val, len))
5188 return -EFAULT;
5189
5190 return 0;
5191 }
5192
5193 /*
5194 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5195 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5196 */
5197 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5198 char __user *optval,
5199 int __user *optlen)
5200 {
5201 struct sctp_assoc_value params;
5202 struct sctp_sock *sp;
5203 struct sctp_association *asoc;
5204
5205 if (len == sizeof(int)) {
5206 pr_warn("Use of int in max_burst socket option deprecated\n");
5207 pr_warn("Use struct sctp_assoc_value instead\n");
5208 params.assoc_id = 0;
5209 } else if (len >= sizeof(struct sctp_assoc_value)) {
5210 len = sizeof(struct sctp_assoc_value);
5211 if (copy_from_user(&params, optval, len))
5212 return -EFAULT;
5213 } else
5214 return -EINVAL;
5215
5216 sp = sctp_sk(sk);
5217
5218 if (params.assoc_id != 0) {
5219 asoc = sctp_id2assoc(sk, params.assoc_id);
5220 if (!asoc)
5221 return -EINVAL;
5222 params.assoc_value = asoc->max_burst;
5223 } else
5224 params.assoc_value = sp->max_burst;
5225
5226 if (len == sizeof(int)) {
5227 if (copy_to_user(optval, &params.assoc_value, len))
5228 return -EFAULT;
5229 } else {
5230 if (copy_to_user(optval, &params, len))
5231 return -EFAULT;
5232 }
5233
5234 return 0;
5235
5236 }
5237
5238 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5239 char __user *optval, int __user *optlen)
5240 {
5241 struct sctp_hmacalgo __user *p = (void __user *)optval;
5242 struct sctp_hmac_algo_param *hmacs;
5243 __u16 data_len = 0;
5244 u32 num_idents;
5245
5246 if (!sctp_auth_enable)
5247 return -EACCES;
5248
5249 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5250 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5251
5252 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5253 return -EINVAL;
5254
5255 len = sizeof(struct sctp_hmacalgo) + data_len;
5256 num_idents = data_len / sizeof(u16);
5257
5258 if (put_user(len, optlen))
5259 return -EFAULT;
5260 if (put_user(num_idents, &p->shmac_num_idents))
5261 return -EFAULT;
5262 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5263 return -EFAULT;
5264 return 0;
5265 }
5266
5267 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5268 char __user *optval, int __user *optlen)
5269 {
5270 struct sctp_authkeyid val;
5271 struct sctp_association *asoc;
5272
5273 if (!sctp_auth_enable)
5274 return -EACCES;
5275
5276 if (len < sizeof(struct sctp_authkeyid))
5277 return -EINVAL;
5278 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5279 return -EFAULT;
5280
5281 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5282 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5283 return -EINVAL;
5284
5285 if (asoc)
5286 val.scact_keynumber = asoc->active_key_id;
5287 else
5288 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5289
5290 len = sizeof(struct sctp_authkeyid);
5291 if (put_user(len, optlen))
5292 return -EFAULT;
5293 if (copy_to_user(optval, &val, len))
5294 return -EFAULT;
5295
5296 return 0;
5297 }
5298
5299 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5300 char __user *optval, int __user *optlen)
5301 {
5302 struct sctp_authchunks __user *p = (void __user *)optval;
5303 struct sctp_authchunks val;
5304 struct sctp_association *asoc;
5305 struct sctp_chunks_param *ch;
5306 u32 num_chunks = 0;
5307 char __user *to;
5308
5309 if (!sctp_auth_enable)
5310 return -EACCES;
5311
5312 if (len < sizeof(struct sctp_authchunks))
5313 return -EINVAL;
5314
5315 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5316 return -EFAULT;
5317
5318 to = p->gauth_chunks;
5319 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5320 if (!asoc)
5321 return -EINVAL;
5322
5323 ch = asoc->peer.peer_chunks;
5324 if (!ch)
5325 goto num;
5326
5327 /* See if the user provided enough room for all the data */
5328 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5329 if (len < num_chunks)
5330 return -EINVAL;
5331
5332 if (copy_to_user(to, ch->chunks, num_chunks))
5333 return -EFAULT;
5334 num:
5335 len = sizeof(struct sctp_authchunks) + num_chunks;
5336 if (put_user(len, optlen)) return -EFAULT;
5337 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5338 return -EFAULT;
5339 return 0;
5340 }
5341
5342 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5343 char __user *optval, int __user *optlen)
5344 {
5345 struct sctp_authchunks __user *p = (void __user *)optval;
5346 struct sctp_authchunks val;
5347 struct sctp_association *asoc;
5348 struct sctp_chunks_param *ch;
5349 u32 num_chunks = 0;
5350 char __user *to;
5351
5352 if (!sctp_auth_enable)
5353 return -EACCES;
5354
5355 if (len < sizeof(struct sctp_authchunks))
5356 return -EINVAL;
5357
5358 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5359 return -EFAULT;
5360
5361 to = p->gauth_chunks;
5362 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5363 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5364 return -EINVAL;
5365
5366 if (asoc)
5367 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5368 else
5369 ch = sctp_sk(sk)->ep->auth_chunk_list;
5370
5371 if (!ch)
5372 goto num;
5373
5374 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5375 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5376 return -EINVAL;
5377
5378 if (copy_to_user(to, ch->chunks, num_chunks))
5379 return -EFAULT;
5380 num:
5381 len = sizeof(struct sctp_authchunks) + num_chunks;
5382 if (put_user(len, optlen))
5383 return -EFAULT;
5384 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5385 return -EFAULT;
5386
5387 return 0;
5388 }
5389
5390 /*
5391 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5392 * This option gets the current number of associations that are attached
5393 * to a one-to-many style socket. The option value is an uint32_t.
5394 */
5395 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5396 char __user *optval, int __user *optlen)
5397 {
5398 struct sctp_sock *sp = sctp_sk(sk);
5399 struct sctp_association *asoc;
5400 u32 val = 0;
5401
5402 if (sctp_style(sk, TCP))
5403 return -EOPNOTSUPP;
5404
5405 if (len < sizeof(u32))
5406 return -EINVAL;
5407
5408 len = sizeof(u32);
5409
5410 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5411 val++;
5412 }
5413
5414 if (put_user(len, optlen))
5415 return -EFAULT;
5416 if (copy_to_user(optval, &val, len))
5417 return -EFAULT;
5418
5419 return 0;
5420 }
5421
5422 /*
5423 * 8.1.23 SCTP_AUTO_ASCONF
5424 * See the corresponding setsockopt entry as description
5425 */
5426 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5427 char __user *optval, int __user *optlen)
5428 {
5429 int val = 0;
5430
5431 if (len < sizeof(int))
5432 return -EINVAL;
5433
5434 len = sizeof(int);
5435 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5436 val = 1;
5437 if (put_user(len, optlen))
5438 return -EFAULT;
5439 if (copy_to_user(optval, &val, len))
5440 return -EFAULT;
5441 return 0;
5442 }
5443
5444 /*
5445 * 8.2.6. Get the Current Identifiers of Associations
5446 * (SCTP_GET_ASSOC_ID_LIST)
5447 *
5448 * This option gets the current list of SCTP association identifiers of
5449 * the SCTP associations handled by a one-to-many style socket.
5450 */
5451 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5452 char __user *optval, int __user *optlen)
5453 {
5454 struct sctp_sock *sp = sctp_sk(sk);
5455 struct sctp_association *asoc;
5456 struct sctp_assoc_ids *ids;
5457 u32 num = 0;
5458
5459 if (sctp_style(sk, TCP))
5460 return -EOPNOTSUPP;
5461
5462 if (len < sizeof(struct sctp_assoc_ids))
5463 return -EINVAL;
5464
5465 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5466 num++;
5467 }
5468
5469 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5470 return -EINVAL;
5471
5472 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5473
5474 ids = kmalloc(len, GFP_KERNEL);
5475 if (unlikely(!ids))
5476 return -ENOMEM;
5477
5478 ids->gaids_number_of_ids = num;
5479 num = 0;
5480 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5481 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5482 }
5483
5484 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5485 kfree(ids);
5486 return -EFAULT;
5487 }
5488
5489 kfree(ids);
5490 return 0;
5491 }
5492
5493 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5494 char __user *optval, int __user *optlen)
5495 {
5496 int retval = 0;
5497 int len;
5498
5499 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5500 sk, optname);
5501
5502 /* I can hardly begin to describe how wrong this is. This is
5503 * so broken as to be worse than useless. The API draft
5504 * REALLY is NOT helpful here... I am not convinced that the
5505 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5506 * are at all well-founded.
5507 */
5508 if (level != SOL_SCTP) {
5509 struct sctp_af *af = sctp_sk(sk)->pf->af;
5510
5511 retval = af->getsockopt(sk, level, optname, optval, optlen);
5512 return retval;
5513 }
5514
5515 if (get_user(len, optlen))
5516 return -EFAULT;
5517
5518 sctp_lock_sock(sk);
5519
5520 switch (optname) {
5521 case SCTP_STATUS:
5522 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5523 break;
5524 case SCTP_DISABLE_FRAGMENTS:
5525 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5526 optlen);
5527 break;
5528 case SCTP_EVENTS:
5529 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5530 break;
5531 case SCTP_AUTOCLOSE:
5532 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5533 break;
5534 case SCTP_SOCKOPT_PEELOFF:
5535 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5536 break;
5537 case SCTP_PEER_ADDR_PARAMS:
5538 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5539 optlen);
5540 break;
5541 case SCTP_DELAYED_SACK:
5542 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5543 optlen);
5544 break;
5545 case SCTP_INITMSG:
5546 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5547 break;
5548 case SCTP_GET_PEER_ADDRS:
5549 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5550 optlen);
5551 break;
5552 case SCTP_GET_LOCAL_ADDRS:
5553 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5554 optlen);
5555 break;
5556 case SCTP_SOCKOPT_CONNECTX3:
5557 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5558 break;
5559 case SCTP_DEFAULT_SEND_PARAM:
5560 retval = sctp_getsockopt_default_send_param(sk, len,
5561 optval, optlen);
5562 break;
5563 case SCTP_PRIMARY_ADDR:
5564 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5565 break;
5566 case SCTP_NODELAY:
5567 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5568 break;
5569 case SCTP_RTOINFO:
5570 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5571 break;
5572 case SCTP_ASSOCINFO:
5573 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5574 break;
5575 case SCTP_I_WANT_MAPPED_V4_ADDR:
5576 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5577 break;
5578 case SCTP_MAXSEG:
5579 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5580 break;
5581 case SCTP_GET_PEER_ADDR_INFO:
5582 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5583 optlen);
5584 break;
5585 case SCTP_ADAPTATION_LAYER:
5586 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5587 optlen);
5588 break;
5589 case SCTP_CONTEXT:
5590 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5591 break;
5592 case SCTP_FRAGMENT_INTERLEAVE:
5593 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5594 optlen);
5595 break;
5596 case SCTP_PARTIAL_DELIVERY_POINT:
5597 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5598 optlen);
5599 break;
5600 case SCTP_MAX_BURST:
5601 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5602 break;
5603 case SCTP_AUTH_KEY:
5604 case SCTP_AUTH_CHUNK:
5605 case SCTP_AUTH_DELETE_KEY:
5606 retval = -EOPNOTSUPP;
5607 break;
5608 case SCTP_HMAC_IDENT:
5609 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5610 break;
5611 case SCTP_AUTH_ACTIVE_KEY:
5612 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5613 break;
5614 case SCTP_PEER_AUTH_CHUNKS:
5615 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5616 optlen);
5617 break;
5618 case SCTP_LOCAL_AUTH_CHUNKS:
5619 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5620 optlen);
5621 break;
5622 case SCTP_GET_ASSOC_NUMBER:
5623 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5624 break;
5625 case SCTP_GET_ASSOC_ID_LIST:
5626 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5627 break;
5628 case SCTP_AUTO_ASCONF:
5629 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5630 break;
5631 default:
5632 retval = -ENOPROTOOPT;
5633 break;
5634 }
5635
5636 sctp_release_sock(sk);
5637 return retval;
5638 }
5639
5640 static void sctp_hash(struct sock *sk)
5641 {
5642 /* STUB */
5643 }
5644
5645 static void sctp_unhash(struct sock *sk)
5646 {
5647 /* STUB */
5648 }
5649
5650 /* Check if port is acceptable. Possibly find first available port.
5651 *
5652 * The port hash table (contained in the 'global' SCTP protocol storage
5653 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5654 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5655 * list (the list number is the port number hashed out, so as you
5656 * would expect from a hash function, all the ports in a given list have
5657 * such a number that hashes out to the same list number; you were
5658 * expecting that, right?); so each list has a set of ports, with a
5659 * link to the socket (struct sock) that uses it, the port number and
5660 * a fastreuse flag (FIXME: NPI ipg).
5661 */
5662 static struct sctp_bind_bucket *sctp_bucket_create(
5663 struct sctp_bind_hashbucket *head, unsigned short snum);
5664
5665 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5666 {
5667 struct sctp_bind_hashbucket *head; /* hash list */
5668 struct sctp_bind_bucket *pp; /* hash list port iterator */
5669 struct hlist_node *node;
5670 unsigned short snum;
5671 int ret;
5672
5673 snum = ntohs(addr->v4.sin_port);
5674
5675 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5676 sctp_local_bh_disable();
5677
5678 if (snum == 0) {
5679 /* Search for an available port. */
5680 int low, high, remaining, index;
5681 unsigned int rover;
5682
5683 inet_get_local_port_range(&low, &high);
5684 remaining = (high - low) + 1;
5685 rover = net_random() % remaining + low;
5686
5687 do {
5688 rover++;
5689 if ((rover < low) || (rover > high))
5690 rover = low;
5691 if (inet_is_reserved_local_port(rover))
5692 continue;
5693 index = sctp_phashfn(rover);
5694 head = &sctp_port_hashtable[index];
5695 sctp_spin_lock(&head->lock);
5696 sctp_for_each_hentry(pp, node, &head->chain)
5697 if (pp->port == rover)
5698 goto next;
5699 break;
5700 next:
5701 sctp_spin_unlock(&head->lock);
5702 } while (--remaining > 0);
5703
5704 /* Exhausted local port range during search? */
5705 ret = 1;
5706 if (remaining <= 0)
5707 goto fail;
5708
5709 /* OK, here is the one we will use. HEAD (the port
5710 * hash table list entry) is non-NULL and we hold it's
5711 * mutex.
5712 */
5713 snum = rover;
5714 } else {
5715 /* We are given an specific port number; we verify
5716 * that it is not being used. If it is used, we will
5717 * exahust the search in the hash list corresponding
5718 * to the port number (snum) - we detect that with the
5719 * port iterator, pp being NULL.
5720 */
5721 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5722 sctp_spin_lock(&head->lock);
5723 sctp_for_each_hentry(pp, node, &head->chain) {
5724 if (pp->port == snum)
5725 goto pp_found;
5726 }
5727 }
5728 pp = NULL;
5729 goto pp_not_found;
5730 pp_found:
5731 if (!hlist_empty(&pp->owner)) {
5732 /* We had a port hash table hit - there is an
5733 * available port (pp != NULL) and it is being
5734 * used by other socket (pp->owner not empty); that other
5735 * socket is going to be sk2.
5736 */
5737 int reuse = sk->sk_reuse;
5738 struct sock *sk2;
5739
5740 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5741 if (pp->fastreuse && sk->sk_reuse &&
5742 sk->sk_state != SCTP_SS_LISTENING)
5743 goto success;
5744
5745 /* Run through the list of sockets bound to the port
5746 * (pp->port) [via the pointers bind_next and
5747 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5748 * we get the endpoint they describe and run through
5749 * the endpoint's list of IP (v4 or v6) addresses,
5750 * comparing each of the addresses with the address of
5751 * the socket sk. If we find a match, then that means
5752 * that this port/socket (sk) combination are already
5753 * in an endpoint.
5754 */
5755 sk_for_each_bound(sk2, node, &pp->owner) {
5756 struct sctp_endpoint *ep2;
5757 ep2 = sctp_sk(sk2)->ep;
5758
5759 if (sk == sk2 ||
5760 (reuse && sk2->sk_reuse &&
5761 sk2->sk_state != SCTP_SS_LISTENING))
5762 continue;
5763
5764 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5765 sctp_sk(sk2), sctp_sk(sk))) {
5766 ret = (long)sk2;
5767 goto fail_unlock;
5768 }
5769 }
5770 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5771 }
5772 pp_not_found:
5773 /* If there was a hash table miss, create a new port. */
5774 ret = 1;
5775 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5776 goto fail_unlock;
5777
5778 /* In either case (hit or miss), make sure fastreuse is 1 only
5779 * if sk->sk_reuse is too (that is, if the caller requested
5780 * SO_REUSEADDR on this socket -sk-).
5781 */
5782 if (hlist_empty(&pp->owner)) {
5783 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5784 pp->fastreuse = 1;
5785 else
5786 pp->fastreuse = 0;
5787 } else if (pp->fastreuse &&
5788 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5789 pp->fastreuse = 0;
5790
5791 /* We are set, so fill up all the data in the hash table
5792 * entry, tie the socket list information with the rest of the
5793 * sockets FIXME: Blurry, NPI (ipg).
5794 */
5795 success:
5796 if (!sctp_sk(sk)->bind_hash) {
5797 inet_sk(sk)->inet_num = snum;
5798 sk_add_bind_node(sk, &pp->owner);
5799 sctp_sk(sk)->bind_hash = pp;
5800 }
5801 ret = 0;
5802
5803 fail_unlock:
5804 sctp_spin_unlock(&head->lock);
5805
5806 fail:
5807 sctp_local_bh_enable();
5808 return ret;
5809 }
5810
5811 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5812 * port is requested.
5813 */
5814 static int sctp_get_port(struct sock *sk, unsigned short snum)
5815 {
5816 long ret;
5817 union sctp_addr addr;
5818 struct sctp_af *af = sctp_sk(sk)->pf->af;
5819
5820 /* Set up a dummy address struct from the sk. */
5821 af->from_sk(&addr, sk);
5822 addr.v4.sin_port = htons(snum);
5823
5824 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5825 ret = sctp_get_port_local(sk, &addr);
5826
5827 return ret ? 1 : 0;
5828 }
5829
5830 /*
5831 * Move a socket to LISTENING state.
5832 */
5833 SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
5834 {
5835 struct sctp_sock *sp = sctp_sk(sk);
5836 struct sctp_endpoint *ep = sp->ep;
5837 struct crypto_hash *tfm = NULL;
5838
5839 /* Allocate HMAC for generating cookie. */
5840 if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
5841 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5842 if (IS_ERR(tfm)) {
5843 net_info_ratelimited("failed to load transform for %s: %ld\n",
5844 sctp_hmac_alg, PTR_ERR(tfm));
5845 return -ENOSYS;
5846 }
5847 sctp_sk(sk)->hmac = tfm;
5848 }
5849
5850 /*
5851 * If a bind() or sctp_bindx() is not called prior to a listen()
5852 * call that allows new associations to be accepted, the system
5853 * picks an ephemeral port and will choose an address set equivalent
5854 * to binding with a wildcard address.
5855 *
5856 * This is not currently spelled out in the SCTP sockets
5857 * extensions draft, but follows the practice as seen in TCP
5858 * sockets.
5859 *
5860 */
5861 sk->sk_state = SCTP_SS_LISTENING;
5862 if (!ep->base.bind_addr.port) {
5863 if (sctp_autobind(sk))
5864 return -EAGAIN;
5865 } else {
5866 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
5867 sk->sk_state = SCTP_SS_CLOSED;
5868 return -EADDRINUSE;
5869 }
5870 }
5871
5872 sk->sk_max_ack_backlog = backlog;
5873 sctp_hash_endpoint(ep);
5874 return 0;
5875 }
5876
5877 /*
5878 * 4.1.3 / 5.1.3 listen()
5879 *
5880 * By default, new associations are not accepted for UDP style sockets.
5881 * An application uses listen() to mark a socket as being able to
5882 * accept new associations.
5883 *
5884 * On TCP style sockets, applications use listen() to ready the SCTP
5885 * endpoint for accepting inbound associations.
5886 *
5887 * On both types of endpoints a backlog of '0' disables listening.
5888 *
5889 * Move a socket to LISTENING state.
5890 */
5891 int sctp_inet_listen(struct socket *sock, int backlog)
5892 {
5893 struct sock *sk = sock->sk;
5894 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5895 int err = -EINVAL;
5896
5897 if (unlikely(backlog < 0))
5898 return err;
5899
5900 sctp_lock_sock(sk);
5901
5902 /* Peeled-off sockets are not allowed to listen(). */
5903 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
5904 goto out;
5905
5906 if (sock->state != SS_UNCONNECTED)
5907 goto out;
5908
5909 /* If backlog is zero, disable listening. */
5910 if (!backlog) {
5911 if (sctp_sstate(sk, CLOSED))
5912 goto out;
5913
5914 err = 0;
5915 sctp_unhash_endpoint(ep);
5916 sk->sk_state = SCTP_SS_CLOSED;
5917 if (sk->sk_reuse)
5918 sctp_sk(sk)->bind_hash->fastreuse = 1;
5919 goto out;
5920 }
5921
5922 /* If we are already listening, just update the backlog */
5923 if (sctp_sstate(sk, LISTENING))
5924 sk->sk_max_ack_backlog = backlog;
5925 else {
5926 err = sctp_listen_start(sk, backlog);
5927 if (err)
5928 goto out;
5929 }
5930
5931 err = 0;
5932 out:
5933 sctp_release_sock(sk);
5934 return err;
5935 }
5936
5937 /*
5938 * This function is done by modeling the current datagram_poll() and the
5939 * tcp_poll(). Note that, based on these implementations, we don't
5940 * lock the socket in this function, even though it seems that,
5941 * ideally, locking or some other mechanisms can be used to ensure
5942 * the integrity of the counters (sndbuf and wmem_alloc) used
5943 * in this place. We assume that we don't need locks either until proven
5944 * otherwise.
5945 *
5946 * Another thing to note is that we include the Async I/O support
5947 * here, again, by modeling the current TCP/UDP code. We don't have
5948 * a good way to test with it yet.
5949 */
5950 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5951 {
5952 struct sock *sk = sock->sk;
5953 struct sctp_sock *sp = sctp_sk(sk);
5954 unsigned int mask;
5955
5956 poll_wait(file, sk_sleep(sk), wait);
5957
5958 /* A TCP-style listening socket becomes readable when the accept queue
5959 * is not empty.
5960 */
5961 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5962 return (!list_empty(&sp->ep->asocs)) ?
5963 (POLLIN | POLLRDNORM) : 0;
5964
5965 mask = 0;
5966
5967 /* Is there any exceptional events? */
5968 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5969 mask |= POLLERR;
5970 if (sk->sk_shutdown & RCV_SHUTDOWN)
5971 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
5972 if (sk->sk_shutdown == SHUTDOWN_MASK)
5973 mask |= POLLHUP;
5974
5975 /* Is it readable? Reconsider this code with TCP-style support. */
5976 if (!skb_queue_empty(&sk->sk_receive_queue))
5977 mask |= POLLIN | POLLRDNORM;
5978
5979 /* The association is either gone or not ready. */
5980 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
5981 return mask;
5982
5983 /* Is it writable? */
5984 if (sctp_writeable(sk)) {
5985 mask |= POLLOUT | POLLWRNORM;
5986 } else {
5987 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
5988 /*
5989 * Since the socket is not locked, the buffer
5990 * might be made available after the writeable check and
5991 * before the bit is set. This could cause a lost I/O
5992 * signal. tcp_poll() has a race breaker for this race
5993 * condition. Based on their implementation, we put
5994 * in the following code to cover it as well.
5995 */
5996 if (sctp_writeable(sk))
5997 mask |= POLLOUT | POLLWRNORM;
5998 }
5999 return mask;
6000 }
6001
6002 /********************************************************************
6003 * 2nd Level Abstractions
6004 ********************************************************************/
6005
6006 static struct sctp_bind_bucket *sctp_bucket_create(
6007 struct sctp_bind_hashbucket *head, unsigned short snum)
6008 {
6009 struct sctp_bind_bucket *pp;
6010
6011 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6012 if (pp) {
6013 SCTP_DBG_OBJCNT_INC(bind_bucket);
6014 pp->port = snum;
6015 pp->fastreuse = 0;
6016 INIT_HLIST_HEAD(&pp->owner);
6017 hlist_add_head(&pp->node, &head->chain);
6018 }
6019 return pp;
6020 }
6021
6022 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6023 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6024 {
6025 if (pp && hlist_empty(&pp->owner)) {
6026 __hlist_del(&pp->node);
6027 kmem_cache_free(sctp_bucket_cachep, pp);
6028 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6029 }
6030 }
6031
6032 /* Release this socket's reference to a local port. */
6033 static inline void __sctp_put_port(struct sock *sk)
6034 {
6035 struct sctp_bind_hashbucket *head =
6036 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->inet_num)];
6037 struct sctp_bind_bucket *pp;
6038
6039 sctp_spin_lock(&head->lock);
6040 pp = sctp_sk(sk)->bind_hash;
6041 __sk_del_bind_node(sk);
6042 sctp_sk(sk)->bind_hash = NULL;
6043 inet_sk(sk)->inet_num = 0;
6044 sctp_bucket_destroy(pp);
6045 sctp_spin_unlock(&head->lock);
6046 }
6047
6048 void sctp_put_port(struct sock *sk)
6049 {
6050 sctp_local_bh_disable();
6051 __sctp_put_port(sk);
6052 sctp_local_bh_enable();
6053 }
6054
6055 /*
6056 * The system picks an ephemeral port and choose an address set equivalent
6057 * to binding with a wildcard address.
6058 * One of those addresses will be the primary address for the association.
6059 * This automatically enables the multihoming capability of SCTP.
6060 */
6061 static int sctp_autobind(struct sock *sk)
6062 {
6063 union sctp_addr autoaddr;
6064 struct sctp_af *af;
6065 __be16 port;
6066
6067 /* Initialize a local sockaddr structure to INADDR_ANY. */
6068 af = sctp_sk(sk)->pf->af;
6069
6070 port = htons(inet_sk(sk)->inet_num);
6071 af->inaddr_any(&autoaddr, port);
6072
6073 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6074 }
6075
6076 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6077 *
6078 * From RFC 2292
6079 * 4.2 The cmsghdr Structure *
6080 *
6081 * When ancillary data is sent or received, any number of ancillary data
6082 * objects can be specified by the msg_control and msg_controllen members of
6083 * the msghdr structure, because each object is preceded by
6084 * a cmsghdr structure defining the object's length (the cmsg_len member).
6085 * Historically Berkeley-derived implementations have passed only one object
6086 * at a time, but this API allows multiple objects to be
6087 * passed in a single call to sendmsg() or recvmsg(). The following example
6088 * shows two ancillary data objects in a control buffer.
6089 *
6090 * |<--------------------------- msg_controllen -------------------------->|
6091 * | |
6092 *
6093 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6094 *
6095 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6096 * | | |
6097 *
6098 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6099 *
6100 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6101 * | | | | |
6102 *
6103 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6104 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6105 *
6106 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6107 *
6108 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6109 * ^
6110 * |
6111 *
6112 * msg_control
6113 * points here
6114 */
6115 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
6116 sctp_cmsgs_t *cmsgs)
6117 {
6118 struct cmsghdr *cmsg;
6119 struct msghdr *my_msg = (struct msghdr *)msg;
6120
6121 for (cmsg = CMSG_FIRSTHDR(msg);
6122 cmsg != NULL;
6123 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6124 if (!CMSG_OK(my_msg, cmsg))
6125 return -EINVAL;
6126
6127 /* Should we parse this header or ignore? */
6128 if (cmsg->cmsg_level != IPPROTO_SCTP)
6129 continue;
6130
6131 /* Strictly check lengths following example in SCM code. */
6132 switch (cmsg->cmsg_type) {
6133 case SCTP_INIT:
6134 /* SCTP Socket API Extension
6135 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6136 *
6137 * This cmsghdr structure provides information for
6138 * initializing new SCTP associations with sendmsg().
6139 * The SCTP_INITMSG socket option uses this same data
6140 * structure. This structure is not used for
6141 * recvmsg().
6142 *
6143 * cmsg_level cmsg_type cmsg_data[]
6144 * ------------ ------------ ----------------------
6145 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6146 */
6147 if (cmsg->cmsg_len !=
6148 CMSG_LEN(sizeof(struct sctp_initmsg)))
6149 return -EINVAL;
6150 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6151 break;
6152
6153 case SCTP_SNDRCV:
6154 /* SCTP Socket API Extension
6155 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6156 *
6157 * This cmsghdr structure specifies SCTP options for
6158 * sendmsg() and describes SCTP header information
6159 * about a received message through recvmsg().
6160 *
6161 * cmsg_level cmsg_type cmsg_data[]
6162 * ------------ ------------ ----------------------
6163 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6164 */
6165 if (cmsg->cmsg_len !=
6166 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6167 return -EINVAL;
6168
6169 cmsgs->info =
6170 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6171
6172 /* Minimally, validate the sinfo_flags. */
6173 if (cmsgs->info->sinfo_flags &
6174 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6175 SCTP_ABORT | SCTP_EOF))
6176 return -EINVAL;
6177 break;
6178
6179 default:
6180 return -EINVAL;
6181 }
6182 }
6183 return 0;
6184 }
6185
6186 /*
6187 * Wait for a packet..
6188 * Note: This function is the same function as in core/datagram.c
6189 * with a few modifications to make lksctp work.
6190 */
6191 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6192 {
6193 int error;
6194 DEFINE_WAIT(wait);
6195
6196 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6197
6198 /* Socket errors? */
6199 error = sock_error(sk);
6200 if (error)
6201 goto out;
6202
6203 if (!skb_queue_empty(&sk->sk_receive_queue))
6204 goto ready;
6205
6206 /* Socket shut down? */
6207 if (sk->sk_shutdown & RCV_SHUTDOWN)
6208 goto out;
6209
6210 /* Sequenced packets can come disconnected. If so we report the
6211 * problem.
6212 */
6213 error = -ENOTCONN;
6214
6215 /* Is there a good reason to think that we may receive some data? */
6216 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6217 goto out;
6218
6219 /* Handle signals. */
6220 if (signal_pending(current))
6221 goto interrupted;
6222
6223 /* Let another process have a go. Since we are going to sleep
6224 * anyway. Note: This may cause odd behaviors if the message
6225 * does not fit in the user's buffer, but this seems to be the
6226 * only way to honor MSG_DONTWAIT realistically.
6227 */
6228 sctp_release_sock(sk);
6229 *timeo_p = schedule_timeout(*timeo_p);
6230 sctp_lock_sock(sk);
6231
6232 ready:
6233 finish_wait(sk_sleep(sk), &wait);
6234 return 0;
6235
6236 interrupted:
6237 error = sock_intr_errno(*timeo_p);
6238
6239 out:
6240 finish_wait(sk_sleep(sk), &wait);
6241 *err = error;
6242 return error;
6243 }
6244
6245 /* Receive a datagram.
6246 * Note: This is pretty much the same routine as in core/datagram.c
6247 * with a few changes to make lksctp work.
6248 */
6249 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6250 int noblock, int *err)
6251 {
6252 int error;
6253 struct sk_buff *skb;
6254 long timeo;
6255
6256 timeo = sock_rcvtimeo(sk, noblock);
6257
6258 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6259 timeo, MAX_SCHEDULE_TIMEOUT);
6260
6261 do {
6262 /* Again only user level code calls this function,
6263 * so nothing interrupt level
6264 * will suddenly eat the receive_queue.
6265 *
6266 * Look at current nfs client by the way...
6267 * However, this function was correct in any case. 8)
6268 */
6269 if (flags & MSG_PEEK) {
6270 spin_lock_bh(&sk->sk_receive_queue.lock);
6271 skb = skb_peek(&sk->sk_receive_queue);
6272 if (skb)
6273 atomic_inc(&skb->users);
6274 spin_unlock_bh(&sk->sk_receive_queue.lock);
6275 } else {
6276 skb = skb_dequeue(&sk->sk_receive_queue);
6277 }
6278
6279 if (skb)
6280 return skb;
6281
6282 /* Caller is allowed not to check sk->sk_err before calling. */
6283 error = sock_error(sk);
6284 if (error)
6285 goto no_packet;
6286
6287 if (sk->sk_shutdown & RCV_SHUTDOWN)
6288 break;
6289
6290 /* User doesn't want to wait. */
6291 error = -EAGAIN;
6292 if (!timeo)
6293 goto no_packet;
6294 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6295
6296 return NULL;
6297
6298 no_packet:
6299 *err = error;
6300 return NULL;
6301 }
6302
6303 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6304 static void __sctp_write_space(struct sctp_association *asoc)
6305 {
6306 struct sock *sk = asoc->base.sk;
6307 struct socket *sock = sk->sk_socket;
6308
6309 if ((sctp_wspace(asoc) > 0) && sock) {
6310 if (waitqueue_active(&asoc->wait))
6311 wake_up_interruptible(&asoc->wait);
6312
6313 if (sctp_writeable(sk)) {
6314 wait_queue_head_t *wq = sk_sleep(sk);
6315
6316 if (wq && waitqueue_active(wq))
6317 wake_up_interruptible(wq);
6318
6319 /* Note that we try to include the Async I/O support
6320 * here by modeling from the current TCP/UDP code.
6321 * We have not tested with it yet.
6322 */
6323 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6324 sock_wake_async(sock,
6325 SOCK_WAKE_SPACE, POLL_OUT);
6326 }
6327 }
6328 }
6329
6330 /* Do accounting for the sndbuf space.
6331 * Decrement the used sndbuf space of the corresponding association by the
6332 * data size which was just transmitted(freed).
6333 */
6334 static void sctp_wfree(struct sk_buff *skb)
6335 {
6336 struct sctp_association *asoc;
6337 struct sctp_chunk *chunk;
6338 struct sock *sk;
6339
6340 /* Get the saved chunk pointer. */
6341 chunk = *((struct sctp_chunk **)(skb->cb));
6342 asoc = chunk->asoc;
6343 sk = asoc->base.sk;
6344 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6345 sizeof(struct sk_buff) +
6346 sizeof(struct sctp_chunk);
6347
6348 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6349
6350 /*
6351 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6352 */
6353 sk->sk_wmem_queued -= skb->truesize;
6354 sk_mem_uncharge(sk, skb->truesize);
6355
6356 sock_wfree(skb);
6357 __sctp_write_space(asoc);
6358
6359 sctp_association_put(asoc);
6360 }
6361
6362 /* Do accounting for the receive space on the socket.
6363 * Accounting for the association is done in ulpevent.c
6364 * We set this as a destructor for the cloned data skbs so that
6365 * accounting is done at the correct time.
6366 */
6367 void sctp_sock_rfree(struct sk_buff *skb)
6368 {
6369 struct sock *sk = skb->sk;
6370 struct sctp_ulpevent *event = sctp_skb2event(skb);
6371
6372 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6373
6374 /*
6375 * Mimic the behavior of sock_rfree
6376 */
6377 sk_mem_uncharge(sk, event->rmem_len);
6378 }
6379
6380
6381 /* Helper function to wait for space in the sndbuf. */
6382 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6383 size_t msg_len)
6384 {
6385 struct sock *sk = asoc->base.sk;
6386 int err = 0;
6387 long current_timeo = *timeo_p;
6388 DEFINE_WAIT(wait);
6389
6390 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6391 asoc, (long)(*timeo_p), msg_len);
6392
6393 /* Increment the association's refcnt. */
6394 sctp_association_hold(asoc);
6395
6396 /* Wait on the association specific sndbuf space. */
6397 for (;;) {
6398 prepare_to_wait_exclusive(&asoc->wait, &wait,
6399 TASK_INTERRUPTIBLE);
6400 if (!*timeo_p)
6401 goto do_nonblock;
6402 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6403 asoc->base.dead)
6404 goto do_error;
6405 if (signal_pending(current))
6406 goto do_interrupted;
6407 if (msg_len <= sctp_wspace(asoc))
6408 break;
6409
6410 /* Let another process have a go. Since we are going
6411 * to sleep anyway.
6412 */
6413 sctp_release_sock(sk);
6414 current_timeo = schedule_timeout(current_timeo);
6415 BUG_ON(sk != asoc->base.sk);
6416 sctp_lock_sock(sk);
6417
6418 *timeo_p = current_timeo;
6419 }
6420
6421 out:
6422 finish_wait(&asoc->wait, &wait);
6423
6424 /* Release the association's refcnt. */
6425 sctp_association_put(asoc);
6426
6427 return err;
6428
6429 do_error:
6430 err = -EPIPE;
6431 goto out;
6432
6433 do_interrupted:
6434 err = sock_intr_errno(*timeo_p);
6435 goto out;
6436
6437 do_nonblock:
6438 err = -EAGAIN;
6439 goto out;
6440 }
6441
6442 void sctp_data_ready(struct sock *sk, int len)
6443 {
6444 struct socket_wq *wq;
6445
6446 rcu_read_lock();
6447 wq = rcu_dereference(sk->sk_wq);
6448 if (wq_has_sleeper(wq))
6449 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6450 POLLRDNORM | POLLRDBAND);
6451 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6452 rcu_read_unlock();
6453 }
6454
6455 /* If socket sndbuf has changed, wake up all per association waiters. */
6456 void sctp_write_space(struct sock *sk)
6457 {
6458 struct sctp_association *asoc;
6459
6460 /* Wake up the tasks in each wait queue. */
6461 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6462 __sctp_write_space(asoc);
6463 }
6464 }
6465
6466 /* Is there any sndbuf space available on the socket?
6467 *
6468 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6469 * associations on the same socket. For a UDP-style socket with
6470 * multiple associations, it is possible for it to be "unwriteable"
6471 * prematurely. I assume that this is acceptable because
6472 * a premature "unwriteable" is better than an accidental "writeable" which
6473 * would cause an unwanted block under certain circumstances. For the 1-1
6474 * UDP-style sockets or TCP-style sockets, this code should work.
6475 * - Daisy
6476 */
6477 static int sctp_writeable(struct sock *sk)
6478 {
6479 int amt = 0;
6480
6481 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6482 if (amt < 0)
6483 amt = 0;
6484 return amt;
6485 }
6486
6487 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6488 * returns immediately with EINPROGRESS.
6489 */
6490 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6491 {
6492 struct sock *sk = asoc->base.sk;
6493 int err = 0;
6494 long current_timeo = *timeo_p;
6495 DEFINE_WAIT(wait);
6496
6497 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6498 (long)(*timeo_p));
6499
6500 /* Increment the association's refcnt. */
6501 sctp_association_hold(asoc);
6502
6503 for (;;) {
6504 prepare_to_wait_exclusive(&asoc->wait, &wait,
6505 TASK_INTERRUPTIBLE);
6506 if (!*timeo_p)
6507 goto do_nonblock;
6508 if (sk->sk_shutdown & RCV_SHUTDOWN)
6509 break;
6510 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6511 asoc->base.dead)
6512 goto do_error;
6513 if (signal_pending(current))
6514 goto do_interrupted;
6515
6516 if (sctp_state(asoc, ESTABLISHED))
6517 break;
6518
6519 /* Let another process have a go. Since we are going
6520 * to sleep anyway.
6521 */
6522 sctp_release_sock(sk);
6523 current_timeo = schedule_timeout(current_timeo);
6524 sctp_lock_sock(sk);
6525
6526 *timeo_p = current_timeo;
6527 }
6528
6529 out:
6530 finish_wait(&asoc->wait, &wait);
6531
6532 /* Release the association's refcnt. */
6533 sctp_association_put(asoc);
6534
6535 return err;
6536
6537 do_error:
6538 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6539 err = -ETIMEDOUT;
6540 else
6541 err = -ECONNREFUSED;
6542 goto out;
6543
6544 do_interrupted:
6545 err = sock_intr_errno(*timeo_p);
6546 goto out;
6547
6548 do_nonblock:
6549 err = -EINPROGRESS;
6550 goto out;
6551 }
6552
6553 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6554 {
6555 struct sctp_endpoint *ep;
6556 int err = 0;
6557 DEFINE_WAIT(wait);
6558
6559 ep = sctp_sk(sk)->ep;
6560
6561
6562 for (;;) {
6563 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6564 TASK_INTERRUPTIBLE);
6565
6566 if (list_empty(&ep->asocs)) {
6567 sctp_release_sock(sk);
6568 timeo = schedule_timeout(timeo);
6569 sctp_lock_sock(sk);
6570 }
6571
6572 err = -EINVAL;
6573 if (!sctp_sstate(sk, LISTENING))
6574 break;
6575
6576 err = 0;
6577 if (!list_empty(&ep->asocs))
6578 break;
6579
6580 err = sock_intr_errno(timeo);
6581 if (signal_pending(current))
6582 break;
6583
6584 err = -EAGAIN;
6585 if (!timeo)
6586 break;
6587 }
6588
6589 finish_wait(sk_sleep(sk), &wait);
6590
6591 return err;
6592 }
6593
6594 static void sctp_wait_for_close(struct sock *sk, long timeout)
6595 {
6596 DEFINE_WAIT(wait);
6597
6598 do {
6599 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6600 if (list_empty(&sctp_sk(sk)->ep->asocs))
6601 break;
6602 sctp_release_sock(sk);
6603 timeout = schedule_timeout(timeout);
6604 sctp_lock_sock(sk);
6605 } while (!signal_pending(current) && timeout);
6606
6607 finish_wait(sk_sleep(sk), &wait);
6608 }
6609
6610 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6611 {
6612 struct sk_buff *frag;
6613
6614 if (!skb->data_len)
6615 goto done;
6616
6617 /* Don't forget the fragments. */
6618 skb_walk_frags(skb, frag)
6619 sctp_skb_set_owner_r_frag(frag, sk);
6620
6621 done:
6622 sctp_skb_set_owner_r(skb, sk);
6623 }
6624
6625 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6626 struct sctp_association *asoc)
6627 {
6628 struct inet_sock *inet = inet_sk(sk);
6629 struct inet_sock *newinet;
6630
6631 newsk->sk_type = sk->sk_type;
6632 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6633 newsk->sk_flags = sk->sk_flags;
6634 newsk->sk_no_check = sk->sk_no_check;
6635 newsk->sk_reuse = sk->sk_reuse;
6636
6637 newsk->sk_shutdown = sk->sk_shutdown;
6638 newsk->sk_destruct = inet_sock_destruct;
6639 newsk->sk_family = sk->sk_family;
6640 newsk->sk_protocol = IPPROTO_SCTP;
6641 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6642 newsk->sk_sndbuf = sk->sk_sndbuf;
6643 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6644 newsk->sk_lingertime = sk->sk_lingertime;
6645 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6646 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6647
6648 newinet = inet_sk(newsk);
6649
6650 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6651 * getsockname() and getpeername()
6652 */
6653 newinet->inet_sport = inet->inet_sport;
6654 newinet->inet_saddr = inet->inet_saddr;
6655 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6656 newinet->inet_dport = htons(asoc->peer.port);
6657 newinet->pmtudisc = inet->pmtudisc;
6658 newinet->inet_id = asoc->next_tsn ^ jiffies;
6659
6660 newinet->uc_ttl = inet->uc_ttl;
6661 newinet->mc_loop = 1;
6662 newinet->mc_ttl = 1;
6663 newinet->mc_index = 0;
6664 newinet->mc_list = NULL;
6665 }
6666
6667 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6668 * and its messages to the newsk.
6669 */
6670 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6671 struct sctp_association *assoc,
6672 sctp_socket_type_t type)
6673 {
6674 struct sctp_sock *oldsp = sctp_sk(oldsk);
6675 struct sctp_sock *newsp = sctp_sk(newsk);
6676 struct sctp_bind_bucket *pp; /* hash list port iterator */
6677 struct sctp_endpoint *newep = newsp->ep;
6678 struct sk_buff *skb, *tmp;
6679 struct sctp_ulpevent *event;
6680 struct sctp_bind_hashbucket *head;
6681 struct list_head tmplist;
6682
6683 /* Migrate socket buffer sizes and all the socket level options to the
6684 * new socket.
6685 */
6686 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6687 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6688 /* Brute force copy old sctp opt. */
6689 if (oldsp->do_auto_asconf) {
6690 memcpy(&tmplist, &newsp->auto_asconf_list, sizeof(tmplist));
6691 inet_sk_copy_descendant(newsk, oldsk);
6692 memcpy(&newsp->auto_asconf_list, &tmplist, sizeof(tmplist));
6693 } else
6694 inet_sk_copy_descendant(newsk, oldsk);
6695
6696 /* Restore the ep value that was overwritten with the above structure
6697 * copy.
6698 */
6699 newsp->ep = newep;
6700 newsp->hmac = NULL;
6701
6702 /* Hook this new socket in to the bind_hash list. */
6703 head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->inet_num)];
6704 sctp_local_bh_disable();
6705 sctp_spin_lock(&head->lock);
6706 pp = sctp_sk(oldsk)->bind_hash;
6707 sk_add_bind_node(newsk, &pp->owner);
6708 sctp_sk(newsk)->bind_hash = pp;
6709 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6710 sctp_spin_unlock(&head->lock);
6711 sctp_local_bh_enable();
6712
6713 /* Copy the bind_addr list from the original endpoint to the new
6714 * endpoint so that we can handle restarts properly
6715 */
6716 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6717 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6718
6719 /* Move any messages in the old socket's receive queue that are for the
6720 * peeled off association to the new socket's receive queue.
6721 */
6722 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6723 event = sctp_skb2event(skb);
6724 if (event->asoc == assoc) {
6725 __skb_unlink(skb, &oldsk->sk_receive_queue);
6726 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6727 sctp_skb_set_owner_r_frag(skb, newsk);
6728 }
6729 }
6730
6731 /* Clean up any messages pending delivery due to partial
6732 * delivery. Three cases:
6733 * 1) No partial deliver; no work.
6734 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6735 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6736 */
6737 skb_queue_head_init(&newsp->pd_lobby);
6738 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6739
6740 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6741 struct sk_buff_head *queue;
6742
6743 /* Decide which queue to move pd_lobby skbs to. */
6744 if (assoc->ulpq.pd_mode) {
6745 queue = &newsp->pd_lobby;
6746 } else
6747 queue = &newsk->sk_receive_queue;
6748
6749 /* Walk through the pd_lobby, looking for skbs that
6750 * need moved to the new socket.
6751 */
6752 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6753 event = sctp_skb2event(skb);
6754 if (event->asoc == assoc) {
6755 __skb_unlink(skb, &oldsp->pd_lobby);
6756 __skb_queue_tail(queue, skb);
6757 sctp_skb_set_owner_r_frag(skb, newsk);
6758 }
6759 }
6760
6761 /* Clear up any skbs waiting for the partial
6762 * delivery to finish.
6763 */
6764 if (assoc->ulpq.pd_mode)
6765 sctp_clear_pd(oldsk, NULL);
6766
6767 }
6768
6769 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6770 sctp_skb_set_owner_r_frag(skb, newsk);
6771
6772 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6773 sctp_skb_set_owner_r_frag(skb, newsk);
6774
6775 /* Set the type of socket to indicate that it is peeled off from the
6776 * original UDP-style socket or created with the accept() call on a
6777 * TCP-style socket..
6778 */
6779 newsp->type = type;
6780
6781 /* Mark the new socket "in-use" by the user so that any packets
6782 * that may arrive on the association after we've moved it are
6783 * queued to the backlog. This prevents a potential race between
6784 * backlog processing on the old socket and new-packet processing
6785 * on the new socket.
6786 *
6787 * The caller has just allocated newsk so we can guarantee that other
6788 * paths won't try to lock it and then oldsk.
6789 */
6790 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6791 sctp_assoc_migrate(assoc, newsk);
6792
6793 /* If the association on the newsk is already closed before accept()
6794 * is called, set RCV_SHUTDOWN flag.
6795 */
6796 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6797 newsk->sk_shutdown |= RCV_SHUTDOWN;
6798
6799 newsk->sk_state = SCTP_SS_ESTABLISHED;
6800 sctp_release_sock(newsk);
6801 }
6802
6803
6804 /* This proto struct describes the ULP interface for SCTP. */
6805 struct proto sctp_prot = {
6806 .name = "SCTP",
6807 .owner = THIS_MODULE,
6808 .close = sctp_close,
6809 .connect = sctp_connect,
6810 .disconnect = sctp_disconnect,
6811 .accept = sctp_accept,
6812 .ioctl = sctp_ioctl,
6813 .init = sctp_init_sock,
6814 .destroy = sctp_destroy_sock,
6815 .shutdown = sctp_shutdown,
6816 .setsockopt = sctp_setsockopt,
6817 .getsockopt = sctp_getsockopt,
6818 .sendmsg = sctp_sendmsg,
6819 .recvmsg = sctp_recvmsg,
6820 .bind = sctp_bind,
6821 .backlog_rcv = sctp_backlog_rcv,
6822 .hash = sctp_hash,
6823 .unhash = sctp_unhash,
6824 .get_port = sctp_get_port,
6825 .obj_size = sizeof(struct sctp_sock),
6826 .sysctl_mem = sysctl_sctp_mem,
6827 .sysctl_rmem = sysctl_sctp_rmem,
6828 .sysctl_wmem = sysctl_sctp_wmem,
6829 .memory_pressure = &sctp_memory_pressure,
6830 .enter_memory_pressure = sctp_enter_memory_pressure,
6831 .memory_allocated = &sctp_memory_allocated,
6832 .sockets_allocated = &sctp_sockets_allocated,
6833 };
6834
6835 #if IS_ENABLED(CONFIG_IPV6)
6836
6837 struct proto sctpv6_prot = {
6838 .name = "SCTPv6",
6839 .owner = THIS_MODULE,
6840 .close = sctp_close,
6841 .connect = sctp_connect,
6842 .disconnect = sctp_disconnect,
6843 .accept = sctp_accept,
6844 .ioctl = sctp_ioctl,
6845 .init = sctp_init_sock,
6846 .destroy = sctp_destroy_sock,
6847 .shutdown = sctp_shutdown,
6848 .setsockopt = sctp_setsockopt,
6849 .getsockopt = sctp_getsockopt,
6850 .sendmsg = sctp_sendmsg,
6851 .recvmsg = sctp_recvmsg,
6852 .bind = sctp_bind,
6853 .backlog_rcv = sctp_backlog_rcv,
6854 .hash = sctp_hash,
6855 .unhash = sctp_unhash,
6856 .get_port = sctp_get_port,
6857 .obj_size = sizeof(struct sctp6_sock),
6858 .sysctl_mem = sysctl_sctp_mem,
6859 .sysctl_rmem = sysctl_sctp_rmem,
6860 .sysctl_wmem = sysctl_sctp_wmem,
6861 .memory_pressure = &sctp_memory_pressure,
6862 .enter_memory_pressure = sctp_enter_memory_pressure,
6863 .memory_allocated = &sctp_memory_allocated,
6864 .sockets_allocated = &sctp_sockets_allocated,
6865 };
6866 #endif /* IS_ENABLED(CONFIG_IPV6) */