Merge branch 'pci/resource' into next
[deliverable/linux.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, see
32 * <http://www.gnu.org/licenses/>.
33 *
34 * Please send any bug reports or fixes you make to the
35 * email address(es):
36 * lksctp developers <linux-sctp@vger.kernel.org>
37 *
38 * Written or modified by:
39 * La Monte H.P. Yarroll <piggy@acm.org>
40 * Narasimha Budihal <narsi@refcode.org>
41 * Karl Knutson <karl@athena.chicago.il.us>
42 * Jon Grimm <jgrimm@us.ibm.com>
43 * Xingang Guo <xingang.guo@intel.com>
44 * Daisy Chang <daisyc@us.ibm.com>
45 * Sridhar Samudrala <samudrala@us.ibm.com>
46 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
47 * Ardelle Fan <ardelle.fan@intel.com>
48 * Ryan Layer <rmlayer@us.ibm.com>
49 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
50 * Kevin Gao <kevin.gao@intel.com>
51 */
52
53 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
54
55 #include <linux/types.h>
56 #include <linux/kernel.h>
57 #include <linux/wait.h>
58 #include <linux/time.h>
59 #include <linux/ip.h>
60 #include <linux/capability.h>
61 #include <linux/fcntl.h>
62 #include <linux/poll.h>
63 #include <linux/init.h>
64 #include <linux/crypto.h>
65 #include <linux/slab.h>
66 #include <linux/file.h>
67
68 #include <net/ip.h>
69 #include <net/icmp.h>
70 #include <net/route.h>
71 #include <net/ipv6.h>
72 #include <net/inet_common.h>
73
74 #include <linux/socket.h> /* for sa_family_t */
75 #include <linux/export.h>
76 #include <net/sock.h>
77 #include <net/sctp/sctp.h>
78 #include <net/sctp/sm.h>
79
80 /* Forward declarations for internal helper functions. */
81 static int sctp_writeable(struct sock *sk);
82 static void sctp_wfree(struct sk_buff *skb);
83 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
84 size_t msg_len);
85 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p);
86 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
87 static int sctp_wait_for_accept(struct sock *sk, long timeo);
88 static void sctp_wait_for_close(struct sock *sk, long timeo);
89 static void sctp_destruct_sock(struct sock *sk);
90 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
91 union sctp_addr *addr, int len);
92 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
93 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
94 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
95 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
96 static int sctp_send_asconf(struct sctp_association *asoc,
97 struct sctp_chunk *chunk);
98 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
99 static int sctp_autobind(struct sock *sk);
100 static void sctp_sock_migrate(struct sock *, struct sock *,
101 struct sctp_association *, sctp_socket_type_t);
102
103 extern struct kmem_cache *sctp_bucket_cachep;
104 extern long sysctl_sctp_mem[3];
105 extern int sysctl_sctp_rmem[3];
106 extern int sysctl_sctp_wmem[3];
107
108 static int sctp_memory_pressure;
109 static atomic_long_t sctp_memory_allocated;
110 struct percpu_counter sctp_sockets_allocated;
111
112 static void sctp_enter_memory_pressure(struct sock *sk)
113 {
114 sctp_memory_pressure = 1;
115 }
116
117
118 /* Get the sndbuf space available at the time on the association. */
119 static inline int sctp_wspace(struct sctp_association *asoc)
120 {
121 int amt;
122
123 if (asoc->ep->sndbuf_policy)
124 amt = asoc->sndbuf_used;
125 else
126 amt = sk_wmem_alloc_get(asoc->base.sk);
127
128 if (amt >= asoc->base.sk->sk_sndbuf) {
129 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
130 amt = 0;
131 else {
132 amt = sk_stream_wspace(asoc->base.sk);
133 if (amt < 0)
134 amt = 0;
135 }
136 } else {
137 amt = asoc->base.sk->sk_sndbuf - amt;
138 }
139 return amt;
140 }
141
142 /* Increment the used sndbuf space count of the corresponding association by
143 * the size of the outgoing data chunk.
144 * Also, set the skb destructor for sndbuf accounting later.
145 *
146 * Since it is always 1-1 between chunk and skb, and also a new skb is always
147 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
148 * destructor in the data chunk skb for the purpose of the sndbuf space
149 * tracking.
150 */
151 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
152 {
153 struct sctp_association *asoc = chunk->asoc;
154 struct sock *sk = asoc->base.sk;
155
156 /* The sndbuf space is tracked per association. */
157 sctp_association_hold(asoc);
158
159 skb_set_owner_w(chunk->skb, sk);
160
161 chunk->skb->destructor = sctp_wfree;
162 /* Save the chunk pointer in skb for sctp_wfree to use later. */
163 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
164
165 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
166 sizeof(struct sk_buff) +
167 sizeof(struct sctp_chunk);
168
169 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
170 sk->sk_wmem_queued += chunk->skb->truesize;
171 sk_mem_charge(sk, chunk->skb->truesize);
172 }
173
174 /* Verify that this is a valid address. */
175 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
176 int len)
177 {
178 struct sctp_af *af;
179
180 /* Verify basic sockaddr. */
181 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
182 if (!af)
183 return -EINVAL;
184
185 /* Is this a valid SCTP address? */
186 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
187 return -EINVAL;
188
189 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
190 return -EINVAL;
191
192 return 0;
193 }
194
195 /* Look up the association by its id. If this is not a UDP-style
196 * socket, the ID field is always ignored.
197 */
198 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
199 {
200 struct sctp_association *asoc = NULL;
201
202 /* If this is not a UDP-style socket, assoc id should be ignored. */
203 if (!sctp_style(sk, UDP)) {
204 /* Return NULL if the socket state is not ESTABLISHED. It
205 * could be a TCP-style listening socket or a socket which
206 * hasn't yet called connect() to establish an association.
207 */
208 if (!sctp_sstate(sk, ESTABLISHED))
209 return NULL;
210
211 /* Get the first and the only association from the list. */
212 if (!list_empty(&sctp_sk(sk)->ep->asocs))
213 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
214 struct sctp_association, asocs);
215 return asoc;
216 }
217
218 /* Otherwise this is a UDP-style socket. */
219 if (!id || (id == (sctp_assoc_t)-1))
220 return NULL;
221
222 spin_lock_bh(&sctp_assocs_id_lock);
223 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
224 spin_unlock_bh(&sctp_assocs_id_lock);
225
226 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
227 return NULL;
228
229 return asoc;
230 }
231
232 /* Look up the transport from an address and an assoc id. If both address and
233 * id are specified, the associations matching the address and the id should be
234 * the same.
235 */
236 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
237 struct sockaddr_storage *addr,
238 sctp_assoc_t id)
239 {
240 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
241 struct sctp_transport *transport;
242 union sctp_addr *laddr = (union sctp_addr *)addr;
243
244 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
245 laddr,
246 &transport);
247
248 if (!addr_asoc)
249 return NULL;
250
251 id_asoc = sctp_id2assoc(sk, id);
252 if (id_asoc && (id_asoc != addr_asoc))
253 return NULL;
254
255 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
256 (union sctp_addr *)addr);
257
258 return transport;
259 }
260
261 /* API 3.1.2 bind() - UDP Style Syntax
262 * The syntax of bind() is,
263 *
264 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
265 *
266 * sd - the socket descriptor returned by socket().
267 * addr - the address structure (struct sockaddr_in or struct
268 * sockaddr_in6 [RFC 2553]),
269 * addr_len - the size of the address structure.
270 */
271 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
272 {
273 int retval = 0;
274
275 lock_sock(sk);
276
277 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
278 addr, addr_len);
279
280 /* Disallow binding twice. */
281 if (!sctp_sk(sk)->ep->base.bind_addr.port)
282 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
283 addr_len);
284 else
285 retval = -EINVAL;
286
287 release_sock(sk);
288
289 return retval;
290 }
291
292 static long sctp_get_port_local(struct sock *, union sctp_addr *);
293
294 /* Verify this is a valid sockaddr. */
295 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
296 union sctp_addr *addr, int len)
297 {
298 struct sctp_af *af;
299
300 /* Check minimum size. */
301 if (len < sizeof (struct sockaddr))
302 return NULL;
303
304 /* V4 mapped address are really of AF_INET family */
305 if (addr->sa.sa_family == AF_INET6 &&
306 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
307 if (!opt->pf->af_supported(AF_INET, opt))
308 return NULL;
309 } else {
310 /* Does this PF support this AF? */
311 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
312 return NULL;
313 }
314
315 /* If we get this far, af is valid. */
316 af = sctp_get_af_specific(addr->sa.sa_family);
317
318 if (len < af->sockaddr_len)
319 return NULL;
320
321 return af;
322 }
323
324 /* Bind a local address either to an endpoint or to an association. */
325 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
326 {
327 struct net *net = sock_net(sk);
328 struct sctp_sock *sp = sctp_sk(sk);
329 struct sctp_endpoint *ep = sp->ep;
330 struct sctp_bind_addr *bp = &ep->base.bind_addr;
331 struct sctp_af *af;
332 unsigned short snum;
333 int ret = 0;
334
335 /* Common sockaddr verification. */
336 af = sctp_sockaddr_af(sp, addr, len);
337 if (!af) {
338 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
339 __func__, sk, addr, len);
340 return -EINVAL;
341 }
342
343 snum = ntohs(addr->v4.sin_port);
344
345 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
346 __func__, sk, &addr->sa, bp->port, snum, len);
347
348 /* PF specific bind() address verification. */
349 if (!sp->pf->bind_verify(sp, addr))
350 return -EADDRNOTAVAIL;
351
352 /* We must either be unbound, or bind to the same port.
353 * It's OK to allow 0 ports if we are already bound.
354 * We'll just inhert an already bound port in this case
355 */
356 if (bp->port) {
357 if (!snum)
358 snum = bp->port;
359 else if (snum != bp->port) {
360 pr_debug("%s: new port %d doesn't match existing port "
361 "%d\n", __func__, snum, bp->port);
362 return -EINVAL;
363 }
364 }
365
366 if (snum && snum < PROT_SOCK &&
367 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
368 return -EACCES;
369
370 /* See if the address matches any of the addresses we may have
371 * already bound before checking against other endpoints.
372 */
373 if (sctp_bind_addr_match(bp, addr, sp))
374 return -EINVAL;
375
376 /* Make sure we are allowed to bind here.
377 * The function sctp_get_port_local() does duplicate address
378 * detection.
379 */
380 addr->v4.sin_port = htons(snum);
381 if ((ret = sctp_get_port_local(sk, addr))) {
382 return -EADDRINUSE;
383 }
384
385 /* Refresh ephemeral port. */
386 if (!bp->port)
387 bp->port = inet_sk(sk)->inet_num;
388
389 /* Add the address to the bind address list.
390 * Use GFP_ATOMIC since BHs will be disabled.
391 */
392 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
393
394 /* Copy back into socket for getsockname() use. */
395 if (!ret) {
396 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
397 af->to_sk_saddr(addr, sk);
398 }
399
400 return ret;
401 }
402
403 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
404 *
405 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
406 * at any one time. If a sender, after sending an ASCONF chunk, decides
407 * it needs to transfer another ASCONF Chunk, it MUST wait until the
408 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
409 * subsequent ASCONF. Note this restriction binds each side, so at any
410 * time two ASCONF may be in-transit on any given association (one sent
411 * from each endpoint).
412 */
413 static int sctp_send_asconf(struct sctp_association *asoc,
414 struct sctp_chunk *chunk)
415 {
416 struct net *net = sock_net(asoc->base.sk);
417 int retval = 0;
418
419 /* If there is an outstanding ASCONF chunk, queue it for later
420 * transmission.
421 */
422 if (asoc->addip_last_asconf) {
423 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
424 goto out;
425 }
426
427 /* Hold the chunk until an ASCONF_ACK is received. */
428 sctp_chunk_hold(chunk);
429 retval = sctp_primitive_ASCONF(net, asoc, chunk);
430 if (retval)
431 sctp_chunk_free(chunk);
432 else
433 asoc->addip_last_asconf = chunk;
434
435 out:
436 return retval;
437 }
438
439 /* Add a list of addresses as bind addresses to local endpoint or
440 * association.
441 *
442 * Basically run through each address specified in the addrs/addrcnt
443 * array/length pair, determine if it is IPv6 or IPv4 and call
444 * sctp_do_bind() on it.
445 *
446 * If any of them fails, then the operation will be reversed and the
447 * ones that were added will be removed.
448 *
449 * Only sctp_setsockopt_bindx() is supposed to call this function.
450 */
451 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
452 {
453 int cnt;
454 int retval = 0;
455 void *addr_buf;
456 struct sockaddr *sa_addr;
457 struct sctp_af *af;
458
459 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
460 addrs, addrcnt);
461
462 addr_buf = addrs;
463 for (cnt = 0; cnt < addrcnt; cnt++) {
464 /* The list may contain either IPv4 or IPv6 address;
465 * determine the address length for walking thru the list.
466 */
467 sa_addr = addr_buf;
468 af = sctp_get_af_specific(sa_addr->sa_family);
469 if (!af) {
470 retval = -EINVAL;
471 goto err_bindx_add;
472 }
473
474 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
475 af->sockaddr_len);
476
477 addr_buf += af->sockaddr_len;
478
479 err_bindx_add:
480 if (retval < 0) {
481 /* Failed. Cleanup the ones that have been added */
482 if (cnt > 0)
483 sctp_bindx_rem(sk, addrs, cnt);
484 return retval;
485 }
486 }
487
488 return retval;
489 }
490
491 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
492 * associations that are part of the endpoint indicating that a list of local
493 * addresses are added to the endpoint.
494 *
495 * If any of the addresses is already in the bind address list of the
496 * association, we do not send the chunk for that association. But it will not
497 * affect other associations.
498 *
499 * Only sctp_setsockopt_bindx() is supposed to call this function.
500 */
501 static int sctp_send_asconf_add_ip(struct sock *sk,
502 struct sockaddr *addrs,
503 int addrcnt)
504 {
505 struct net *net = sock_net(sk);
506 struct sctp_sock *sp;
507 struct sctp_endpoint *ep;
508 struct sctp_association *asoc;
509 struct sctp_bind_addr *bp;
510 struct sctp_chunk *chunk;
511 struct sctp_sockaddr_entry *laddr;
512 union sctp_addr *addr;
513 union sctp_addr saveaddr;
514 void *addr_buf;
515 struct sctp_af *af;
516 struct list_head *p;
517 int i;
518 int retval = 0;
519
520 if (!net->sctp.addip_enable)
521 return retval;
522
523 sp = sctp_sk(sk);
524 ep = sp->ep;
525
526 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
527 __func__, sk, addrs, addrcnt);
528
529 list_for_each_entry(asoc, &ep->asocs, asocs) {
530 if (!asoc->peer.asconf_capable)
531 continue;
532
533 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
534 continue;
535
536 if (!sctp_state(asoc, ESTABLISHED))
537 continue;
538
539 /* Check if any address in the packed array of addresses is
540 * in the bind address list of the association. If so,
541 * do not send the asconf chunk to its peer, but continue with
542 * other associations.
543 */
544 addr_buf = addrs;
545 for (i = 0; i < addrcnt; i++) {
546 addr = addr_buf;
547 af = sctp_get_af_specific(addr->v4.sin_family);
548 if (!af) {
549 retval = -EINVAL;
550 goto out;
551 }
552
553 if (sctp_assoc_lookup_laddr(asoc, addr))
554 break;
555
556 addr_buf += af->sockaddr_len;
557 }
558 if (i < addrcnt)
559 continue;
560
561 /* Use the first valid address in bind addr list of
562 * association as Address Parameter of ASCONF CHUNK.
563 */
564 bp = &asoc->base.bind_addr;
565 p = bp->address_list.next;
566 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
567 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
568 addrcnt, SCTP_PARAM_ADD_IP);
569 if (!chunk) {
570 retval = -ENOMEM;
571 goto out;
572 }
573
574 /* Add the new addresses to the bind address list with
575 * use_as_src set to 0.
576 */
577 addr_buf = addrs;
578 for (i = 0; i < addrcnt; i++) {
579 addr = addr_buf;
580 af = sctp_get_af_specific(addr->v4.sin_family);
581 memcpy(&saveaddr, addr, af->sockaddr_len);
582 retval = sctp_add_bind_addr(bp, &saveaddr,
583 SCTP_ADDR_NEW, GFP_ATOMIC);
584 addr_buf += af->sockaddr_len;
585 }
586 if (asoc->src_out_of_asoc_ok) {
587 struct sctp_transport *trans;
588
589 list_for_each_entry(trans,
590 &asoc->peer.transport_addr_list, transports) {
591 /* Clear the source and route cache */
592 dst_release(trans->dst);
593 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
594 2*asoc->pathmtu, 4380));
595 trans->ssthresh = asoc->peer.i.a_rwnd;
596 trans->rto = asoc->rto_initial;
597 sctp_max_rto(asoc, trans);
598 trans->rtt = trans->srtt = trans->rttvar = 0;
599 sctp_transport_route(trans, NULL,
600 sctp_sk(asoc->base.sk));
601 }
602 }
603 retval = sctp_send_asconf(asoc, chunk);
604 }
605
606 out:
607 return retval;
608 }
609
610 /* Remove a list of addresses from bind addresses list. Do not remove the
611 * last address.
612 *
613 * Basically run through each address specified in the addrs/addrcnt
614 * array/length pair, determine if it is IPv6 or IPv4 and call
615 * sctp_del_bind() on it.
616 *
617 * If any of them fails, then the operation will be reversed and the
618 * ones that were removed will be added back.
619 *
620 * At least one address has to be left; if only one address is
621 * available, the operation will return -EBUSY.
622 *
623 * Only sctp_setsockopt_bindx() is supposed to call this function.
624 */
625 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
626 {
627 struct sctp_sock *sp = sctp_sk(sk);
628 struct sctp_endpoint *ep = sp->ep;
629 int cnt;
630 struct sctp_bind_addr *bp = &ep->base.bind_addr;
631 int retval = 0;
632 void *addr_buf;
633 union sctp_addr *sa_addr;
634 struct sctp_af *af;
635
636 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
637 __func__, sk, addrs, addrcnt);
638
639 addr_buf = addrs;
640 for (cnt = 0; cnt < addrcnt; cnt++) {
641 /* If the bind address list is empty or if there is only one
642 * bind address, there is nothing more to be removed (we need
643 * at least one address here).
644 */
645 if (list_empty(&bp->address_list) ||
646 (sctp_list_single_entry(&bp->address_list))) {
647 retval = -EBUSY;
648 goto err_bindx_rem;
649 }
650
651 sa_addr = addr_buf;
652 af = sctp_get_af_specific(sa_addr->sa.sa_family);
653 if (!af) {
654 retval = -EINVAL;
655 goto err_bindx_rem;
656 }
657
658 if (!af->addr_valid(sa_addr, sp, NULL)) {
659 retval = -EADDRNOTAVAIL;
660 goto err_bindx_rem;
661 }
662
663 if (sa_addr->v4.sin_port &&
664 sa_addr->v4.sin_port != htons(bp->port)) {
665 retval = -EINVAL;
666 goto err_bindx_rem;
667 }
668
669 if (!sa_addr->v4.sin_port)
670 sa_addr->v4.sin_port = htons(bp->port);
671
672 /* FIXME - There is probably a need to check if sk->sk_saddr and
673 * sk->sk_rcv_addr are currently set to one of the addresses to
674 * be removed. This is something which needs to be looked into
675 * when we are fixing the outstanding issues with multi-homing
676 * socket routing and failover schemes. Refer to comments in
677 * sctp_do_bind(). -daisy
678 */
679 retval = sctp_del_bind_addr(bp, sa_addr);
680
681 addr_buf += af->sockaddr_len;
682 err_bindx_rem:
683 if (retval < 0) {
684 /* Failed. Add the ones that has been removed back */
685 if (cnt > 0)
686 sctp_bindx_add(sk, addrs, cnt);
687 return retval;
688 }
689 }
690
691 return retval;
692 }
693
694 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
695 * the associations that are part of the endpoint indicating that a list of
696 * local addresses are removed from the endpoint.
697 *
698 * If any of the addresses is already in the bind address list of the
699 * association, we do not send the chunk for that association. But it will not
700 * affect other associations.
701 *
702 * Only sctp_setsockopt_bindx() is supposed to call this function.
703 */
704 static int sctp_send_asconf_del_ip(struct sock *sk,
705 struct sockaddr *addrs,
706 int addrcnt)
707 {
708 struct net *net = sock_net(sk);
709 struct sctp_sock *sp;
710 struct sctp_endpoint *ep;
711 struct sctp_association *asoc;
712 struct sctp_transport *transport;
713 struct sctp_bind_addr *bp;
714 struct sctp_chunk *chunk;
715 union sctp_addr *laddr;
716 void *addr_buf;
717 struct sctp_af *af;
718 struct sctp_sockaddr_entry *saddr;
719 int i;
720 int retval = 0;
721 int stored = 0;
722
723 chunk = NULL;
724 if (!net->sctp.addip_enable)
725 return retval;
726
727 sp = sctp_sk(sk);
728 ep = sp->ep;
729
730 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
731 __func__, sk, addrs, addrcnt);
732
733 list_for_each_entry(asoc, &ep->asocs, asocs) {
734
735 if (!asoc->peer.asconf_capable)
736 continue;
737
738 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
739 continue;
740
741 if (!sctp_state(asoc, ESTABLISHED))
742 continue;
743
744 /* Check if any address in the packed array of addresses is
745 * not present in the bind address list of the association.
746 * If so, do not send the asconf chunk to its peer, but
747 * continue with other associations.
748 */
749 addr_buf = addrs;
750 for (i = 0; i < addrcnt; i++) {
751 laddr = addr_buf;
752 af = sctp_get_af_specific(laddr->v4.sin_family);
753 if (!af) {
754 retval = -EINVAL;
755 goto out;
756 }
757
758 if (!sctp_assoc_lookup_laddr(asoc, laddr))
759 break;
760
761 addr_buf += af->sockaddr_len;
762 }
763 if (i < addrcnt)
764 continue;
765
766 /* Find one address in the association's bind address list
767 * that is not in the packed array of addresses. This is to
768 * make sure that we do not delete all the addresses in the
769 * association.
770 */
771 bp = &asoc->base.bind_addr;
772 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
773 addrcnt, sp);
774 if ((laddr == NULL) && (addrcnt == 1)) {
775 if (asoc->asconf_addr_del_pending)
776 continue;
777 asoc->asconf_addr_del_pending =
778 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
779 if (asoc->asconf_addr_del_pending == NULL) {
780 retval = -ENOMEM;
781 goto out;
782 }
783 asoc->asconf_addr_del_pending->sa.sa_family =
784 addrs->sa_family;
785 asoc->asconf_addr_del_pending->v4.sin_port =
786 htons(bp->port);
787 if (addrs->sa_family == AF_INET) {
788 struct sockaddr_in *sin;
789
790 sin = (struct sockaddr_in *)addrs;
791 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
792 } else if (addrs->sa_family == AF_INET6) {
793 struct sockaddr_in6 *sin6;
794
795 sin6 = (struct sockaddr_in6 *)addrs;
796 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
797 }
798
799 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
800 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
801 asoc->asconf_addr_del_pending);
802
803 asoc->src_out_of_asoc_ok = 1;
804 stored = 1;
805 goto skip_mkasconf;
806 }
807
808 if (laddr == NULL)
809 return -EINVAL;
810
811 /* We do not need RCU protection throughout this loop
812 * because this is done under a socket lock from the
813 * setsockopt call.
814 */
815 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
816 SCTP_PARAM_DEL_IP);
817 if (!chunk) {
818 retval = -ENOMEM;
819 goto out;
820 }
821
822 skip_mkasconf:
823 /* Reset use_as_src flag for the addresses in the bind address
824 * list that are to be deleted.
825 */
826 addr_buf = addrs;
827 for (i = 0; i < addrcnt; i++) {
828 laddr = addr_buf;
829 af = sctp_get_af_specific(laddr->v4.sin_family);
830 list_for_each_entry(saddr, &bp->address_list, list) {
831 if (sctp_cmp_addr_exact(&saddr->a, laddr))
832 saddr->state = SCTP_ADDR_DEL;
833 }
834 addr_buf += af->sockaddr_len;
835 }
836
837 /* Update the route and saddr entries for all the transports
838 * as some of the addresses in the bind address list are
839 * about to be deleted and cannot be used as source addresses.
840 */
841 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
842 transports) {
843 dst_release(transport->dst);
844 sctp_transport_route(transport, NULL,
845 sctp_sk(asoc->base.sk));
846 }
847
848 if (stored)
849 /* We don't need to transmit ASCONF */
850 continue;
851 retval = sctp_send_asconf(asoc, chunk);
852 }
853 out:
854 return retval;
855 }
856
857 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
858 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
859 {
860 struct sock *sk = sctp_opt2sk(sp);
861 union sctp_addr *addr;
862 struct sctp_af *af;
863
864 /* It is safe to write port space in caller. */
865 addr = &addrw->a;
866 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
867 af = sctp_get_af_specific(addr->sa.sa_family);
868 if (!af)
869 return -EINVAL;
870 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
871 return -EINVAL;
872
873 if (addrw->state == SCTP_ADDR_NEW)
874 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
875 else
876 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
877 }
878
879 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
880 *
881 * API 8.1
882 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
883 * int flags);
884 *
885 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
886 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
887 * or IPv6 addresses.
888 *
889 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
890 * Section 3.1.2 for this usage.
891 *
892 * addrs is a pointer to an array of one or more socket addresses. Each
893 * address is contained in its appropriate structure (i.e. struct
894 * sockaddr_in or struct sockaddr_in6) the family of the address type
895 * must be used to distinguish the address length (note that this
896 * representation is termed a "packed array" of addresses). The caller
897 * specifies the number of addresses in the array with addrcnt.
898 *
899 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
900 * -1, and sets errno to the appropriate error code.
901 *
902 * For SCTP, the port given in each socket address must be the same, or
903 * sctp_bindx() will fail, setting errno to EINVAL.
904 *
905 * The flags parameter is formed from the bitwise OR of zero or more of
906 * the following currently defined flags:
907 *
908 * SCTP_BINDX_ADD_ADDR
909 *
910 * SCTP_BINDX_REM_ADDR
911 *
912 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
913 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
914 * addresses from the association. The two flags are mutually exclusive;
915 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
916 * not remove all addresses from an association; sctp_bindx() will
917 * reject such an attempt with EINVAL.
918 *
919 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
920 * additional addresses with an endpoint after calling bind(). Or use
921 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
922 * socket is associated with so that no new association accepted will be
923 * associated with those addresses. If the endpoint supports dynamic
924 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
925 * endpoint to send the appropriate message to the peer to change the
926 * peers address lists.
927 *
928 * Adding and removing addresses from a connected association is
929 * optional functionality. Implementations that do not support this
930 * functionality should return EOPNOTSUPP.
931 *
932 * Basically do nothing but copying the addresses from user to kernel
933 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
934 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
935 * from userspace.
936 *
937 * We don't use copy_from_user() for optimization: we first do the
938 * sanity checks (buffer size -fast- and access check-healthy
939 * pointer); if all of those succeed, then we can alloc the memory
940 * (expensive operation) needed to copy the data to kernel. Then we do
941 * the copying without checking the user space area
942 * (__copy_from_user()).
943 *
944 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
945 * it.
946 *
947 * sk The sk of the socket
948 * addrs The pointer to the addresses in user land
949 * addrssize Size of the addrs buffer
950 * op Operation to perform (add or remove, see the flags of
951 * sctp_bindx)
952 *
953 * Returns 0 if ok, <0 errno code on error.
954 */
955 static int sctp_setsockopt_bindx(struct sock *sk,
956 struct sockaddr __user *addrs,
957 int addrs_size, int op)
958 {
959 struct sockaddr *kaddrs;
960 int err;
961 int addrcnt = 0;
962 int walk_size = 0;
963 struct sockaddr *sa_addr;
964 void *addr_buf;
965 struct sctp_af *af;
966
967 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
968 __func__, sk, addrs, addrs_size, op);
969
970 if (unlikely(addrs_size <= 0))
971 return -EINVAL;
972
973 /* Check the user passed a healthy pointer. */
974 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
975 return -EFAULT;
976
977 /* Alloc space for the address array in kernel memory. */
978 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
979 if (unlikely(!kaddrs))
980 return -ENOMEM;
981
982 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
983 kfree(kaddrs);
984 return -EFAULT;
985 }
986
987 /* Walk through the addrs buffer and count the number of addresses. */
988 addr_buf = kaddrs;
989 while (walk_size < addrs_size) {
990 if (walk_size + sizeof(sa_family_t) > addrs_size) {
991 kfree(kaddrs);
992 return -EINVAL;
993 }
994
995 sa_addr = addr_buf;
996 af = sctp_get_af_specific(sa_addr->sa_family);
997
998 /* If the address family is not supported or if this address
999 * causes the address buffer to overflow return EINVAL.
1000 */
1001 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1002 kfree(kaddrs);
1003 return -EINVAL;
1004 }
1005 addrcnt++;
1006 addr_buf += af->sockaddr_len;
1007 walk_size += af->sockaddr_len;
1008 }
1009
1010 /* Do the work. */
1011 switch (op) {
1012 case SCTP_BINDX_ADD_ADDR:
1013 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1014 if (err)
1015 goto out;
1016 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1017 break;
1018
1019 case SCTP_BINDX_REM_ADDR:
1020 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1021 if (err)
1022 goto out;
1023 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1024 break;
1025
1026 default:
1027 err = -EINVAL;
1028 break;
1029 }
1030
1031 out:
1032 kfree(kaddrs);
1033
1034 return err;
1035 }
1036
1037 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1038 *
1039 * Common routine for handling connect() and sctp_connectx().
1040 * Connect will come in with just a single address.
1041 */
1042 static int __sctp_connect(struct sock *sk,
1043 struct sockaddr *kaddrs,
1044 int addrs_size,
1045 sctp_assoc_t *assoc_id)
1046 {
1047 struct net *net = sock_net(sk);
1048 struct sctp_sock *sp;
1049 struct sctp_endpoint *ep;
1050 struct sctp_association *asoc = NULL;
1051 struct sctp_association *asoc2;
1052 struct sctp_transport *transport;
1053 union sctp_addr to;
1054 struct sctp_af *af;
1055 sctp_scope_t scope;
1056 long timeo;
1057 int err = 0;
1058 int addrcnt = 0;
1059 int walk_size = 0;
1060 union sctp_addr *sa_addr = NULL;
1061 void *addr_buf;
1062 unsigned short port;
1063 unsigned int f_flags = 0;
1064
1065 sp = sctp_sk(sk);
1066 ep = sp->ep;
1067
1068 /* connect() cannot be done on a socket that is already in ESTABLISHED
1069 * state - UDP-style peeled off socket or a TCP-style socket that
1070 * is already connected.
1071 * It cannot be done even on a TCP-style listening socket.
1072 */
1073 if (sctp_sstate(sk, ESTABLISHED) ||
1074 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1075 err = -EISCONN;
1076 goto out_free;
1077 }
1078
1079 /* Walk through the addrs buffer and count the number of addresses. */
1080 addr_buf = kaddrs;
1081 while (walk_size < addrs_size) {
1082 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1083 err = -EINVAL;
1084 goto out_free;
1085 }
1086
1087 sa_addr = addr_buf;
1088 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1089
1090 /* If the address family is not supported or if this address
1091 * causes the address buffer to overflow return EINVAL.
1092 */
1093 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1094 err = -EINVAL;
1095 goto out_free;
1096 }
1097
1098 port = ntohs(sa_addr->v4.sin_port);
1099
1100 /* Save current address so we can work with it */
1101 memcpy(&to, sa_addr, af->sockaddr_len);
1102
1103 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1104 if (err)
1105 goto out_free;
1106
1107 /* Make sure the destination port is correctly set
1108 * in all addresses.
1109 */
1110 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1111 err = -EINVAL;
1112 goto out_free;
1113 }
1114
1115 /* Check if there already is a matching association on the
1116 * endpoint (other than the one created here).
1117 */
1118 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1119 if (asoc2 && asoc2 != asoc) {
1120 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1121 err = -EISCONN;
1122 else
1123 err = -EALREADY;
1124 goto out_free;
1125 }
1126
1127 /* If we could not find a matching association on the endpoint,
1128 * make sure that there is no peeled-off association matching
1129 * the peer address even on another socket.
1130 */
1131 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1132 err = -EADDRNOTAVAIL;
1133 goto out_free;
1134 }
1135
1136 if (!asoc) {
1137 /* If a bind() or sctp_bindx() is not called prior to
1138 * an sctp_connectx() call, the system picks an
1139 * ephemeral port and will choose an address set
1140 * equivalent to binding with a wildcard address.
1141 */
1142 if (!ep->base.bind_addr.port) {
1143 if (sctp_autobind(sk)) {
1144 err = -EAGAIN;
1145 goto out_free;
1146 }
1147 } else {
1148 /*
1149 * If an unprivileged user inherits a 1-many
1150 * style socket with open associations on a
1151 * privileged port, it MAY be permitted to
1152 * accept new associations, but it SHOULD NOT
1153 * be permitted to open new associations.
1154 */
1155 if (ep->base.bind_addr.port < PROT_SOCK &&
1156 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1157 err = -EACCES;
1158 goto out_free;
1159 }
1160 }
1161
1162 scope = sctp_scope(&to);
1163 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1164 if (!asoc) {
1165 err = -ENOMEM;
1166 goto out_free;
1167 }
1168
1169 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1170 GFP_KERNEL);
1171 if (err < 0) {
1172 goto out_free;
1173 }
1174
1175 }
1176
1177 /* Prime the peer's transport structures. */
1178 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1179 SCTP_UNKNOWN);
1180 if (!transport) {
1181 err = -ENOMEM;
1182 goto out_free;
1183 }
1184
1185 addrcnt++;
1186 addr_buf += af->sockaddr_len;
1187 walk_size += af->sockaddr_len;
1188 }
1189
1190 /* In case the user of sctp_connectx() wants an association
1191 * id back, assign one now.
1192 */
1193 if (assoc_id) {
1194 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1195 if (err < 0)
1196 goto out_free;
1197 }
1198
1199 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1200 if (err < 0) {
1201 goto out_free;
1202 }
1203
1204 /* Initialize sk's dport and daddr for getpeername() */
1205 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1206 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1207 af->to_sk_daddr(sa_addr, sk);
1208 sk->sk_err = 0;
1209
1210 /* in-kernel sockets don't generally have a file allocated to them
1211 * if all they do is call sock_create_kern().
1212 */
1213 if (sk->sk_socket->file)
1214 f_flags = sk->sk_socket->file->f_flags;
1215
1216 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1217
1218 err = sctp_wait_for_connect(asoc, &timeo);
1219 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1220 *assoc_id = asoc->assoc_id;
1221
1222 /* Don't free association on exit. */
1223 asoc = NULL;
1224
1225 out_free:
1226 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1227 __func__, asoc, kaddrs, err);
1228
1229 if (asoc) {
1230 /* sctp_primitive_ASSOCIATE may have added this association
1231 * To the hash table, try to unhash it, just in case, its a noop
1232 * if it wasn't hashed so we're safe
1233 */
1234 sctp_unhash_established(asoc);
1235 sctp_association_free(asoc);
1236 }
1237 return err;
1238 }
1239
1240 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1241 *
1242 * API 8.9
1243 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1244 * sctp_assoc_t *asoc);
1245 *
1246 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1247 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1248 * or IPv6 addresses.
1249 *
1250 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1251 * Section 3.1.2 for this usage.
1252 *
1253 * addrs is a pointer to an array of one or more socket addresses. Each
1254 * address is contained in its appropriate structure (i.e. struct
1255 * sockaddr_in or struct sockaddr_in6) the family of the address type
1256 * must be used to distengish the address length (note that this
1257 * representation is termed a "packed array" of addresses). The caller
1258 * specifies the number of addresses in the array with addrcnt.
1259 *
1260 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1261 * the association id of the new association. On failure, sctp_connectx()
1262 * returns -1, and sets errno to the appropriate error code. The assoc_id
1263 * is not touched by the kernel.
1264 *
1265 * For SCTP, the port given in each socket address must be the same, or
1266 * sctp_connectx() will fail, setting errno to EINVAL.
1267 *
1268 * An application can use sctp_connectx to initiate an association with
1269 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1270 * allows a caller to specify multiple addresses at which a peer can be
1271 * reached. The way the SCTP stack uses the list of addresses to set up
1272 * the association is implementation dependent. This function only
1273 * specifies that the stack will try to make use of all the addresses in
1274 * the list when needed.
1275 *
1276 * Note that the list of addresses passed in is only used for setting up
1277 * the association. It does not necessarily equal the set of addresses
1278 * the peer uses for the resulting association. If the caller wants to
1279 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1280 * retrieve them after the association has been set up.
1281 *
1282 * Basically do nothing but copying the addresses from user to kernel
1283 * land and invoking either sctp_connectx(). This is used for tunneling
1284 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1285 *
1286 * We don't use copy_from_user() for optimization: we first do the
1287 * sanity checks (buffer size -fast- and access check-healthy
1288 * pointer); if all of those succeed, then we can alloc the memory
1289 * (expensive operation) needed to copy the data to kernel. Then we do
1290 * the copying without checking the user space area
1291 * (__copy_from_user()).
1292 *
1293 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1294 * it.
1295 *
1296 * sk The sk of the socket
1297 * addrs The pointer to the addresses in user land
1298 * addrssize Size of the addrs buffer
1299 *
1300 * Returns >=0 if ok, <0 errno code on error.
1301 */
1302 static int __sctp_setsockopt_connectx(struct sock *sk,
1303 struct sockaddr __user *addrs,
1304 int addrs_size,
1305 sctp_assoc_t *assoc_id)
1306 {
1307 int err = 0;
1308 struct sockaddr *kaddrs;
1309
1310 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1311 __func__, sk, addrs, addrs_size);
1312
1313 if (unlikely(addrs_size <= 0))
1314 return -EINVAL;
1315
1316 /* Check the user passed a healthy pointer. */
1317 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1318 return -EFAULT;
1319
1320 /* Alloc space for the address array in kernel memory. */
1321 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1322 if (unlikely(!kaddrs))
1323 return -ENOMEM;
1324
1325 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1326 err = -EFAULT;
1327 } else {
1328 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1329 }
1330
1331 kfree(kaddrs);
1332
1333 return err;
1334 }
1335
1336 /*
1337 * This is an older interface. It's kept for backward compatibility
1338 * to the option that doesn't provide association id.
1339 */
1340 static int sctp_setsockopt_connectx_old(struct sock *sk,
1341 struct sockaddr __user *addrs,
1342 int addrs_size)
1343 {
1344 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1345 }
1346
1347 /*
1348 * New interface for the API. The since the API is done with a socket
1349 * option, to make it simple we feed back the association id is as a return
1350 * indication to the call. Error is always negative and association id is
1351 * always positive.
1352 */
1353 static int sctp_setsockopt_connectx(struct sock *sk,
1354 struct sockaddr __user *addrs,
1355 int addrs_size)
1356 {
1357 sctp_assoc_t assoc_id = 0;
1358 int err = 0;
1359
1360 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1361
1362 if (err)
1363 return err;
1364 else
1365 return assoc_id;
1366 }
1367
1368 /*
1369 * New (hopefully final) interface for the API.
1370 * We use the sctp_getaddrs_old structure so that use-space library
1371 * can avoid any unnecessary allocations. The only defferent part
1372 * is that we store the actual length of the address buffer into the
1373 * addrs_num structure member. That way we can re-use the existing
1374 * code.
1375 */
1376 static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1377 char __user *optval,
1378 int __user *optlen)
1379 {
1380 struct sctp_getaddrs_old param;
1381 sctp_assoc_t assoc_id = 0;
1382 int err = 0;
1383
1384 if (len < sizeof(param))
1385 return -EINVAL;
1386
1387 if (copy_from_user(&param, optval, sizeof(param)))
1388 return -EFAULT;
1389
1390 err = __sctp_setsockopt_connectx(sk,
1391 (struct sockaddr __user *)param.addrs,
1392 param.addr_num, &assoc_id);
1393
1394 if (err == 0 || err == -EINPROGRESS) {
1395 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1396 return -EFAULT;
1397 if (put_user(sizeof(assoc_id), optlen))
1398 return -EFAULT;
1399 }
1400
1401 return err;
1402 }
1403
1404 /* API 3.1.4 close() - UDP Style Syntax
1405 * Applications use close() to perform graceful shutdown (as described in
1406 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1407 * by a UDP-style socket.
1408 *
1409 * The syntax is
1410 *
1411 * ret = close(int sd);
1412 *
1413 * sd - the socket descriptor of the associations to be closed.
1414 *
1415 * To gracefully shutdown a specific association represented by the
1416 * UDP-style socket, an application should use the sendmsg() call,
1417 * passing no user data, but including the appropriate flag in the
1418 * ancillary data (see Section xxxx).
1419 *
1420 * If sd in the close() call is a branched-off socket representing only
1421 * one association, the shutdown is performed on that association only.
1422 *
1423 * 4.1.6 close() - TCP Style Syntax
1424 *
1425 * Applications use close() to gracefully close down an association.
1426 *
1427 * The syntax is:
1428 *
1429 * int close(int sd);
1430 *
1431 * sd - the socket descriptor of the association to be closed.
1432 *
1433 * After an application calls close() on a socket descriptor, no further
1434 * socket operations will succeed on that descriptor.
1435 *
1436 * API 7.1.4 SO_LINGER
1437 *
1438 * An application using the TCP-style socket can use this option to
1439 * perform the SCTP ABORT primitive. The linger option structure is:
1440 *
1441 * struct linger {
1442 * int l_onoff; // option on/off
1443 * int l_linger; // linger time
1444 * };
1445 *
1446 * To enable the option, set l_onoff to 1. If the l_linger value is set
1447 * to 0, calling close() is the same as the ABORT primitive. If the
1448 * value is set to a negative value, the setsockopt() call will return
1449 * an error. If the value is set to a positive value linger_time, the
1450 * close() can be blocked for at most linger_time ms. If the graceful
1451 * shutdown phase does not finish during this period, close() will
1452 * return but the graceful shutdown phase continues in the system.
1453 */
1454 static void sctp_close(struct sock *sk, long timeout)
1455 {
1456 struct net *net = sock_net(sk);
1457 struct sctp_endpoint *ep;
1458 struct sctp_association *asoc;
1459 struct list_head *pos, *temp;
1460 unsigned int data_was_unread;
1461
1462 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1463
1464 lock_sock(sk);
1465 sk->sk_shutdown = SHUTDOWN_MASK;
1466 sk->sk_state = SCTP_SS_CLOSING;
1467
1468 ep = sctp_sk(sk)->ep;
1469
1470 /* Clean up any skbs sitting on the receive queue. */
1471 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1472 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1473
1474 /* Walk all associations on an endpoint. */
1475 list_for_each_safe(pos, temp, &ep->asocs) {
1476 asoc = list_entry(pos, struct sctp_association, asocs);
1477
1478 if (sctp_style(sk, TCP)) {
1479 /* A closed association can still be in the list if
1480 * it belongs to a TCP-style listening socket that is
1481 * not yet accepted. If so, free it. If not, send an
1482 * ABORT or SHUTDOWN based on the linger options.
1483 */
1484 if (sctp_state(asoc, CLOSED)) {
1485 sctp_unhash_established(asoc);
1486 sctp_association_free(asoc);
1487 continue;
1488 }
1489 }
1490
1491 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1492 !skb_queue_empty(&asoc->ulpq.reasm) ||
1493 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1494 struct sctp_chunk *chunk;
1495
1496 chunk = sctp_make_abort_user(asoc, NULL, 0);
1497 if (chunk)
1498 sctp_primitive_ABORT(net, asoc, chunk);
1499 } else
1500 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1501 }
1502
1503 /* On a TCP-style socket, block for at most linger_time if set. */
1504 if (sctp_style(sk, TCP) && timeout)
1505 sctp_wait_for_close(sk, timeout);
1506
1507 /* This will run the backlog queue. */
1508 release_sock(sk);
1509
1510 /* Supposedly, no process has access to the socket, but
1511 * the net layers still may.
1512 */
1513 local_bh_disable();
1514 bh_lock_sock(sk);
1515
1516 /* Hold the sock, since sk_common_release() will put sock_put()
1517 * and we have just a little more cleanup.
1518 */
1519 sock_hold(sk);
1520 sk_common_release(sk);
1521
1522 bh_unlock_sock(sk);
1523 local_bh_enable();
1524
1525 sock_put(sk);
1526
1527 SCTP_DBG_OBJCNT_DEC(sock);
1528 }
1529
1530 /* Handle EPIPE error. */
1531 static int sctp_error(struct sock *sk, int flags, int err)
1532 {
1533 if (err == -EPIPE)
1534 err = sock_error(sk) ? : -EPIPE;
1535 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1536 send_sig(SIGPIPE, current, 0);
1537 return err;
1538 }
1539
1540 /* API 3.1.3 sendmsg() - UDP Style Syntax
1541 *
1542 * An application uses sendmsg() and recvmsg() calls to transmit data to
1543 * and receive data from its peer.
1544 *
1545 * ssize_t sendmsg(int socket, const struct msghdr *message,
1546 * int flags);
1547 *
1548 * socket - the socket descriptor of the endpoint.
1549 * message - pointer to the msghdr structure which contains a single
1550 * user message and possibly some ancillary data.
1551 *
1552 * See Section 5 for complete description of the data
1553 * structures.
1554 *
1555 * flags - flags sent or received with the user message, see Section
1556 * 5 for complete description of the flags.
1557 *
1558 * Note: This function could use a rewrite especially when explicit
1559 * connect support comes in.
1560 */
1561 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1562
1563 static int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1564
1565 static int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1566 struct msghdr *msg, size_t msg_len)
1567 {
1568 struct net *net = sock_net(sk);
1569 struct sctp_sock *sp;
1570 struct sctp_endpoint *ep;
1571 struct sctp_association *new_asoc = NULL, *asoc = NULL;
1572 struct sctp_transport *transport, *chunk_tp;
1573 struct sctp_chunk *chunk;
1574 union sctp_addr to;
1575 struct sockaddr *msg_name = NULL;
1576 struct sctp_sndrcvinfo default_sinfo;
1577 struct sctp_sndrcvinfo *sinfo;
1578 struct sctp_initmsg *sinit;
1579 sctp_assoc_t associd = 0;
1580 sctp_cmsgs_t cmsgs = { NULL };
1581 int err;
1582 sctp_scope_t scope;
1583 long timeo;
1584 __u16 sinfo_flags = 0;
1585 struct sctp_datamsg *datamsg;
1586 int msg_flags = msg->msg_flags;
1587
1588 err = 0;
1589 sp = sctp_sk(sk);
1590 ep = sp->ep;
1591
1592 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1593 msg, msg_len, ep);
1594
1595 /* We cannot send a message over a TCP-style listening socket. */
1596 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1597 err = -EPIPE;
1598 goto out_nounlock;
1599 }
1600
1601 /* Parse out the SCTP CMSGs. */
1602 err = sctp_msghdr_parse(msg, &cmsgs);
1603 if (err) {
1604 pr_debug("%s: msghdr parse err:%x\n", __func__, 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 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
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 pr_debug("%s: about to look up association\n", __func__);
1667
1668 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 pr_debug("%s: just looked up association:%p\n", __func__, 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 pr_debug("%s: shutting down association:%p\n",
1710 __func__, asoc);
1711
1712 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1713 err = 0;
1714 goto out_unlock;
1715 }
1716 if (sinfo_flags & SCTP_ABORT) {
1717
1718 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1719 if (!chunk) {
1720 err = -ENOMEM;
1721 goto out_unlock;
1722 }
1723
1724 pr_debug("%s: aborting association:%p\n",
1725 __func__, asoc);
1726
1727 sctp_primitive_ABORT(net, asoc, chunk);
1728 err = 0;
1729 goto out_unlock;
1730 }
1731 }
1732
1733 /* Do we need to create the association? */
1734 if (!asoc) {
1735 pr_debug("%s: there is no association yet\n", __func__);
1736
1737 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1738 err = -EINVAL;
1739 goto out_unlock;
1740 }
1741
1742 /* Check for invalid stream against the stream counts,
1743 * either the default or the user specified stream counts.
1744 */
1745 if (sinfo) {
1746 if (!sinit || !sinit->sinit_num_ostreams) {
1747 /* Check against the defaults. */
1748 if (sinfo->sinfo_stream >=
1749 sp->initmsg.sinit_num_ostreams) {
1750 err = -EINVAL;
1751 goto out_unlock;
1752 }
1753 } else {
1754 /* Check against the requested. */
1755 if (sinfo->sinfo_stream >=
1756 sinit->sinit_num_ostreams) {
1757 err = -EINVAL;
1758 goto out_unlock;
1759 }
1760 }
1761 }
1762
1763 /*
1764 * API 3.1.2 bind() - UDP Style Syntax
1765 * If a bind() or sctp_bindx() is not called prior to a
1766 * sendmsg() call that initiates a new association, the
1767 * system picks an ephemeral port and will choose an address
1768 * set equivalent to binding with a wildcard address.
1769 */
1770 if (!ep->base.bind_addr.port) {
1771 if (sctp_autobind(sk)) {
1772 err = -EAGAIN;
1773 goto out_unlock;
1774 }
1775 } else {
1776 /*
1777 * If an unprivileged user inherits a one-to-many
1778 * style socket with open associations on a privileged
1779 * port, it MAY be permitted to accept new associations,
1780 * but it SHOULD NOT be permitted to open new
1781 * associations.
1782 */
1783 if (ep->base.bind_addr.port < PROT_SOCK &&
1784 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1785 err = -EACCES;
1786 goto out_unlock;
1787 }
1788 }
1789
1790 scope = sctp_scope(&to);
1791 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1792 if (!new_asoc) {
1793 err = -ENOMEM;
1794 goto out_unlock;
1795 }
1796 asoc = new_asoc;
1797 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1798 if (err < 0) {
1799 err = -ENOMEM;
1800 goto out_free;
1801 }
1802
1803 /* If the SCTP_INIT ancillary data is specified, set all
1804 * the association init values accordingly.
1805 */
1806 if (sinit) {
1807 if (sinit->sinit_num_ostreams) {
1808 asoc->c.sinit_num_ostreams =
1809 sinit->sinit_num_ostreams;
1810 }
1811 if (sinit->sinit_max_instreams) {
1812 asoc->c.sinit_max_instreams =
1813 sinit->sinit_max_instreams;
1814 }
1815 if (sinit->sinit_max_attempts) {
1816 asoc->max_init_attempts
1817 = sinit->sinit_max_attempts;
1818 }
1819 if (sinit->sinit_max_init_timeo) {
1820 asoc->max_init_timeo =
1821 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1822 }
1823 }
1824
1825 /* Prime the peer's transport structures. */
1826 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1827 if (!transport) {
1828 err = -ENOMEM;
1829 goto out_free;
1830 }
1831 }
1832
1833 /* ASSERT: we have a valid association at this point. */
1834 pr_debug("%s: we have a valid association\n", __func__);
1835
1836 if (!sinfo) {
1837 /* If the user didn't specify SNDRCVINFO, make up one with
1838 * some defaults.
1839 */
1840 memset(&default_sinfo, 0, sizeof(default_sinfo));
1841 default_sinfo.sinfo_stream = asoc->default_stream;
1842 default_sinfo.sinfo_flags = asoc->default_flags;
1843 default_sinfo.sinfo_ppid = asoc->default_ppid;
1844 default_sinfo.sinfo_context = asoc->default_context;
1845 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1846 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1847 sinfo = &default_sinfo;
1848 }
1849
1850 /* API 7.1.7, the sndbuf size per association bounds the
1851 * maximum size of data that can be sent in a single send call.
1852 */
1853 if (msg_len > sk->sk_sndbuf) {
1854 err = -EMSGSIZE;
1855 goto out_free;
1856 }
1857
1858 if (asoc->pmtu_pending)
1859 sctp_assoc_pending_pmtu(sk, asoc);
1860
1861 /* If fragmentation is disabled and the message length exceeds the
1862 * association fragmentation point, return EMSGSIZE. The I-D
1863 * does not specify what this error is, but this looks like
1864 * a great fit.
1865 */
1866 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1867 err = -EMSGSIZE;
1868 goto out_free;
1869 }
1870
1871 /* Check for invalid stream. */
1872 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1873 err = -EINVAL;
1874 goto out_free;
1875 }
1876
1877 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1878 if (!sctp_wspace(asoc)) {
1879 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1880 if (err)
1881 goto out_free;
1882 }
1883
1884 /* If an address is passed with the sendto/sendmsg call, it is used
1885 * to override the primary destination address in the TCP model, or
1886 * when SCTP_ADDR_OVER flag is set in the UDP model.
1887 */
1888 if ((sctp_style(sk, TCP) && msg_name) ||
1889 (sinfo_flags & SCTP_ADDR_OVER)) {
1890 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1891 if (!chunk_tp) {
1892 err = -EINVAL;
1893 goto out_free;
1894 }
1895 } else
1896 chunk_tp = NULL;
1897
1898 /* Auto-connect, if we aren't connected already. */
1899 if (sctp_state(asoc, CLOSED)) {
1900 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1901 if (err < 0)
1902 goto out_free;
1903
1904 pr_debug("%s: we associated primitively\n", __func__);
1905 }
1906
1907 /* Break the message into multiple chunks of maximum size. */
1908 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1909 if (IS_ERR(datamsg)) {
1910 err = PTR_ERR(datamsg);
1911 goto out_free;
1912 }
1913
1914 /* Now send the (possibly) fragmented message. */
1915 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1916 sctp_chunk_hold(chunk);
1917
1918 /* Do accounting for the write space. */
1919 sctp_set_owner_w(chunk);
1920
1921 chunk->transport = chunk_tp;
1922 }
1923
1924 /* Send it to the lower layers. Note: all chunks
1925 * must either fail or succeed. The lower layer
1926 * works that way today. Keep it that way or this
1927 * breaks.
1928 */
1929 err = sctp_primitive_SEND(net, asoc, datamsg);
1930 /* Did the lower layer accept the chunk? */
1931 if (err) {
1932 sctp_datamsg_free(datamsg);
1933 goto out_free;
1934 }
1935
1936 pr_debug("%s: we sent primitively\n", __func__);
1937
1938 sctp_datamsg_put(datamsg);
1939 err = msg_len;
1940
1941 /* If we are already past ASSOCIATE, the lower
1942 * layers are responsible for association cleanup.
1943 */
1944 goto out_unlock;
1945
1946 out_free:
1947 if (new_asoc) {
1948 sctp_unhash_established(asoc);
1949 sctp_association_free(asoc);
1950 }
1951 out_unlock:
1952 release_sock(sk);
1953
1954 out_nounlock:
1955 return sctp_error(sk, msg_flags, err);
1956
1957 #if 0
1958 do_sock_err:
1959 if (msg_len)
1960 err = msg_len;
1961 else
1962 err = sock_error(sk);
1963 goto out;
1964
1965 do_interrupted:
1966 if (msg_len)
1967 err = msg_len;
1968 goto out;
1969 #endif /* 0 */
1970 }
1971
1972 /* This is an extended version of skb_pull() that removes the data from the
1973 * start of a skb even when data is spread across the list of skb's in the
1974 * frag_list. len specifies the total amount of data that needs to be removed.
1975 * when 'len' bytes could be removed from the skb, it returns 0.
1976 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1977 * could not be removed.
1978 */
1979 static int sctp_skb_pull(struct sk_buff *skb, int len)
1980 {
1981 struct sk_buff *list;
1982 int skb_len = skb_headlen(skb);
1983 int rlen;
1984
1985 if (len <= skb_len) {
1986 __skb_pull(skb, len);
1987 return 0;
1988 }
1989 len -= skb_len;
1990 __skb_pull(skb, skb_len);
1991
1992 skb_walk_frags(skb, list) {
1993 rlen = sctp_skb_pull(list, len);
1994 skb->len -= (len-rlen);
1995 skb->data_len -= (len-rlen);
1996
1997 if (!rlen)
1998 return 0;
1999
2000 len = rlen;
2001 }
2002
2003 return len;
2004 }
2005
2006 /* API 3.1.3 recvmsg() - UDP Style Syntax
2007 *
2008 * ssize_t recvmsg(int socket, struct msghdr *message,
2009 * int flags);
2010 *
2011 * socket - the socket descriptor of the endpoint.
2012 * message - pointer to the msghdr structure which contains a single
2013 * user message and possibly some ancillary data.
2014 *
2015 * See Section 5 for complete description of the data
2016 * structures.
2017 *
2018 * flags - flags sent or received with the user message, see Section
2019 * 5 for complete description of the flags.
2020 */
2021 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2022
2023 static int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2024 struct msghdr *msg, size_t len, int noblock,
2025 int flags, int *addr_len)
2026 {
2027 struct sctp_ulpevent *event = NULL;
2028 struct sctp_sock *sp = sctp_sk(sk);
2029 struct sk_buff *skb;
2030 int copied;
2031 int err = 0;
2032 int skb_len;
2033
2034 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2035 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2036 addr_len);
2037
2038 lock_sock(sk);
2039
2040 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2041 err = -ENOTCONN;
2042 goto out;
2043 }
2044
2045 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2046 if (!skb)
2047 goto out;
2048
2049 /* Get the total length of the skb including any skb's in the
2050 * frag_list.
2051 */
2052 skb_len = skb->len;
2053
2054 copied = skb_len;
2055 if (copied > len)
2056 copied = len;
2057
2058 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2059
2060 event = sctp_skb2event(skb);
2061
2062 if (err)
2063 goto out_free;
2064
2065 sock_recv_ts_and_drops(msg, sk, skb);
2066 if (sctp_ulpevent_is_notification(event)) {
2067 msg->msg_flags |= MSG_NOTIFICATION;
2068 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2069 } else {
2070 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2071 }
2072
2073 /* Check if we allow SCTP_SNDRCVINFO. */
2074 if (sp->subscribe.sctp_data_io_event)
2075 sctp_ulpevent_read_sndrcvinfo(event, msg);
2076 #if 0
2077 /* FIXME: we should be calling IP/IPv6 layers. */
2078 if (sk->sk_protinfo.af_inet.cmsg_flags)
2079 ip_cmsg_recv(msg, skb);
2080 #endif
2081
2082 err = copied;
2083
2084 /* If skb's length exceeds the user's buffer, update the skb and
2085 * push it back to the receive_queue so that the next call to
2086 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2087 */
2088 if (skb_len > copied) {
2089 msg->msg_flags &= ~MSG_EOR;
2090 if (flags & MSG_PEEK)
2091 goto out_free;
2092 sctp_skb_pull(skb, copied);
2093 skb_queue_head(&sk->sk_receive_queue, skb);
2094
2095 /* When only partial message is copied to the user, increase
2096 * rwnd by that amount. If all the data in the skb is read,
2097 * rwnd is updated when the event is freed.
2098 */
2099 if (!sctp_ulpevent_is_notification(event))
2100 sctp_assoc_rwnd_increase(event->asoc, copied);
2101 goto out;
2102 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2103 (event->msg_flags & MSG_EOR))
2104 msg->msg_flags |= MSG_EOR;
2105 else
2106 msg->msg_flags &= ~MSG_EOR;
2107
2108 out_free:
2109 if (flags & MSG_PEEK) {
2110 /* Release the skb reference acquired after peeking the skb in
2111 * sctp_skb_recv_datagram().
2112 */
2113 kfree_skb(skb);
2114 } else {
2115 /* Free the event which includes releasing the reference to
2116 * the owner of the skb, freeing the skb and updating the
2117 * rwnd.
2118 */
2119 sctp_ulpevent_free(event);
2120 }
2121 out:
2122 release_sock(sk);
2123 return err;
2124 }
2125
2126 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2127 *
2128 * This option is a on/off flag. If enabled no SCTP message
2129 * fragmentation will be performed. Instead if a message being sent
2130 * exceeds the current PMTU size, the message will NOT be sent and
2131 * instead a error will be indicated to the user.
2132 */
2133 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2134 char __user *optval,
2135 unsigned int optlen)
2136 {
2137 int val;
2138
2139 if (optlen < sizeof(int))
2140 return -EINVAL;
2141
2142 if (get_user(val, (int __user *)optval))
2143 return -EFAULT;
2144
2145 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2146
2147 return 0;
2148 }
2149
2150 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2151 unsigned int optlen)
2152 {
2153 struct sctp_association *asoc;
2154 struct sctp_ulpevent *event;
2155
2156 if (optlen > sizeof(struct sctp_event_subscribe))
2157 return -EINVAL;
2158 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2159 return -EFAULT;
2160
2161 /*
2162 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2163 * if there is no data to be sent or retransmit, the stack will
2164 * immediately send up this notification.
2165 */
2166 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2167 &sctp_sk(sk)->subscribe)) {
2168 asoc = sctp_id2assoc(sk, 0);
2169
2170 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2171 event = sctp_ulpevent_make_sender_dry_event(asoc,
2172 GFP_ATOMIC);
2173 if (!event)
2174 return -ENOMEM;
2175
2176 sctp_ulpq_tail_event(&asoc->ulpq, event);
2177 }
2178 }
2179
2180 return 0;
2181 }
2182
2183 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2184 *
2185 * This socket option is applicable to the UDP-style socket only. When
2186 * set it will cause associations that are idle for more than the
2187 * specified number of seconds to automatically close. An association
2188 * being idle is defined an association that has NOT sent or received
2189 * user data. The special value of '0' indicates that no automatic
2190 * close of any associations should be performed. The option expects an
2191 * integer defining the number of seconds of idle time before an
2192 * association is closed.
2193 */
2194 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2195 unsigned int optlen)
2196 {
2197 struct sctp_sock *sp = sctp_sk(sk);
2198 struct net *net = sock_net(sk);
2199
2200 /* Applicable to UDP-style socket only */
2201 if (sctp_style(sk, TCP))
2202 return -EOPNOTSUPP;
2203 if (optlen != sizeof(int))
2204 return -EINVAL;
2205 if (copy_from_user(&sp->autoclose, optval, optlen))
2206 return -EFAULT;
2207
2208 if (sp->autoclose > net->sctp.max_autoclose)
2209 sp->autoclose = net->sctp.max_autoclose;
2210
2211 return 0;
2212 }
2213
2214 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2215 *
2216 * Applications can enable or disable heartbeats for any peer address of
2217 * an association, modify an address's heartbeat interval, force a
2218 * heartbeat to be sent immediately, and adjust the address's maximum
2219 * number of retransmissions sent before an address is considered
2220 * unreachable. The following structure is used to access and modify an
2221 * address's parameters:
2222 *
2223 * struct sctp_paddrparams {
2224 * sctp_assoc_t spp_assoc_id;
2225 * struct sockaddr_storage spp_address;
2226 * uint32_t spp_hbinterval;
2227 * uint16_t spp_pathmaxrxt;
2228 * uint32_t spp_pathmtu;
2229 * uint32_t spp_sackdelay;
2230 * uint32_t spp_flags;
2231 * };
2232 *
2233 * spp_assoc_id - (one-to-many style socket) This is filled in the
2234 * application, and identifies the association for
2235 * this query.
2236 * spp_address - This specifies which address is of interest.
2237 * spp_hbinterval - This contains the value of the heartbeat interval,
2238 * in milliseconds. If a value of zero
2239 * is present in this field then no changes are to
2240 * be made to this parameter.
2241 * spp_pathmaxrxt - This contains the maximum number of
2242 * retransmissions before this address shall be
2243 * considered unreachable. If a value of zero
2244 * is present in this field then no changes are to
2245 * be made to this parameter.
2246 * spp_pathmtu - When Path MTU discovery is disabled the value
2247 * specified here will be the "fixed" path mtu.
2248 * Note that if the spp_address field is empty
2249 * then all associations on this address will
2250 * have this fixed path mtu set upon them.
2251 *
2252 * spp_sackdelay - When delayed sack is enabled, this value specifies
2253 * the number of milliseconds that sacks will be delayed
2254 * for. This value will apply to all addresses of an
2255 * association if the spp_address field is empty. Note
2256 * also, that if delayed sack is enabled and this
2257 * value is set to 0, no change is made to the last
2258 * recorded delayed sack timer value.
2259 *
2260 * spp_flags - These flags are used to control various features
2261 * on an association. The flag field may contain
2262 * zero or more of the following options.
2263 *
2264 * SPP_HB_ENABLE - Enable heartbeats on the
2265 * specified address. Note that if the address
2266 * field is empty all addresses for the association
2267 * have heartbeats enabled upon them.
2268 *
2269 * SPP_HB_DISABLE - Disable heartbeats on the
2270 * speicifed address. Note that if the address
2271 * field is empty all addresses for the association
2272 * will have their heartbeats disabled. Note also
2273 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2274 * mutually exclusive, only one of these two should
2275 * be specified. Enabling both fields will have
2276 * undetermined results.
2277 *
2278 * SPP_HB_DEMAND - Request a user initiated heartbeat
2279 * to be made immediately.
2280 *
2281 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2282 * heartbeat delayis to be set to the value of 0
2283 * milliseconds.
2284 *
2285 * SPP_PMTUD_ENABLE - This field will enable PMTU
2286 * discovery upon the specified address. Note that
2287 * if the address feild is empty then all addresses
2288 * on the association are effected.
2289 *
2290 * SPP_PMTUD_DISABLE - This field will disable PMTU
2291 * discovery upon the specified address. Note that
2292 * if the address feild is empty then all addresses
2293 * on the association are effected. Not also that
2294 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2295 * exclusive. Enabling both will have undetermined
2296 * results.
2297 *
2298 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2299 * on delayed sack. The time specified in spp_sackdelay
2300 * is used to specify the sack delay for this address. Note
2301 * that if spp_address is empty then all addresses will
2302 * enable delayed sack and take on the sack delay
2303 * value specified in spp_sackdelay.
2304 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2305 * off delayed sack. If the spp_address field is blank then
2306 * delayed sack is disabled for the entire association. Note
2307 * also that this field is mutually exclusive to
2308 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2309 * results.
2310 */
2311 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2312 struct sctp_transport *trans,
2313 struct sctp_association *asoc,
2314 struct sctp_sock *sp,
2315 int hb_change,
2316 int pmtud_change,
2317 int sackdelay_change)
2318 {
2319 int error;
2320
2321 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2322 struct net *net = sock_net(trans->asoc->base.sk);
2323
2324 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2325 if (error)
2326 return error;
2327 }
2328
2329 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2330 * this field is ignored. Note also that a value of zero indicates
2331 * the current setting should be left unchanged.
2332 */
2333 if (params->spp_flags & SPP_HB_ENABLE) {
2334
2335 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2336 * set. This lets us use 0 value when this flag
2337 * is set.
2338 */
2339 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2340 params->spp_hbinterval = 0;
2341
2342 if (params->spp_hbinterval ||
2343 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2344 if (trans) {
2345 trans->hbinterval =
2346 msecs_to_jiffies(params->spp_hbinterval);
2347 } else if (asoc) {
2348 asoc->hbinterval =
2349 msecs_to_jiffies(params->spp_hbinterval);
2350 } else {
2351 sp->hbinterval = params->spp_hbinterval;
2352 }
2353 }
2354 }
2355
2356 if (hb_change) {
2357 if (trans) {
2358 trans->param_flags =
2359 (trans->param_flags & ~SPP_HB) | hb_change;
2360 } else if (asoc) {
2361 asoc->param_flags =
2362 (asoc->param_flags & ~SPP_HB) | hb_change;
2363 } else {
2364 sp->param_flags =
2365 (sp->param_flags & ~SPP_HB) | hb_change;
2366 }
2367 }
2368
2369 /* When Path MTU discovery is disabled the value specified here will
2370 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2371 * include the flag SPP_PMTUD_DISABLE for this field to have any
2372 * effect).
2373 */
2374 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2375 if (trans) {
2376 trans->pathmtu = params->spp_pathmtu;
2377 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2378 } else if (asoc) {
2379 asoc->pathmtu = params->spp_pathmtu;
2380 sctp_frag_point(asoc, params->spp_pathmtu);
2381 } else {
2382 sp->pathmtu = params->spp_pathmtu;
2383 }
2384 }
2385
2386 if (pmtud_change) {
2387 if (trans) {
2388 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2389 (params->spp_flags & SPP_PMTUD_ENABLE);
2390 trans->param_flags =
2391 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2392 if (update) {
2393 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2394 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2395 }
2396 } else if (asoc) {
2397 asoc->param_flags =
2398 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2399 } else {
2400 sp->param_flags =
2401 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2402 }
2403 }
2404
2405 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2406 * value of this field is ignored. Note also that a value of zero
2407 * indicates the current setting should be left unchanged.
2408 */
2409 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2410 if (trans) {
2411 trans->sackdelay =
2412 msecs_to_jiffies(params->spp_sackdelay);
2413 } else if (asoc) {
2414 asoc->sackdelay =
2415 msecs_to_jiffies(params->spp_sackdelay);
2416 } else {
2417 sp->sackdelay = params->spp_sackdelay;
2418 }
2419 }
2420
2421 if (sackdelay_change) {
2422 if (trans) {
2423 trans->param_flags =
2424 (trans->param_flags & ~SPP_SACKDELAY) |
2425 sackdelay_change;
2426 } else if (asoc) {
2427 asoc->param_flags =
2428 (asoc->param_flags & ~SPP_SACKDELAY) |
2429 sackdelay_change;
2430 } else {
2431 sp->param_flags =
2432 (sp->param_flags & ~SPP_SACKDELAY) |
2433 sackdelay_change;
2434 }
2435 }
2436
2437 /* Note that a value of zero indicates the current setting should be
2438 left unchanged.
2439 */
2440 if (params->spp_pathmaxrxt) {
2441 if (trans) {
2442 trans->pathmaxrxt = params->spp_pathmaxrxt;
2443 } else if (asoc) {
2444 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2445 } else {
2446 sp->pathmaxrxt = params->spp_pathmaxrxt;
2447 }
2448 }
2449
2450 return 0;
2451 }
2452
2453 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2454 char __user *optval,
2455 unsigned int optlen)
2456 {
2457 struct sctp_paddrparams params;
2458 struct sctp_transport *trans = NULL;
2459 struct sctp_association *asoc = NULL;
2460 struct sctp_sock *sp = sctp_sk(sk);
2461 int error;
2462 int hb_change, pmtud_change, sackdelay_change;
2463
2464 if (optlen != sizeof(struct sctp_paddrparams))
2465 return -EINVAL;
2466
2467 if (copy_from_user(&params, optval, optlen))
2468 return -EFAULT;
2469
2470 /* Validate flags and value parameters. */
2471 hb_change = params.spp_flags & SPP_HB;
2472 pmtud_change = params.spp_flags & SPP_PMTUD;
2473 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2474
2475 if (hb_change == SPP_HB ||
2476 pmtud_change == SPP_PMTUD ||
2477 sackdelay_change == SPP_SACKDELAY ||
2478 params.spp_sackdelay > 500 ||
2479 (params.spp_pathmtu &&
2480 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2481 return -EINVAL;
2482
2483 /* If an address other than INADDR_ANY is specified, and
2484 * no transport is found, then the request is invalid.
2485 */
2486 if (!sctp_is_any(sk, (union sctp_addr *)&params.spp_address)) {
2487 trans = sctp_addr_id2transport(sk, &params.spp_address,
2488 params.spp_assoc_id);
2489 if (!trans)
2490 return -EINVAL;
2491 }
2492
2493 /* Get association, if assoc_id != 0 and the socket is a one
2494 * to many style socket, and an association was not found, then
2495 * the id was invalid.
2496 */
2497 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2498 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2499 return -EINVAL;
2500
2501 /* Heartbeat demand can only be sent on a transport or
2502 * association, but not a socket.
2503 */
2504 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2505 return -EINVAL;
2506
2507 /* Process parameters. */
2508 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2509 hb_change, pmtud_change,
2510 sackdelay_change);
2511
2512 if (error)
2513 return error;
2514
2515 /* If changes are for association, also apply parameters to each
2516 * transport.
2517 */
2518 if (!trans && asoc) {
2519 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2520 transports) {
2521 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2522 hb_change, pmtud_change,
2523 sackdelay_change);
2524 }
2525 }
2526
2527 return 0;
2528 }
2529
2530 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2531 {
2532 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2533 }
2534
2535 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2536 {
2537 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2538 }
2539
2540 /*
2541 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2542 *
2543 * This option will effect the way delayed acks are performed. This
2544 * option allows you to get or set the delayed ack time, in
2545 * milliseconds. It also allows changing the delayed ack frequency.
2546 * Changing the frequency to 1 disables the delayed sack algorithm. If
2547 * the assoc_id is 0, then this sets or gets the endpoints default
2548 * values. If the assoc_id field is non-zero, then the set or get
2549 * effects the specified association for the one to many model (the
2550 * assoc_id field is ignored by the one to one model). Note that if
2551 * sack_delay or sack_freq are 0 when setting this option, then the
2552 * current values will remain unchanged.
2553 *
2554 * struct sctp_sack_info {
2555 * sctp_assoc_t sack_assoc_id;
2556 * uint32_t sack_delay;
2557 * uint32_t sack_freq;
2558 * };
2559 *
2560 * sack_assoc_id - This parameter, indicates which association the user
2561 * is performing an action upon. Note that if this field's value is
2562 * zero then the endpoints default value is changed (effecting future
2563 * associations only).
2564 *
2565 * sack_delay - This parameter contains the number of milliseconds that
2566 * the user is requesting the delayed ACK timer be set to. Note that
2567 * this value is defined in the standard to be between 200 and 500
2568 * milliseconds.
2569 *
2570 * sack_freq - This parameter contains the number of packets that must
2571 * be received before a sack is sent without waiting for the delay
2572 * timer to expire. The default value for this is 2, setting this
2573 * value to 1 will disable the delayed sack algorithm.
2574 */
2575
2576 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2577 char __user *optval, unsigned int optlen)
2578 {
2579 struct sctp_sack_info params;
2580 struct sctp_transport *trans = NULL;
2581 struct sctp_association *asoc = NULL;
2582 struct sctp_sock *sp = sctp_sk(sk);
2583
2584 if (optlen == sizeof(struct sctp_sack_info)) {
2585 if (copy_from_user(&params, optval, optlen))
2586 return -EFAULT;
2587
2588 if (params.sack_delay == 0 && params.sack_freq == 0)
2589 return 0;
2590 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2591 pr_warn_ratelimited(DEPRECATED
2592 "%s (pid %d) "
2593 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2594 "Use struct sctp_sack_info instead\n",
2595 current->comm, task_pid_nr(current));
2596 if (copy_from_user(&params, optval, optlen))
2597 return -EFAULT;
2598
2599 if (params.sack_delay == 0)
2600 params.sack_freq = 1;
2601 else
2602 params.sack_freq = 0;
2603 } else
2604 return -EINVAL;
2605
2606 /* Validate value parameter. */
2607 if (params.sack_delay > 500)
2608 return -EINVAL;
2609
2610 /* Get association, if sack_assoc_id != 0 and the socket is a one
2611 * to many style socket, and an association was not found, then
2612 * the id was invalid.
2613 */
2614 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2615 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2616 return -EINVAL;
2617
2618 if (params.sack_delay) {
2619 if (asoc) {
2620 asoc->sackdelay =
2621 msecs_to_jiffies(params.sack_delay);
2622 asoc->param_flags =
2623 sctp_spp_sackdelay_enable(asoc->param_flags);
2624 } else {
2625 sp->sackdelay = params.sack_delay;
2626 sp->param_flags =
2627 sctp_spp_sackdelay_enable(sp->param_flags);
2628 }
2629 }
2630
2631 if (params.sack_freq == 1) {
2632 if (asoc) {
2633 asoc->param_flags =
2634 sctp_spp_sackdelay_disable(asoc->param_flags);
2635 } else {
2636 sp->param_flags =
2637 sctp_spp_sackdelay_disable(sp->param_flags);
2638 }
2639 } else if (params.sack_freq > 1) {
2640 if (asoc) {
2641 asoc->sackfreq = params.sack_freq;
2642 asoc->param_flags =
2643 sctp_spp_sackdelay_enable(asoc->param_flags);
2644 } else {
2645 sp->sackfreq = params.sack_freq;
2646 sp->param_flags =
2647 sctp_spp_sackdelay_enable(sp->param_flags);
2648 }
2649 }
2650
2651 /* If change is for association, also apply to each transport. */
2652 if (asoc) {
2653 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2654 transports) {
2655 if (params.sack_delay) {
2656 trans->sackdelay =
2657 msecs_to_jiffies(params.sack_delay);
2658 trans->param_flags =
2659 sctp_spp_sackdelay_enable(trans->param_flags);
2660 }
2661 if (params.sack_freq == 1) {
2662 trans->param_flags =
2663 sctp_spp_sackdelay_disable(trans->param_flags);
2664 } else if (params.sack_freq > 1) {
2665 trans->sackfreq = params.sack_freq;
2666 trans->param_flags =
2667 sctp_spp_sackdelay_enable(trans->param_flags);
2668 }
2669 }
2670 }
2671
2672 return 0;
2673 }
2674
2675 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2676 *
2677 * Applications can specify protocol parameters for the default association
2678 * initialization. The option name argument to setsockopt() and getsockopt()
2679 * is SCTP_INITMSG.
2680 *
2681 * Setting initialization parameters is effective only on an unconnected
2682 * socket (for UDP-style sockets only future associations are effected
2683 * by the change). With TCP-style sockets, this option is inherited by
2684 * sockets derived from a listener socket.
2685 */
2686 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2687 {
2688 struct sctp_initmsg sinit;
2689 struct sctp_sock *sp = sctp_sk(sk);
2690
2691 if (optlen != sizeof(struct sctp_initmsg))
2692 return -EINVAL;
2693 if (copy_from_user(&sinit, optval, optlen))
2694 return -EFAULT;
2695
2696 if (sinit.sinit_num_ostreams)
2697 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2698 if (sinit.sinit_max_instreams)
2699 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2700 if (sinit.sinit_max_attempts)
2701 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2702 if (sinit.sinit_max_init_timeo)
2703 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2704
2705 return 0;
2706 }
2707
2708 /*
2709 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2710 *
2711 * Applications that wish to use the sendto() system call may wish to
2712 * specify a default set of parameters that would normally be supplied
2713 * through the inclusion of ancillary data. This socket option allows
2714 * such an application to set the default sctp_sndrcvinfo structure.
2715 * The application that wishes to use this socket option simply passes
2716 * in to this call the sctp_sndrcvinfo structure defined in Section
2717 * 5.2.2) The input parameters accepted by this call include
2718 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2719 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2720 * to this call if the caller is using the UDP model.
2721 */
2722 static int sctp_setsockopt_default_send_param(struct sock *sk,
2723 char __user *optval,
2724 unsigned int optlen)
2725 {
2726 struct sctp_sndrcvinfo info;
2727 struct sctp_association *asoc;
2728 struct sctp_sock *sp = sctp_sk(sk);
2729
2730 if (optlen != sizeof(struct sctp_sndrcvinfo))
2731 return -EINVAL;
2732 if (copy_from_user(&info, optval, optlen))
2733 return -EFAULT;
2734
2735 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2736 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2737 return -EINVAL;
2738
2739 if (asoc) {
2740 asoc->default_stream = info.sinfo_stream;
2741 asoc->default_flags = info.sinfo_flags;
2742 asoc->default_ppid = info.sinfo_ppid;
2743 asoc->default_context = info.sinfo_context;
2744 asoc->default_timetolive = info.sinfo_timetolive;
2745 } else {
2746 sp->default_stream = info.sinfo_stream;
2747 sp->default_flags = info.sinfo_flags;
2748 sp->default_ppid = info.sinfo_ppid;
2749 sp->default_context = info.sinfo_context;
2750 sp->default_timetolive = info.sinfo_timetolive;
2751 }
2752
2753 return 0;
2754 }
2755
2756 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2757 *
2758 * Requests that the local SCTP stack use the enclosed peer address as
2759 * the association primary. The enclosed address must be one of the
2760 * association peer's addresses.
2761 */
2762 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2763 unsigned int optlen)
2764 {
2765 struct sctp_prim prim;
2766 struct sctp_transport *trans;
2767
2768 if (optlen != sizeof(struct sctp_prim))
2769 return -EINVAL;
2770
2771 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2772 return -EFAULT;
2773
2774 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2775 if (!trans)
2776 return -EINVAL;
2777
2778 sctp_assoc_set_primary(trans->asoc, trans);
2779
2780 return 0;
2781 }
2782
2783 /*
2784 * 7.1.5 SCTP_NODELAY
2785 *
2786 * Turn on/off any Nagle-like algorithm. This means that packets are
2787 * generally sent as soon as possible and no unnecessary delays are
2788 * introduced, at the cost of more packets in the network. Expects an
2789 * integer boolean flag.
2790 */
2791 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2792 unsigned int optlen)
2793 {
2794 int val;
2795
2796 if (optlen < sizeof(int))
2797 return -EINVAL;
2798 if (get_user(val, (int __user *)optval))
2799 return -EFAULT;
2800
2801 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2802 return 0;
2803 }
2804
2805 /*
2806 *
2807 * 7.1.1 SCTP_RTOINFO
2808 *
2809 * The protocol parameters used to initialize and bound retransmission
2810 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2811 * and modify these parameters.
2812 * All parameters are time values, in milliseconds. A value of 0, when
2813 * modifying the parameters, indicates that the current value should not
2814 * be changed.
2815 *
2816 */
2817 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2818 {
2819 struct sctp_rtoinfo rtoinfo;
2820 struct sctp_association *asoc;
2821 unsigned long rto_min, rto_max;
2822 struct sctp_sock *sp = sctp_sk(sk);
2823
2824 if (optlen != sizeof (struct sctp_rtoinfo))
2825 return -EINVAL;
2826
2827 if (copy_from_user(&rtoinfo, optval, optlen))
2828 return -EFAULT;
2829
2830 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2831
2832 /* Set the values to the specific association */
2833 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2834 return -EINVAL;
2835
2836 rto_max = rtoinfo.srto_max;
2837 rto_min = rtoinfo.srto_min;
2838
2839 if (rto_max)
2840 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
2841 else
2842 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
2843
2844 if (rto_min)
2845 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
2846 else
2847 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
2848
2849 if (rto_min > rto_max)
2850 return -EINVAL;
2851
2852 if (asoc) {
2853 if (rtoinfo.srto_initial != 0)
2854 asoc->rto_initial =
2855 msecs_to_jiffies(rtoinfo.srto_initial);
2856 asoc->rto_max = rto_max;
2857 asoc->rto_min = rto_min;
2858 } else {
2859 /* If there is no association or the association-id = 0
2860 * set the values to the endpoint.
2861 */
2862 if (rtoinfo.srto_initial != 0)
2863 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2864 sp->rtoinfo.srto_max = rto_max;
2865 sp->rtoinfo.srto_min = rto_min;
2866 }
2867
2868 return 0;
2869 }
2870
2871 /*
2872 *
2873 * 7.1.2 SCTP_ASSOCINFO
2874 *
2875 * This option is used to tune the maximum retransmission attempts
2876 * of the association.
2877 * Returns an error if the new association retransmission value is
2878 * greater than the sum of the retransmission value of the peer.
2879 * See [SCTP] for more information.
2880 *
2881 */
2882 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2883 {
2884
2885 struct sctp_assocparams assocparams;
2886 struct sctp_association *asoc;
2887
2888 if (optlen != sizeof(struct sctp_assocparams))
2889 return -EINVAL;
2890 if (copy_from_user(&assocparams, optval, optlen))
2891 return -EFAULT;
2892
2893 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2894
2895 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2896 return -EINVAL;
2897
2898 /* Set the values to the specific association */
2899 if (asoc) {
2900 if (assocparams.sasoc_asocmaxrxt != 0) {
2901 __u32 path_sum = 0;
2902 int paths = 0;
2903 struct sctp_transport *peer_addr;
2904
2905 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2906 transports) {
2907 path_sum += peer_addr->pathmaxrxt;
2908 paths++;
2909 }
2910
2911 /* Only validate asocmaxrxt if we have more than
2912 * one path/transport. We do this because path
2913 * retransmissions are only counted when we have more
2914 * then one path.
2915 */
2916 if (paths > 1 &&
2917 assocparams.sasoc_asocmaxrxt > path_sum)
2918 return -EINVAL;
2919
2920 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2921 }
2922
2923 if (assocparams.sasoc_cookie_life != 0)
2924 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
2925 } else {
2926 /* Set the values to the endpoint */
2927 struct sctp_sock *sp = sctp_sk(sk);
2928
2929 if (assocparams.sasoc_asocmaxrxt != 0)
2930 sp->assocparams.sasoc_asocmaxrxt =
2931 assocparams.sasoc_asocmaxrxt;
2932 if (assocparams.sasoc_cookie_life != 0)
2933 sp->assocparams.sasoc_cookie_life =
2934 assocparams.sasoc_cookie_life;
2935 }
2936 return 0;
2937 }
2938
2939 /*
2940 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2941 *
2942 * This socket option is a boolean flag which turns on or off mapped V4
2943 * addresses. If this option is turned on and the socket is type
2944 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2945 * If this option is turned off, then no mapping will be done of V4
2946 * addresses and a user will receive both PF_INET6 and PF_INET type
2947 * addresses on the socket.
2948 */
2949 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2950 {
2951 int val;
2952 struct sctp_sock *sp = sctp_sk(sk);
2953
2954 if (optlen < sizeof(int))
2955 return -EINVAL;
2956 if (get_user(val, (int __user *)optval))
2957 return -EFAULT;
2958 if (val)
2959 sp->v4mapped = 1;
2960 else
2961 sp->v4mapped = 0;
2962
2963 return 0;
2964 }
2965
2966 /*
2967 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2968 * This option will get or set the maximum size to put in any outgoing
2969 * SCTP DATA chunk. If a message is larger than this size it will be
2970 * fragmented by SCTP into the specified size. Note that the underlying
2971 * SCTP implementation may fragment into smaller sized chunks when the
2972 * PMTU of the underlying association is smaller than the value set by
2973 * the user. The default value for this option is '0' which indicates
2974 * the user is NOT limiting fragmentation and only the PMTU will effect
2975 * SCTP's choice of DATA chunk size. Note also that values set larger
2976 * than the maximum size of an IP datagram will effectively let SCTP
2977 * control fragmentation (i.e. the same as setting this option to 0).
2978 *
2979 * The following structure is used to access and modify this parameter:
2980 *
2981 * struct sctp_assoc_value {
2982 * sctp_assoc_t assoc_id;
2983 * uint32_t assoc_value;
2984 * };
2985 *
2986 * assoc_id: This parameter is ignored for one-to-one style sockets.
2987 * For one-to-many style sockets this parameter indicates which
2988 * association the user is performing an action upon. Note that if
2989 * this field's value is zero then the endpoints default value is
2990 * changed (effecting future associations only).
2991 * assoc_value: This parameter specifies the maximum size in bytes.
2992 */
2993 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2994 {
2995 struct sctp_assoc_value params;
2996 struct sctp_association *asoc;
2997 struct sctp_sock *sp = sctp_sk(sk);
2998 int val;
2999
3000 if (optlen == sizeof(int)) {
3001 pr_warn_ratelimited(DEPRECATED
3002 "%s (pid %d) "
3003 "Use of int in maxseg socket option.\n"
3004 "Use struct sctp_assoc_value instead\n",
3005 current->comm, task_pid_nr(current));
3006 if (copy_from_user(&val, optval, optlen))
3007 return -EFAULT;
3008 params.assoc_id = 0;
3009 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3010 if (copy_from_user(&params, optval, optlen))
3011 return -EFAULT;
3012 val = params.assoc_value;
3013 } else
3014 return -EINVAL;
3015
3016 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3017 return -EINVAL;
3018
3019 asoc = sctp_id2assoc(sk, params.assoc_id);
3020 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3021 return -EINVAL;
3022
3023 if (asoc) {
3024 if (val == 0) {
3025 val = asoc->pathmtu;
3026 val -= sp->pf->af->net_header_len;
3027 val -= sizeof(struct sctphdr) +
3028 sizeof(struct sctp_data_chunk);
3029 }
3030 asoc->user_frag = val;
3031 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3032 } else {
3033 sp->user_frag = val;
3034 }
3035
3036 return 0;
3037 }
3038
3039
3040 /*
3041 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3042 *
3043 * Requests that the peer mark the enclosed address as the association
3044 * primary. The enclosed address must be one of the association's
3045 * locally bound addresses. The following structure is used to make a
3046 * set primary request:
3047 */
3048 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3049 unsigned int optlen)
3050 {
3051 struct net *net = sock_net(sk);
3052 struct sctp_sock *sp;
3053 struct sctp_association *asoc = NULL;
3054 struct sctp_setpeerprim prim;
3055 struct sctp_chunk *chunk;
3056 struct sctp_af *af;
3057 int err;
3058
3059 sp = sctp_sk(sk);
3060
3061 if (!net->sctp.addip_enable)
3062 return -EPERM;
3063
3064 if (optlen != sizeof(struct sctp_setpeerprim))
3065 return -EINVAL;
3066
3067 if (copy_from_user(&prim, optval, optlen))
3068 return -EFAULT;
3069
3070 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3071 if (!asoc)
3072 return -EINVAL;
3073
3074 if (!asoc->peer.asconf_capable)
3075 return -EPERM;
3076
3077 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3078 return -EPERM;
3079
3080 if (!sctp_state(asoc, ESTABLISHED))
3081 return -ENOTCONN;
3082
3083 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3084 if (!af)
3085 return -EINVAL;
3086
3087 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3088 return -EADDRNOTAVAIL;
3089
3090 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3091 return -EADDRNOTAVAIL;
3092
3093 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3094 chunk = sctp_make_asconf_set_prim(asoc,
3095 (union sctp_addr *)&prim.sspp_addr);
3096 if (!chunk)
3097 return -ENOMEM;
3098
3099 err = sctp_send_asconf(asoc, chunk);
3100
3101 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3102
3103 return err;
3104 }
3105
3106 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3107 unsigned int optlen)
3108 {
3109 struct sctp_setadaptation adaptation;
3110
3111 if (optlen != sizeof(struct sctp_setadaptation))
3112 return -EINVAL;
3113 if (copy_from_user(&adaptation, optval, optlen))
3114 return -EFAULT;
3115
3116 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3117
3118 return 0;
3119 }
3120
3121 /*
3122 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3123 *
3124 * The context field in the sctp_sndrcvinfo structure is normally only
3125 * used when a failed message is retrieved holding the value that was
3126 * sent down on the actual send call. This option allows the setting of
3127 * a default context on an association basis that will be received on
3128 * reading messages from the peer. This is especially helpful in the
3129 * one-2-many model for an application to keep some reference to an
3130 * internal state machine that is processing messages on the
3131 * association. Note that the setting of this value only effects
3132 * received messages from the peer and does not effect the value that is
3133 * saved with outbound messages.
3134 */
3135 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3136 unsigned int optlen)
3137 {
3138 struct sctp_assoc_value params;
3139 struct sctp_sock *sp;
3140 struct sctp_association *asoc;
3141
3142 if (optlen != sizeof(struct sctp_assoc_value))
3143 return -EINVAL;
3144 if (copy_from_user(&params, optval, optlen))
3145 return -EFAULT;
3146
3147 sp = sctp_sk(sk);
3148
3149 if (params.assoc_id != 0) {
3150 asoc = sctp_id2assoc(sk, params.assoc_id);
3151 if (!asoc)
3152 return -EINVAL;
3153 asoc->default_rcv_context = params.assoc_value;
3154 } else {
3155 sp->default_rcv_context = params.assoc_value;
3156 }
3157
3158 return 0;
3159 }
3160
3161 /*
3162 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3163 *
3164 * This options will at a minimum specify if the implementation is doing
3165 * fragmented interleave. Fragmented interleave, for a one to many
3166 * socket, is when subsequent calls to receive a message may return
3167 * parts of messages from different associations. Some implementations
3168 * may allow you to turn this value on or off. If so, when turned off,
3169 * no fragment interleave will occur (which will cause a head of line
3170 * blocking amongst multiple associations sharing the same one to many
3171 * socket). When this option is turned on, then each receive call may
3172 * come from a different association (thus the user must receive data
3173 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3174 * association each receive belongs to.
3175 *
3176 * This option takes a boolean value. A non-zero value indicates that
3177 * fragmented interleave is on. A value of zero indicates that
3178 * fragmented interleave is off.
3179 *
3180 * Note that it is important that an implementation that allows this
3181 * option to be turned on, have it off by default. Otherwise an unaware
3182 * application using the one to many model may become confused and act
3183 * incorrectly.
3184 */
3185 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3186 char __user *optval,
3187 unsigned int optlen)
3188 {
3189 int val;
3190
3191 if (optlen != sizeof(int))
3192 return -EINVAL;
3193 if (get_user(val, (int __user *)optval))
3194 return -EFAULT;
3195
3196 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3197
3198 return 0;
3199 }
3200
3201 /*
3202 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3203 * (SCTP_PARTIAL_DELIVERY_POINT)
3204 *
3205 * This option will set or get the SCTP partial delivery point. This
3206 * point is the size of a message where the partial delivery API will be
3207 * invoked to help free up rwnd space for the peer. Setting this to a
3208 * lower value will cause partial deliveries to happen more often. The
3209 * calls argument is an integer that sets or gets the partial delivery
3210 * point. Note also that the call will fail if the user attempts to set
3211 * this value larger than the socket receive buffer size.
3212 *
3213 * Note that any single message having a length smaller than or equal to
3214 * the SCTP partial delivery point will be delivered in one single read
3215 * call as long as the user provided buffer is large enough to hold the
3216 * message.
3217 */
3218 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3219 char __user *optval,
3220 unsigned int optlen)
3221 {
3222 u32 val;
3223
3224 if (optlen != sizeof(u32))
3225 return -EINVAL;
3226 if (get_user(val, (int __user *)optval))
3227 return -EFAULT;
3228
3229 /* Note: We double the receive buffer from what the user sets
3230 * it to be, also initial rwnd is based on rcvbuf/2.
3231 */
3232 if (val > (sk->sk_rcvbuf >> 1))
3233 return -EINVAL;
3234
3235 sctp_sk(sk)->pd_point = val;
3236
3237 return 0; /* is this the right error code? */
3238 }
3239
3240 /*
3241 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3242 *
3243 * This option will allow a user to change the maximum burst of packets
3244 * that can be emitted by this association. Note that the default value
3245 * is 4, and some implementations may restrict this setting so that it
3246 * can only be lowered.
3247 *
3248 * NOTE: This text doesn't seem right. Do this on a socket basis with
3249 * future associations inheriting the socket value.
3250 */
3251 static int sctp_setsockopt_maxburst(struct sock *sk,
3252 char __user *optval,
3253 unsigned int optlen)
3254 {
3255 struct sctp_assoc_value params;
3256 struct sctp_sock *sp;
3257 struct sctp_association *asoc;
3258 int val;
3259 int assoc_id = 0;
3260
3261 if (optlen == sizeof(int)) {
3262 pr_warn_ratelimited(DEPRECATED
3263 "%s (pid %d) "
3264 "Use of int in max_burst socket option deprecated.\n"
3265 "Use struct sctp_assoc_value instead\n",
3266 current->comm, task_pid_nr(current));
3267 if (copy_from_user(&val, optval, optlen))
3268 return -EFAULT;
3269 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3270 if (copy_from_user(&params, optval, optlen))
3271 return -EFAULT;
3272 val = params.assoc_value;
3273 assoc_id = params.assoc_id;
3274 } else
3275 return -EINVAL;
3276
3277 sp = sctp_sk(sk);
3278
3279 if (assoc_id != 0) {
3280 asoc = sctp_id2assoc(sk, assoc_id);
3281 if (!asoc)
3282 return -EINVAL;
3283 asoc->max_burst = val;
3284 } else
3285 sp->max_burst = val;
3286
3287 return 0;
3288 }
3289
3290 /*
3291 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3292 *
3293 * This set option adds a chunk type that the user is requesting to be
3294 * received only in an authenticated way. Changes to the list of chunks
3295 * will only effect future associations on the socket.
3296 */
3297 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3298 char __user *optval,
3299 unsigned int optlen)
3300 {
3301 struct net *net = sock_net(sk);
3302 struct sctp_authchunk val;
3303
3304 if (!net->sctp.auth_enable)
3305 return -EACCES;
3306
3307 if (optlen != sizeof(struct sctp_authchunk))
3308 return -EINVAL;
3309 if (copy_from_user(&val, optval, optlen))
3310 return -EFAULT;
3311
3312 switch (val.sauth_chunk) {
3313 case SCTP_CID_INIT:
3314 case SCTP_CID_INIT_ACK:
3315 case SCTP_CID_SHUTDOWN_COMPLETE:
3316 case SCTP_CID_AUTH:
3317 return -EINVAL;
3318 }
3319
3320 /* add this chunk id to the endpoint */
3321 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3322 }
3323
3324 /*
3325 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3326 *
3327 * This option gets or sets the list of HMAC algorithms that the local
3328 * endpoint requires the peer to use.
3329 */
3330 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3331 char __user *optval,
3332 unsigned int optlen)
3333 {
3334 struct net *net = sock_net(sk);
3335 struct sctp_hmacalgo *hmacs;
3336 u32 idents;
3337 int err;
3338
3339 if (!net->sctp.auth_enable)
3340 return -EACCES;
3341
3342 if (optlen < sizeof(struct sctp_hmacalgo))
3343 return -EINVAL;
3344
3345 hmacs = memdup_user(optval, optlen);
3346 if (IS_ERR(hmacs))
3347 return PTR_ERR(hmacs);
3348
3349 idents = hmacs->shmac_num_idents;
3350 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3351 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3352 err = -EINVAL;
3353 goto out;
3354 }
3355
3356 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3357 out:
3358 kfree(hmacs);
3359 return err;
3360 }
3361
3362 /*
3363 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3364 *
3365 * This option will set a shared secret key which is used to build an
3366 * association shared key.
3367 */
3368 static int sctp_setsockopt_auth_key(struct sock *sk,
3369 char __user *optval,
3370 unsigned int optlen)
3371 {
3372 struct net *net = sock_net(sk);
3373 struct sctp_authkey *authkey;
3374 struct sctp_association *asoc;
3375 int ret;
3376
3377 if (!net->sctp.auth_enable)
3378 return -EACCES;
3379
3380 if (optlen <= sizeof(struct sctp_authkey))
3381 return -EINVAL;
3382
3383 authkey = memdup_user(optval, optlen);
3384 if (IS_ERR(authkey))
3385 return PTR_ERR(authkey);
3386
3387 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3388 ret = -EINVAL;
3389 goto out;
3390 }
3391
3392 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3393 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3394 ret = -EINVAL;
3395 goto out;
3396 }
3397
3398 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3399 out:
3400 kzfree(authkey);
3401 return ret;
3402 }
3403
3404 /*
3405 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3406 *
3407 * This option will get or set the active shared key to be used to build
3408 * the association shared key.
3409 */
3410 static int sctp_setsockopt_active_key(struct sock *sk,
3411 char __user *optval,
3412 unsigned int optlen)
3413 {
3414 struct net *net = sock_net(sk);
3415 struct sctp_authkeyid val;
3416 struct sctp_association *asoc;
3417
3418 if (!net->sctp.auth_enable)
3419 return -EACCES;
3420
3421 if (optlen != sizeof(struct sctp_authkeyid))
3422 return -EINVAL;
3423 if (copy_from_user(&val, optval, optlen))
3424 return -EFAULT;
3425
3426 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3427 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3428 return -EINVAL;
3429
3430 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3431 val.scact_keynumber);
3432 }
3433
3434 /*
3435 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3436 *
3437 * This set option will delete a shared secret key from use.
3438 */
3439 static int sctp_setsockopt_del_key(struct sock *sk,
3440 char __user *optval,
3441 unsigned int optlen)
3442 {
3443 struct net *net = sock_net(sk);
3444 struct sctp_authkeyid val;
3445 struct sctp_association *asoc;
3446
3447 if (!net->sctp.auth_enable)
3448 return -EACCES;
3449
3450 if (optlen != sizeof(struct sctp_authkeyid))
3451 return -EINVAL;
3452 if (copy_from_user(&val, optval, optlen))
3453 return -EFAULT;
3454
3455 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3456 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3457 return -EINVAL;
3458
3459 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3460 val.scact_keynumber);
3461
3462 }
3463
3464 /*
3465 * 8.1.23 SCTP_AUTO_ASCONF
3466 *
3467 * This option will enable or disable the use of the automatic generation of
3468 * ASCONF chunks to add and delete addresses to an existing association. Note
3469 * that this option has two caveats namely: a) it only affects sockets that
3470 * are bound to all addresses available to the SCTP stack, and b) the system
3471 * administrator may have an overriding control that turns the ASCONF feature
3472 * off no matter what setting the socket option may have.
3473 * This option expects an integer boolean flag, where a non-zero value turns on
3474 * the option, and a zero value turns off the option.
3475 * Note. In this implementation, socket operation overrides default parameter
3476 * being set by sysctl as well as FreeBSD implementation
3477 */
3478 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3479 unsigned int optlen)
3480 {
3481 int val;
3482 struct sctp_sock *sp = sctp_sk(sk);
3483
3484 if (optlen < sizeof(int))
3485 return -EINVAL;
3486 if (get_user(val, (int __user *)optval))
3487 return -EFAULT;
3488 if (!sctp_is_ep_boundall(sk) && val)
3489 return -EINVAL;
3490 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3491 return 0;
3492
3493 if (val == 0 && sp->do_auto_asconf) {
3494 list_del(&sp->auto_asconf_list);
3495 sp->do_auto_asconf = 0;
3496 } else if (val && !sp->do_auto_asconf) {
3497 list_add_tail(&sp->auto_asconf_list,
3498 &sock_net(sk)->sctp.auto_asconf_splist);
3499 sp->do_auto_asconf = 1;
3500 }
3501 return 0;
3502 }
3503
3504
3505 /*
3506 * SCTP_PEER_ADDR_THLDS
3507 *
3508 * This option allows us to alter the partially failed threshold for one or all
3509 * transports in an association. See Section 6.1 of:
3510 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3511 */
3512 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3513 char __user *optval,
3514 unsigned int optlen)
3515 {
3516 struct sctp_paddrthlds val;
3517 struct sctp_transport *trans;
3518 struct sctp_association *asoc;
3519
3520 if (optlen < sizeof(struct sctp_paddrthlds))
3521 return -EINVAL;
3522 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3523 sizeof(struct sctp_paddrthlds)))
3524 return -EFAULT;
3525
3526
3527 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3528 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3529 if (!asoc)
3530 return -ENOENT;
3531 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3532 transports) {
3533 if (val.spt_pathmaxrxt)
3534 trans->pathmaxrxt = val.spt_pathmaxrxt;
3535 trans->pf_retrans = val.spt_pathpfthld;
3536 }
3537
3538 if (val.spt_pathmaxrxt)
3539 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3540 asoc->pf_retrans = val.spt_pathpfthld;
3541 } else {
3542 trans = sctp_addr_id2transport(sk, &val.spt_address,
3543 val.spt_assoc_id);
3544 if (!trans)
3545 return -ENOENT;
3546
3547 if (val.spt_pathmaxrxt)
3548 trans->pathmaxrxt = val.spt_pathmaxrxt;
3549 trans->pf_retrans = val.spt_pathpfthld;
3550 }
3551
3552 return 0;
3553 }
3554
3555 /* API 6.2 setsockopt(), getsockopt()
3556 *
3557 * Applications use setsockopt() and getsockopt() to set or retrieve
3558 * socket options. Socket options are used to change the default
3559 * behavior of sockets calls. They are described in Section 7.
3560 *
3561 * The syntax is:
3562 *
3563 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3564 * int __user *optlen);
3565 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3566 * int optlen);
3567 *
3568 * sd - the socket descript.
3569 * level - set to IPPROTO_SCTP for all SCTP options.
3570 * optname - the option name.
3571 * optval - the buffer to store the value of the option.
3572 * optlen - the size of the buffer.
3573 */
3574 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3575 char __user *optval, unsigned int optlen)
3576 {
3577 int retval = 0;
3578
3579 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3580
3581 /* I can hardly begin to describe how wrong this is. This is
3582 * so broken as to be worse than useless. The API draft
3583 * REALLY is NOT helpful here... I am not convinced that the
3584 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3585 * are at all well-founded.
3586 */
3587 if (level != SOL_SCTP) {
3588 struct sctp_af *af = sctp_sk(sk)->pf->af;
3589 retval = af->setsockopt(sk, level, optname, optval, optlen);
3590 goto out_nounlock;
3591 }
3592
3593 lock_sock(sk);
3594
3595 switch (optname) {
3596 case SCTP_SOCKOPT_BINDX_ADD:
3597 /* 'optlen' is the size of the addresses buffer. */
3598 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3599 optlen, SCTP_BINDX_ADD_ADDR);
3600 break;
3601
3602 case SCTP_SOCKOPT_BINDX_REM:
3603 /* 'optlen' is the size of the addresses buffer. */
3604 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3605 optlen, SCTP_BINDX_REM_ADDR);
3606 break;
3607
3608 case SCTP_SOCKOPT_CONNECTX_OLD:
3609 /* 'optlen' is the size of the addresses buffer. */
3610 retval = sctp_setsockopt_connectx_old(sk,
3611 (struct sockaddr __user *)optval,
3612 optlen);
3613 break;
3614
3615 case SCTP_SOCKOPT_CONNECTX:
3616 /* 'optlen' is the size of the addresses buffer. */
3617 retval = sctp_setsockopt_connectx(sk,
3618 (struct sockaddr __user *)optval,
3619 optlen);
3620 break;
3621
3622 case SCTP_DISABLE_FRAGMENTS:
3623 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3624 break;
3625
3626 case SCTP_EVENTS:
3627 retval = sctp_setsockopt_events(sk, optval, optlen);
3628 break;
3629
3630 case SCTP_AUTOCLOSE:
3631 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3632 break;
3633
3634 case SCTP_PEER_ADDR_PARAMS:
3635 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3636 break;
3637
3638 case SCTP_DELAYED_SACK:
3639 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3640 break;
3641 case SCTP_PARTIAL_DELIVERY_POINT:
3642 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3643 break;
3644
3645 case SCTP_INITMSG:
3646 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3647 break;
3648 case SCTP_DEFAULT_SEND_PARAM:
3649 retval = sctp_setsockopt_default_send_param(sk, optval,
3650 optlen);
3651 break;
3652 case SCTP_PRIMARY_ADDR:
3653 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3654 break;
3655 case SCTP_SET_PEER_PRIMARY_ADDR:
3656 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3657 break;
3658 case SCTP_NODELAY:
3659 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3660 break;
3661 case SCTP_RTOINFO:
3662 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3663 break;
3664 case SCTP_ASSOCINFO:
3665 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3666 break;
3667 case SCTP_I_WANT_MAPPED_V4_ADDR:
3668 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3669 break;
3670 case SCTP_MAXSEG:
3671 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3672 break;
3673 case SCTP_ADAPTATION_LAYER:
3674 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3675 break;
3676 case SCTP_CONTEXT:
3677 retval = sctp_setsockopt_context(sk, optval, optlen);
3678 break;
3679 case SCTP_FRAGMENT_INTERLEAVE:
3680 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3681 break;
3682 case SCTP_MAX_BURST:
3683 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3684 break;
3685 case SCTP_AUTH_CHUNK:
3686 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3687 break;
3688 case SCTP_HMAC_IDENT:
3689 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3690 break;
3691 case SCTP_AUTH_KEY:
3692 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3693 break;
3694 case SCTP_AUTH_ACTIVE_KEY:
3695 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3696 break;
3697 case SCTP_AUTH_DELETE_KEY:
3698 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3699 break;
3700 case SCTP_AUTO_ASCONF:
3701 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3702 break;
3703 case SCTP_PEER_ADDR_THLDS:
3704 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3705 break;
3706 default:
3707 retval = -ENOPROTOOPT;
3708 break;
3709 }
3710
3711 release_sock(sk);
3712
3713 out_nounlock:
3714 return retval;
3715 }
3716
3717 /* API 3.1.6 connect() - UDP Style Syntax
3718 *
3719 * An application may use the connect() call in the UDP model to initiate an
3720 * association without sending data.
3721 *
3722 * The syntax is:
3723 *
3724 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3725 *
3726 * sd: the socket descriptor to have a new association added to.
3727 *
3728 * nam: the address structure (either struct sockaddr_in or struct
3729 * sockaddr_in6 defined in RFC2553 [7]).
3730 *
3731 * len: the size of the address.
3732 */
3733 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
3734 int addr_len)
3735 {
3736 int err = 0;
3737 struct sctp_af *af;
3738
3739 lock_sock(sk);
3740
3741 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
3742 addr, addr_len);
3743
3744 /* Validate addr_len before calling common connect/connectx routine. */
3745 af = sctp_get_af_specific(addr->sa_family);
3746 if (!af || addr_len < af->sockaddr_len) {
3747 err = -EINVAL;
3748 } else {
3749 /* Pass correct addr len to common routine (so it knows there
3750 * is only one address being passed.
3751 */
3752 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3753 }
3754
3755 release_sock(sk);
3756 return err;
3757 }
3758
3759 /* FIXME: Write comments. */
3760 static int sctp_disconnect(struct sock *sk, int flags)
3761 {
3762 return -EOPNOTSUPP; /* STUB */
3763 }
3764
3765 /* 4.1.4 accept() - TCP Style Syntax
3766 *
3767 * Applications use accept() call to remove an established SCTP
3768 * association from the accept queue of the endpoint. A new socket
3769 * descriptor will be returned from accept() to represent the newly
3770 * formed association.
3771 */
3772 static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3773 {
3774 struct sctp_sock *sp;
3775 struct sctp_endpoint *ep;
3776 struct sock *newsk = NULL;
3777 struct sctp_association *asoc;
3778 long timeo;
3779 int error = 0;
3780
3781 lock_sock(sk);
3782
3783 sp = sctp_sk(sk);
3784 ep = sp->ep;
3785
3786 if (!sctp_style(sk, TCP)) {
3787 error = -EOPNOTSUPP;
3788 goto out;
3789 }
3790
3791 if (!sctp_sstate(sk, LISTENING)) {
3792 error = -EINVAL;
3793 goto out;
3794 }
3795
3796 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3797
3798 error = sctp_wait_for_accept(sk, timeo);
3799 if (error)
3800 goto out;
3801
3802 /* We treat the list of associations on the endpoint as the accept
3803 * queue and pick the first association on the list.
3804 */
3805 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3806
3807 newsk = sp->pf->create_accept_sk(sk, asoc);
3808 if (!newsk) {
3809 error = -ENOMEM;
3810 goto out;
3811 }
3812
3813 /* Populate the fields of the newsk from the oldsk and migrate the
3814 * asoc to the newsk.
3815 */
3816 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3817
3818 out:
3819 release_sock(sk);
3820 *err = error;
3821 return newsk;
3822 }
3823
3824 /* The SCTP ioctl handler. */
3825 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3826 {
3827 int rc = -ENOTCONN;
3828
3829 lock_sock(sk);
3830
3831 /*
3832 * SEQPACKET-style sockets in LISTENING state are valid, for
3833 * SCTP, so only discard TCP-style sockets in LISTENING state.
3834 */
3835 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3836 goto out;
3837
3838 switch (cmd) {
3839 case SIOCINQ: {
3840 struct sk_buff *skb;
3841 unsigned int amount = 0;
3842
3843 skb = skb_peek(&sk->sk_receive_queue);
3844 if (skb != NULL) {
3845 /*
3846 * We will only return the amount of this packet since
3847 * that is all that will be read.
3848 */
3849 amount = skb->len;
3850 }
3851 rc = put_user(amount, (int __user *)arg);
3852 break;
3853 }
3854 default:
3855 rc = -ENOIOCTLCMD;
3856 break;
3857 }
3858 out:
3859 release_sock(sk);
3860 return rc;
3861 }
3862
3863 /* This is the function which gets called during socket creation to
3864 * initialized the SCTP-specific portion of the sock.
3865 * The sock structure should already be zero-filled memory.
3866 */
3867 static int sctp_init_sock(struct sock *sk)
3868 {
3869 struct net *net = sock_net(sk);
3870 struct sctp_sock *sp;
3871
3872 pr_debug("%s: sk:%p\n", __func__, sk);
3873
3874 sp = sctp_sk(sk);
3875
3876 /* Initialize the SCTP per socket area. */
3877 switch (sk->sk_type) {
3878 case SOCK_SEQPACKET:
3879 sp->type = SCTP_SOCKET_UDP;
3880 break;
3881 case SOCK_STREAM:
3882 sp->type = SCTP_SOCKET_TCP;
3883 break;
3884 default:
3885 return -ESOCKTNOSUPPORT;
3886 }
3887
3888 /* Initialize default send parameters. These parameters can be
3889 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3890 */
3891 sp->default_stream = 0;
3892 sp->default_ppid = 0;
3893 sp->default_flags = 0;
3894 sp->default_context = 0;
3895 sp->default_timetolive = 0;
3896
3897 sp->default_rcv_context = 0;
3898 sp->max_burst = net->sctp.max_burst;
3899
3900 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
3901
3902 /* Initialize default setup parameters. These parameters
3903 * can be modified with the SCTP_INITMSG socket option or
3904 * overridden by the SCTP_INIT CMSG.
3905 */
3906 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3907 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3908 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
3909 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
3910
3911 /* Initialize default RTO related parameters. These parameters can
3912 * be modified for with the SCTP_RTOINFO socket option.
3913 */
3914 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
3915 sp->rtoinfo.srto_max = net->sctp.rto_max;
3916 sp->rtoinfo.srto_min = net->sctp.rto_min;
3917
3918 /* Initialize default association related parameters. These parameters
3919 * can be modified with the SCTP_ASSOCINFO socket option.
3920 */
3921 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
3922 sp->assocparams.sasoc_number_peer_destinations = 0;
3923 sp->assocparams.sasoc_peer_rwnd = 0;
3924 sp->assocparams.sasoc_local_rwnd = 0;
3925 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
3926
3927 /* Initialize default event subscriptions. By default, all the
3928 * options are off.
3929 */
3930 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3931
3932 /* Default Peer Address Parameters. These defaults can
3933 * be modified via SCTP_PEER_ADDR_PARAMS
3934 */
3935 sp->hbinterval = net->sctp.hb_interval;
3936 sp->pathmaxrxt = net->sctp.max_retrans_path;
3937 sp->pathmtu = 0; /* allow default discovery */
3938 sp->sackdelay = net->sctp.sack_timeout;
3939 sp->sackfreq = 2;
3940 sp->param_flags = SPP_HB_ENABLE |
3941 SPP_PMTUD_ENABLE |
3942 SPP_SACKDELAY_ENABLE;
3943
3944 /* If enabled no SCTP message fragmentation will be performed.
3945 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3946 */
3947 sp->disable_fragments = 0;
3948
3949 /* Enable Nagle algorithm by default. */
3950 sp->nodelay = 0;
3951
3952 /* Enable by default. */
3953 sp->v4mapped = 1;
3954
3955 /* Auto-close idle associations after the configured
3956 * number of seconds. A value of 0 disables this
3957 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3958 * for UDP-style sockets only.
3959 */
3960 sp->autoclose = 0;
3961
3962 /* User specified fragmentation limit. */
3963 sp->user_frag = 0;
3964
3965 sp->adaptation_ind = 0;
3966
3967 sp->pf = sctp_get_pf_specific(sk->sk_family);
3968
3969 /* Control variables for partial data delivery. */
3970 atomic_set(&sp->pd_mode, 0);
3971 skb_queue_head_init(&sp->pd_lobby);
3972 sp->frag_interleave = 0;
3973
3974 /* Create a per socket endpoint structure. Even if we
3975 * change the data structure relationships, this may still
3976 * be useful for storing pre-connect address information.
3977 */
3978 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
3979 if (!sp->ep)
3980 return -ENOMEM;
3981
3982 sp->hmac = NULL;
3983
3984 sk->sk_destruct = sctp_destruct_sock;
3985
3986 SCTP_DBG_OBJCNT_INC(sock);
3987
3988 local_bh_disable();
3989 percpu_counter_inc(&sctp_sockets_allocated);
3990 sock_prot_inuse_add(net, sk->sk_prot, 1);
3991 if (net->sctp.default_auto_asconf) {
3992 list_add_tail(&sp->auto_asconf_list,
3993 &net->sctp.auto_asconf_splist);
3994 sp->do_auto_asconf = 1;
3995 } else
3996 sp->do_auto_asconf = 0;
3997 local_bh_enable();
3998
3999 return 0;
4000 }
4001
4002 /* Cleanup any SCTP per socket resources. */
4003 static void sctp_destroy_sock(struct sock *sk)
4004 {
4005 struct sctp_sock *sp;
4006
4007 pr_debug("%s: sk:%p\n", __func__, sk);
4008
4009 /* Release our hold on the endpoint. */
4010 sp = sctp_sk(sk);
4011 /* This could happen during socket init, thus we bail out
4012 * early, since the rest of the below is not setup either.
4013 */
4014 if (sp->ep == NULL)
4015 return;
4016
4017 if (sp->do_auto_asconf) {
4018 sp->do_auto_asconf = 0;
4019 list_del(&sp->auto_asconf_list);
4020 }
4021 sctp_endpoint_free(sp->ep);
4022 local_bh_disable();
4023 percpu_counter_dec(&sctp_sockets_allocated);
4024 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4025 local_bh_enable();
4026 }
4027
4028 /* Triggered when there are no references on the socket anymore */
4029 static void sctp_destruct_sock(struct sock *sk)
4030 {
4031 struct sctp_sock *sp = sctp_sk(sk);
4032
4033 /* Free up the HMAC transform. */
4034 crypto_free_hash(sp->hmac);
4035
4036 inet_sock_destruct(sk);
4037 }
4038
4039 /* API 4.1.7 shutdown() - TCP Style Syntax
4040 * int shutdown(int socket, int how);
4041 *
4042 * sd - the socket descriptor of the association to be closed.
4043 * how - Specifies the type of shutdown. The values are
4044 * as follows:
4045 * SHUT_RD
4046 * Disables further receive operations. No SCTP
4047 * protocol action is taken.
4048 * SHUT_WR
4049 * Disables further send operations, and initiates
4050 * the SCTP shutdown sequence.
4051 * SHUT_RDWR
4052 * Disables further send and receive operations
4053 * and initiates the SCTP shutdown sequence.
4054 */
4055 static void sctp_shutdown(struct sock *sk, int how)
4056 {
4057 struct net *net = sock_net(sk);
4058 struct sctp_endpoint *ep;
4059 struct sctp_association *asoc;
4060
4061 if (!sctp_style(sk, TCP))
4062 return;
4063
4064 if (how & SEND_SHUTDOWN) {
4065 ep = sctp_sk(sk)->ep;
4066 if (!list_empty(&ep->asocs)) {
4067 asoc = list_entry(ep->asocs.next,
4068 struct sctp_association, asocs);
4069 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4070 }
4071 }
4072 }
4073
4074 /* 7.2.1 Association Status (SCTP_STATUS)
4075
4076 * Applications can retrieve current status information about an
4077 * association, including association state, peer receiver window size,
4078 * number of unacked data chunks, and number of data chunks pending
4079 * receipt. This information is read-only.
4080 */
4081 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4082 char __user *optval,
4083 int __user *optlen)
4084 {
4085 struct sctp_status status;
4086 struct sctp_association *asoc = NULL;
4087 struct sctp_transport *transport;
4088 sctp_assoc_t associd;
4089 int retval = 0;
4090
4091 if (len < sizeof(status)) {
4092 retval = -EINVAL;
4093 goto out;
4094 }
4095
4096 len = sizeof(status);
4097 if (copy_from_user(&status, optval, len)) {
4098 retval = -EFAULT;
4099 goto out;
4100 }
4101
4102 associd = status.sstat_assoc_id;
4103 asoc = sctp_id2assoc(sk, associd);
4104 if (!asoc) {
4105 retval = -EINVAL;
4106 goto out;
4107 }
4108
4109 transport = asoc->peer.primary_path;
4110
4111 status.sstat_assoc_id = sctp_assoc2id(asoc);
4112 status.sstat_state = asoc->state;
4113 status.sstat_rwnd = asoc->peer.rwnd;
4114 status.sstat_unackdata = asoc->unack_data;
4115
4116 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4117 status.sstat_instrms = asoc->c.sinit_max_instreams;
4118 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4119 status.sstat_fragmentation_point = asoc->frag_point;
4120 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4121 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4122 transport->af_specific->sockaddr_len);
4123 /* Map ipv4 address into v4-mapped-on-v6 address. */
4124 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4125 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4126 status.sstat_primary.spinfo_state = transport->state;
4127 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4128 status.sstat_primary.spinfo_srtt = transport->srtt;
4129 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4130 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4131
4132 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4133 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4134
4135 if (put_user(len, optlen)) {
4136 retval = -EFAULT;
4137 goto out;
4138 }
4139
4140 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4141 __func__, len, status.sstat_state, status.sstat_rwnd,
4142 status.sstat_assoc_id);
4143
4144 if (copy_to_user(optval, &status, len)) {
4145 retval = -EFAULT;
4146 goto out;
4147 }
4148
4149 out:
4150 return retval;
4151 }
4152
4153
4154 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4155 *
4156 * Applications can retrieve information about a specific peer address
4157 * of an association, including its reachability state, congestion
4158 * window, and retransmission timer values. This information is
4159 * read-only.
4160 */
4161 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4162 char __user *optval,
4163 int __user *optlen)
4164 {
4165 struct sctp_paddrinfo pinfo;
4166 struct sctp_transport *transport;
4167 int retval = 0;
4168
4169 if (len < sizeof(pinfo)) {
4170 retval = -EINVAL;
4171 goto out;
4172 }
4173
4174 len = sizeof(pinfo);
4175 if (copy_from_user(&pinfo, optval, len)) {
4176 retval = -EFAULT;
4177 goto out;
4178 }
4179
4180 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4181 pinfo.spinfo_assoc_id);
4182 if (!transport)
4183 return -EINVAL;
4184
4185 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4186 pinfo.spinfo_state = transport->state;
4187 pinfo.spinfo_cwnd = transport->cwnd;
4188 pinfo.spinfo_srtt = transport->srtt;
4189 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4190 pinfo.spinfo_mtu = transport->pathmtu;
4191
4192 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4193 pinfo.spinfo_state = SCTP_ACTIVE;
4194
4195 if (put_user(len, optlen)) {
4196 retval = -EFAULT;
4197 goto out;
4198 }
4199
4200 if (copy_to_user(optval, &pinfo, len)) {
4201 retval = -EFAULT;
4202 goto out;
4203 }
4204
4205 out:
4206 return retval;
4207 }
4208
4209 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4210 *
4211 * This option is a on/off flag. If enabled no SCTP message
4212 * fragmentation will be performed. Instead if a message being sent
4213 * exceeds the current PMTU size, the message will NOT be sent and
4214 * instead a error will be indicated to the user.
4215 */
4216 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4217 char __user *optval, int __user *optlen)
4218 {
4219 int val;
4220
4221 if (len < sizeof(int))
4222 return -EINVAL;
4223
4224 len = sizeof(int);
4225 val = (sctp_sk(sk)->disable_fragments == 1);
4226 if (put_user(len, optlen))
4227 return -EFAULT;
4228 if (copy_to_user(optval, &val, len))
4229 return -EFAULT;
4230 return 0;
4231 }
4232
4233 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4234 *
4235 * This socket option is used to specify various notifications and
4236 * ancillary data the user wishes to receive.
4237 */
4238 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4239 int __user *optlen)
4240 {
4241 if (len <= 0)
4242 return -EINVAL;
4243 if (len > sizeof(struct sctp_event_subscribe))
4244 len = sizeof(struct sctp_event_subscribe);
4245 if (put_user(len, optlen))
4246 return -EFAULT;
4247 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4248 return -EFAULT;
4249 return 0;
4250 }
4251
4252 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4253 *
4254 * This socket option is applicable to the UDP-style socket only. When
4255 * set it will cause associations that are idle for more than the
4256 * specified number of seconds to automatically close. An association
4257 * being idle is defined an association that has NOT sent or received
4258 * user data. The special value of '0' indicates that no automatic
4259 * close of any associations should be performed. The option expects an
4260 * integer defining the number of seconds of idle time before an
4261 * association is closed.
4262 */
4263 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4264 {
4265 /* Applicable to UDP-style socket only */
4266 if (sctp_style(sk, TCP))
4267 return -EOPNOTSUPP;
4268 if (len < sizeof(int))
4269 return -EINVAL;
4270 len = sizeof(int);
4271 if (put_user(len, optlen))
4272 return -EFAULT;
4273 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4274 return -EFAULT;
4275 return 0;
4276 }
4277
4278 /* Helper routine to branch off an association to a new socket. */
4279 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4280 {
4281 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4282 struct socket *sock;
4283 struct sctp_af *af;
4284 int err = 0;
4285
4286 if (!asoc)
4287 return -EINVAL;
4288
4289 /* An association cannot be branched off from an already peeled-off
4290 * socket, nor is this supported for tcp style sockets.
4291 */
4292 if (!sctp_style(sk, UDP))
4293 return -EINVAL;
4294
4295 /* Create a new socket. */
4296 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4297 if (err < 0)
4298 return err;
4299
4300 sctp_copy_sock(sock->sk, sk, asoc);
4301
4302 /* Make peeled-off sockets more like 1-1 accepted sockets.
4303 * Set the daddr and initialize id to something more random
4304 */
4305 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4306 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4307
4308 /* Populate the fields of the newsk from the oldsk and migrate the
4309 * asoc to the newsk.
4310 */
4311 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4312
4313 *sockp = sock;
4314
4315 return err;
4316 }
4317 EXPORT_SYMBOL(sctp_do_peeloff);
4318
4319 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4320 {
4321 sctp_peeloff_arg_t peeloff;
4322 struct socket *newsock;
4323 struct file *newfile;
4324 int retval = 0;
4325
4326 if (len < sizeof(sctp_peeloff_arg_t))
4327 return -EINVAL;
4328 len = sizeof(sctp_peeloff_arg_t);
4329 if (copy_from_user(&peeloff, optval, len))
4330 return -EFAULT;
4331
4332 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4333 if (retval < 0)
4334 goto out;
4335
4336 /* Map the socket to an unused fd that can be returned to the user. */
4337 retval = get_unused_fd_flags(0);
4338 if (retval < 0) {
4339 sock_release(newsock);
4340 goto out;
4341 }
4342
4343 newfile = sock_alloc_file(newsock, 0, NULL);
4344 if (unlikely(IS_ERR(newfile))) {
4345 put_unused_fd(retval);
4346 sock_release(newsock);
4347 return PTR_ERR(newfile);
4348 }
4349
4350 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4351 retval);
4352
4353 /* Return the fd mapped to the new socket. */
4354 if (put_user(len, optlen)) {
4355 fput(newfile);
4356 put_unused_fd(retval);
4357 return -EFAULT;
4358 }
4359 peeloff.sd = retval;
4360 if (copy_to_user(optval, &peeloff, len)) {
4361 fput(newfile);
4362 put_unused_fd(retval);
4363 return -EFAULT;
4364 }
4365 fd_install(retval, newfile);
4366 out:
4367 return retval;
4368 }
4369
4370 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4371 *
4372 * Applications can enable or disable heartbeats for any peer address of
4373 * an association, modify an address's heartbeat interval, force a
4374 * heartbeat to be sent immediately, and adjust the address's maximum
4375 * number of retransmissions sent before an address is considered
4376 * unreachable. The following structure is used to access and modify an
4377 * address's parameters:
4378 *
4379 * struct sctp_paddrparams {
4380 * sctp_assoc_t spp_assoc_id;
4381 * struct sockaddr_storage spp_address;
4382 * uint32_t spp_hbinterval;
4383 * uint16_t spp_pathmaxrxt;
4384 * uint32_t spp_pathmtu;
4385 * uint32_t spp_sackdelay;
4386 * uint32_t spp_flags;
4387 * };
4388 *
4389 * spp_assoc_id - (one-to-many style socket) This is filled in the
4390 * application, and identifies the association for
4391 * this query.
4392 * spp_address - This specifies which address is of interest.
4393 * spp_hbinterval - This contains the value of the heartbeat interval,
4394 * in milliseconds. If a value of zero
4395 * is present in this field then no changes are to
4396 * be made to this parameter.
4397 * spp_pathmaxrxt - This contains the maximum number of
4398 * retransmissions before this address shall be
4399 * considered unreachable. If a value of zero
4400 * is present in this field then no changes are to
4401 * be made to this parameter.
4402 * spp_pathmtu - When Path MTU discovery is disabled the value
4403 * specified here will be the "fixed" path mtu.
4404 * Note that if the spp_address field is empty
4405 * then all associations on this address will
4406 * have this fixed path mtu set upon them.
4407 *
4408 * spp_sackdelay - When delayed sack is enabled, this value specifies
4409 * the number of milliseconds that sacks will be delayed
4410 * for. This value will apply to all addresses of an
4411 * association if the spp_address field is empty. Note
4412 * also, that if delayed sack is enabled and this
4413 * value is set to 0, no change is made to the last
4414 * recorded delayed sack timer value.
4415 *
4416 * spp_flags - These flags are used to control various features
4417 * on an association. The flag field may contain
4418 * zero or more of the following options.
4419 *
4420 * SPP_HB_ENABLE - Enable heartbeats on the
4421 * specified address. Note that if the address
4422 * field is empty all addresses for the association
4423 * have heartbeats enabled upon them.
4424 *
4425 * SPP_HB_DISABLE - Disable heartbeats on the
4426 * speicifed address. Note that if the address
4427 * field is empty all addresses for the association
4428 * will have their heartbeats disabled. Note also
4429 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4430 * mutually exclusive, only one of these two should
4431 * be specified. Enabling both fields will have
4432 * undetermined results.
4433 *
4434 * SPP_HB_DEMAND - Request a user initiated heartbeat
4435 * to be made immediately.
4436 *
4437 * SPP_PMTUD_ENABLE - This field will enable PMTU
4438 * discovery upon the specified address. Note that
4439 * if the address feild is empty then all addresses
4440 * on the association are effected.
4441 *
4442 * SPP_PMTUD_DISABLE - This field will disable PMTU
4443 * discovery upon the specified address. Note that
4444 * if the address feild is empty then all addresses
4445 * on the association are effected. Not also that
4446 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4447 * exclusive. Enabling both will have undetermined
4448 * results.
4449 *
4450 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4451 * on delayed sack. The time specified in spp_sackdelay
4452 * is used to specify the sack delay for this address. Note
4453 * that if spp_address is empty then all addresses will
4454 * enable delayed sack and take on the sack delay
4455 * value specified in spp_sackdelay.
4456 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4457 * off delayed sack. If the spp_address field is blank then
4458 * delayed sack is disabled for the entire association. Note
4459 * also that this field is mutually exclusive to
4460 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4461 * results.
4462 */
4463 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4464 char __user *optval, int __user *optlen)
4465 {
4466 struct sctp_paddrparams params;
4467 struct sctp_transport *trans = NULL;
4468 struct sctp_association *asoc = NULL;
4469 struct sctp_sock *sp = sctp_sk(sk);
4470
4471 if (len < sizeof(struct sctp_paddrparams))
4472 return -EINVAL;
4473 len = sizeof(struct sctp_paddrparams);
4474 if (copy_from_user(&params, optval, len))
4475 return -EFAULT;
4476
4477 /* If an address other than INADDR_ANY is specified, and
4478 * no transport is found, then the request is invalid.
4479 */
4480 if (!sctp_is_any(sk, (union sctp_addr *)&params.spp_address)) {
4481 trans = sctp_addr_id2transport(sk, &params.spp_address,
4482 params.spp_assoc_id);
4483 if (!trans) {
4484 pr_debug("%s: failed no transport\n", __func__);
4485 return -EINVAL;
4486 }
4487 }
4488
4489 /* Get association, if assoc_id != 0 and the socket is a one
4490 * to many style socket, and an association was not found, then
4491 * the id was invalid.
4492 */
4493 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4494 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4495 pr_debug("%s: failed no association\n", __func__);
4496 return -EINVAL;
4497 }
4498
4499 if (trans) {
4500 /* Fetch transport values. */
4501 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4502 params.spp_pathmtu = trans->pathmtu;
4503 params.spp_pathmaxrxt = trans->pathmaxrxt;
4504 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4505
4506 /*draft-11 doesn't say what to return in spp_flags*/
4507 params.spp_flags = trans->param_flags;
4508 } else if (asoc) {
4509 /* Fetch association values. */
4510 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4511 params.spp_pathmtu = asoc->pathmtu;
4512 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4513 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4514
4515 /*draft-11 doesn't say what to return in spp_flags*/
4516 params.spp_flags = asoc->param_flags;
4517 } else {
4518 /* Fetch socket values. */
4519 params.spp_hbinterval = sp->hbinterval;
4520 params.spp_pathmtu = sp->pathmtu;
4521 params.spp_sackdelay = sp->sackdelay;
4522 params.spp_pathmaxrxt = sp->pathmaxrxt;
4523
4524 /*draft-11 doesn't say what to return in spp_flags*/
4525 params.spp_flags = sp->param_flags;
4526 }
4527
4528 if (copy_to_user(optval, &params, len))
4529 return -EFAULT;
4530
4531 if (put_user(len, optlen))
4532 return -EFAULT;
4533
4534 return 0;
4535 }
4536
4537 /*
4538 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4539 *
4540 * This option will effect the way delayed acks are performed. This
4541 * option allows you to get or set the delayed ack time, in
4542 * milliseconds. It also allows changing the delayed ack frequency.
4543 * Changing the frequency to 1 disables the delayed sack algorithm. If
4544 * the assoc_id is 0, then this sets or gets the endpoints default
4545 * values. If the assoc_id field is non-zero, then the set or get
4546 * effects the specified association for the one to many model (the
4547 * assoc_id field is ignored by the one to one model). Note that if
4548 * sack_delay or sack_freq are 0 when setting this option, then the
4549 * current values will remain unchanged.
4550 *
4551 * struct sctp_sack_info {
4552 * sctp_assoc_t sack_assoc_id;
4553 * uint32_t sack_delay;
4554 * uint32_t sack_freq;
4555 * };
4556 *
4557 * sack_assoc_id - This parameter, indicates which association the user
4558 * is performing an action upon. Note that if this field's value is
4559 * zero then the endpoints default value is changed (effecting future
4560 * associations only).
4561 *
4562 * sack_delay - This parameter contains the number of milliseconds that
4563 * the user is requesting the delayed ACK timer be set to. Note that
4564 * this value is defined in the standard to be between 200 and 500
4565 * milliseconds.
4566 *
4567 * sack_freq - This parameter contains the number of packets that must
4568 * be received before a sack is sent without waiting for the delay
4569 * timer to expire. The default value for this is 2, setting this
4570 * value to 1 will disable the delayed sack algorithm.
4571 */
4572 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4573 char __user *optval,
4574 int __user *optlen)
4575 {
4576 struct sctp_sack_info params;
4577 struct sctp_association *asoc = NULL;
4578 struct sctp_sock *sp = sctp_sk(sk);
4579
4580 if (len >= sizeof(struct sctp_sack_info)) {
4581 len = sizeof(struct sctp_sack_info);
4582
4583 if (copy_from_user(&params, optval, len))
4584 return -EFAULT;
4585 } else if (len == sizeof(struct sctp_assoc_value)) {
4586 pr_warn_ratelimited(DEPRECATED
4587 "%s (pid %d) "
4588 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
4589 "Use struct sctp_sack_info instead\n",
4590 current->comm, task_pid_nr(current));
4591 if (copy_from_user(&params, optval, len))
4592 return -EFAULT;
4593 } else
4594 return -EINVAL;
4595
4596 /* Get association, if sack_assoc_id != 0 and the socket is a one
4597 * to many style socket, and an association was not found, then
4598 * the id was invalid.
4599 */
4600 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4601 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4602 return -EINVAL;
4603
4604 if (asoc) {
4605 /* Fetch association values. */
4606 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4607 params.sack_delay = jiffies_to_msecs(
4608 asoc->sackdelay);
4609 params.sack_freq = asoc->sackfreq;
4610
4611 } else {
4612 params.sack_delay = 0;
4613 params.sack_freq = 1;
4614 }
4615 } else {
4616 /* Fetch socket values. */
4617 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4618 params.sack_delay = sp->sackdelay;
4619 params.sack_freq = sp->sackfreq;
4620 } else {
4621 params.sack_delay = 0;
4622 params.sack_freq = 1;
4623 }
4624 }
4625
4626 if (copy_to_user(optval, &params, len))
4627 return -EFAULT;
4628
4629 if (put_user(len, optlen))
4630 return -EFAULT;
4631
4632 return 0;
4633 }
4634
4635 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4636 *
4637 * Applications can specify protocol parameters for the default association
4638 * initialization. The option name argument to setsockopt() and getsockopt()
4639 * is SCTP_INITMSG.
4640 *
4641 * Setting initialization parameters is effective only on an unconnected
4642 * socket (for UDP-style sockets only future associations are effected
4643 * by the change). With TCP-style sockets, this option is inherited by
4644 * sockets derived from a listener socket.
4645 */
4646 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4647 {
4648 if (len < sizeof(struct sctp_initmsg))
4649 return -EINVAL;
4650 len = sizeof(struct sctp_initmsg);
4651 if (put_user(len, optlen))
4652 return -EFAULT;
4653 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4654 return -EFAULT;
4655 return 0;
4656 }
4657
4658
4659 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4660 char __user *optval, int __user *optlen)
4661 {
4662 struct sctp_association *asoc;
4663 int cnt = 0;
4664 struct sctp_getaddrs getaddrs;
4665 struct sctp_transport *from;
4666 void __user *to;
4667 union sctp_addr temp;
4668 struct sctp_sock *sp = sctp_sk(sk);
4669 int addrlen;
4670 size_t space_left;
4671 int bytes_copied;
4672
4673 if (len < sizeof(struct sctp_getaddrs))
4674 return -EINVAL;
4675
4676 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4677 return -EFAULT;
4678
4679 /* For UDP-style sockets, id specifies the association to query. */
4680 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4681 if (!asoc)
4682 return -EINVAL;
4683
4684 to = optval + offsetof(struct sctp_getaddrs, addrs);
4685 space_left = len - offsetof(struct sctp_getaddrs, addrs);
4686
4687 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4688 transports) {
4689 memcpy(&temp, &from->ipaddr, sizeof(temp));
4690 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4691 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4692 if (space_left < addrlen)
4693 return -ENOMEM;
4694 if (copy_to_user(to, &temp, addrlen))
4695 return -EFAULT;
4696 to += addrlen;
4697 cnt++;
4698 space_left -= addrlen;
4699 }
4700
4701 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4702 return -EFAULT;
4703 bytes_copied = ((char __user *)to) - optval;
4704 if (put_user(bytes_copied, optlen))
4705 return -EFAULT;
4706
4707 return 0;
4708 }
4709
4710 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4711 size_t space_left, int *bytes_copied)
4712 {
4713 struct sctp_sockaddr_entry *addr;
4714 union sctp_addr temp;
4715 int cnt = 0;
4716 int addrlen;
4717 struct net *net = sock_net(sk);
4718
4719 rcu_read_lock();
4720 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
4721 if (!addr->valid)
4722 continue;
4723
4724 if ((PF_INET == sk->sk_family) &&
4725 (AF_INET6 == addr->a.sa.sa_family))
4726 continue;
4727 if ((PF_INET6 == sk->sk_family) &&
4728 inet_v6_ipv6only(sk) &&
4729 (AF_INET == addr->a.sa.sa_family))
4730 continue;
4731 memcpy(&temp, &addr->a, sizeof(temp));
4732 if (!temp.v4.sin_port)
4733 temp.v4.sin_port = htons(port);
4734
4735 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4736 &temp);
4737 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4738 if (space_left < addrlen) {
4739 cnt = -ENOMEM;
4740 break;
4741 }
4742 memcpy(to, &temp, addrlen);
4743
4744 to += addrlen;
4745 cnt++;
4746 space_left -= addrlen;
4747 *bytes_copied += addrlen;
4748 }
4749 rcu_read_unlock();
4750
4751 return cnt;
4752 }
4753
4754
4755 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4756 char __user *optval, int __user *optlen)
4757 {
4758 struct sctp_bind_addr *bp;
4759 struct sctp_association *asoc;
4760 int cnt = 0;
4761 struct sctp_getaddrs getaddrs;
4762 struct sctp_sockaddr_entry *addr;
4763 void __user *to;
4764 union sctp_addr temp;
4765 struct sctp_sock *sp = sctp_sk(sk);
4766 int addrlen;
4767 int err = 0;
4768 size_t space_left;
4769 int bytes_copied = 0;
4770 void *addrs;
4771 void *buf;
4772
4773 if (len < sizeof(struct sctp_getaddrs))
4774 return -EINVAL;
4775
4776 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4777 return -EFAULT;
4778
4779 /*
4780 * For UDP-style sockets, id specifies the association to query.
4781 * If the id field is set to the value '0' then the locally bound
4782 * addresses are returned without regard to any particular
4783 * association.
4784 */
4785 if (0 == getaddrs.assoc_id) {
4786 bp = &sctp_sk(sk)->ep->base.bind_addr;
4787 } else {
4788 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4789 if (!asoc)
4790 return -EINVAL;
4791 bp = &asoc->base.bind_addr;
4792 }
4793
4794 to = optval + offsetof(struct sctp_getaddrs, addrs);
4795 space_left = len - offsetof(struct sctp_getaddrs, addrs);
4796
4797 addrs = kmalloc(space_left, GFP_KERNEL);
4798 if (!addrs)
4799 return -ENOMEM;
4800
4801 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4802 * addresses from the global local address list.
4803 */
4804 if (sctp_list_single_entry(&bp->address_list)) {
4805 addr = list_entry(bp->address_list.next,
4806 struct sctp_sockaddr_entry, list);
4807 if (sctp_is_any(sk, &addr->a)) {
4808 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4809 space_left, &bytes_copied);
4810 if (cnt < 0) {
4811 err = cnt;
4812 goto out;
4813 }
4814 goto copy_getaddrs;
4815 }
4816 }
4817
4818 buf = addrs;
4819 /* Protection on the bound address list is not needed since
4820 * in the socket option context we hold a socket lock and
4821 * thus the bound address list can't change.
4822 */
4823 list_for_each_entry(addr, &bp->address_list, list) {
4824 memcpy(&temp, &addr->a, sizeof(temp));
4825 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4826 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4827 if (space_left < addrlen) {
4828 err = -ENOMEM; /*fixme: right error?*/
4829 goto out;
4830 }
4831 memcpy(buf, &temp, addrlen);
4832 buf += addrlen;
4833 bytes_copied += addrlen;
4834 cnt++;
4835 space_left -= addrlen;
4836 }
4837
4838 copy_getaddrs:
4839 if (copy_to_user(to, addrs, bytes_copied)) {
4840 err = -EFAULT;
4841 goto out;
4842 }
4843 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4844 err = -EFAULT;
4845 goto out;
4846 }
4847 if (put_user(bytes_copied, optlen))
4848 err = -EFAULT;
4849 out:
4850 kfree(addrs);
4851 return err;
4852 }
4853
4854 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4855 *
4856 * Requests that the local SCTP stack use the enclosed peer address as
4857 * the association primary. The enclosed address must be one of the
4858 * association peer's addresses.
4859 */
4860 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4861 char __user *optval, int __user *optlen)
4862 {
4863 struct sctp_prim prim;
4864 struct sctp_association *asoc;
4865 struct sctp_sock *sp = sctp_sk(sk);
4866
4867 if (len < sizeof(struct sctp_prim))
4868 return -EINVAL;
4869
4870 len = sizeof(struct sctp_prim);
4871
4872 if (copy_from_user(&prim, optval, len))
4873 return -EFAULT;
4874
4875 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4876 if (!asoc)
4877 return -EINVAL;
4878
4879 if (!asoc->peer.primary_path)
4880 return -ENOTCONN;
4881
4882 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4883 asoc->peer.primary_path->af_specific->sockaddr_len);
4884
4885 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4886 (union sctp_addr *)&prim.ssp_addr);
4887
4888 if (put_user(len, optlen))
4889 return -EFAULT;
4890 if (copy_to_user(optval, &prim, len))
4891 return -EFAULT;
4892
4893 return 0;
4894 }
4895
4896 /*
4897 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4898 *
4899 * Requests that the local endpoint set the specified Adaptation Layer
4900 * Indication parameter for all future INIT and INIT-ACK exchanges.
4901 */
4902 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4903 char __user *optval, int __user *optlen)
4904 {
4905 struct sctp_setadaptation adaptation;
4906
4907 if (len < sizeof(struct sctp_setadaptation))
4908 return -EINVAL;
4909
4910 len = sizeof(struct sctp_setadaptation);
4911
4912 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4913
4914 if (put_user(len, optlen))
4915 return -EFAULT;
4916 if (copy_to_user(optval, &adaptation, len))
4917 return -EFAULT;
4918
4919 return 0;
4920 }
4921
4922 /*
4923 *
4924 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4925 *
4926 * Applications that wish to use the sendto() system call may wish to
4927 * specify a default set of parameters that would normally be supplied
4928 * through the inclusion of ancillary data. This socket option allows
4929 * such an application to set the default sctp_sndrcvinfo structure.
4930
4931
4932 * The application that wishes to use this socket option simply passes
4933 * in to this call the sctp_sndrcvinfo structure defined in Section
4934 * 5.2.2) The input parameters accepted by this call include
4935 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4936 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4937 * to this call if the caller is using the UDP model.
4938 *
4939 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4940 */
4941 static int sctp_getsockopt_default_send_param(struct sock *sk,
4942 int len, char __user *optval,
4943 int __user *optlen)
4944 {
4945 struct sctp_sndrcvinfo info;
4946 struct sctp_association *asoc;
4947 struct sctp_sock *sp = sctp_sk(sk);
4948
4949 if (len < sizeof(struct sctp_sndrcvinfo))
4950 return -EINVAL;
4951
4952 len = sizeof(struct sctp_sndrcvinfo);
4953
4954 if (copy_from_user(&info, optval, len))
4955 return -EFAULT;
4956
4957 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4958 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4959 return -EINVAL;
4960
4961 if (asoc) {
4962 info.sinfo_stream = asoc->default_stream;
4963 info.sinfo_flags = asoc->default_flags;
4964 info.sinfo_ppid = asoc->default_ppid;
4965 info.sinfo_context = asoc->default_context;
4966 info.sinfo_timetolive = asoc->default_timetolive;
4967 } else {
4968 info.sinfo_stream = sp->default_stream;
4969 info.sinfo_flags = sp->default_flags;
4970 info.sinfo_ppid = sp->default_ppid;
4971 info.sinfo_context = sp->default_context;
4972 info.sinfo_timetolive = sp->default_timetolive;
4973 }
4974
4975 if (put_user(len, optlen))
4976 return -EFAULT;
4977 if (copy_to_user(optval, &info, len))
4978 return -EFAULT;
4979
4980 return 0;
4981 }
4982
4983 /*
4984 *
4985 * 7.1.5 SCTP_NODELAY
4986 *
4987 * Turn on/off any Nagle-like algorithm. This means that packets are
4988 * generally sent as soon as possible and no unnecessary delays are
4989 * introduced, at the cost of more packets in the network. Expects an
4990 * integer boolean flag.
4991 */
4992
4993 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4994 char __user *optval, int __user *optlen)
4995 {
4996 int val;
4997
4998 if (len < sizeof(int))
4999 return -EINVAL;
5000
5001 len = sizeof(int);
5002 val = (sctp_sk(sk)->nodelay == 1);
5003 if (put_user(len, optlen))
5004 return -EFAULT;
5005 if (copy_to_user(optval, &val, len))
5006 return -EFAULT;
5007 return 0;
5008 }
5009
5010 /*
5011 *
5012 * 7.1.1 SCTP_RTOINFO
5013 *
5014 * The protocol parameters used to initialize and bound retransmission
5015 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5016 * and modify these parameters.
5017 * All parameters are time values, in milliseconds. A value of 0, when
5018 * modifying the parameters, indicates that the current value should not
5019 * be changed.
5020 *
5021 */
5022 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5023 char __user *optval,
5024 int __user *optlen) {
5025 struct sctp_rtoinfo rtoinfo;
5026 struct sctp_association *asoc;
5027
5028 if (len < sizeof (struct sctp_rtoinfo))
5029 return -EINVAL;
5030
5031 len = sizeof(struct sctp_rtoinfo);
5032
5033 if (copy_from_user(&rtoinfo, optval, len))
5034 return -EFAULT;
5035
5036 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5037
5038 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5039 return -EINVAL;
5040
5041 /* Values corresponding to the specific association. */
5042 if (asoc) {
5043 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5044 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5045 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5046 } else {
5047 /* Values corresponding to the endpoint. */
5048 struct sctp_sock *sp = sctp_sk(sk);
5049
5050 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5051 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5052 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5053 }
5054
5055 if (put_user(len, optlen))
5056 return -EFAULT;
5057
5058 if (copy_to_user(optval, &rtoinfo, len))
5059 return -EFAULT;
5060
5061 return 0;
5062 }
5063
5064 /*
5065 *
5066 * 7.1.2 SCTP_ASSOCINFO
5067 *
5068 * This option is used to tune the maximum retransmission attempts
5069 * of the association.
5070 * Returns an error if the new association retransmission value is
5071 * greater than the sum of the retransmission value of the peer.
5072 * See [SCTP] for more information.
5073 *
5074 */
5075 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5076 char __user *optval,
5077 int __user *optlen)
5078 {
5079
5080 struct sctp_assocparams assocparams;
5081 struct sctp_association *asoc;
5082 struct list_head *pos;
5083 int cnt = 0;
5084
5085 if (len < sizeof (struct sctp_assocparams))
5086 return -EINVAL;
5087
5088 len = sizeof(struct sctp_assocparams);
5089
5090 if (copy_from_user(&assocparams, optval, len))
5091 return -EFAULT;
5092
5093 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5094
5095 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5096 return -EINVAL;
5097
5098 /* Values correspoinding to the specific association */
5099 if (asoc) {
5100 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5101 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5102 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5103 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5104
5105 list_for_each(pos, &asoc->peer.transport_addr_list) {
5106 cnt++;
5107 }
5108
5109 assocparams.sasoc_number_peer_destinations = cnt;
5110 } else {
5111 /* Values corresponding to the endpoint */
5112 struct sctp_sock *sp = sctp_sk(sk);
5113
5114 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5115 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5116 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5117 assocparams.sasoc_cookie_life =
5118 sp->assocparams.sasoc_cookie_life;
5119 assocparams.sasoc_number_peer_destinations =
5120 sp->assocparams.
5121 sasoc_number_peer_destinations;
5122 }
5123
5124 if (put_user(len, optlen))
5125 return -EFAULT;
5126
5127 if (copy_to_user(optval, &assocparams, len))
5128 return -EFAULT;
5129
5130 return 0;
5131 }
5132
5133 /*
5134 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5135 *
5136 * This socket option is a boolean flag which turns on or off mapped V4
5137 * addresses. If this option is turned on and the socket is type
5138 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5139 * If this option is turned off, then no mapping will be done of V4
5140 * addresses and a user will receive both PF_INET6 and PF_INET type
5141 * addresses on the socket.
5142 */
5143 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5144 char __user *optval, int __user *optlen)
5145 {
5146 int val;
5147 struct sctp_sock *sp = sctp_sk(sk);
5148
5149 if (len < sizeof(int))
5150 return -EINVAL;
5151
5152 len = sizeof(int);
5153 val = sp->v4mapped;
5154 if (put_user(len, optlen))
5155 return -EFAULT;
5156 if (copy_to_user(optval, &val, len))
5157 return -EFAULT;
5158
5159 return 0;
5160 }
5161
5162 /*
5163 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5164 * (chapter and verse is quoted at sctp_setsockopt_context())
5165 */
5166 static int sctp_getsockopt_context(struct sock *sk, int len,
5167 char __user *optval, int __user *optlen)
5168 {
5169 struct sctp_assoc_value params;
5170 struct sctp_sock *sp;
5171 struct sctp_association *asoc;
5172
5173 if (len < sizeof(struct sctp_assoc_value))
5174 return -EINVAL;
5175
5176 len = sizeof(struct sctp_assoc_value);
5177
5178 if (copy_from_user(&params, optval, len))
5179 return -EFAULT;
5180
5181 sp = sctp_sk(sk);
5182
5183 if (params.assoc_id != 0) {
5184 asoc = sctp_id2assoc(sk, params.assoc_id);
5185 if (!asoc)
5186 return -EINVAL;
5187 params.assoc_value = asoc->default_rcv_context;
5188 } else {
5189 params.assoc_value = sp->default_rcv_context;
5190 }
5191
5192 if (put_user(len, optlen))
5193 return -EFAULT;
5194 if (copy_to_user(optval, &params, len))
5195 return -EFAULT;
5196
5197 return 0;
5198 }
5199
5200 /*
5201 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5202 * This option will get or set the maximum size to put in any outgoing
5203 * SCTP DATA chunk. If a message is larger than this size it will be
5204 * fragmented by SCTP into the specified size. Note that the underlying
5205 * SCTP implementation may fragment into smaller sized chunks when the
5206 * PMTU of the underlying association is smaller than the value set by
5207 * the user. The default value for this option is '0' which indicates
5208 * the user is NOT limiting fragmentation and only the PMTU will effect
5209 * SCTP's choice of DATA chunk size. Note also that values set larger
5210 * than the maximum size of an IP datagram will effectively let SCTP
5211 * control fragmentation (i.e. the same as setting this option to 0).
5212 *
5213 * The following structure is used to access and modify this parameter:
5214 *
5215 * struct sctp_assoc_value {
5216 * sctp_assoc_t assoc_id;
5217 * uint32_t assoc_value;
5218 * };
5219 *
5220 * assoc_id: This parameter is ignored for one-to-one style sockets.
5221 * For one-to-many style sockets this parameter indicates which
5222 * association the user is performing an action upon. Note that if
5223 * this field's value is zero then the endpoints default value is
5224 * changed (effecting future associations only).
5225 * assoc_value: This parameter specifies the maximum size in bytes.
5226 */
5227 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5228 char __user *optval, int __user *optlen)
5229 {
5230 struct sctp_assoc_value params;
5231 struct sctp_association *asoc;
5232
5233 if (len == sizeof(int)) {
5234 pr_warn_ratelimited(DEPRECATED
5235 "%s (pid %d) "
5236 "Use of int in maxseg socket option.\n"
5237 "Use struct sctp_assoc_value instead\n",
5238 current->comm, task_pid_nr(current));
5239 params.assoc_id = 0;
5240 } else if (len >= sizeof(struct sctp_assoc_value)) {
5241 len = sizeof(struct sctp_assoc_value);
5242 if (copy_from_user(&params, optval, sizeof(params)))
5243 return -EFAULT;
5244 } else
5245 return -EINVAL;
5246
5247 asoc = sctp_id2assoc(sk, params.assoc_id);
5248 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5249 return -EINVAL;
5250
5251 if (asoc)
5252 params.assoc_value = asoc->frag_point;
5253 else
5254 params.assoc_value = sctp_sk(sk)->user_frag;
5255
5256 if (put_user(len, optlen))
5257 return -EFAULT;
5258 if (len == sizeof(int)) {
5259 if (copy_to_user(optval, &params.assoc_value, len))
5260 return -EFAULT;
5261 } else {
5262 if (copy_to_user(optval, &params, len))
5263 return -EFAULT;
5264 }
5265
5266 return 0;
5267 }
5268
5269 /*
5270 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5271 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5272 */
5273 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5274 char __user *optval, int __user *optlen)
5275 {
5276 int val;
5277
5278 if (len < sizeof(int))
5279 return -EINVAL;
5280
5281 len = sizeof(int);
5282
5283 val = sctp_sk(sk)->frag_interleave;
5284 if (put_user(len, optlen))
5285 return -EFAULT;
5286 if (copy_to_user(optval, &val, len))
5287 return -EFAULT;
5288
5289 return 0;
5290 }
5291
5292 /*
5293 * 7.1.25. Set or Get the sctp partial delivery point
5294 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5295 */
5296 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5297 char __user *optval,
5298 int __user *optlen)
5299 {
5300 u32 val;
5301
5302 if (len < sizeof(u32))
5303 return -EINVAL;
5304
5305 len = sizeof(u32);
5306
5307 val = sctp_sk(sk)->pd_point;
5308 if (put_user(len, optlen))
5309 return -EFAULT;
5310 if (copy_to_user(optval, &val, len))
5311 return -EFAULT;
5312
5313 return 0;
5314 }
5315
5316 /*
5317 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5318 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5319 */
5320 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5321 char __user *optval,
5322 int __user *optlen)
5323 {
5324 struct sctp_assoc_value params;
5325 struct sctp_sock *sp;
5326 struct sctp_association *asoc;
5327
5328 if (len == sizeof(int)) {
5329 pr_warn_ratelimited(DEPRECATED
5330 "%s (pid %d) "
5331 "Use of int in max_burst socket option.\n"
5332 "Use struct sctp_assoc_value instead\n",
5333 current->comm, task_pid_nr(current));
5334 params.assoc_id = 0;
5335 } else if (len >= sizeof(struct sctp_assoc_value)) {
5336 len = sizeof(struct sctp_assoc_value);
5337 if (copy_from_user(&params, optval, len))
5338 return -EFAULT;
5339 } else
5340 return -EINVAL;
5341
5342 sp = sctp_sk(sk);
5343
5344 if (params.assoc_id != 0) {
5345 asoc = sctp_id2assoc(sk, params.assoc_id);
5346 if (!asoc)
5347 return -EINVAL;
5348 params.assoc_value = asoc->max_burst;
5349 } else
5350 params.assoc_value = sp->max_burst;
5351
5352 if (len == sizeof(int)) {
5353 if (copy_to_user(optval, &params.assoc_value, len))
5354 return -EFAULT;
5355 } else {
5356 if (copy_to_user(optval, &params, len))
5357 return -EFAULT;
5358 }
5359
5360 return 0;
5361
5362 }
5363
5364 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5365 char __user *optval, int __user *optlen)
5366 {
5367 struct net *net = sock_net(sk);
5368 struct sctp_hmacalgo __user *p = (void __user *)optval;
5369 struct sctp_hmac_algo_param *hmacs;
5370 __u16 data_len = 0;
5371 u32 num_idents;
5372
5373 if (!net->sctp.auth_enable)
5374 return -EACCES;
5375
5376 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5377 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5378
5379 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5380 return -EINVAL;
5381
5382 len = sizeof(struct sctp_hmacalgo) + data_len;
5383 num_idents = data_len / sizeof(u16);
5384
5385 if (put_user(len, optlen))
5386 return -EFAULT;
5387 if (put_user(num_idents, &p->shmac_num_idents))
5388 return -EFAULT;
5389 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5390 return -EFAULT;
5391 return 0;
5392 }
5393
5394 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5395 char __user *optval, int __user *optlen)
5396 {
5397 struct net *net = sock_net(sk);
5398 struct sctp_authkeyid val;
5399 struct sctp_association *asoc;
5400
5401 if (!net->sctp.auth_enable)
5402 return -EACCES;
5403
5404 if (len < sizeof(struct sctp_authkeyid))
5405 return -EINVAL;
5406 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5407 return -EFAULT;
5408
5409 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5410 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5411 return -EINVAL;
5412
5413 if (asoc)
5414 val.scact_keynumber = asoc->active_key_id;
5415 else
5416 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5417
5418 len = sizeof(struct sctp_authkeyid);
5419 if (put_user(len, optlen))
5420 return -EFAULT;
5421 if (copy_to_user(optval, &val, len))
5422 return -EFAULT;
5423
5424 return 0;
5425 }
5426
5427 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5428 char __user *optval, int __user *optlen)
5429 {
5430 struct net *net = sock_net(sk);
5431 struct sctp_authchunks __user *p = (void __user *)optval;
5432 struct sctp_authchunks val;
5433 struct sctp_association *asoc;
5434 struct sctp_chunks_param *ch;
5435 u32 num_chunks = 0;
5436 char __user *to;
5437
5438 if (!net->sctp.auth_enable)
5439 return -EACCES;
5440
5441 if (len < sizeof(struct sctp_authchunks))
5442 return -EINVAL;
5443
5444 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5445 return -EFAULT;
5446
5447 to = p->gauth_chunks;
5448 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5449 if (!asoc)
5450 return -EINVAL;
5451
5452 ch = asoc->peer.peer_chunks;
5453 if (!ch)
5454 goto num;
5455
5456 /* See if the user provided enough room for all the data */
5457 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5458 if (len < num_chunks)
5459 return -EINVAL;
5460
5461 if (copy_to_user(to, ch->chunks, num_chunks))
5462 return -EFAULT;
5463 num:
5464 len = sizeof(struct sctp_authchunks) + num_chunks;
5465 if (put_user(len, optlen))
5466 return -EFAULT;
5467 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5468 return -EFAULT;
5469 return 0;
5470 }
5471
5472 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5473 char __user *optval, int __user *optlen)
5474 {
5475 struct net *net = sock_net(sk);
5476 struct sctp_authchunks __user *p = (void __user *)optval;
5477 struct sctp_authchunks val;
5478 struct sctp_association *asoc;
5479 struct sctp_chunks_param *ch;
5480 u32 num_chunks = 0;
5481 char __user *to;
5482
5483 if (!net->sctp.auth_enable)
5484 return -EACCES;
5485
5486 if (len < sizeof(struct sctp_authchunks))
5487 return -EINVAL;
5488
5489 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5490 return -EFAULT;
5491
5492 to = p->gauth_chunks;
5493 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5494 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5495 return -EINVAL;
5496
5497 if (asoc)
5498 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
5499 else
5500 ch = sctp_sk(sk)->ep->auth_chunk_list;
5501
5502 if (!ch)
5503 goto num;
5504
5505 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5506 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5507 return -EINVAL;
5508
5509 if (copy_to_user(to, ch->chunks, num_chunks))
5510 return -EFAULT;
5511 num:
5512 len = sizeof(struct sctp_authchunks) + num_chunks;
5513 if (put_user(len, optlen))
5514 return -EFAULT;
5515 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5516 return -EFAULT;
5517
5518 return 0;
5519 }
5520
5521 /*
5522 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5523 * This option gets the current number of associations that are attached
5524 * to a one-to-many style socket. The option value is an uint32_t.
5525 */
5526 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5527 char __user *optval, int __user *optlen)
5528 {
5529 struct sctp_sock *sp = sctp_sk(sk);
5530 struct sctp_association *asoc;
5531 u32 val = 0;
5532
5533 if (sctp_style(sk, TCP))
5534 return -EOPNOTSUPP;
5535
5536 if (len < sizeof(u32))
5537 return -EINVAL;
5538
5539 len = sizeof(u32);
5540
5541 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5542 val++;
5543 }
5544
5545 if (put_user(len, optlen))
5546 return -EFAULT;
5547 if (copy_to_user(optval, &val, len))
5548 return -EFAULT;
5549
5550 return 0;
5551 }
5552
5553 /*
5554 * 8.1.23 SCTP_AUTO_ASCONF
5555 * See the corresponding setsockopt entry as description
5556 */
5557 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5558 char __user *optval, int __user *optlen)
5559 {
5560 int val = 0;
5561
5562 if (len < sizeof(int))
5563 return -EINVAL;
5564
5565 len = sizeof(int);
5566 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5567 val = 1;
5568 if (put_user(len, optlen))
5569 return -EFAULT;
5570 if (copy_to_user(optval, &val, len))
5571 return -EFAULT;
5572 return 0;
5573 }
5574
5575 /*
5576 * 8.2.6. Get the Current Identifiers of Associations
5577 * (SCTP_GET_ASSOC_ID_LIST)
5578 *
5579 * This option gets the current list of SCTP association identifiers of
5580 * the SCTP associations handled by a one-to-many style socket.
5581 */
5582 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5583 char __user *optval, int __user *optlen)
5584 {
5585 struct sctp_sock *sp = sctp_sk(sk);
5586 struct sctp_association *asoc;
5587 struct sctp_assoc_ids *ids;
5588 u32 num = 0;
5589
5590 if (sctp_style(sk, TCP))
5591 return -EOPNOTSUPP;
5592
5593 if (len < sizeof(struct sctp_assoc_ids))
5594 return -EINVAL;
5595
5596 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5597 num++;
5598 }
5599
5600 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5601 return -EINVAL;
5602
5603 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5604
5605 ids = kmalloc(len, GFP_KERNEL);
5606 if (unlikely(!ids))
5607 return -ENOMEM;
5608
5609 ids->gaids_number_of_ids = num;
5610 num = 0;
5611 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5612 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5613 }
5614
5615 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5616 kfree(ids);
5617 return -EFAULT;
5618 }
5619
5620 kfree(ids);
5621 return 0;
5622 }
5623
5624 /*
5625 * SCTP_PEER_ADDR_THLDS
5626 *
5627 * This option allows us to fetch the partially failed threshold for one or all
5628 * transports in an association. See Section 6.1 of:
5629 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5630 */
5631 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
5632 char __user *optval,
5633 int len,
5634 int __user *optlen)
5635 {
5636 struct sctp_paddrthlds val;
5637 struct sctp_transport *trans;
5638 struct sctp_association *asoc;
5639
5640 if (len < sizeof(struct sctp_paddrthlds))
5641 return -EINVAL;
5642 len = sizeof(struct sctp_paddrthlds);
5643 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
5644 return -EFAULT;
5645
5646 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
5647 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
5648 if (!asoc)
5649 return -ENOENT;
5650
5651 val.spt_pathpfthld = asoc->pf_retrans;
5652 val.spt_pathmaxrxt = asoc->pathmaxrxt;
5653 } else {
5654 trans = sctp_addr_id2transport(sk, &val.spt_address,
5655 val.spt_assoc_id);
5656 if (!trans)
5657 return -ENOENT;
5658
5659 val.spt_pathmaxrxt = trans->pathmaxrxt;
5660 val.spt_pathpfthld = trans->pf_retrans;
5661 }
5662
5663 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
5664 return -EFAULT;
5665
5666 return 0;
5667 }
5668
5669 /*
5670 * SCTP_GET_ASSOC_STATS
5671 *
5672 * This option retrieves local per endpoint statistics. It is modeled
5673 * after OpenSolaris' implementation
5674 */
5675 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
5676 char __user *optval,
5677 int __user *optlen)
5678 {
5679 struct sctp_assoc_stats sas;
5680 struct sctp_association *asoc = NULL;
5681
5682 /* User must provide at least the assoc id */
5683 if (len < sizeof(sctp_assoc_t))
5684 return -EINVAL;
5685
5686 /* Allow the struct to grow and fill in as much as possible */
5687 len = min_t(size_t, len, sizeof(sas));
5688
5689 if (copy_from_user(&sas, optval, len))
5690 return -EFAULT;
5691
5692 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
5693 if (!asoc)
5694 return -EINVAL;
5695
5696 sas.sas_rtxchunks = asoc->stats.rtxchunks;
5697 sas.sas_gapcnt = asoc->stats.gapcnt;
5698 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
5699 sas.sas_osacks = asoc->stats.osacks;
5700 sas.sas_isacks = asoc->stats.isacks;
5701 sas.sas_octrlchunks = asoc->stats.octrlchunks;
5702 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
5703 sas.sas_oodchunks = asoc->stats.oodchunks;
5704 sas.sas_iodchunks = asoc->stats.iodchunks;
5705 sas.sas_ouodchunks = asoc->stats.ouodchunks;
5706 sas.sas_iuodchunks = asoc->stats.iuodchunks;
5707 sas.sas_idupchunks = asoc->stats.idupchunks;
5708 sas.sas_opackets = asoc->stats.opackets;
5709 sas.sas_ipackets = asoc->stats.ipackets;
5710
5711 /* New high max rto observed, will return 0 if not a single
5712 * RTO update took place. obs_rto_ipaddr will be bogus
5713 * in such a case
5714 */
5715 sas.sas_maxrto = asoc->stats.max_obs_rto;
5716 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
5717 sizeof(struct sockaddr_storage));
5718
5719 /* Mark beginning of a new observation period */
5720 asoc->stats.max_obs_rto = asoc->rto_min;
5721
5722 if (put_user(len, optlen))
5723 return -EFAULT;
5724
5725 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
5726
5727 if (copy_to_user(optval, &sas, len))
5728 return -EFAULT;
5729
5730 return 0;
5731 }
5732
5733 static int sctp_getsockopt(struct sock *sk, int level, int optname,
5734 char __user *optval, int __user *optlen)
5735 {
5736 int retval = 0;
5737 int len;
5738
5739 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
5740
5741 /* I can hardly begin to describe how wrong this is. This is
5742 * so broken as to be worse than useless. The API draft
5743 * REALLY is NOT helpful here... I am not convinced that the
5744 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5745 * are at all well-founded.
5746 */
5747 if (level != SOL_SCTP) {
5748 struct sctp_af *af = sctp_sk(sk)->pf->af;
5749
5750 retval = af->getsockopt(sk, level, optname, optval, optlen);
5751 return retval;
5752 }
5753
5754 if (get_user(len, optlen))
5755 return -EFAULT;
5756
5757 lock_sock(sk);
5758
5759 switch (optname) {
5760 case SCTP_STATUS:
5761 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5762 break;
5763 case SCTP_DISABLE_FRAGMENTS:
5764 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5765 optlen);
5766 break;
5767 case SCTP_EVENTS:
5768 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5769 break;
5770 case SCTP_AUTOCLOSE:
5771 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5772 break;
5773 case SCTP_SOCKOPT_PEELOFF:
5774 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5775 break;
5776 case SCTP_PEER_ADDR_PARAMS:
5777 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5778 optlen);
5779 break;
5780 case SCTP_DELAYED_SACK:
5781 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5782 optlen);
5783 break;
5784 case SCTP_INITMSG:
5785 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5786 break;
5787 case SCTP_GET_PEER_ADDRS:
5788 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5789 optlen);
5790 break;
5791 case SCTP_GET_LOCAL_ADDRS:
5792 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5793 optlen);
5794 break;
5795 case SCTP_SOCKOPT_CONNECTX3:
5796 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5797 break;
5798 case SCTP_DEFAULT_SEND_PARAM:
5799 retval = sctp_getsockopt_default_send_param(sk, len,
5800 optval, optlen);
5801 break;
5802 case SCTP_PRIMARY_ADDR:
5803 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5804 break;
5805 case SCTP_NODELAY:
5806 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5807 break;
5808 case SCTP_RTOINFO:
5809 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5810 break;
5811 case SCTP_ASSOCINFO:
5812 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5813 break;
5814 case SCTP_I_WANT_MAPPED_V4_ADDR:
5815 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5816 break;
5817 case SCTP_MAXSEG:
5818 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5819 break;
5820 case SCTP_GET_PEER_ADDR_INFO:
5821 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5822 optlen);
5823 break;
5824 case SCTP_ADAPTATION_LAYER:
5825 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5826 optlen);
5827 break;
5828 case SCTP_CONTEXT:
5829 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5830 break;
5831 case SCTP_FRAGMENT_INTERLEAVE:
5832 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5833 optlen);
5834 break;
5835 case SCTP_PARTIAL_DELIVERY_POINT:
5836 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5837 optlen);
5838 break;
5839 case SCTP_MAX_BURST:
5840 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5841 break;
5842 case SCTP_AUTH_KEY:
5843 case SCTP_AUTH_CHUNK:
5844 case SCTP_AUTH_DELETE_KEY:
5845 retval = -EOPNOTSUPP;
5846 break;
5847 case SCTP_HMAC_IDENT:
5848 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5849 break;
5850 case SCTP_AUTH_ACTIVE_KEY:
5851 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5852 break;
5853 case SCTP_PEER_AUTH_CHUNKS:
5854 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5855 optlen);
5856 break;
5857 case SCTP_LOCAL_AUTH_CHUNKS:
5858 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5859 optlen);
5860 break;
5861 case SCTP_GET_ASSOC_NUMBER:
5862 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5863 break;
5864 case SCTP_GET_ASSOC_ID_LIST:
5865 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5866 break;
5867 case SCTP_AUTO_ASCONF:
5868 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5869 break;
5870 case SCTP_PEER_ADDR_THLDS:
5871 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
5872 break;
5873 case SCTP_GET_ASSOC_STATS:
5874 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
5875 break;
5876 default:
5877 retval = -ENOPROTOOPT;
5878 break;
5879 }
5880
5881 release_sock(sk);
5882 return retval;
5883 }
5884
5885 static void sctp_hash(struct sock *sk)
5886 {
5887 /* STUB */
5888 }
5889
5890 static void sctp_unhash(struct sock *sk)
5891 {
5892 /* STUB */
5893 }
5894
5895 /* Check if port is acceptable. Possibly find first available port.
5896 *
5897 * The port hash table (contained in the 'global' SCTP protocol storage
5898 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5899 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5900 * list (the list number is the port number hashed out, so as you
5901 * would expect from a hash function, all the ports in a given list have
5902 * such a number that hashes out to the same list number; you were
5903 * expecting that, right?); so each list has a set of ports, with a
5904 * link to the socket (struct sock) that uses it, the port number and
5905 * a fastreuse flag (FIXME: NPI ipg).
5906 */
5907 static struct sctp_bind_bucket *sctp_bucket_create(
5908 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
5909
5910 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5911 {
5912 struct sctp_bind_hashbucket *head; /* hash list */
5913 struct sctp_bind_bucket *pp;
5914 unsigned short snum;
5915 int ret;
5916
5917 snum = ntohs(addr->v4.sin_port);
5918
5919 pr_debug("%s: begins, snum:%d\n", __func__, snum);
5920
5921 local_bh_disable();
5922
5923 if (snum == 0) {
5924 /* Search for an available port. */
5925 int low, high, remaining, index;
5926 unsigned int rover;
5927
5928 inet_get_local_port_range(sock_net(sk), &low, &high);
5929 remaining = (high - low) + 1;
5930 rover = prandom_u32() % remaining + low;
5931
5932 do {
5933 rover++;
5934 if ((rover < low) || (rover > high))
5935 rover = low;
5936 if (inet_is_reserved_local_port(rover))
5937 continue;
5938 index = sctp_phashfn(sock_net(sk), rover);
5939 head = &sctp_port_hashtable[index];
5940 spin_lock(&head->lock);
5941 sctp_for_each_hentry(pp, &head->chain)
5942 if ((pp->port == rover) &&
5943 net_eq(sock_net(sk), pp->net))
5944 goto next;
5945 break;
5946 next:
5947 spin_unlock(&head->lock);
5948 } while (--remaining > 0);
5949
5950 /* Exhausted local port range during search? */
5951 ret = 1;
5952 if (remaining <= 0)
5953 goto fail;
5954
5955 /* OK, here is the one we will use. HEAD (the port
5956 * hash table list entry) is non-NULL and we hold it's
5957 * mutex.
5958 */
5959 snum = rover;
5960 } else {
5961 /* We are given an specific port number; we verify
5962 * that it is not being used. If it is used, we will
5963 * exahust the search in the hash list corresponding
5964 * to the port number (snum) - we detect that with the
5965 * port iterator, pp being NULL.
5966 */
5967 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
5968 spin_lock(&head->lock);
5969 sctp_for_each_hentry(pp, &head->chain) {
5970 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
5971 goto pp_found;
5972 }
5973 }
5974 pp = NULL;
5975 goto pp_not_found;
5976 pp_found:
5977 if (!hlist_empty(&pp->owner)) {
5978 /* We had a port hash table hit - there is an
5979 * available port (pp != NULL) and it is being
5980 * used by other socket (pp->owner not empty); that other
5981 * socket is going to be sk2.
5982 */
5983 int reuse = sk->sk_reuse;
5984 struct sock *sk2;
5985
5986 pr_debug("%s: found a possible match\n", __func__);
5987
5988 if (pp->fastreuse && sk->sk_reuse &&
5989 sk->sk_state != SCTP_SS_LISTENING)
5990 goto success;
5991
5992 /* Run through the list of sockets bound to the port
5993 * (pp->port) [via the pointers bind_next and
5994 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5995 * we get the endpoint they describe and run through
5996 * the endpoint's list of IP (v4 or v6) addresses,
5997 * comparing each of the addresses with the address of
5998 * the socket sk. If we find a match, then that means
5999 * that this port/socket (sk) combination are already
6000 * in an endpoint.
6001 */
6002 sk_for_each_bound(sk2, &pp->owner) {
6003 struct sctp_endpoint *ep2;
6004 ep2 = sctp_sk(sk2)->ep;
6005
6006 if (sk == sk2 ||
6007 (reuse && sk2->sk_reuse &&
6008 sk2->sk_state != SCTP_SS_LISTENING))
6009 continue;
6010
6011 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
6012 sctp_sk(sk2), sctp_sk(sk))) {
6013 ret = (long)sk2;
6014 goto fail_unlock;
6015 }
6016 }
6017
6018 pr_debug("%s: found a match\n", __func__);
6019 }
6020 pp_not_found:
6021 /* If there was a hash table miss, create a new port. */
6022 ret = 1;
6023 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
6024 goto fail_unlock;
6025
6026 /* In either case (hit or miss), make sure fastreuse is 1 only
6027 * if sk->sk_reuse is too (that is, if the caller requested
6028 * SO_REUSEADDR on this socket -sk-).
6029 */
6030 if (hlist_empty(&pp->owner)) {
6031 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
6032 pp->fastreuse = 1;
6033 else
6034 pp->fastreuse = 0;
6035 } else if (pp->fastreuse &&
6036 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6037 pp->fastreuse = 0;
6038
6039 /* We are set, so fill up all the data in the hash table
6040 * entry, tie the socket list information with the rest of the
6041 * sockets FIXME: Blurry, NPI (ipg).
6042 */
6043 success:
6044 if (!sctp_sk(sk)->bind_hash) {
6045 inet_sk(sk)->inet_num = snum;
6046 sk_add_bind_node(sk, &pp->owner);
6047 sctp_sk(sk)->bind_hash = pp;
6048 }
6049 ret = 0;
6050
6051 fail_unlock:
6052 spin_unlock(&head->lock);
6053
6054 fail:
6055 local_bh_enable();
6056 return ret;
6057 }
6058
6059 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6060 * port is requested.
6061 */
6062 static int sctp_get_port(struct sock *sk, unsigned short snum)
6063 {
6064 union sctp_addr addr;
6065 struct sctp_af *af = sctp_sk(sk)->pf->af;
6066
6067 /* Set up a dummy address struct from the sk. */
6068 af->from_sk(&addr, sk);
6069 addr.v4.sin_port = htons(snum);
6070
6071 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6072 return !!sctp_get_port_local(sk, &addr);
6073 }
6074
6075 /*
6076 * Move a socket to LISTENING state.
6077 */
6078 static int sctp_listen_start(struct sock *sk, int backlog)
6079 {
6080 struct sctp_sock *sp = sctp_sk(sk);
6081 struct sctp_endpoint *ep = sp->ep;
6082 struct crypto_hash *tfm = NULL;
6083 char alg[32];
6084
6085 /* Allocate HMAC for generating cookie. */
6086 if (!sp->hmac && sp->sctp_hmac_alg) {
6087 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6088 tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC);
6089 if (IS_ERR(tfm)) {
6090 net_info_ratelimited("failed to load transform for %s: %ld\n",
6091 sp->sctp_hmac_alg, PTR_ERR(tfm));
6092 return -ENOSYS;
6093 }
6094 sctp_sk(sk)->hmac = tfm;
6095 }
6096
6097 /*
6098 * If a bind() or sctp_bindx() is not called prior to a listen()
6099 * call that allows new associations to be accepted, the system
6100 * picks an ephemeral port and will choose an address set equivalent
6101 * to binding with a wildcard address.
6102 *
6103 * This is not currently spelled out in the SCTP sockets
6104 * extensions draft, but follows the practice as seen in TCP
6105 * sockets.
6106 *
6107 */
6108 sk->sk_state = SCTP_SS_LISTENING;
6109 if (!ep->base.bind_addr.port) {
6110 if (sctp_autobind(sk))
6111 return -EAGAIN;
6112 } else {
6113 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6114 sk->sk_state = SCTP_SS_CLOSED;
6115 return -EADDRINUSE;
6116 }
6117 }
6118
6119 sk->sk_max_ack_backlog = backlog;
6120 sctp_hash_endpoint(ep);
6121 return 0;
6122 }
6123
6124 /*
6125 * 4.1.3 / 5.1.3 listen()
6126 *
6127 * By default, new associations are not accepted for UDP style sockets.
6128 * An application uses listen() to mark a socket as being able to
6129 * accept new associations.
6130 *
6131 * On TCP style sockets, applications use listen() to ready the SCTP
6132 * endpoint for accepting inbound associations.
6133 *
6134 * On both types of endpoints a backlog of '0' disables listening.
6135 *
6136 * Move a socket to LISTENING state.
6137 */
6138 int sctp_inet_listen(struct socket *sock, int backlog)
6139 {
6140 struct sock *sk = sock->sk;
6141 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6142 int err = -EINVAL;
6143
6144 if (unlikely(backlog < 0))
6145 return err;
6146
6147 lock_sock(sk);
6148
6149 /* Peeled-off sockets are not allowed to listen(). */
6150 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6151 goto out;
6152
6153 if (sock->state != SS_UNCONNECTED)
6154 goto out;
6155
6156 /* If backlog is zero, disable listening. */
6157 if (!backlog) {
6158 if (sctp_sstate(sk, CLOSED))
6159 goto out;
6160
6161 err = 0;
6162 sctp_unhash_endpoint(ep);
6163 sk->sk_state = SCTP_SS_CLOSED;
6164 if (sk->sk_reuse)
6165 sctp_sk(sk)->bind_hash->fastreuse = 1;
6166 goto out;
6167 }
6168
6169 /* If we are already listening, just update the backlog */
6170 if (sctp_sstate(sk, LISTENING))
6171 sk->sk_max_ack_backlog = backlog;
6172 else {
6173 err = sctp_listen_start(sk, backlog);
6174 if (err)
6175 goto out;
6176 }
6177
6178 err = 0;
6179 out:
6180 release_sock(sk);
6181 return err;
6182 }
6183
6184 /*
6185 * This function is done by modeling the current datagram_poll() and the
6186 * tcp_poll(). Note that, based on these implementations, we don't
6187 * lock the socket in this function, even though it seems that,
6188 * ideally, locking or some other mechanisms can be used to ensure
6189 * the integrity of the counters (sndbuf and wmem_alloc) used
6190 * in this place. We assume that we don't need locks either until proven
6191 * otherwise.
6192 *
6193 * Another thing to note is that we include the Async I/O support
6194 * here, again, by modeling the current TCP/UDP code. We don't have
6195 * a good way to test with it yet.
6196 */
6197 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6198 {
6199 struct sock *sk = sock->sk;
6200 struct sctp_sock *sp = sctp_sk(sk);
6201 unsigned int mask;
6202
6203 poll_wait(file, sk_sleep(sk), wait);
6204
6205 /* A TCP-style listening socket becomes readable when the accept queue
6206 * is not empty.
6207 */
6208 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6209 return (!list_empty(&sp->ep->asocs)) ?
6210 (POLLIN | POLLRDNORM) : 0;
6211
6212 mask = 0;
6213
6214 /* Is there any exceptional events? */
6215 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6216 mask |= POLLERR |
6217 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
6218 if (sk->sk_shutdown & RCV_SHUTDOWN)
6219 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6220 if (sk->sk_shutdown == SHUTDOWN_MASK)
6221 mask |= POLLHUP;
6222
6223 /* Is it readable? Reconsider this code with TCP-style support. */
6224 if (!skb_queue_empty(&sk->sk_receive_queue))
6225 mask |= POLLIN | POLLRDNORM;
6226
6227 /* The association is either gone or not ready. */
6228 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6229 return mask;
6230
6231 /* Is it writable? */
6232 if (sctp_writeable(sk)) {
6233 mask |= POLLOUT | POLLWRNORM;
6234 } else {
6235 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6236 /*
6237 * Since the socket is not locked, the buffer
6238 * might be made available after the writeable check and
6239 * before the bit is set. This could cause a lost I/O
6240 * signal. tcp_poll() has a race breaker for this race
6241 * condition. Based on their implementation, we put
6242 * in the following code to cover it as well.
6243 */
6244 if (sctp_writeable(sk))
6245 mask |= POLLOUT | POLLWRNORM;
6246 }
6247 return mask;
6248 }
6249
6250 /********************************************************************
6251 * 2nd Level Abstractions
6252 ********************************************************************/
6253
6254 static struct sctp_bind_bucket *sctp_bucket_create(
6255 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6256 {
6257 struct sctp_bind_bucket *pp;
6258
6259 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6260 if (pp) {
6261 SCTP_DBG_OBJCNT_INC(bind_bucket);
6262 pp->port = snum;
6263 pp->fastreuse = 0;
6264 INIT_HLIST_HEAD(&pp->owner);
6265 pp->net = net;
6266 hlist_add_head(&pp->node, &head->chain);
6267 }
6268 return pp;
6269 }
6270
6271 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6272 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6273 {
6274 if (pp && hlist_empty(&pp->owner)) {
6275 __hlist_del(&pp->node);
6276 kmem_cache_free(sctp_bucket_cachep, pp);
6277 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6278 }
6279 }
6280
6281 /* Release this socket's reference to a local port. */
6282 static inline void __sctp_put_port(struct sock *sk)
6283 {
6284 struct sctp_bind_hashbucket *head =
6285 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6286 inet_sk(sk)->inet_num)];
6287 struct sctp_bind_bucket *pp;
6288
6289 spin_lock(&head->lock);
6290 pp = sctp_sk(sk)->bind_hash;
6291 __sk_del_bind_node(sk);
6292 sctp_sk(sk)->bind_hash = NULL;
6293 inet_sk(sk)->inet_num = 0;
6294 sctp_bucket_destroy(pp);
6295 spin_unlock(&head->lock);
6296 }
6297
6298 void sctp_put_port(struct sock *sk)
6299 {
6300 local_bh_disable();
6301 __sctp_put_port(sk);
6302 local_bh_enable();
6303 }
6304
6305 /*
6306 * The system picks an ephemeral port and choose an address set equivalent
6307 * to binding with a wildcard address.
6308 * One of those addresses will be the primary address for the association.
6309 * This automatically enables the multihoming capability of SCTP.
6310 */
6311 static int sctp_autobind(struct sock *sk)
6312 {
6313 union sctp_addr autoaddr;
6314 struct sctp_af *af;
6315 __be16 port;
6316
6317 /* Initialize a local sockaddr structure to INADDR_ANY. */
6318 af = sctp_sk(sk)->pf->af;
6319
6320 port = htons(inet_sk(sk)->inet_num);
6321 af->inaddr_any(&autoaddr, port);
6322
6323 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6324 }
6325
6326 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6327 *
6328 * From RFC 2292
6329 * 4.2 The cmsghdr Structure *
6330 *
6331 * When ancillary data is sent or received, any number of ancillary data
6332 * objects can be specified by the msg_control and msg_controllen members of
6333 * the msghdr structure, because each object is preceded by
6334 * a cmsghdr structure defining the object's length (the cmsg_len member).
6335 * Historically Berkeley-derived implementations have passed only one object
6336 * at a time, but this API allows multiple objects to be
6337 * passed in a single call to sendmsg() or recvmsg(). The following example
6338 * shows two ancillary data objects in a control buffer.
6339 *
6340 * |<--------------------------- msg_controllen -------------------------->|
6341 * | |
6342 *
6343 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6344 *
6345 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6346 * | | |
6347 *
6348 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6349 *
6350 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6351 * | | | | |
6352 *
6353 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6354 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6355 *
6356 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6357 *
6358 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6359 * ^
6360 * |
6361 *
6362 * msg_control
6363 * points here
6364 */
6365 static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
6366 {
6367 struct cmsghdr *cmsg;
6368 struct msghdr *my_msg = (struct msghdr *)msg;
6369
6370 for (cmsg = CMSG_FIRSTHDR(msg);
6371 cmsg != NULL;
6372 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6373 if (!CMSG_OK(my_msg, cmsg))
6374 return -EINVAL;
6375
6376 /* Should we parse this header or ignore? */
6377 if (cmsg->cmsg_level != IPPROTO_SCTP)
6378 continue;
6379
6380 /* Strictly check lengths following example in SCM code. */
6381 switch (cmsg->cmsg_type) {
6382 case SCTP_INIT:
6383 /* SCTP Socket API Extension
6384 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6385 *
6386 * This cmsghdr structure provides information for
6387 * initializing new SCTP associations with sendmsg().
6388 * The SCTP_INITMSG socket option uses this same data
6389 * structure. This structure is not used for
6390 * recvmsg().
6391 *
6392 * cmsg_level cmsg_type cmsg_data[]
6393 * ------------ ------------ ----------------------
6394 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6395 */
6396 if (cmsg->cmsg_len !=
6397 CMSG_LEN(sizeof(struct sctp_initmsg)))
6398 return -EINVAL;
6399 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6400 break;
6401
6402 case SCTP_SNDRCV:
6403 /* SCTP Socket API Extension
6404 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6405 *
6406 * This cmsghdr structure specifies SCTP options for
6407 * sendmsg() and describes SCTP header information
6408 * about a received message through recvmsg().
6409 *
6410 * cmsg_level cmsg_type cmsg_data[]
6411 * ------------ ------------ ----------------------
6412 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6413 */
6414 if (cmsg->cmsg_len !=
6415 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6416 return -EINVAL;
6417
6418 cmsgs->info =
6419 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6420
6421 /* Minimally, validate the sinfo_flags. */
6422 if (cmsgs->info->sinfo_flags &
6423 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6424 SCTP_ABORT | SCTP_EOF))
6425 return -EINVAL;
6426 break;
6427
6428 default:
6429 return -EINVAL;
6430 }
6431 }
6432 return 0;
6433 }
6434
6435 /*
6436 * Wait for a packet..
6437 * Note: This function is the same function as in core/datagram.c
6438 * with a few modifications to make lksctp work.
6439 */
6440 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
6441 {
6442 int error;
6443 DEFINE_WAIT(wait);
6444
6445 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6446
6447 /* Socket errors? */
6448 error = sock_error(sk);
6449 if (error)
6450 goto out;
6451
6452 if (!skb_queue_empty(&sk->sk_receive_queue))
6453 goto ready;
6454
6455 /* Socket shut down? */
6456 if (sk->sk_shutdown & RCV_SHUTDOWN)
6457 goto out;
6458
6459 /* Sequenced packets can come disconnected. If so we report the
6460 * problem.
6461 */
6462 error = -ENOTCONN;
6463
6464 /* Is there a good reason to think that we may receive some data? */
6465 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6466 goto out;
6467
6468 /* Handle signals. */
6469 if (signal_pending(current))
6470 goto interrupted;
6471
6472 /* Let another process have a go. Since we are going to sleep
6473 * anyway. Note: This may cause odd behaviors if the message
6474 * does not fit in the user's buffer, but this seems to be the
6475 * only way to honor MSG_DONTWAIT realistically.
6476 */
6477 release_sock(sk);
6478 *timeo_p = schedule_timeout(*timeo_p);
6479 lock_sock(sk);
6480
6481 ready:
6482 finish_wait(sk_sleep(sk), &wait);
6483 return 0;
6484
6485 interrupted:
6486 error = sock_intr_errno(*timeo_p);
6487
6488 out:
6489 finish_wait(sk_sleep(sk), &wait);
6490 *err = error;
6491 return error;
6492 }
6493
6494 /* Receive a datagram.
6495 * Note: This is pretty much the same routine as in core/datagram.c
6496 * with a few changes to make lksctp work.
6497 */
6498 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6499 int noblock, int *err)
6500 {
6501 int error;
6502 struct sk_buff *skb;
6503 long timeo;
6504
6505 timeo = sock_rcvtimeo(sk, noblock);
6506
6507 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
6508 MAX_SCHEDULE_TIMEOUT);
6509
6510 do {
6511 /* Again only user level code calls this function,
6512 * so nothing interrupt level
6513 * will suddenly eat the receive_queue.
6514 *
6515 * Look at current nfs client by the way...
6516 * However, this function was correct in any case. 8)
6517 */
6518 if (flags & MSG_PEEK) {
6519 spin_lock_bh(&sk->sk_receive_queue.lock);
6520 skb = skb_peek(&sk->sk_receive_queue);
6521 if (skb)
6522 atomic_inc(&skb->users);
6523 spin_unlock_bh(&sk->sk_receive_queue.lock);
6524 } else {
6525 skb = skb_dequeue(&sk->sk_receive_queue);
6526 }
6527
6528 if (skb)
6529 return skb;
6530
6531 /* Caller is allowed not to check sk->sk_err before calling. */
6532 error = sock_error(sk);
6533 if (error)
6534 goto no_packet;
6535
6536 if (sk->sk_shutdown & RCV_SHUTDOWN)
6537 break;
6538
6539 /* User doesn't want to wait. */
6540 error = -EAGAIN;
6541 if (!timeo)
6542 goto no_packet;
6543 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6544
6545 return NULL;
6546
6547 no_packet:
6548 *err = error;
6549 return NULL;
6550 }
6551
6552 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6553 static void __sctp_write_space(struct sctp_association *asoc)
6554 {
6555 struct sock *sk = asoc->base.sk;
6556 struct socket *sock = sk->sk_socket;
6557
6558 if ((sctp_wspace(asoc) > 0) && sock) {
6559 if (waitqueue_active(&asoc->wait))
6560 wake_up_interruptible(&asoc->wait);
6561
6562 if (sctp_writeable(sk)) {
6563 wait_queue_head_t *wq = sk_sleep(sk);
6564
6565 if (wq && waitqueue_active(wq))
6566 wake_up_interruptible(wq);
6567
6568 /* Note that we try to include the Async I/O support
6569 * here by modeling from the current TCP/UDP code.
6570 * We have not tested with it yet.
6571 */
6572 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6573 sock_wake_async(sock,
6574 SOCK_WAKE_SPACE, POLL_OUT);
6575 }
6576 }
6577 }
6578
6579 /* Do accounting for the sndbuf space.
6580 * Decrement the used sndbuf space of the corresponding association by the
6581 * data size which was just transmitted(freed).
6582 */
6583 static void sctp_wfree(struct sk_buff *skb)
6584 {
6585 struct sctp_association *asoc;
6586 struct sctp_chunk *chunk;
6587 struct sock *sk;
6588
6589 /* Get the saved chunk pointer. */
6590 chunk = *((struct sctp_chunk **)(skb->cb));
6591 asoc = chunk->asoc;
6592 sk = asoc->base.sk;
6593 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6594 sizeof(struct sk_buff) +
6595 sizeof(struct sctp_chunk);
6596
6597 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6598
6599 /*
6600 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6601 */
6602 sk->sk_wmem_queued -= skb->truesize;
6603 sk_mem_uncharge(sk, skb->truesize);
6604
6605 sock_wfree(skb);
6606 __sctp_write_space(asoc);
6607
6608 sctp_association_put(asoc);
6609 }
6610
6611 /* Do accounting for the receive space on the socket.
6612 * Accounting for the association is done in ulpevent.c
6613 * We set this as a destructor for the cloned data skbs so that
6614 * accounting is done at the correct time.
6615 */
6616 void sctp_sock_rfree(struct sk_buff *skb)
6617 {
6618 struct sock *sk = skb->sk;
6619 struct sctp_ulpevent *event = sctp_skb2event(skb);
6620
6621 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6622
6623 /*
6624 * Mimic the behavior of sock_rfree
6625 */
6626 sk_mem_uncharge(sk, event->rmem_len);
6627 }
6628
6629
6630 /* Helper function to wait for space in the sndbuf. */
6631 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6632 size_t msg_len)
6633 {
6634 struct sock *sk = asoc->base.sk;
6635 int err = 0;
6636 long current_timeo = *timeo_p;
6637 DEFINE_WAIT(wait);
6638
6639 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
6640 *timeo_p, msg_len);
6641
6642 /* Increment the association's refcnt. */
6643 sctp_association_hold(asoc);
6644
6645 /* Wait on the association specific sndbuf space. */
6646 for (;;) {
6647 prepare_to_wait_exclusive(&asoc->wait, &wait,
6648 TASK_INTERRUPTIBLE);
6649 if (!*timeo_p)
6650 goto do_nonblock;
6651 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6652 asoc->base.dead)
6653 goto do_error;
6654 if (signal_pending(current))
6655 goto do_interrupted;
6656 if (msg_len <= sctp_wspace(asoc))
6657 break;
6658
6659 /* Let another process have a go. Since we are going
6660 * to sleep anyway.
6661 */
6662 release_sock(sk);
6663 current_timeo = schedule_timeout(current_timeo);
6664 BUG_ON(sk != asoc->base.sk);
6665 lock_sock(sk);
6666
6667 *timeo_p = current_timeo;
6668 }
6669
6670 out:
6671 finish_wait(&asoc->wait, &wait);
6672
6673 /* Release the association's refcnt. */
6674 sctp_association_put(asoc);
6675
6676 return err;
6677
6678 do_error:
6679 err = -EPIPE;
6680 goto out;
6681
6682 do_interrupted:
6683 err = sock_intr_errno(*timeo_p);
6684 goto out;
6685
6686 do_nonblock:
6687 err = -EAGAIN;
6688 goto out;
6689 }
6690
6691 void sctp_data_ready(struct sock *sk, int len)
6692 {
6693 struct socket_wq *wq;
6694
6695 rcu_read_lock();
6696 wq = rcu_dereference(sk->sk_wq);
6697 if (wq_has_sleeper(wq))
6698 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6699 POLLRDNORM | POLLRDBAND);
6700 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6701 rcu_read_unlock();
6702 }
6703
6704 /* If socket sndbuf has changed, wake up all per association waiters. */
6705 void sctp_write_space(struct sock *sk)
6706 {
6707 struct sctp_association *asoc;
6708
6709 /* Wake up the tasks in each wait queue. */
6710 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6711 __sctp_write_space(asoc);
6712 }
6713 }
6714
6715 /* Is there any sndbuf space available on the socket?
6716 *
6717 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6718 * associations on the same socket. For a UDP-style socket with
6719 * multiple associations, it is possible for it to be "unwriteable"
6720 * prematurely. I assume that this is acceptable because
6721 * a premature "unwriteable" is better than an accidental "writeable" which
6722 * would cause an unwanted block under certain circumstances. For the 1-1
6723 * UDP-style sockets or TCP-style sockets, this code should work.
6724 * - Daisy
6725 */
6726 static int sctp_writeable(struct sock *sk)
6727 {
6728 int amt = 0;
6729
6730 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6731 if (amt < 0)
6732 amt = 0;
6733 return amt;
6734 }
6735
6736 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6737 * returns immediately with EINPROGRESS.
6738 */
6739 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6740 {
6741 struct sock *sk = asoc->base.sk;
6742 int err = 0;
6743 long current_timeo = *timeo_p;
6744 DEFINE_WAIT(wait);
6745
6746 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
6747
6748 /* Increment the association's refcnt. */
6749 sctp_association_hold(asoc);
6750
6751 for (;;) {
6752 prepare_to_wait_exclusive(&asoc->wait, &wait,
6753 TASK_INTERRUPTIBLE);
6754 if (!*timeo_p)
6755 goto do_nonblock;
6756 if (sk->sk_shutdown & RCV_SHUTDOWN)
6757 break;
6758 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6759 asoc->base.dead)
6760 goto do_error;
6761 if (signal_pending(current))
6762 goto do_interrupted;
6763
6764 if (sctp_state(asoc, ESTABLISHED))
6765 break;
6766
6767 /* Let another process have a go. Since we are going
6768 * to sleep anyway.
6769 */
6770 release_sock(sk);
6771 current_timeo = schedule_timeout(current_timeo);
6772 lock_sock(sk);
6773
6774 *timeo_p = current_timeo;
6775 }
6776
6777 out:
6778 finish_wait(&asoc->wait, &wait);
6779
6780 /* Release the association's refcnt. */
6781 sctp_association_put(asoc);
6782
6783 return err;
6784
6785 do_error:
6786 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6787 err = -ETIMEDOUT;
6788 else
6789 err = -ECONNREFUSED;
6790 goto out;
6791
6792 do_interrupted:
6793 err = sock_intr_errno(*timeo_p);
6794 goto out;
6795
6796 do_nonblock:
6797 err = -EINPROGRESS;
6798 goto out;
6799 }
6800
6801 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6802 {
6803 struct sctp_endpoint *ep;
6804 int err = 0;
6805 DEFINE_WAIT(wait);
6806
6807 ep = sctp_sk(sk)->ep;
6808
6809
6810 for (;;) {
6811 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6812 TASK_INTERRUPTIBLE);
6813
6814 if (list_empty(&ep->asocs)) {
6815 release_sock(sk);
6816 timeo = schedule_timeout(timeo);
6817 lock_sock(sk);
6818 }
6819
6820 err = -EINVAL;
6821 if (!sctp_sstate(sk, LISTENING))
6822 break;
6823
6824 err = 0;
6825 if (!list_empty(&ep->asocs))
6826 break;
6827
6828 err = sock_intr_errno(timeo);
6829 if (signal_pending(current))
6830 break;
6831
6832 err = -EAGAIN;
6833 if (!timeo)
6834 break;
6835 }
6836
6837 finish_wait(sk_sleep(sk), &wait);
6838
6839 return err;
6840 }
6841
6842 static void sctp_wait_for_close(struct sock *sk, long timeout)
6843 {
6844 DEFINE_WAIT(wait);
6845
6846 do {
6847 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6848 if (list_empty(&sctp_sk(sk)->ep->asocs))
6849 break;
6850 release_sock(sk);
6851 timeout = schedule_timeout(timeout);
6852 lock_sock(sk);
6853 } while (!signal_pending(current) && timeout);
6854
6855 finish_wait(sk_sleep(sk), &wait);
6856 }
6857
6858 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6859 {
6860 struct sk_buff *frag;
6861
6862 if (!skb->data_len)
6863 goto done;
6864
6865 /* Don't forget the fragments. */
6866 skb_walk_frags(skb, frag)
6867 sctp_skb_set_owner_r_frag(frag, sk);
6868
6869 done:
6870 sctp_skb_set_owner_r(skb, sk);
6871 }
6872
6873 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6874 struct sctp_association *asoc)
6875 {
6876 struct inet_sock *inet = inet_sk(sk);
6877 struct inet_sock *newinet;
6878
6879 newsk->sk_type = sk->sk_type;
6880 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6881 newsk->sk_flags = sk->sk_flags;
6882 newsk->sk_no_check = sk->sk_no_check;
6883 newsk->sk_reuse = sk->sk_reuse;
6884
6885 newsk->sk_shutdown = sk->sk_shutdown;
6886 newsk->sk_destruct = sctp_destruct_sock;
6887 newsk->sk_family = sk->sk_family;
6888 newsk->sk_protocol = IPPROTO_SCTP;
6889 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6890 newsk->sk_sndbuf = sk->sk_sndbuf;
6891 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6892 newsk->sk_lingertime = sk->sk_lingertime;
6893 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6894 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6895
6896 newinet = inet_sk(newsk);
6897
6898 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6899 * getsockname() and getpeername()
6900 */
6901 newinet->inet_sport = inet->inet_sport;
6902 newinet->inet_saddr = inet->inet_saddr;
6903 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6904 newinet->inet_dport = htons(asoc->peer.port);
6905 newinet->pmtudisc = inet->pmtudisc;
6906 newinet->inet_id = asoc->next_tsn ^ jiffies;
6907
6908 newinet->uc_ttl = inet->uc_ttl;
6909 newinet->mc_loop = 1;
6910 newinet->mc_ttl = 1;
6911 newinet->mc_index = 0;
6912 newinet->mc_list = NULL;
6913 }
6914
6915 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6916 * and its messages to the newsk.
6917 */
6918 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6919 struct sctp_association *assoc,
6920 sctp_socket_type_t type)
6921 {
6922 struct sctp_sock *oldsp = sctp_sk(oldsk);
6923 struct sctp_sock *newsp = sctp_sk(newsk);
6924 struct sctp_bind_bucket *pp; /* hash list port iterator */
6925 struct sctp_endpoint *newep = newsp->ep;
6926 struct sk_buff *skb, *tmp;
6927 struct sctp_ulpevent *event;
6928 struct sctp_bind_hashbucket *head;
6929 struct list_head tmplist;
6930
6931 /* Migrate socket buffer sizes and all the socket level options to the
6932 * new socket.
6933 */
6934 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6935 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6936 /* Brute force copy old sctp opt. */
6937 if (oldsp->do_auto_asconf) {
6938 memcpy(&tmplist, &newsp->auto_asconf_list, sizeof(tmplist));
6939 inet_sk_copy_descendant(newsk, oldsk);
6940 memcpy(&newsp->auto_asconf_list, &tmplist, sizeof(tmplist));
6941 } else
6942 inet_sk_copy_descendant(newsk, oldsk);
6943
6944 /* Restore the ep value that was overwritten with the above structure
6945 * copy.
6946 */
6947 newsp->ep = newep;
6948 newsp->hmac = NULL;
6949
6950 /* Hook this new socket in to the bind_hash list. */
6951 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
6952 inet_sk(oldsk)->inet_num)];
6953 local_bh_disable();
6954 spin_lock(&head->lock);
6955 pp = sctp_sk(oldsk)->bind_hash;
6956 sk_add_bind_node(newsk, &pp->owner);
6957 sctp_sk(newsk)->bind_hash = pp;
6958 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6959 spin_unlock(&head->lock);
6960 local_bh_enable();
6961
6962 /* Copy the bind_addr list from the original endpoint to the new
6963 * endpoint so that we can handle restarts properly
6964 */
6965 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6966 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6967
6968 /* Move any messages in the old socket's receive queue that are for the
6969 * peeled off association to the new socket's receive queue.
6970 */
6971 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6972 event = sctp_skb2event(skb);
6973 if (event->asoc == assoc) {
6974 __skb_unlink(skb, &oldsk->sk_receive_queue);
6975 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6976 sctp_skb_set_owner_r_frag(skb, newsk);
6977 }
6978 }
6979
6980 /* Clean up any messages pending delivery due to partial
6981 * delivery. Three cases:
6982 * 1) No partial deliver; no work.
6983 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6984 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6985 */
6986 skb_queue_head_init(&newsp->pd_lobby);
6987 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6988
6989 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6990 struct sk_buff_head *queue;
6991
6992 /* Decide which queue to move pd_lobby skbs to. */
6993 if (assoc->ulpq.pd_mode) {
6994 queue = &newsp->pd_lobby;
6995 } else
6996 queue = &newsk->sk_receive_queue;
6997
6998 /* Walk through the pd_lobby, looking for skbs that
6999 * need moved to the new socket.
7000 */
7001 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
7002 event = sctp_skb2event(skb);
7003 if (event->asoc == assoc) {
7004 __skb_unlink(skb, &oldsp->pd_lobby);
7005 __skb_queue_tail(queue, skb);
7006 sctp_skb_set_owner_r_frag(skb, newsk);
7007 }
7008 }
7009
7010 /* Clear up any skbs waiting for the partial
7011 * delivery to finish.
7012 */
7013 if (assoc->ulpq.pd_mode)
7014 sctp_clear_pd(oldsk, NULL);
7015
7016 }
7017
7018 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
7019 sctp_skb_set_owner_r_frag(skb, newsk);
7020
7021 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
7022 sctp_skb_set_owner_r_frag(skb, newsk);
7023
7024 /* Set the type of socket to indicate that it is peeled off from the
7025 * original UDP-style socket or created with the accept() call on a
7026 * TCP-style socket..
7027 */
7028 newsp->type = type;
7029
7030 /* Mark the new socket "in-use" by the user so that any packets
7031 * that may arrive on the association after we've moved it are
7032 * queued to the backlog. This prevents a potential race between
7033 * backlog processing on the old socket and new-packet processing
7034 * on the new socket.
7035 *
7036 * The caller has just allocated newsk so we can guarantee that other
7037 * paths won't try to lock it and then oldsk.
7038 */
7039 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7040 sctp_assoc_migrate(assoc, newsk);
7041
7042 /* If the association on the newsk is already closed before accept()
7043 * is called, set RCV_SHUTDOWN flag.
7044 */
7045 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
7046 newsk->sk_shutdown |= RCV_SHUTDOWN;
7047
7048 newsk->sk_state = SCTP_SS_ESTABLISHED;
7049 release_sock(newsk);
7050 }
7051
7052
7053 /* This proto struct describes the ULP interface for SCTP. */
7054 struct proto sctp_prot = {
7055 .name = "SCTP",
7056 .owner = THIS_MODULE,
7057 .close = sctp_close,
7058 .connect = sctp_connect,
7059 .disconnect = sctp_disconnect,
7060 .accept = sctp_accept,
7061 .ioctl = sctp_ioctl,
7062 .init = sctp_init_sock,
7063 .destroy = sctp_destroy_sock,
7064 .shutdown = sctp_shutdown,
7065 .setsockopt = sctp_setsockopt,
7066 .getsockopt = sctp_getsockopt,
7067 .sendmsg = sctp_sendmsg,
7068 .recvmsg = sctp_recvmsg,
7069 .bind = sctp_bind,
7070 .backlog_rcv = sctp_backlog_rcv,
7071 .hash = sctp_hash,
7072 .unhash = sctp_unhash,
7073 .get_port = sctp_get_port,
7074 .obj_size = sizeof(struct sctp_sock),
7075 .sysctl_mem = sysctl_sctp_mem,
7076 .sysctl_rmem = sysctl_sctp_rmem,
7077 .sysctl_wmem = sysctl_sctp_wmem,
7078 .memory_pressure = &sctp_memory_pressure,
7079 .enter_memory_pressure = sctp_enter_memory_pressure,
7080 .memory_allocated = &sctp_memory_allocated,
7081 .sockets_allocated = &sctp_sockets_allocated,
7082 };
7083
7084 #if IS_ENABLED(CONFIG_IPV6)
7085
7086 struct proto sctpv6_prot = {
7087 .name = "SCTPv6",
7088 .owner = THIS_MODULE,
7089 .close = sctp_close,
7090 .connect = sctp_connect,
7091 .disconnect = sctp_disconnect,
7092 .accept = sctp_accept,
7093 .ioctl = sctp_ioctl,
7094 .init = sctp_init_sock,
7095 .destroy = sctp_destroy_sock,
7096 .shutdown = sctp_shutdown,
7097 .setsockopt = sctp_setsockopt,
7098 .getsockopt = sctp_getsockopt,
7099 .sendmsg = sctp_sendmsg,
7100 .recvmsg = sctp_recvmsg,
7101 .bind = sctp_bind,
7102 .backlog_rcv = sctp_backlog_rcv,
7103 .hash = sctp_hash,
7104 .unhash = sctp_unhash,
7105 .get_port = sctp_get_port,
7106 .obj_size = sizeof(struct sctp6_sock),
7107 .sysctl_mem = sysctl_sctp_mem,
7108 .sysctl_rmem = sysctl_sctp_rmem,
7109 .sysctl_wmem = sysctl_sctp_wmem,
7110 .memory_pressure = &sctp_memory_pressure,
7111 .enter_memory_pressure = sctp_enter_memory_pressure,
7112 .memory_allocated = &sctp_memory_allocated,
7113 .sockets_allocated = &sctp_sockets_allocated,
7114 };
7115 #endif /* IS_ENABLED(CONFIG_IPV6) */
This page took 0.289748 seconds and 5 git commands to generate.