sctp: fix checkpatch errors with //commen
[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 sctp_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 sctp_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 sctp_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 sctp_release_sock(sk);
1509
1510 /* Supposedly, no process has access to the socket, but
1511 * the net layers still may.
1512 */
1513 sctp_local_bh_disable();
1514 sctp_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 sctp_bh_unlock_sock(sk);
1523 sctp_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 sctp_lock_sock(sk);
1669
1670 /* If a msg_name has been specified, assume this is to be used. */
1671 if (msg_name) {
1672 /* Look for a matching association on the endpoint. */
1673 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1674 if (!asoc) {
1675 /* If we could not find a matching association on the
1676 * endpoint, make sure that it is not a TCP-style
1677 * socket that already has an association or there is
1678 * no peeled-off association on another socket.
1679 */
1680 if ((sctp_style(sk, TCP) &&
1681 sctp_sstate(sk, ESTABLISHED)) ||
1682 sctp_endpoint_is_peeled_off(ep, &to)) {
1683 err = -EADDRNOTAVAIL;
1684 goto out_unlock;
1685 }
1686 }
1687 } else {
1688 asoc = sctp_id2assoc(sk, associd);
1689 if (!asoc) {
1690 err = -EPIPE;
1691 goto out_unlock;
1692 }
1693 }
1694
1695 if (asoc) {
1696 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->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 sctp_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 sctp_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 sctp_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 /*
2531 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2532 *
2533 * This option will effect the way delayed acks are performed. This
2534 * option allows you to get or set the delayed ack time, in
2535 * milliseconds. It also allows changing the delayed ack frequency.
2536 * Changing the frequency to 1 disables the delayed sack algorithm. If
2537 * the assoc_id is 0, then this sets or gets the endpoints default
2538 * values. If the assoc_id field is non-zero, then the set or get
2539 * effects the specified association for the one to many model (the
2540 * assoc_id field is ignored by the one to one model). Note that if
2541 * sack_delay or sack_freq are 0 when setting this option, then the
2542 * current values will remain unchanged.
2543 *
2544 * struct sctp_sack_info {
2545 * sctp_assoc_t sack_assoc_id;
2546 * uint32_t sack_delay;
2547 * uint32_t sack_freq;
2548 * };
2549 *
2550 * sack_assoc_id - This parameter, indicates which association the user
2551 * is performing an action upon. Note that if this field's value is
2552 * zero then the endpoints default value is changed (effecting future
2553 * associations only).
2554 *
2555 * sack_delay - This parameter contains the number of milliseconds that
2556 * the user is requesting the delayed ACK timer be set to. Note that
2557 * this value is defined in the standard to be between 200 and 500
2558 * milliseconds.
2559 *
2560 * sack_freq - This parameter contains the number of packets that must
2561 * be received before a sack is sent without waiting for the delay
2562 * timer to expire. The default value for this is 2, setting this
2563 * value to 1 will disable the delayed sack algorithm.
2564 */
2565
2566 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2567 char __user *optval, unsigned int optlen)
2568 {
2569 struct sctp_sack_info params;
2570 struct sctp_transport *trans = NULL;
2571 struct sctp_association *asoc = NULL;
2572 struct sctp_sock *sp = sctp_sk(sk);
2573
2574 if (optlen == sizeof(struct sctp_sack_info)) {
2575 if (copy_from_user(&params, optval, optlen))
2576 return -EFAULT;
2577
2578 if (params.sack_delay == 0 && params.sack_freq == 0)
2579 return 0;
2580 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2581 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2582 pr_warn("Use struct sctp_sack_info instead\n");
2583 if (copy_from_user(&params, optval, optlen))
2584 return -EFAULT;
2585
2586 if (params.sack_delay == 0)
2587 params.sack_freq = 1;
2588 else
2589 params.sack_freq = 0;
2590 } else
2591 return -EINVAL;
2592
2593 /* Validate value parameter. */
2594 if (params.sack_delay > 500)
2595 return -EINVAL;
2596
2597 /* Get association, if sack_assoc_id != 0 and the socket is a one
2598 * to many style socket, and an association was not found, then
2599 * the id was invalid.
2600 */
2601 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2602 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2603 return -EINVAL;
2604
2605 if (params.sack_delay) {
2606 if (asoc) {
2607 asoc->sackdelay =
2608 msecs_to_jiffies(params.sack_delay);
2609 asoc->param_flags =
2610 (asoc->param_flags & ~SPP_SACKDELAY) |
2611 SPP_SACKDELAY_ENABLE;
2612 } else {
2613 sp->sackdelay = params.sack_delay;
2614 sp->param_flags =
2615 (sp->param_flags & ~SPP_SACKDELAY) |
2616 SPP_SACKDELAY_ENABLE;
2617 }
2618 }
2619
2620 if (params.sack_freq == 1) {
2621 if (asoc) {
2622 asoc->param_flags =
2623 (asoc->param_flags & ~SPP_SACKDELAY) |
2624 SPP_SACKDELAY_DISABLE;
2625 } else {
2626 sp->param_flags =
2627 (sp->param_flags & ~SPP_SACKDELAY) |
2628 SPP_SACKDELAY_DISABLE;
2629 }
2630 } else if (params.sack_freq > 1) {
2631 if (asoc) {
2632 asoc->sackfreq = params.sack_freq;
2633 asoc->param_flags =
2634 (asoc->param_flags & ~SPP_SACKDELAY) |
2635 SPP_SACKDELAY_ENABLE;
2636 } else {
2637 sp->sackfreq = params.sack_freq;
2638 sp->param_flags =
2639 (sp->param_flags & ~SPP_SACKDELAY) |
2640 SPP_SACKDELAY_ENABLE;
2641 }
2642 }
2643
2644 /* If change is for association, also apply to each transport. */
2645 if (asoc) {
2646 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2647 transports) {
2648 if (params.sack_delay) {
2649 trans->sackdelay =
2650 msecs_to_jiffies(params.sack_delay);
2651 trans->param_flags =
2652 (trans->param_flags & ~SPP_SACKDELAY) |
2653 SPP_SACKDELAY_ENABLE;
2654 }
2655 if (params.sack_freq == 1) {
2656 trans->param_flags =
2657 (trans->param_flags & ~SPP_SACKDELAY) |
2658 SPP_SACKDELAY_DISABLE;
2659 } else if (params.sack_freq > 1) {
2660 trans->sackfreq = params.sack_freq;
2661 trans->param_flags =
2662 (trans->param_flags & ~SPP_SACKDELAY) |
2663 SPP_SACKDELAY_ENABLE;
2664 }
2665 }
2666 }
2667
2668 return 0;
2669 }
2670
2671 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2672 *
2673 * Applications can specify protocol parameters for the default association
2674 * initialization. The option name argument to setsockopt() and getsockopt()
2675 * is SCTP_INITMSG.
2676 *
2677 * Setting initialization parameters is effective only on an unconnected
2678 * socket (for UDP-style sockets only future associations are effected
2679 * by the change). With TCP-style sockets, this option is inherited by
2680 * sockets derived from a listener socket.
2681 */
2682 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2683 {
2684 struct sctp_initmsg sinit;
2685 struct sctp_sock *sp = sctp_sk(sk);
2686
2687 if (optlen != sizeof(struct sctp_initmsg))
2688 return -EINVAL;
2689 if (copy_from_user(&sinit, optval, optlen))
2690 return -EFAULT;
2691
2692 if (sinit.sinit_num_ostreams)
2693 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2694 if (sinit.sinit_max_instreams)
2695 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2696 if (sinit.sinit_max_attempts)
2697 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2698 if (sinit.sinit_max_init_timeo)
2699 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2700
2701 return 0;
2702 }
2703
2704 /*
2705 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2706 *
2707 * Applications that wish to use the sendto() system call may wish to
2708 * specify a default set of parameters that would normally be supplied
2709 * through the inclusion of ancillary data. This socket option allows
2710 * such an application to set the default sctp_sndrcvinfo structure.
2711 * The application that wishes to use this socket option simply passes
2712 * in to this call the sctp_sndrcvinfo structure defined in Section
2713 * 5.2.2) The input parameters accepted by this call include
2714 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2715 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2716 * to this call if the caller is using the UDP model.
2717 */
2718 static int sctp_setsockopt_default_send_param(struct sock *sk,
2719 char __user *optval,
2720 unsigned int optlen)
2721 {
2722 struct sctp_sndrcvinfo info;
2723 struct sctp_association *asoc;
2724 struct sctp_sock *sp = sctp_sk(sk);
2725
2726 if (optlen != sizeof(struct sctp_sndrcvinfo))
2727 return -EINVAL;
2728 if (copy_from_user(&info, optval, optlen))
2729 return -EFAULT;
2730
2731 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2732 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2733 return -EINVAL;
2734
2735 if (asoc) {
2736 asoc->default_stream = info.sinfo_stream;
2737 asoc->default_flags = info.sinfo_flags;
2738 asoc->default_ppid = info.sinfo_ppid;
2739 asoc->default_context = info.sinfo_context;
2740 asoc->default_timetolive = info.sinfo_timetolive;
2741 } else {
2742 sp->default_stream = info.sinfo_stream;
2743 sp->default_flags = info.sinfo_flags;
2744 sp->default_ppid = info.sinfo_ppid;
2745 sp->default_context = info.sinfo_context;
2746 sp->default_timetolive = info.sinfo_timetolive;
2747 }
2748
2749 return 0;
2750 }
2751
2752 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2753 *
2754 * Requests that the local SCTP stack use the enclosed peer address as
2755 * the association primary. The enclosed address must be one of the
2756 * association peer's addresses.
2757 */
2758 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2759 unsigned int optlen)
2760 {
2761 struct sctp_prim prim;
2762 struct sctp_transport *trans;
2763
2764 if (optlen != sizeof(struct sctp_prim))
2765 return -EINVAL;
2766
2767 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2768 return -EFAULT;
2769
2770 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2771 if (!trans)
2772 return -EINVAL;
2773
2774 sctp_assoc_set_primary(trans->asoc, trans);
2775
2776 return 0;
2777 }
2778
2779 /*
2780 * 7.1.5 SCTP_NODELAY
2781 *
2782 * Turn on/off any Nagle-like algorithm. This means that packets are
2783 * generally sent as soon as possible and no unnecessary delays are
2784 * introduced, at the cost of more packets in the network. Expects an
2785 * integer boolean flag.
2786 */
2787 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2788 unsigned int optlen)
2789 {
2790 int val;
2791
2792 if (optlen < sizeof(int))
2793 return -EINVAL;
2794 if (get_user(val, (int __user *)optval))
2795 return -EFAULT;
2796
2797 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2798 return 0;
2799 }
2800
2801 /*
2802 *
2803 * 7.1.1 SCTP_RTOINFO
2804 *
2805 * The protocol parameters used to initialize and bound retransmission
2806 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2807 * and modify these parameters.
2808 * All parameters are time values, in milliseconds. A value of 0, when
2809 * modifying the parameters, indicates that the current value should not
2810 * be changed.
2811 *
2812 */
2813 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2814 {
2815 struct sctp_rtoinfo rtoinfo;
2816 struct sctp_association *asoc;
2817 unsigned long rto_min, rto_max;
2818 struct sctp_sock *sp = sctp_sk(sk);
2819
2820 if (optlen != sizeof (struct sctp_rtoinfo))
2821 return -EINVAL;
2822
2823 if (copy_from_user(&rtoinfo, optval, optlen))
2824 return -EFAULT;
2825
2826 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2827
2828 /* Set the values to the specific association */
2829 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2830 return -EINVAL;
2831
2832 rto_max = rtoinfo.srto_max;
2833 rto_min = rtoinfo.srto_min;
2834
2835 if (rto_max)
2836 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
2837 else
2838 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
2839
2840 if (rto_min)
2841 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
2842 else
2843 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
2844
2845 if (rto_min > rto_max)
2846 return -EINVAL;
2847
2848 if (asoc) {
2849 if (rtoinfo.srto_initial != 0)
2850 asoc->rto_initial =
2851 msecs_to_jiffies(rtoinfo.srto_initial);
2852 asoc->rto_max = rto_max;
2853 asoc->rto_min = rto_min;
2854 } else {
2855 /* If there is no association or the association-id = 0
2856 * set the values to the endpoint.
2857 */
2858 if (rtoinfo.srto_initial != 0)
2859 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2860 sp->rtoinfo.srto_max = rto_max;
2861 sp->rtoinfo.srto_min = rto_min;
2862 }
2863
2864 return 0;
2865 }
2866
2867 /*
2868 *
2869 * 7.1.2 SCTP_ASSOCINFO
2870 *
2871 * This option is used to tune the maximum retransmission attempts
2872 * of the association.
2873 * Returns an error if the new association retransmission value is
2874 * greater than the sum of the retransmission value of the peer.
2875 * See [SCTP] for more information.
2876 *
2877 */
2878 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2879 {
2880
2881 struct sctp_assocparams assocparams;
2882 struct sctp_association *asoc;
2883
2884 if (optlen != sizeof(struct sctp_assocparams))
2885 return -EINVAL;
2886 if (copy_from_user(&assocparams, optval, optlen))
2887 return -EFAULT;
2888
2889 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2890
2891 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2892 return -EINVAL;
2893
2894 /* Set the values to the specific association */
2895 if (asoc) {
2896 if (assocparams.sasoc_asocmaxrxt != 0) {
2897 __u32 path_sum = 0;
2898 int paths = 0;
2899 struct sctp_transport *peer_addr;
2900
2901 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2902 transports) {
2903 path_sum += peer_addr->pathmaxrxt;
2904 paths++;
2905 }
2906
2907 /* Only validate asocmaxrxt if we have more than
2908 * one path/transport. We do this because path
2909 * retransmissions are only counted when we have more
2910 * then one path.
2911 */
2912 if (paths > 1 &&
2913 assocparams.sasoc_asocmaxrxt > path_sum)
2914 return -EINVAL;
2915
2916 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2917 }
2918
2919 if (assocparams.sasoc_cookie_life != 0)
2920 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
2921 } else {
2922 /* Set the values to the endpoint */
2923 struct sctp_sock *sp = sctp_sk(sk);
2924
2925 if (assocparams.sasoc_asocmaxrxt != 0)
2926 sp->assocparams.sasoc_asocmaxrxt =
2927 assocparams.sasoc_asocmaxrxt;
2928 if (assocparams.sasoc_cookie_life != 0)
2929 sp->assocparams.sasoc_cookie_life =
2930 assocparams.sasoc_cookie_life;
2931 }
2932 return 0;
2933 }
2934
2935 /*
2936 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2937 *
2938 * This socket option is a boolean flag which turns on or off mapped V4
2939 * addresses. If this option is turned on and the socket is type
2940 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2941 * If this option is turned off, then no mapping will be done of V4
2942 * addresses and a user will receive both PF_INET6 and PF_INET type
2943 * addresses on the socket.
2944 */
2945 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2946 {
2947 int val;
2948 struct sctp_sock *sp = sctp_sk(sk);
2949
2950 if (optlen < sizeof(int))
2951 return -EINVAL;
2952 if (get_user(val, (int __user *)optval))
2953 return -EFAULT;
2954 if (val)
2955 sp->v4mapped = 1;
2956 else
2957 sp->v4mapped = 0;
2958
2959 return 0;
2960 }
2961
2962 /*
2963 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2964 * This option will get or set the maximum size to put in any outgoing
2965 * SCTP DATA chunk. If a message is larger than this size it will be
2966 * fragmented by SCTP into the specified size. Note that the underlying
2967 * SCTP implementation may fragment into smaller sized chunks when the
2968 * PMTU of the underlying association is smaller than the value set by
2969 * the user. The default value for this option is '0' which indicates
2970 * the user is NOT limiting fragmentation and only the PMTU will effect
2971 * SCTP's choice of DATA chunk size. Note also that values set larger
2972 * than the maximum size of an IP datagram will effectively let SCTP
2973 * control fragmentation (i.e. the same as setting this option to 0).
2974 *
2975 * The following structure is used to access and modify this parameter:
2976 *
2977 * struct sctp_assoc_value {
2978 * sctp_assoc_t assoc_id;
2979 * uint32_t assoc_value;
2980 * };
2981 *
2982 * assoc_id: This parameter is ignored for one-to-one style sockets.
2983 * For one-to-many style sockets this parameter indicates which
2984 * association the user is performing an action upon. Note that if
2985 * this field's value is zero then the endpoints default value is
2986 * changed (effecting future associations only).
2987 * assoc_value: This parameter specifies the maximum size in bytes.
2988 */
2989 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2990 {
2991 struct sctp_assoc_value params;
2992 struct sctp_association *asoc;
2993 struct sctp_sock *sp = sctp_sk(sk);
2994 int val;
2995
2996 if (optlen == sizeof(int)) {
2997 pr_warn("Use of int in maxseg socket option deprecated\n");
2998 pr_warn("Use struct sctp_assoc_value instead\n");
2999 if (copy_from_user(&val, optval, optlen))
3000 return -EFAULT;
3001 params.assoc_id = 0;
3002 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3003 if (copy_from_user(&params, optval, optlen))
3004 return -EFAULT;
3005 val = params.assoc_value;
3006 } else
3007 return -EINVAL;
3008
3009 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3010 return -EINVAL;
3011
3012 asoc = sctp_id2assoc(sk, params.assoc_id);
3013 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3014 return -EINVAL;
3015
3016 if (asoc) {
3017 if (val == 0) {
3018 val = asoc->pathmtu;
3019 val -= sp->pf->af->net_header_len;
3020 val -= sizeof(struct sctphdr) +
3021 sizeof(struct sctp_data_chunk);
3022 }
3023 asoc->user_frag = val;
3024 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3025 } else {
3026 sp->user_frag = val;
3027 }
3028
3029 return 0;
3030 }
3031
3032
3033 /*
3034 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3035 *
3036 * Requests that the peer mark the enclosed address as the association
3037 * primary. The enclosed address must be one of the association's
3038 * locally bound addresses. The following structure is used to make a
3039 * set primary request:
3040 */
3041 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3042 unsigned int optlen)
3043 {
3044 struct net *net = sock_net(sk);
3045 struct sctp_sock *sp;
3046 struct sctp_association *asoc = NULL;
3047 struct sctp_setpeerprim prim;
3048 struct sctp_chunk *chunk;
3049 struct sctp_af *af;
3050 int err;
3051
3052 sp = sctp_sk(sk);
3053
3054 if (!net->sctp.addip_enable)
3055 return -EPERM;
3056
3057 if (optlen != sizeof(struct sctp_setpeerprim))
3058 return -EINVAL;
3059
3060 if (copy_from_user(&prim, optval, optlen))
3061 return -EFAULT;
3062
3063 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3064 if (!asoc)
3065 return -EINVAL;
3066
3067 if (!asoc->peer.asconf_capable)
3068 return -EPERM;
3069
3070 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3071 return -EPERM;
3072
3073 if (!sctp_state(asoc, ESTABLISHED))
3074 return -ENOTCONN;
3075
3076 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3077 if (!af)
3078 return -EINVAL;
3079
3080 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3081 return -EADDRNOTAVAIL;
3082
3083 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3084 return -EADDRNOTAVAIL;
3085
3086 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3087 chunk = sctp_make_asconf_set_prim(asoc,
3088 (union sctp_addr *)&prim.sspp_addr);
3089 if (!chunk)
3090 return -ENOMEM;
3091
3092 err = sctp_send_asconf(asoc, chunk);
3093
3094 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3095
3096 return err;
3097 }
3098
3099 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3100 unsigned int optlen)
3101 {
3102 struct sctp_setadaptation adaptation;
3103
3104 if (optlen != sizeof(struct sctp_setadaptation))
3105 return -EINVAL;
3106 if (copy_from_user(&adaptation, optval, optlen))
3107 return -EFAULT;
3108
3109 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3110
3111 return 0;
3112 }
3113
3114 /*
3115 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3116 *
3117 * The context field in the sctp_sndrcvinfo structure is normally only
3118 * used when a failed message is retrieved holding the value that was
3119 * sent down on the actual send call. This option allows the setting of
3120 * a default context on an association basis that will be received on
3121 * reading messages from the peer. This is especially helpful in the
3122 * one-2-many model for an application to keep some reference to an
3123 * internal state machine that is processing messages on the
3124 * association. Note that the setting of this value only effects
3125 * received messages from the peer and does not effect the value that is
3126 * saved with outbound messages.
3127 */
3128 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3129 unsigned int optlen)
3130 {
3131 struct sctp_assoc_value params;
3132 struct sctp_sock *sp;
3133 struct sctp_association *asoc;
3134
3135 if (optlen != sizeof(struct sctp_assoc_value))
3136 return -EINVAL;
3137 if (copy_from_user(&params, optval, optlen))
3138 return -EFAULT;
3139
3140 sp = sctp_sk(sk);
3141
3142 if (params.assoc_id != 0) {
3143 asoc = sctp_id2assoc(sk, params.assoc_id);
3144 if (!asoc)
3145 return -EINVAL;
3146 asoc->default_rcv_context = params.assoc_value;
3147 } else {
3148 sp->default_rcv_context = params.assoc_value;
3149 }
3150
3151 return 0;
3152 }
3153
3154 /*
3155 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3156 *
3157 * This options will at a minimum specify if the implementation is doing
3158 * fragmented interleave. Fragmented interleave, for a one to many
3159 * socket, is when subsequent calls to receive a message may return
3160 * parts of messages from different associations. Some implementations
3161 * may allow you to turn this value on or off. If so, when turned off,
3162 * no fragment interleave will occur (which will cause a head of line
3163 * blocking amongst multiple associations sharing the same one to many
3164 * socket). When this option is turned on, then each receive call may
3165 * come from a different association (thus the user must receive data
3166 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3167 * association each receive belongs to.
3168 *
3169 * This option takes a boolean value. A non-zero value indicates that
3170 * fragmented interleave is on. A value of zero indicates that
3171 * fragmented interleave is off.
3172 *
3173 * Note that it is important that an implementation that allows this
3174 * option to be turned on, have it off by default. Otherwise an unaware
3175 * application using the one to many model may become confused and act
3176 * incorrectly.
3177 */
3178 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3179 char __user *optval,
3180 unsigned int optlen)
3181 {
3182 int val;
3183
3184 if (optlen != sizeof(int))
3185 return -EINVAL;
3186 if (get_user(val, (int __user *)optval))
3187 return -EFAULT;
3188
3189 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3190
3191 return 0;
3192 }
3193
3194 /*
3195 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3196 * (SCTP_PARTIAL_DELIVERY_POINT)
3197 *
3198 * This option will set or get the SCTP partial delivery point. This
3199 * point is the size of a message where the partial delivery API will be
3200 * invoked to help free up rwnd space for the peer. Setting this to a
3201 * lower value will cause partial deliveries to happen more often. The
3202 * calls argument is an integer that sets or gets the partial delivery
3203 * point. Note also that the call will fail if the user attempts to set
3204 * this value larger than the socket receive buffer size.
3205 *
3206 * Note that any single message having a length smaller than or equal to
3207 * the SCTP partial delivery point will be delivered in one single read
3208 * call as long as the user provided buffer is large enough to hold the
3209 * message.
3210 */
3211 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3212 char __user *optval,
3213 unsigned int optlen)
3214 {
3215 u32 val;
3216
3217 if (optlen != sizeof(u32))
3218 return -EINVAL;
3219 if (get_user(val, (int __user *)optval))
3220 return -EFAULT;
3221
3222 /* Note: We double the receive buffer from what the user sets
3223 * it to be, also initial rwnd is based on rcvbuf/2.
3224 */
3225 if (val > (sk->sk_rcvbuf >> 1))
3226 return -EINVAL;
3227
3228 sctp_sk(sk)->pd_point = val;
3229
3230 return 0; /* is this the right error code? */
3231 }
3232
3233 /*
3234 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3235 *
3236 * This option will allow a user to change the maximum burst of packets
3237 * that can be emitted by this association. Note that the default value
3238 * is 4, and some implementations may restrict this setting so that it
3239 * can only be lowered.
3240 *
3241 * NOTE: This text doesn't seem right. Do this on a socket basis with
3242 * future associations inheriting the socket value.
3243 */
3244 static int sctp_setsockopt_maxburst(struct sock *sk,
3245 char __user *optval,
3246 unsigned int optlen)
3247 {
3248 struct sctp_assoc_value params;
3249 struct sctp_sock *sp;
3250 struct sctp_association *asoc;
3251 int val;
3252 int assoc_id = 0;
3253
3254 if (optlen == sizeof(int)) {
3255 pr_warn("Use of int in max_burst socket option deprecated\n");
3256 pr_warn("Use struct sctp_assoc_value instead\n");
3257 if (copy_from_user(&val, optval, optlen))
3258 return -EFAULT;
3259 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3260 if (copy_from_user(&params, optval, optlen))
3261 return -EFAULT;
3262 val = params.assoc_value;
3263 assoc_id = params.assoc_id;
3264 } else
3265 return -EINVAL;
3266
3267 sp = sctp_sk(sk);
3268
3269 if (assoc_id != 0) {
3270 asoc = sctp_id2assoc(sk, assoc_id);
3271 if (!asoc)
3272 return -EINVAL;
3273 asoc->max_burst = val;
3274 } else
3275 sp->max_burst = val;
3276
3277 return 0;
3278 }
3279
3280 /*
3281 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3282 *
3283 * This set option adds a chunk type that the user is requesting to be
3284 * received only in an authenticated way. Changes to the list of chunks
3285 * will only effect future associations on the socket.
3286 */
3287 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3288 char __user *optval,
3289 unsigned int optlen)
3290 {
3291 struct net *net = sock_net(sk);
3292 struct sctp_authchunk val;
3293
3294 if (!net->sctp.auth_enable)
3295 return -EACCES;
3296
3297 if (optlen != sizeof(struct sctp_authchunk))
3298 return -EINVAL;
3299 if (copy_from_user(&val, optval, optlen))
3300 return -EFAULT;
3301
3302 switch (val.sauth_chunk) {
3303 case SCTP_CID_INIT:
3304 case SCTP_CID_INIT_ACK:
3305 case SCTP_CID_SHUTDOWN_COMPLETE:
3306 case SCTP_CID_AUTH:
3307 return -EINVAL;
3308 }
3309
3310 /* add this chunk id to the endpoint */
3311 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3312 }
3313
3314 /*
3315 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3316 *
3317 * This option gets or sets the list of HMAC algorithms that the local
3318 * endpoint requires the peer to use.
3319 */
3320 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3321 char __user *optval,
3322 unsigned int optlen)
3323 {
3324 struct net *net = sock_net(sk);
3325 struct sctp_hmacalgo *hmacs;
3326 u32 idents;
3327 int err;
3328
3329 if (!net->sctp.auth_enable)
3330 return -EACCES;
3331
3332 if (optlen < sizeof(struct sctp_hmacalgo))
3333 return -EINVAL;
3334
3335 hmacs = memdup_user(optval, optlen);
3336 if (IS_ERR(hmacs))
3337 return PTR_ERR(hmacs);
3338
3339 idents = hmacs->shmac_num_idents;
3340 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3341 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3342 err = -EINVAL;
3343 goto out;
3344 }
3345
3346 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3347 out:
3348 kfree(hmacs);
3349 return err;
3350 }
3351
3352 /*
3353 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3354 *
3355 * This option will set a shared secret key which is used to build an
3356 * association shared key.
3357 */
3358 static int sctp_setsockopt_auth_key(struct sock *sk,
3359 char __user *optval,
3360 unsigned int optlen)
3361 {
3362 struct net *net = sock_net(sk);
3363 struct sctp_authkey *authkey;
3364 struct sctp_association *asoc;
3365 int ret;
3366
3367 if (!net->sctp.auth_enable)
3368 return -EACCES;
3369
3370 if (optlen <= sizeof(struct sctp_authkey))
3371 return -EINVAL;
3372
3373 authkey = memdup_user(optval, optlen);
3374 if (IS_ERR(authkey))
3375 return PTR_ERR(authkey);
3376
3377 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3378 ret = -EINVAL;
3379 goto out;
3380 }
3381
3382 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3383 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3384 ret = -EINVAL;
3385 goto out;
3386 }
3387
3388 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3389 out:
3390 kzfree(authkey);
3391 return ret;
3392 }
3393
3394 /*
3395 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3396 *
3397 * This option will get or set the active shared key to be used to build
3398 * the association shared key.
3399 */
3400 static int sctp_setsockopt_active_key(struct sock *sk,
3401 char __user *optval,
3402 unsigned int optlen)
3403 {
3404 struct net *net = sock_net(sk);
3405 struct sctp_authkeyid val;
3406 struct sctp_association *asoc;
3407
3408 if (!net->sctp.auth_enable)
3409 return -EACCES;
3410
3411 if (optlen != sizeof(struct sctp_authkeyid))
3412 return -EINVAL;
3413 if (copy_from_user(&val, optval, optlen))
3414 return -EFAULT;
3415
3416 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3417 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3418 return -EINVAL;
3419
3420 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3421 val.scact_keynumber);
3422 }
3423
3424 /*
3425 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3426 *
3427 * This set option will delete a shared secret key from use.
3428 */
3429 static int sctp_setsockopt_del_key(struct sock *sk,
3430 char __user *optval,
3431 unsigned int optlen)
3432 {
3433 struct net *net = sock_net(sk);
3434 struct sctp_authkeyid val;
3435 struct sctp_association *asoc;
3436
3437 if (!net->sctp.auth_enable)
3438 return -EACCES;
3439
3440 if (optlen != sizeof(struct sctp_authkeyid))
3441 return -EINVAL;
3442 if (copy_from_user(&val, optval, optlen))
3443 return -EFAULT;
3444
3445 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3446 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3447 return -EINVAL;
3448
3449 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3450 val.scact_keynumber);
3451
3452 }
3453
3454 /*
3455 * 8.1.23 SCTP_AUTO_ASCONF
3456 *
3457 * This option will enable or disable the use of the automatic generation of
3458 * ASCONF chunks to add and delete addresses to an existing association. Note
3459 * that this option has two caveats namely: a) it only affects sockets that
3460 * are bound to all addresses available to the SCTP stack, and b) the system
3461 * administrator may have an overriding control that turns the ASCONF feature
3462 * off no matter what setting the socket option may have.
3463 * This option expects an integer boolean flag, where a non-zero value turns on
3464 * the option, and a zero value turns off the option.
3465 * Note. In this implementation, socket operation overrides default parameter
3466 * being set by sysctl as well as FreeBSD implementation
3467 */
3468 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3469 unsigned int optlen)
3470 {
3471 int val;
3472 struct sctp_sock *sp = sctp_sk(sk);
3473
3474 if (optlen < sizeof(int))
3475 return -EINVAL;
3476 if (get_user(val, (int __user *)optval))
3477 return -EFAULT;
3478 if (!sctp_is_ep_boundall(sk) && val)
3479 return -EINVAL;
3480 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3481 return 0;
3482
3483 if (val == 0 && sp->do_auto_asconf) {
3484 list_del(&sp->auto_asconf_list);
3485 sp->do_auto_asconf = 0;
3486 } else if (val && !sp->do_auto_asconf) {
3487 list_add_tail(&sp->auto_asconf_list,
3488 &sock_net(sk)->sctp.auto_asconf_splist);
3489 sp->do_auto_asconf = 1;
3490 }
3491 return 0;
3492 }
3493
3494
3495 /*
3496 * SCTP_PEER_ADDR_THLDS
3497 *
3498 * This option allows us to alter the partially failed threshold for one or all
3499 * transports in an association. See Section 6.1 of:
3500 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3501 */
3502 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3503 char __user *optval,
3504 unsigned int optlen)
3505 {
3506 struct sctp_paddrthlds val;
3507 struct sctp_transport *trans;
3508 struct sctp_association *asoc;
3509
3510 if (optlen < sizeof(struct sctp_paddrthlds))
3511 return -EINVAL;
3512 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3513 sizeof(struct sctp_paddrthlds)))
3514 return -EFAULT;
3515
3516
3517 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3518 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3519 if (!asoc)
3520 return -ENOENT;
3521 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3522 transports) {
3523 if (val.spt_pathmaxrxt)
3524 trans->pathmaxrxt = val.spt_pathmaxrxt;
3525 trans->pf_retrans = val.spt_pathpfthld;
3526 }
3527
3528 if (val.spt_pathmaxrxt)
3529 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3530 asoc->pf_retrans = val.spt_pathpfthld;
3531 } else {
3532 trans = sctp_addr_id2transport(sk, &val.spt_address,
3533 val.spt_assoc_id);
3534 if (!trans)
3535 return -ENOENT;
3536
3537 if (val.spt_pathmaxrxt)
3538 trans->pathmaxrxt = val.spt_pathmaxrxt;
3539 trans->pf_retrans = val.spt_pathpfthld;
3540 }
3541
3542 return 0;
3543 }
3544
3545 /* API 6.2 setsockopt(), getsockopt()
3546 *
3547 * Applications use setsockopt() and getsockopt() to set or retrieve
3548 * socket options. Socket options are used to change the default
3549 * behavior of sockets calls. They are described in Section 7.
3550 *
3551 * The syntax is:
3552 *
3553 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3554 * int __user *optlen);
3555 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3556 * int optlen);
3557 *
3558 * sd - the socket descript.
3559 * level - set to IPPROTO_SCTP for all SCTP options.
3560 * optname - the option name.
3561 * optval - the buffer to store the value of the option.
3562 * optlen - the size of the buffer.
3563 */
3564 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3565 char __user *optval, unsigned int optlen)
3566 {
3567 int retval = 0;
3568
3569 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3570
3571 /* I can hardly begin to describe how wrong this is. This is
3572 * so broken as to be worse than useless. The API draft
3573 * REALLY is NOT helpful here... I am not convinced that the
3574 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3575 * are at all well-founded.
3576 */
3577 if (level != SOL_SCTP) {
3578 struct sctp_af *af = sctp_sk(sk)->pf->af;
3579 retval = af->setsockopt(sk, level, optname, optval, optlen);
3580 goto out_nounlock;
3581 }
3582
3583 sctp_lock_sock(sk);
3584
3585 switch (optname) {
3586 case SCTP_SOCKOPT_BINDX_ADD:
3587 /* 'optlen' is the size of the addresses buffer. */
3588 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3589 optlen, SCTP_BINDX_ADD_ADDR);
3590 break;
3591
3592 case SCTP_SOCKOPT_BINDX_REM:
3593 /* 'optlen' is the size of the addresses buffer. */
3594 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3595 optlen, SCTP_BINDX_REM_ADDR);
3596 break;
3597
3598 case SCTP_SOCKOPT_CONNECTX_OLD:
3599 /* 'optlen' is the size of the addresses buffer. */
3600 retval = sctp_setsockopt_connectx_old(sk,
3601 (struct sockaddr __user *)optval,
3602 optlen);
3603 break;
3604
3605 case SCTP_SOCKOPT_CONNECTX:
3606 /* 'optlen' is the size of the addresses buffer. */
3607 retval = sctp_setsockopt_connectx(sk,
3608 (struct sockaddr __user *)optval,
3609 optlen);
3610 break;
3611
3612 case SCTP_DISABLE_FRAGMENTS:
3613 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3614 break;
3615
3616 case SCTP_EVENTS:
3617 retval = sctp_setsockopt_events(sk, optval, optlen);
3618 break;
3619
3620 case SCTP_AUTOCLOSE:
3621 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3622 break;
3623
3624 case SCTP_PEER_ADDR_PARAMS:
3625 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3626 break;
3627
3628 case SCTP_DELAYED_SACK:
3629 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3630 break;
3631 case SCTP_PARTIAL_DELIVERY_POINT:
3632 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3633 break;
3634
3635 case SCTP_INITMSG:
3636 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3637 break;
3638 case SCTP_DEFAULT_SEND_PARAM:
3639 retval = sctp_setsockopt_default_send_param(sk, optval,
3640 optlen);
3641 break;
3642 case SCTP_PRIMARY_ADDR:
3643 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3644 break;
3645 case SCTP_SET_PEER_PRIMARY_ADDR:
3646 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3647 break;
3648 case SCTP_NODELAY:
3649 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3650 break;
3651 case SCTP_RTOINFO:
3652 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3653 break;
3654 case SCTP_ASSOCINFO:
3655 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3656 break;
3657 case SCTP_I_WANT_MAPPED_V4_ADDR:
3658 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3659 break;
3660 case SCTP_MAXSEG:
3661 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3662 break;
3663 case SCTP_ADAPTATION_LAYER:
3664 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3665 break;
3666 case SCTP_CONTEXT:
3667 retval = sctp_setsockopt_context(sk, optval, optlen);
3668 break;
3669 case SCTP_FRAGMENT_INTERLEAVE:
3670 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3671 break;
3672 case SCTP_MAX_BURST:
3673 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3674 break;
3675 case SCTP_AUTH_CHUNK:
3676 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3677 break;
3678 case SCTP_HMAC_IDENT:
3679 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3680 break;
3681 case SCTP_AUTH_KEY:
3682 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3683 break;
3684 case SCTP_AUTH_ACTIVE_KEY:
3685 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3686 break;
3687 case SCTP_AUTH_DELETE_KEY:
3688 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3689 break;
3690 case SCTP_AUTO_ASCONF:
3691 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3692 break;
3693 case SCTP_PEER_ADDR_THLDS:
3694 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3695 break;
3696 default:
3697 retval = -ENOPROTOOPT;
3698 break;
3699 }
3700
3701 sctp_release_sock(sk);
3702
3703 out_nounlock:
3704 return retval;
3705 }
3706
3707 /* API 3.1.6 connect() - UDP Style Syntax
3708 *
3709 * An application may use the connect() call in the UDP model to initiate an
3710 * association without sending data.
3711 *
3712 * The syntax is:
3713 *
3714 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3715 *
3716 * sd: the socket descriptor to have a new association added to.
3717 *
3718 * nam: the address structure (either struct sockaddr_in or struct
3719 * sockaddr_in6 defined in RFC2553 [7]).
3720 *
3721 * len: the size of the address.
3722 */
3723 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
3724 int addr_len)
3725 {
3726 int err = 0;
3727 struct sctp_af *af;
3728
3729 sctp_lock_sock(sk);
3730
3731 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
3732 addr, addr_len);
3733
3734 /* Validate addr_len before calling common connect/connectx routine. */
3735 af = sctp_get_af_specific(addr->sa_family);
3736 if (!af || addr_len < af->sockaddr_len) {
3737 err = -EINVAL;
3738 } else {
3739 /* Pass correct addr len to common routine (so it knows there
3740 * is only one address being passed.
3741 */
3742 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3743 }
3744
3745 sctp_release_sock(sk);
3746 return err;
3747 }
3748
3749 /* FIXME: Write comments. */
3750 static int sctp_disconnect(struct sock *sk, int flags)
3751 {
3752 return -EOPNOTSUPP; /* STUB */
3753 }
3754
3755 /* 4.1.4 accept() - TCP Style Syntax
3756 *
3757 * Applications use accept() call to remove an established SCTP
3758 * association from the accept queue of the endpoint. A new socket
3759 * descriptor will be returned from accept() to represent the newly
3760 * formed association.
3761 */
3762 static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3763 {
3764 struct sctp_sock *sp;
3765 struct sctp_endpoint *ep;
3766 struct sock *newsk = NULL;
3767 struct sctp_association *asoc;
3768 long timeo;
3769 int error = 0;
3770
3771 sctp_lock_sock(sk);
3772
3773 sp = sctp_sk(sk);
3774 ep = sp->ep;
3775
3776 if (!sctp_style(sk, TCP)) {
3777 error = -EOPNOTSUPP;
3778 goto out;
3779 }
3780
3781 if (!sctp_sstate(sk, LISTENING)) {
3782 error = -EINVAL;
3783 goto out;
3784 }
3785
3786 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3787
3788 error = sctp_wait_for_accept(sk, timeo);
3789 if (error)
3790 goto out;
3791
3792 /* We treat the list of associations on the endpoint as the accept
3793 * queue and pick the first association on the list.
3794 */
3795 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3796
3797 newsk = sp->pf->create_accept_sk(sk, asoc);
3798 if (!newsk) {
3799 error = -ENOMEM;
3800 goto out;
3801 }
3802
3803 /* Populate the fields of the newsk from the oldsk and migrate the
3804 * asoc to the newsk.
3805 */
3806 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3807
3808 out:
3809 sctp_release_sock(sk);
3810 *err = error;
3811 return newsk;
3812 }
3813
3814 /* The SCTP ioctl handler. */
3815 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3816 {
3817 int rc = -ENOTCONN;
3818
3819 sctp_lock_sock(sk);
3820
3821 /*
3822 * SEQPACKET-style sockets in LISTENING state are valid, for
3823 * SCTP, so only discard TCP-style sockets in LISTENING state.
3824 */
3825 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3826 goto out;
3827
3828 switch (cmd) {
3829 case SIOCINQ: {
3830 struct sk_buff *skb;
3831 unsigned int amount = 0;
3832
3833 skb = skb_peek(&sk->sk_receive_queue);
3834 if (skb != NULL) {
3835 /*
3836 * We will only return the amount of this packet since
3837 * that is all that will be read.
3838 */
3839 amount = skb->len;
3840 }
3841 rc = put_user(amount, (int __user *)arg);
3842 break;
3843 }
3844 default:
3845 rc = -ENOIOCTLCMD;
3846 break;
3847 }
3848 out:
3849 sctp_release_sock(sk);
3850 return rc;
3851 }
3852
3853 /* This is the function which gets called during socket creation to
3854 * initialized the SCTP-specific portion of the sock.
3855 * The sock structure should already be zero-filled memory.
3856 */
3857 static int sctp_init_sock(struct sock *sk)
3858 {
3859 struct net *net = sock_net(sk);
3860 struct sctp_sock *sp;
3861
3862 pr_debug("%s: sk:%p\n", __func__, sk);
3863
3864 sp = sctp_sk(sk);
3865
3866 /* Initialize the SCTP per socket area. */
3867 switch (sk->sk_type) {
3868 case SOCK_SEQPACKET:
3869 sp->type = SCTP_SOCKET_UDP;
3870 break;
3871 case SOCK_STREAM:
3872 sp->type = SCTP_SOCKET_TCP;
3873 break;
3874 default:
3875 return -ESOCKTNOSUPPORT;
3876 }
3877
3878 /* Initialize default send parameters. These parameters can be
3879 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3880 */
3881 sp->default_stream = 0;
3882 sp->default_ppid = 0;
3883 sp->default_flags = 0;
3884 sp->default_context = 0;
3885 sp->default_timetolive = 0;
3886
3887 sp->default_rcv_context = 0;
3888 sp->max_burst = net->sctp.max_burst;
3889
3890 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
3891
3892 /* Initialize default setup parameters. These parameters
3893 * can be modified with the SCTP_INITMSG socket option or
3894 * overridden by the SCTP_INIT CMSG.
3895 */
3896 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3897 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3898 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
3899 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
3900
3901 /* Initialize default RTO related parameters. These parameters can
3902 * be modified for with the SCTP_RTOINFO socket option.
3903 */
3904 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
3905 sp->rtoinfo.srto_max = net->sctp.rto_max;
3906 sp->rtoinfo.srto_min = net->sctp.rto_min;
3907
3908 /* Initialize default association related parameters. These parameters
3909 * can be modified with the SCTP_ASSOCINFO socket option.
3910 */
3911 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
3912 sp->assocparams.sasoc_number_peer_destinations = 0;
3913 sp->assocparams.sasoc_peer_rwnd = 0;
3914 sp->assocparams.sasoc_local_rwnd = 0;
3915 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
3916
3917 /* Initialize default event subscriptions. By default, all the
3918 * options are off.
3919 */
3920 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3921
3922 /* Default Peer Address Parameters. These defaults can
3923 * be modified via SCTP_PEER_ADDR_PARAMS
3924 */
3925 sp->hbinterval = net->sctp.hb_interval;
3926 sp->pathmaxrxt = net->sctp.max_retrans_path;
3927 sp->pathmtu = 0; /* allow default discovery */
3928 sp->sackdelay = net->sctp.sack_timeout;
3929 sp->sackfreq = 2;
3930 sp->param_flags = SPP_HB_ENABLE |
3931 SPP_PMTUD_ENABLE |
3932 SPP_SACKDELAY_ENABLE;
3933
3934 /* If enabled no SCTP message fragmentation will be performed.
3935 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3936 */
3937 sp->disable_fragments = 0;
3938
3939 /* Enable Nagle algorithm by default. */
3940 sp->nodelay = 0;
3941
3942 /* Enable by default. */
3943 sp->v4mapped = 1;
3944
3945 /* Auto-close idle associations after the configured
3946 * number of seconds. A value of 0 disables this
3947 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3948 * for UDP-style sockets only.
3949 */
3950 sp->autoclose = 0;
3951
3952 /* User specified fragmentation limit. */
3953 sp->user_frag = 0;
3954
3955 sp->adaptation_ind = 0;
3956
3957 sp->pf = sctp_get_pf_specific(sk->sk_family);
3958
3959 /* Control variables for partial data delivery. */
3960 atomic_set(&sp->pd_mode, 0);
3961 skb_queue_head_init(&sp->pd_lobby);
3962 sp->frag_interleave = 0;
3963
3964 /* Create a per socket endpoint structure. Even if we
3965 * change the data structure relationships, this may still
3966 * be useful for storing pre-connect address information.
3967 */
3968 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
3969 if (!sp->ep)
3970 return -ENOMEM;
3971
3972 sp->hmac = NULL;
3973
3974 sk->sk_destruct = sctp_destruct_sock;
3975
3976 SCTP_DBG_OBJCNT_INC(sock);
3977
3978 local_bh_disable();
3979 percpu_counter_inc(&sctp_sockets_allocated);
3980 sock_prot_inuse_add(net, sk->sk_prot, 1);
3981 if (net->sctp.default_auto_asconf) {
3982 list_add_tail(&sp->auto_asconf_list,
3983 &net->sctp.auto_asconf_splist);
3984 sp->do_auto_asconf = 1;
3985 } else
3986 sp->do_auto_asconf = 0;
3987 local_bh_enable();
3988
3989 return 0;
3990 }
3991
3992 /* Cleanup any SCTP per socket resources. */
3993 static void sctp_destroy_sock(struct sock *sk)
3994 {
3995 struct sctp_sock *sp;
3996
3997 pr_debug("%s: sk:%p\n", __func__, sk);
3998
3999 /* Release our hold on the endpoint. */
4000 sp = sctp_sk(sk);
4001 /* This could happen during socket init, thus we bail out
4002 * early, since the rest of the below is not setup either.
4003 */
4004 if (sp->ep == NULL)
4005 return;
4006
4007 if (sp->do_auto_asconf) {
4008 sp->do_auto_asconf = 0;
4009 list_del(&sp->auto_asconf_list);
4010 }
4011 sctp_endpoint_free(sp->ep);
4012 local_bh_disable();
4013 percpu_counter_dec(&sctp_sockets_allocated);
4014 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4015 local_bh_enable();
4016 }
4017
4018 /* Triggered when there are no references on the socket anymore */
4019 static void sctp_destruct_sock(struct sock *sk)
4020 {
4021 struct sctp_sock *sp = sctp_sk(sk);
4022
4023 /* Free up the HMAC transform. */
4024 crypto_free_hash(sp->hmac);
4025
4026 inet_sock_destruct(sk);
4027 }
4028
4029 /* API 4.1.7 shutdown() - TCP Style Syntax
4030 * int shutdown(int socket, int how);
4031 *
4032 * sd - the socket descriptor of the association to be closed.
4033 * how - Specifies the type of shutdown. The values are
4034 * as follows:
4035 * SHUT_RD
4036 * Disables further receive operations. No SCTP
4037 * protocol action is taken.
4038 * SHUT_WR
4039 * Disables further send operations, and initiates
4040 * the SCTP shutdown sequence.
4041 * SHUT_RDWR
4042 * Disables further send and receive operations
4043 * and initiates the SCTP shutdown sequence.
4044 */
4045 static void sctp_shutdown(struct sock *sk, int how)
4046 {
4047 struct net *net = sock_net(sk);
4048 struct sctp_endpoint *ep;
4049 struct sctp_association *asoc;
4050
4051 if (!sctp_style(sk, TCP))
4052 return;
4053
4054 if (how & SEND_SHUTDOWN) {
4055 ep = sctp_sk(sk)->ep;
4056 if (!list_empty(&ep->asocs)) {
4057 asoc = list_entry(ep->asocs.next,
4058 struct sctp_association, asocs);
4059 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4060 }
4061 }
4062 }
4063
4064 /* 7.2.1 Association Status (SCTP_STATUS)
4065
4066 * Applications can retrieve current status information about an
4067 * association, including association state, peer receiver window size,
4068 * number of unacked data chunks, and number of data chunks pending
4069 * receipt. This information is read-only.
4070 */
4071 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4072 char __user *optval,
4073 int __user *optlen)
4074 {
4075 struct sctp_status status;
4076 struct sctp_association *asoc = NULL;
4077 struct sctp_transport *transport;
4078 sctp_assoc_t associd;
4079 int retval = 0;
4080
4081 if (len < sizeof(status)) {
4082 retval = -EINVAL;
4083 goto out;
4084 }
4085
4086 len = sizeof(status);
4087 if (copy_from_user(&status, optval, len)) {
4088 retval = -EFAULT;
4089 goto out;
4090 }
4091
4092 associd = status.sstat_assoc_id;
4093 asoc = sctp_id2assoc(sk, associd);
4094 if (!asoc) {
4095 retval = -EINVAL;
4096 goto out;
4097 }
4098
4099 transport = asoc->peer.primary_path;
4100
4101 status.sstat_assoc_id = sctp_assoc2id(asoc);
4102 status.sstat_state = asoc->state;
4103 status.sstat_rwnd = asoc->peer.rwnd;
4104 status.sstat_unackdata = asoc->unack_data;
4105
4106 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4107 status.sstat_instrms = asoc->c.sinit_max_instreams;
4108 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4109 status.sstat_fragmentation_point = asoc->frag_point;
4110 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4111 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4112 transport->af_specific->sockaddr_len);
4113 /* Map ipv4 address into v4-mapped-on-v6 address. */
4114 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4115 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4116 status.sstat_primary.spinfo_state = transport->state;
4117 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4118 status.sstat_primary.spinfo_srtt = transport->srtt;
4119 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4120 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4121
4122 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4123 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4124
4125 if (put_user(len, optlen)) {
4126 retval = -EFAULT;
4127 goto out;
4128 }
4129
4130 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4131 __func__, len, status.sstat_state, status.sstat_rwnd,
4132 status.sstat_assoc_id);
4133
4134 if (copy_to_user(optval, &status, len)) {
4135 retval = -EFAULT;
4136 goto out;
4137 }
4138
4139 out:
4140 return retval;
4141 }
4142
4143
4144 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4145 *
4146 * Applications can retrieve information about a specific peer address
4147 * of an association, including its reachability state, congestion
4148 * window, and retransmission timer values. This information is
4149 * read-only.
4150 */
4151 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4152 char __user *optval,
4153 int __user *optlen)
4154 {
4155 struct sctp_paddrinfo pinfo;
4156 struct sctp_transport *transport;
4157 int retval = 0;
4158
4159 if (len < sizeof(pinfo)) {
4160 retval = -EINVAL;
4161 goto out;
4162 }
4163
4164 len = sizeof(pinfo);
4165 if (copy_from_user(&pinfo, optval, len)) {
4166 retval = -EFAULT;
4167 goto out;
4168 }
4169
4170 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4171 pinfo.spinfo_assoc_id);
4172 if (!transport)
4173 return -EINVAL;
4174
4175 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4176 pinfo.spinfo_state = transport->state;
4177 pinfo.spinfo_cwnd = transport->cwnd;
4178 pinfo.spinfo_srtt = transport->srtt;
4179 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4180 pinfo.spinfo_mtu = transport->pathmtu;
4181
4182 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4183 pinfo.spinfo_state = SCTP_ACTIVE;
4184
4185 if (put_user(len, optlen)) {
4186 retval = -EFAULT;
4187 goto out;
4188 }
4189
4190 if (copy_to_user(optval, &pinfo, len)) {
4191 retval = -EFAULT;
4192 goto out;
4193 }
4194
4195 out:
4196 return retval;
4197 }
4198
4199 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4200 *
4201 * This option is a on/off flag. If enabled no SCTP message
4202 * fragmentation will be performed. Instead if a message being sent
4203 * exceeds the current PMTU size, the message will NOT be sent and
4204 * instead a error will be indicated to the user.
4205 */
4206 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4207 char __user *optval, int __user *optlen)
4208 {
4209 int val;
4210
4211 if (len < sizeof(int))
4212 return -EINVAL;
4213
4214 len = sizeof(int);
4215 val = (sctp_sk(sk)->disable_fragments == 1);
4216 if (put_user(len, optlen))
4217 return -EFAULT;
4218 if (copy_to_user(optval, &val, len))
4219 return -EFAULT;
4220 return 0;
4221 }
4222
4223 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4224 *
4225 * This socket option is used to specify various notifications and
4226 * ancillary data the user wishes to receive.
4227 */
4228 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4229 int __user *optlen)
4230 {
4231 if (len <= 0)
4232 return -EINVAL;
4233 if (len > sizeof(struct sctp_event_subscribe))
4234 len = sizeof(struct sctp_event_subscribe);
4235 if (put_user(len, optlen))
4236 return -EFAULT;
4237 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4238 return -EFAULT;
4239 return 0;
4240 }
4241
4242 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4243 *
4244 * This socket option is applicable to the UDP-style socket only. When
4245 * set it will cause associations that are idle for more than the
4246 * specified number of seconds to automatically close. An association
4247 * being idle is defined an association that has NOT sent or received
4248 * user data. The special value of '0' indicates that no automatic
4249 * close of any associations should be performed. The option expects an
4250 * integer defining the number of seconds of idle time before an
4251 * association is closed.
4252 */
4253 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4254 {
4255 /* Applicable to UDP-style socket only */
4256 if (sctp_style(sk, TCP))
4257 return -EOPNOTSUPP;
4258 if (len < sizeof(int))
4259 return -EINVAL;
4260 len = sizeof(int);
4261 if (put_user(len, optlen))
4262 return -EFAULT;
4263 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4264 return -EFAULT;
4265 return 0;
4266 }
4267
4268 /* Helper routine to branch off an association to a new socket. */
4269 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4270 {
4271 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4272 struct socket *sock;
4273 struct sctp_af *af;
4274 int err = 0;
4275
4276 if (!asoc)
4277 return -EINVAL;
4278
4279 /* An association cannot be branched off from an already peeled-off
4280 * socket, nor is this supported for tcp style sockets.
4281 */
4282 if (!sctp_style(sk, UDP))
4283 return -EINVAL;
4284
4285 /* Create a new socket. */
4286 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4287 if (err < 0)
4288 return err;
4289
4290 sctp_copy_sock(sock->sk, sk, asoc);
4291
4292 /* Make peeled-off sockets more like 1-1 accepted sockets.
4293 * Set the daddr and initialize id to something more random
4294 */
4295 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4296 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4297
4298 /* Populate the fields of the newsk from the oldsk and migrate the
4299 * asoc to the newsk.
4300 */
4301 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4302
4303 *sockp = sock;
4304
4305 return err;
4306 }
4307 EXPORT_SYMBOL(sctp_do_peeloff);
4308
4309 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4310 {
4311 sctp_peeloff_arg_t peeloff;
4312 struct socket *newsock;
4313 struct file *newfile;
4314 int retval = 0;
4315
4316 if (len < sizeof(sctp_peeloff_arg_t))
4317 return -EINVAL;
4318 len = sizeof(sctp_peeloff_arg_t);
4319 if (copy_from_user(&peeloff, optval, len))
4320 return -EFAULT;
4321
4322 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4323 if (retval < 0)
4324 goto out;
4325
4326 /* Map the socket to an unused fd that can be returned to the user. */
4327 retval = get_unused_fd_flags(0);
4328 if (retval < 0) {
4329 sock_release(newsock);
4330 goto out;
4331 }
4332
4333 newfile = sock_alloc_file(newsock, 0, NULL);
4334 if (unlikely(IS_ERR(newfile))) {
4335 put_unused_fd(retval);
4336 sock_release(newsock);
4337 return PTR_ERR(newfile);
4338 }
4339
4340 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4341 retval);
4342
4343 /* Return the fd mapped to the new socket. */
4344 if (put_user(len, optlen)) {
4345 fput(newfile);
4346 put_unused_fd(retval);
4347 return -EFAULT;
4348 }
4349 peeloff.sd = retval;
4350 if (copy_to_user(optval, &peeloff, len)) {
4351 fput(newfile);
4352 put_unused_fd(retval);
4353 return -EFAULT;
4354 }
4355 fd_install(retval, newfile);
4356 out:
4357 return retval;
4358 }
4359
4360 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4361 *
4362 * Applications can enable or disable heartbeats for any peer address of
4363 * an association, modify an address's heartbeat interval, force a
4364 * heartbeat to be sent immediately, and adjust the address's maximum
4365 * number of retransmissions sent before an address is considered
4366 * unreachable. The following structure is used to access and modify an
4367 * address's parameters:
4368 *
4369 * struct sctp_paddrparams {
4370 * sctp_assoc_t spp_assoc_id;
4371 * struct sockaddr_storage spp_address;
4372 * uint32_t spp_hbinterval;
4373 * uint16_t spp_pathmaxrxt;
4374 * uint32_t spp_pathmtu;
4375 * uint32_t spp_sackdelay;
4376 * uint32_t spp_flags;
4377 * };
4378 *
4379 * spp_assoc_id - (one-to-many style socket) This is filled in the
4380 * application, and identifies the association for
4381 * this query.
4382 * spp_address - This specifies which address is of interest.
4383 * spp_hbinterval - This contains the value of the heartbeat interval,
4384 * in milliseconds. If a value of zero
4385 * is present in this field then no changes are to
4386 * be made to this parameter.
4387 * spp_pathmaxrxt - This contains the maximum number of
4388 * retransmissions before this address shall be
4389 * considered unreachable. If a value of zero
4390 * is present in this field then no changes are to
4391 * be made to this parameter.
4392 * spp_pathmtu - When Path MTU discovery is disabled the value
4393 * specified here will be the "fixed" path mtu.
4394 * Note that if the spp_address field is empty
4395 * then all associations on this address will
4396 * have this fixed path mtu set upon them.
4397 *
4398 * spp_sackdelay - When delayed sack is enabled, this value specifies
4399 * the number of milliseconds that sacks will be delayed
4400 * for. This value will apply to all addresses of an
4401 * association if the spp_address field is empty. Note
4402 * also, that if delayed sack is enabled and this
4403 * value is set to 0, no change is made to the last
4404 * recorded delayed sack timer value.
4405 *
4406 * spp_flags - These flags are used to control various features
4407 * on an association. The flag field may contain
4408 * zero or more of the following options.
4409 *
4410 * SPP_HB_ENABLE - Enable heartbeats on the
4411 * specified address. Note that if the address
4412 * field is empty all addresses for the association
4413 * have heartbeats enabled upon them.
4414 *
4415 * SPP_HB_DISABLE - Disable heartbeats on the
4416 * speicifed address. Note that if the address
4417 * field is empty all addresses for the association
4418 * will have their heartbeats disabled. Note also
4419 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4420 * mutually exclusive, only one of these two should
4421 * be specified. Enabling both fields will have
4422 * undetermined results.
4423 *
4424 * SPP_HB_DEMAND - Request a user initiated heartbeat
4425 * to be made immediately.
4426 *
4427 * SPP_PMTUD_ENABLE - This field will enable PMTU
4428 * discovery upon the specified address. Note that
4429 * if the address feild is empty then all addresses
4430 * on the association are effected.
4431 *
4432 * SPP_PMTUD_DISABLE - This field will disable PMTU
4433 * discovery upon the specified address. Note that
4434 * if the address feild is empty then all addresses
4435 * on the association are effected. Not also that
4436 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4437 * exclusive. Enabling both will have undetermined
4438 * results.
4439 *
4440 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4441 * on delayed sack. The time specified in spp_sackdelay
4442 * is used to specify the sack delay for this address. Note
4443 * that if spp_address is empty then all addresses will
4444 * enable delayed sack and take on the sack delay
4445 * value specified in spp_sackdelay.
4446 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4447 * off delayed sack. If the spp_address field is blank then
4448 * delayed sack is disabled for the entire association. Note
4449 * also that this field is mutually exclusive to
4450 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4451 * results.
4452 */
4453 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4454 char __user *optval, int __user *optlen)
4455 {
4456 struct sctp_paddrparams params;
4457 struct sctp_transport *trans = NULL;
4458 struct sctp_association *asoc = NULL;
4459 struct sctp_sock *sp = sctp_sk(sk);
4460
4461 if (len < sizeof(struct sctp_paddrparams))
4462 return -EINVAL;
4463 len = sizeof(struct sctp_paddrparams);
4464 if (copy_from_user(&params, optval, len))
4465 return -EFAULT;
4466
4467 /* If an address other than INADDR_ANY is specified, and
4468 * no transport is found, then the request is invalid.
4469 */
4470 if (!sctp_is_any(sk, (union sctp_addr *)&params.spp_address)) {
4471 trans = sctp_addr_id2transport(sk, &params.spp_address,
4472 params.spp_assoc_id);
4473 if (!trans) {
4474 pr_debug("%s: failed no transport\n", __func__);
4475 return -EINVAL;
4476 }
4477 }
4478
4479 /* Get association, if assoc_id != 0 and the socket is a one
4480 * to many style socket, and an association was not found, then
4481 * the id was invalid.
4482 */
4483 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4484 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4485 pr_debug("%s: failed no association\n", __func__);
4486 return -EINVAL;
4487 }
4488
4489 if (trans) {
4490 /* Fetch transport values. */
4491 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4492 params.spp_pathmtu = trans->pathmtu;
4493 params.spp_pathmaxrxt = trans->pathmaxrxt;
4494 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4495
4496 /*draft-11 doesn't say what to return in spp_flags*/
4497 params.spp_flags = trans->param_flags;
4498 } else if (asoc) {
4499 /* Fetch association values. */
4500 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4501 params.spp_pathmtu = asoc->pathmtu;
4502 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4503 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4504
4505 /*draft-11 doesn't say what to return in spp_flags*/
4506 params.spp_flags = asoc->param_flags;
4507 } else {
4508 /* Fetch socket values. */
4509 params.spp_hbinterval = sp->hbinterval;
4510 params.spp_pathmtu = sp->pathmtu;
4511 params.spp_sackdelay = sp->sackdelay;
4512 params.spp_pathmaxrxt = sp->pathmaxrxt;
4513
4514 /*draft-11 doesn't say what to return in spp_flags*/
4515 params.spp_flags = sp->param_flags;
4516 }
4517
4518 if (copy_to_user(optval, &params, len))
4519 return -EFAULT;
4520
4521 if (put_user(len, optlen))
4522 return -EFAULT;
4523
4524 return 0;
4525 }
4526
4527 /*
4528 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4529 *
4530 * This option will effect the way delayed acks are performed. This
4531 * option allows you to get or set the delayed ack time, in
4532 * milliseconds. It also allows changing the delayed ack frequency.
4533 * Changing the frequency to 1 disables the delayed sack algorithm. If
4534 * the assoc_id is 0, then this sets or gets the endpoints default
4535 * values. If the assoc_id field is non-zero, then the set or get
4536 * effects the specified association for the one to many model (the
4537 * assoc_id field is ignored by the one to one model). Note that if
4538 * sack_delay or sack_freq are 0 when setting this option, then the
4539 * current values will remain unchanged.
4540 *
4541 * struct sctp_sack_info {
4542 * sctp_assoc_t sack_assoc_id;
4543 * uint32_t sack_delay;
4544 * uint32_t sack_freq;
4545 * };
4546 *
4547 * sack_assoc_id - This parameter, indicates which association the user
4548 * is performing an action upon. Note that if this field's value is
4549 * zero then the endpoints default value is changed (effecting future
4550 * associations only).
4551 *
4552 * sack_delay - This parameter contains the number of milliseconds that
4553 * the user is requesting the delayed ACK timer be set to. Note that
4554 * this value is defined in the standard to be between 200 and 500
4555 * milliseconds.
4556 *
4557 * sack_freq - This parameter contains the number of packets that must
4558 * be received before a sack is sent without waiting for the delay
4559 * timer to expire. The default value for this is 2, setting this
4560 * value to 1 will disable the delayed sack algorithm.
4561 */
4562 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4563 char __user *optval,
4564 int __user *optlen)
4565 {
4566 struct sctp_sack_info params;
4567 struct sctp_association *asoc = NULL;
4568 struct sctp_sock *sp = sctp_sk(sk);
4569
4570 if (len >= sizeof(struct sctp_sack_info)) {
4571 len = sizeof(struct sctp_sack_info);
4572
4573 if (copy_from_user(&params, optval, len))
4574 return -EFAULT;
4575 } else if (len == sizeof(struct sctp_assoc_value)) {
4576 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4577 pr_warn("Use struct sctp_sack_info instead\n");
4578 if (copy_from_user(&params, optval, len))
4579 return -EFAULT;
4580 } else
4581 return -EINVAL;
4582
4583 /* Get association, if sack_assoc_id != 0 and the socket is a one
4584 * to many style socket, and an association was not found, then
4585 * the id was invalid.
4586 */
4587 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4588 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4589 return -EINVAL;
4590
4591 if (asoc) {
4592 /* Fetch association values. */
4593 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4594 params.sack_delay = jiffies_to_msecs(
4595 asoc->sackdelay);
4596 params.sack_freq = asoc->sackfreq;
4597
4598 } else {
4599 params.sack_delay = 0;
4600 params.sack_freq = 1;
4601 }
4602 } else {
4603 /* Fetch socket values. */
4604 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4605 params.sack_delay = sp->sackdelay;
4606 params.sack_freq = sp->sackfreq;
4607 } else {
4608 params.sack_delay = 0;
4609 params.sack_freq = 1;
4610 }
4611 }
4612
4613 if (copy_to_user(optval, &params, len))
4614 return -EFAULT;
4615
4616 if (put_user(len, optlen))
4617 return -EFAULT;
4618
4619 return 0;
4620 }
4621
4622 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4623 *
4624 * Applications can specify protocol parameters for the default association
4625 * initialization. The option name argument to setsockopt() and getsockopt()
4626 * is SCTP_INITMSG.
4627 *
4628 * Setting initialization parameters is effective only on an unconnected
4629 * socket (for UDP-style sockets only future associations are effected
4630 * by the change). With TCP-style sockets, this option is inherited by
4631 * sockets derived from a listener socket.
4632 */
4633 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4634 {
4635 if (len < sizeof(struct sctp_initmsg))
4636 return -EINVAL;
4637 len = sizeof(struct sctp_initmsg);
4638 if (put_user(len, optlen))
4639 return -EFAULT;
4640 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4641 return -EFAULT;
4642 return 0;
4643 }
4644
4645
4646 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4647 char __user *optval, int __user *optlen)
4648 {
4649 struct sctp_association *asoc;
4650 int cnt = 0;
4651 struct sctp_getaddrs getaddrs;
4652 struct sctp_transport *from;
4653 void __user *to;
4654 union sctp_addr temp;
4655 struct sctp_sock *sp = sctp_sk(sk);
4656 int addrlen;
4657 size_t space_left;
4658 int bytes_copied;
4659
4660 if (len < sizeof(struct sctp_getaddrs))
4661 return -EINVAL;
4662
4663 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4664 return -EFAULT;
4665
4666 /* For UDP-style sockets, id specifies the association to query. */
4667 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4668 if (!asoc)
4669 return -EINVAL;
4670
4671 to = optval + offsetof(struct sctp_getaddrs, addrs);
4672 space_left = len - offsetof(struct sctp_getaddrs, addrs);
4673
4674 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4675 transports) {
4676 memcpy(&temp, &from->ipaddr, sizeof(temp));
4677 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4678 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4679 if (space_left < addrlen)
4680 return -ENOMEM;
4681 if (copy_to_user(to, &temp, addrlen))
4682 return -EFAULT;
4683 to += addrlen;
4684 cnt++;
4685 space_left -= addrlen;
4686 }
4687
4688 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4689 return -EFAULT;
4690 bytes_copied = ((char __user *)to) - optval;
4691 if (put_user(bytes_copied, optlen))
4692 return -EFAULT;
4693
4694 return 0;
4695 }
4696
4697 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4698 size_t space_left, int *bytes_copied)
4699 {
4700 struct sctp_sockaddr_entry *addr;
4701 union sctp_addr temp;
4702 int cnt = 0;
4703 int addrlen;
4704 struct net *net = sock_net(sk);
4705
4706 rcu_read_lock();
4707 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
4708 if (!addr->valid)
4709 continue;
4710
4711 if ((PF_INET == sk->sk_family) &&
4712 (AF_INET6 == addr->a.sa.sa_family))
4713 continue;
4714 if ((PF_INET6 == sk->sk_family) &&
4715 inet_v6_ipv6only(sk) &&
4716 (AF_INET == addr->a.sa.sa_family))
4717 continue;
4718 memcpy(&temp, &addr->a, sizeof(temp));
4719 if (!temp.v4.sin_port)
4720 temp.v4.sin_port = htons(port);
4721
4722 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4723 &temp);
4724 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4725 if (space_left < addrlen) {
4726 cnt = -ENOMEM;
4727 break;
4728 }
4729 memcpy(to, &temp, addrlen);
4730
4731 to += addrlen;
4732 cnt++;
4733 space_left -= addrlen;
4734 *bytes_copied += addrlen;
4735 }
4736 rcu_read_unlock();
4737
4738 return cnt;
4739 }
4740
4741
4742 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4743 char __user *optval, int __user *optlen)
4744 {
4745 struct sctp_bind_addr *bp;
4746 struct sctp_association *asoc;
4747 int cnt = 0;
4748 struct sctp_getaddrs getaddrs;
4749 struct sctp_sockaddr_entry *addr;
4750 void __user *to;
4751 union sctp_addr temp;
4752 struct sctp_sock *sp = sctp_sk(sk);
4753 int addrlen;
4754 int err = 0;
4755 size_t space_left;
4756 int bytes_copied = 0;
4757 void *addrs;
4758 void *buf;
4759
4760 if (len < sizeof(struct sctp_getaddrs))
4761 return -EINVAL;
4762
4763 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4764 return -EFAULT;
4765
4766 /*
4767 * For UDP-style sockets, id specifies the association to query.
4768 * If the id field is set to the value '0' then the locally bound
4769 * addresses are returned without regard to any particular
4770 * association.
4771 */
4772 if (0 == getaddrs.assoc_id) {
4773 bp = &sctp_sk(sk)->ep->base.bind_addr;
4774 } else {
4775 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4776 if (!asoc)
4777 return -EINVAL;
4778 bp = &asoc->base.bind_addr;
4779 }
4780
4781 to = optval + offsetof(struct sctp_getaddrs, addrs);
4782 space_left = len - offsetof(struct sctp_getaddrs, addrs);
4783
4784 addrs = kmalloc(space_left, GFP_KERNEL);
4785 if (!addrs)
4786 return -ENOMEM;
4787
4788 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4789 * addresses from the global local address list.
4790 */
4791 if (sctp_list_single_entry(&bp->address_list)) {
4792 addr = list_entry(bp->address_list.next,
4793 struct sctp_sockaddr_entry, list);
4794 if (sctp_is_any(sk, &addr->a)) {
4795 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4796 space_left, &bytes_copied);
4797 if (cnt < 0) {
4798 err = cnt;
4799 goto out;
4800 }
4801 goto copy_getaddrs;
4802 }
4803 }
4804
4805 buf = addrs;
4806 /* Protection on the bound address list is not needed since
4807 * in the socket option context we hold a socket lock and
4808 * thus the bound address list can't change.
4809 */
4810 list_for_each_entry(addr, &bp->address_list, list) {
4811 memcpy(&temp, &addr->a, sizeof(temp));
4812 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4813 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4814 if (space_left < addrlen) {
4815 err = -ENOMEM; /*fixme: right error?*/
4816 goto out;
4817 }
4818 memcpy(buf, &temp, addrlen);
4819 buf += addrlen;
4820 bytes_copied += addrlen;
4821 cnt++;
4822 space_left -= addrlen;
4823 }
4824
4825 copy_getaddrs:
4826 if (copy_to_user(to, addrs, bytes_copied)) {
4827 err = -EFAULT;
4828 goto out;
4829 }
4830 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4831 err = -EFAULT;
4832 goto out;
4833 }
4834 if (put_user(bytes_copied, optlen))
4835 err = -EFAULT;
4836 out:
4837 kfree(addrs);
4838 return err;
4839 }
4840
4841 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4842 *
4843 * Requests that the local SCTP stack use the enclosed peer address as
4844 * the association primary. The enclosed address must be one of the
4845 * association peer's addresses.
4846 */
4847 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4848 char __user *optval, int __user *optlen)
4849 {
4850 struct sctp_prim prim;
4851 struct sctp_association *asoc;
4852 struct sctp_sock *sp = sctp_sk(sk);
4853
4854 if (len < sizeof(struct sctp_prim))
4855 return -EINVAL;
4856
4857 len = sizeof(struct sctp_prim);
4858
4859 if (copy_from_user(&prim, optval, len))
4860 return -EFAULT;
4861
4862 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4863 if (!asoc)
4864 return -EINVAL;
4865
4866 if (!asoc->peer.primary_path)
4867 return -ENOTCONN;
4868
4869 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4870 asoc->peer.primary_path->af_specific->sockaddr_len);
4871
4872 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4873 (union sctp_addr *)&prim.ssp_addr);
4874
4875 if (put_user(len, optlen))
4876 return -EFAULT;
4877 if (copy_to_user(optval, &prim, len))
4878 return -EFAULT;
4879
4880 return 0;
4881 }
4882
4883 /*
4884 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4885 *
4886 * Requests that the local endpoint set the specified Adaptation Layer
4887 * Indication parameter for all future INIT and INIT-ACK exchanges.
4888 */
4889 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4890 char __user *optval, int __user *optlen)
4891 {
4892 struct sctp_setadaptation adaptation;
4893
4894 if (len < sizeof(struct sctp_setadaptation))
4895 return -EINVAL;
4896
4897 len = sizeof(struct sctp_setadaptation);
4898
4899 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4900
4901 if (put_user(len, optlen))
4902 return -EFAULT;
4903 if (copy_to_user(optval, &adaptation, len))
4904 return -EFAULT;
4905
4906 return 0;
4907 }
4908
4909 /*
4910 *
4911 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4912 *
4913 * Applications that wish to use the sendto() system call may wish to
4914 * specify a default set of parameters that would normally be supplied
4915 * through the inclusion of ancillary data. This socket option allows
4916 * such an application to set the default sctp_sndrcvinfo structure.
4917
4918
4919 * The application that wishes to use this socket option simply passes
4920 * in to this call the sctp_sndrcvinfo structure defined in Section
4921 * 5.2.2) The input parameters accepted by this call include
4922 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4923 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4924 * to this call if the caller is using the UDP model.
4925 *
4926 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4927 */
4928 static int sctp_getsockopt_default_send_param(struct sock *sk,
4929 int len, char __user *optval,
4930 int __user *optlen)
4931 {
4932 struct sctp_sndrcvinfo info;
4933 struct sctp_association *asoc;
4934 struct sctp_sock *sp = sctp_sk(sk);
4935
4936 if (len < sizeof(struct sctp_sndrcvinfo))
4937 return -EINVAL;
4938
4939 len = sizeof(struct sctp_sndrcvinfo);
4940
4941 if (copy_from_user(&info, optval, len))
4942 return -EFAULT;
4943
4944 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4945 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4946 return -EINVAL;
4947
4948 if (asoc) {
4949 info.sinfo_stream = asoc->default_stream;
4950 info.sinfo_flags = asoc->default_flags;
4951 info.sinfo_ppid = asoc->default_ppid;
4952 info.sinfo_context = asoc->default_context;
4953 info.sinfo_timetolive = asoc->default_timetolive;
4954 } else {
4955 info.sinfo_stream = sp->default_stream;
4956 info.sinfo_flags = sp->default_flags;
4957 info.sinfo_ppid = sp->default_ppid;
4958 info.sinfo_context = sp->default_context;
4959 info.sinfo_timetolive = sp->default_timetolive;
4960 }
4961
4962 if (put_user(len, optlen))
4963 return -EFAULT;
4964 if (copy_to_user(optval, &info, len))
4965 return -EFAULT;
4966
4967 return 0;
4968 }
4969
4970 /*
4971 *
4972 * 7.1.5 SCTP_NODELAY
4973 *
4974 * Turn on/off any Nagle-like algorithm. This means that packets are
4975 * generally sent as soon as possible and no unnecessary delays are
4976 * introduced, at the cost of more packets in the network. Expects an
4977 * integer boolean flag.
4978 */
4979
4980 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4981 char __user *optval, int __user *optlen)
4982 {
4983 int val;
4984
4985 if (len < sizeof(int))
4986 return -EINVAL;
4987
4988 len = sizeof(int);
4989 val = (sctp_sk(sk)->nodelay == 1);
4990 if (put_user(len, optlen))
4991 return -EFAULT;
4992 if (copy_to_user(optval, &val, len))
4993 return -EFAULT;
4994 return 0;
4995 }
4996
4997 /*
4998 *
4999 * 7.1.1 SCTP_RTOINFO
5000 *
5001 * The protocol parameters used to initialize and bound retransmission
5002 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5003 * and modify these parameters.
5004 * All parameters are time values, in milliseconds. A value of 0, when
5005 * modifying the parameters, indicates that the current value should not
5006 * be changed.
5007 *
5008 */
5009 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5010 char __user *optval,
5011 int __user *optlen) {
5012 struct sctp_rtoinfo rtoinfo;
5013 struct sctp_association *asoc;
5014
5015 if (len < sizeof (struct sctp_rtoinfo))
5016 return -EINVAL;
5017
5018 len = sizeof(struct sctp_rtoinfo);
5019
5020 if (copy_from_user(&rtoinfo, optval, len))
5021 return -EFAULT;
5022
5023 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5024
5025 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5026 return -EINVAL;
5027
5028 /* Values corresponding to the specific association. */
5029 if (asoc) {
5030 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5031 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5032 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5033 } else {
5034 /* Values corresponding to the endpoint. */
5035 struct sctp_sock *sp = sctp_sk(sk);
5036
5037 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5038 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5039 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5040 }
5041
5042 if (put_user(len, optlen))
5043 return -EFAULT;
5044
5045 if (copy_to_user(optval, &rtoinfo, len))
5046 return -EFAULT;
5047
5048 return 0;
5049 }
5050
5051 /*
5052 *
5053 * 7.1.2 SCTP_ASSOCINFO
5054 *
5055 * This option is used to tune the maximum retransmission attempts
5056 * of the association.
5057 * Returns an error if the new association retransmission value is
5058 * greater than the sum of the retransmission value of the peer.
5059 * See [SCTP] for more information.
5060 *
5061 */
5062 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5063 char __user *optval,
5064 int __user *optlen)
5065 {
5066
5067 struct sctp_assocparams assocparams;
5068 struct sctp_association *asoc;
5069 struct list_head *pos;
5070 int cnt = 0;
5071
5072 if (len < sizeof (struct sctp_assocparams))
5073 return -EINVAL;
5074
5075 len = sizeof(struct sctp_assocparams);
5076
5077 if (copy_from_user(&assocparams, optval, len))
5078 return -EFAULT;
5079
5080 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5081
5082 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5083 return -EINVAL;
5084
5085 /* Values correspoinding to the specific association */
5086 if (asoc) {
5087 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5088 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5089 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5090 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5091
5092 list_for_each(pos, &asoc->peer.transport_addr_list) {
5093 cnt++;
5094 }
5095
5096 assocparams.sasoc_number_peer_destinations = cnt;
5097 } else {
5098 /* Values corresponding to the endpoint */
5099 struct sctp_sock *sp = sctp_sk(sk);
5100
5101 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5102 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5103 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5104 assocparams.sasoc_cookie_life =
5105 sp->assocparams.sasoc_cookie_life;
5106 assocparams.sasoc_number_peer_destinations =
5107 sp->assocparams.
5108 sasoc_number_peer_destinations;
5109 }
5110
5111 if (put_user(len, optlen))
5112 return -EFAULT;
5113
5114 if (copy_to_user(optval, &assocparams, len))
5115 return -EFAULT;
5116
5117 return 0;
5118 }
5119
5120 /*
5121 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5122 *
5123 * This socket option is a boolean flag which turns on or off mapped V4
5124 * addresses. If this option is turned on and the socket is type
5125 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5126 * If this option is turned off, then no mapping will be done of V4
5127 * addresses and a user will receive both PF_INET6 and PF_INET type
5128 * addresses on the socket.
5129 */
5130 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5131 char __user *optval, int __user *optlen)
5132 {
5133 int val;
5134 struct sctp_sock *sp = sctp_sk(sk);
5135
5136 if (len < sizeof(int))
5137 return -EINVAL;
5138
5139 len = sizeof(int);
5140 val = sp->v4mapped;
5141 if (put_user(len, optlen))
5142 return -EFAULT;
5143 if (copy_to_user(optval, &val, len))
5144 return -EFAULT;
5145
5146 return 0;
5147 }
5148
5149 /*
5150 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5151 * (chapter and verse is quoted at sctp_setsockopt_context())
5152 */
5153 static int sctp_getsockopt_context(struct sock *sk, int len,
5154 char __user *optval, int __user *optlen)
5155 {
5156 struct sctp_assoc_value params;
5157 struct sctp_sock *sp;
5158 struct sctp_association *asoc;
5159
5160 if (len < sizeof(struct sctp_assoc_value))
5161 return -EINVAL;
5162
5163 len = sizeof(struct sctp_assoc_value);
5164
5165 if (copy_from_user(&params, optval, len))
5166 return -EFAULT;
5167
5168 sp = sctp_sk(sk);
5169
5170 if (params.assoc_id != 0) {
5171 asoc = sctp_id2assoc(sk, params.assoc_id);
5172 if (!asoc)
5173 return -EINVAL;
5174 params.assoc_value = asoc->default_rcv_context;
5175 } else {
5176 params.assoc_value = sp->default_rcv_context;
5177 }
5178
5179 if (put_user(len, optlen))
5180 return -EFAULT;
5181 if (copy_to_user(optval, &params, len))
5182 return -EFAULT;
5183
5184 return 0;
5185 }
5186
5187 /*
5188 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5189 * This option will get or set the maximum size to put in any outgoing
5190 * SCTP DATA chunk. If a message is larger than this size it will be
5191 * fragmented by SCTP into the specified size. Note that the underlying
5192 * SCTP implementation may fragment into smaller sized chunks when the
5193 * PMTU of the underlying association is smaller than the value set by
5194 * the user. The default value for this option is '0' which indicates
5195 * the user is NOT limiting fragmentation and only the PMTU will effect
5196 * SCTP's choice of DATA chunk size. Note also that values set larger
5197 * than the maximum size of an IP datagram will effectively let SCTP
5198 * control fragmentation (i.e. the same as setting this option to 0).
5199 *
5200 * The following structure is used to access and modify this parameter:
5201 *
5202 * struct sctp_assoc_value {
5203 * sctp_assoc_t assoc_id;
5204 * uint32_t assoc_value;
5205 * };
5206 *
5207 * assoc_id: This parameter is ignored for one-to-one style sockets.
5208 * For one-to-many style sockets this parameter indicates which
5209 * association the user is performing an action upon. Note that if
5210 * this field's value is zero then the endpoints default value is
5211 * changed (effecting future associations only).
5212 * assoc_value: This parameter specifies the maximum size in bytes.
5213 */
5214 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5215 char __user *optval, int __user *optlen)
5216 {
5217 struct sctp_assoc_value params;
5218 struct sctp_association *asoc;
5219
5220 if (len == sizeof(int)) {
5221 pr_warn("Use of int in maxseg socket option deprecated\n");
5222 pr_warn("Use struct sctp_assoc_value instead\n");
5223 params.assoc_id = 0;
5224 } else if (len >= sizeof(struct sctp_assoc_value)) {
5225 len = sizeof(struct sctp_assoc_value);
5226 if (copy_from_user(&params, optval, sizeof(params)))
5227 return -EFAULT;
5228 } else
5229 return -EINVAL;
5230
5231 asoc = sctp_id2assoc(sk, params.assoc_id);
5232 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5233 return -EINVAL;
5234
5235 if (asoc)
5236 params.assoc_value = asoc->frag_point;
5237 else
5238 params.assoc_value = sctp_sk(sk)->user_frag;
5239
5240 if (put_user(len, optlen))
5241 return -EFAULT;
5242 if (len == sizeof(int)) {
5243 if (copy_to_user(optval, &params.assoc_value, len))
5244 return -EFAULT;
5245 } else {
5246 if (copy_to_user(optval, &params, len))
5247 return -EFAULT;
5248 }
5249
5250 return 0;
5251 }
5252
5253 /*
5254 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5255 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5256 */
5257 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5258 char __user *optval, int __user *optlen)
5259 {
5260 int val;
5261
5262 if (len < sizeof(int))
5263 return -EINVAL;
5264
5265 len = sizeof(int);
5266
5267 val = sctp_sk(sk)->frag_interleave;
5268 if (put_user(len, optlen))
5269 return -EFAULT;
5270 if (copy_to_user(optval, &val, len))
5271 return -EFAULT;
5272
5273 return 0;
5274 }
5275
5276 /*
5277 * 7.1.25. Set or Get the sctp partial delivery point
5278 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5279 */
5280 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5281 char __user *optval,
5282 int __user *optlen)
5283 {
5284 u32 val;
5285
5286 if (len < sizeof(u32))
5287 return -EINVAL;
5288
5289 len = sizeof(u32);
5290
5291 val = sctp_sk(sk)->pd_point;
5292 if (put_user(len, optlen))
5293 return -EFAULT;
5294 if (copy_to_user(optval, &val, len))
5295 return -EFAULT;
5296
5297 return 0;
5298 }
5299
5300 /*
5301 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5302 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5303 */
5304 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5305 char __user *optval,
5306 int __user *optlen)
5307 {
5308 struct sctp_assoc_value params;
5309 struct sctp_sock *sp;
5310 struct sctp_association *asoc;
5311
5312 if (len == sizeof(int)) {
5313 pr_warn("Use of int in max_burst socket option deprecated\n");
5314 pr_warn("Use struct sctp_assoc_value instead\n");
5315 params.assoc_id = 0;
5316 } else if (len >= sizeof(struct sctp_assoc_value)) {
5317 len = sizeof(struct sctp_assoc_value);
5318 if (copy_from_user(&params, optval, len))
5319 return -EFAULT;
5320 } else
5321 return -EINVAL;
5322
5323 sp = sctp_sk(sk);
5324
5325 if (params.assoc_id != 0) {
5326 asoc = sctp_id2assoc(sk, params.assoc_id);
5327 if (!asoc)
5328 return -EINVAL;
5329 params.assoc_value = asoc->max_burst;
5330 } else
5331 params.assoc_value = sp->max_burst;
5332
5333 if (len == sizeof(int)) {
5334 if (copy_to_user(optval, &params.assoc_value, len))
5335 return -EFAULT;
5336 } else {
5337 if (copy_to_user(optval, &params, len))
5338 return -EFAULT;
5339 }
5340
5341 return 0;
5342
5343 }
5344
5345 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5346 char __user *optval, int __user *optlen)
5347 {
5348 struct net *net = sock_net(sk);
5349 struct sctp_hmacalgo __user *p = (void __user *)optval;
5350 struct sctp_hmac_algo_param *hmacs;
5351 __u16 data_len = 0;
5352 u32 num_idents;
5353
5354 if (!net->sctp.auth_enable)
5355 return -EACCES;
5356
5357 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5358 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5359
5360 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5361 return -EINVAL;
5362
5363 len = sizeof(struct sctp_hmacalgo) + data_len;
5364 num_idents = data_len / sizeof(u16);
5365
5366 if (put_user(len, optlen))
5367 return -EFAULT;
5368 if (put_user(num_idents, &p->shmac_num_idents))
5369 return -EFAULT;
5370 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5371 return -EFAULT;
5372 return 0;
5373 }
5374
5375 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5376 char __user *optval, int __user *optlen)
5377 {
5378 struct net *net = sock_net(sk);
5379 struct sctp_authkeyid val;
5380 struct sctp_association *asoc;
5381
5382 if (!net->sctp.auth_enable)
5383 return -EACCES;
5384
5385 if (len < sizeof(struct sctp_authkeyid))
5386 return -EINVAL;
5387 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5388 return -EFAULT;
5389
5390 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5391 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5392 return -EINVAL;
5393
5394 if (asoc)
5395 val.scact_keynumber = asoc->active_key_id;
5396 else
5397 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5398
5399 len = sizeof(struct sctp_authkeyid);
5400 if (put_user(len, optlen))
5401 return -EFAULT;
5402 if (copy_to_user(optval, &val, len))
5403 return -EFAULT;
5404
5405 return 0;
5406 }
5407
5408 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5409 char __user *optval, int __user *optlen)
5410 {
5411 struct net *net = sock_net(sk);
5412 struct sctp_authchunks __user *p = (void __user *)optval;
5413 struct sctp_authchunks val;
5414 struct sctp_association *asoc;
5415 struct sctp_chunks_param *ch;
5416 u32 num_chunks = 0;
5417 char __user *to;
5418
5419 if (!net->sctp.auth_enable)
5420 return -EACCES;
5421
5422 if (len < sizeof(struct sctp_authchunks))
5423 return -EINVAL;
5424
5425 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5426 return -EFAULT;
5427
5428 to = p->gauth_chunks;
5429 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5430 if (!asoc)
5431 return -EINVAL;
5432
5433 ch = asoc->peer.peer_chunks;
5434 if (!ch)
5435 goto num;
5436
5437 /* See if the user provided enough room for all the data */
5438 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5439 if (len < num_chunks)
5440 return -EINVAL;
5441
5442 if (copy_to_user(to, ch->chunks, num_chunks))
5443 return -EFAULT;
5444 num:
5445 len = sizeof(struct sctp_authchunks) + num_chunks;
5446 if (put_user(len, optlen))
5447 return -EFAULT;
5448 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5449 return -EFAULT;
5450 return 0;
5451 }
5452
5453 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5454 char __user *optval, int __user *optlen)
5455 {
5456 struct net *net = sock_net(sk);
5457 struct sctp_authchunks __user *p = (void __user *)optval;
5458 struct sctp_authchunks val;
5459 struct sctp_association *asoc;
5460 struct sctp_chunks_param *ch;
5461 u32 num_chunks = 0;
5462 char __user *to;
5463
5464 if (!net->sctp.auth_enable)
5465 return -EACCES;
5466
5467 if (len < sizeof(struct sctp_authchunks))
5468 return -EINVAL;
5469
5470 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5471 return -EFAULT;
5472
5473 to = p->gauth_chunks;
5474 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5475 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5476 return -EINVAL;
5477
5478 if (asoc)
5479 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
5480 else
5481 ch = sctp_sk(sk)->ep->auth_chunk_list;
5482
5483 if (!ch)
5484 goto num;
5485
5486 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5487 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5488 return -EINVAL;
5489
5490 if (copy_to_user(to, ch->chunks, num_chunks))
5491 return -EFAULT;
5492 num:
5493 len = sizeof(struct sctp_authchunks) + num_chunks;
5494 if (put_user(len, optlen))
5495 return -EFAULT;
5496 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5497 return -EFAULT;
5498
5499 return 0;
5500 }
5501
5502 /*
5503 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5504 * This option gets the current number of associations that are attached
5505 * to a one-to-many style socket. The option value is an uint32_t.
5506 */
5507 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5508 char __user *optval, int __user *optlen)
5509 {
5510 struct sctp_sock *sp = sctp_sk(sk);
5511 struct sctp_association *asoc;
5512 u32 val = 0;
5513
5514 if (sctp_style(sk, TCP))
5515 return -EOPNOTSUPP;
5516
5517 if (len < sizeof(u32))
5518 return -EINVAL;
5519
5520 len = sizeof(u32);
5521
5522 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5523 val++;
5524 }
5525
5526 if (put_user(len, optlen))
5527 return -EFAULT;
5528 if (copy_to_user(optval, &val, len))
5529 return -EFAULT;
5530
5531 return 0;
5532 }
5533
5534 /*
5535 * 8.1.23 SCTP_AUTO_ASCONF
5536 * See the corresponding setsockopt entry as description
5537 */
5538 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5539 char __user *optval, int __user *optlen)
5540 {
5541 int val = 0;
5542
5543 if (len < sizeof(int))
5544 return -EINVAL;
5545
5546 len = sizeof(int);
5547 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5548 val = 1;
5549 if (put_user(len, optlen))
5550 return -EFAULT;
5551 if (copy_to_user(optval, &val, len))
5552 return -EFAULT;
5553 return 0;
5554 }
5555
5556 /*
5557 * 8.2.6. Get the Current Identifiers of Associations
5558 * (SCTP_GET_ASSOC_ID_LIST)
5559 *
5560 * This option gets the current list of SCTP association identifiers of
5561 * the SCTP associations handled by a one-to-many style socket.
5562 */
5563 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5564 char __user *optval, int __user *optlen)
5565 {
5566 struct sctp_sock *sp = sctp_sk(sk);
5567 struct sctp_association *asoc;
5568 struct sctp_assoc_ids *ids;
5569 u32 num = 0;
5570
5571 if (sctp_style(sk, TCP))
5572 return -EOPNOTSUPP;
5573
5574 if (len < sizeof(struct sctp_assoc_ids))
5575 return -EINVAL;
5576
5577 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5578 num++;
5579 }
5580
5581 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5582 return -EINVAL;
5583
5584 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5585
5586 ids = kmalloc(len, GFP_KERNEL);
5587 if (unlikely(!ids))
5588 return -ENOMEM;
5589
5590 ids->gaids_number_of_ids = num;
5591 num = 0;
5592 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5593 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5594 }
5595
5596 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5597 kfree(ids);
5598 return -EFAULT;
5599 }
5600
5601 kfree(ids);
5602 return 0;
5603 }
5604
5605 /*
5606 * SCTP_PEER_ADDR_THLDS
5607 *
5608 * This option allows us to fetch the partially failed threshold for one or all
5609 * transports in an association. See Section 6.1 of:
5610 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5611 */
5612 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
5613 char __user *optval,
5614 int len,
5615 int __user *optlen)
5616 {
5617 struct sctp_paddrthlds val;
5618 struct sctp_transport *trans;
5619 struct sctp_association *asoc;
5620
5621 if (len < sizeof(struct sctp_paddrthlds))
5622 return -EINVAL;
5623 len = sizeof(struct sctp_paddrthlds);
5624 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
5625 return -EFAULT;
5626
5627 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
5628 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
5629 if (!asoc)
5630 return -ENOENT;
5631
5632 val.spt_pathpfthld = asoc->pf_retrans;
5633 val.spt_pathmaxrxt = asoc->pathmaxrxt;
5634 } else {
5635 trans = sctp_addr_id2transport(sk, &val.spt_address,
5636 val.spt_assoc_id);
5637 if (!trans)
5638 return -ENOENT;
5639
5640 val.spt_pathmaxrxt = trans->pathmaxrxt;
5641 val.spt_pathpfthld = trans->pf_retrans;
5642 }
5643
5644 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
5645 return -EFAULT;
5646
5647 return 0;
5648 }
5649
5650 /*
5651 * SCTP_GET_ASSOC_STATS
5652 *
5653 * This option retrieves local per endpoint statistics. It is modeled
5654 * after OpenSolaris' implementation
5655 */
5656 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
5657 char __user *optval,
5658 int __user *optlen)
5659 {
5660 struct sctp_assoc_stats sas;
5661 struct sctp_association *asoc = NULL;
5662
5663 /* User must provide at least the assoc id */
5664 if (len < sizeof(sctp_assoc_t))
5665 return -EINVAL;
5666
5667 /* Allow the struct to grow and fill in as much as possible */
5668 len = min_t(size_t, len, sizeof(sas));
5669
5670 if (copy_from_user(&sas, optval, len))
5671 return -EFAULT;
5672
5673 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
5674 if (!asoc)
5675 return -EINVAL;
5676
5677 sas.sas_rtxchunks = asoc->stats.rtxchunks;
5678 sas.sas_gapcnt = asoc->stats.gapcnt;
5679 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
5680 sas.sas_osacks = asoc->stats.osacks;
5681 sas.sas_isacks = asoc->stats.isacks;
5682 sas.sas_octrlchunks = asoc->stats.octrlchunks;
5683 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
5684 sas.sas_oodchunks = asoc->stats.oodchunks;
5685 sas.sas_iodchunks = asoc->stats.iodchunks;
5686 sas.sas_ouodchunks = asoc->stats.ouodchunks;
5687 sas.sas_iuodchunks = asoc->stats.iuodchunks;
5688 sas.sas_idupchunks = asoc->stats.idupchunks;
5689 sas.sas_opackets = asoc->stats.opackets;
5690 sas.sas_ipackets = asoc->stats.ipackets;
5691
5692 /* New high max rto observed, will return 0 if not a single
5693 * RTO update took place. obs_rto_ipaddr will be bogus
5694 * in such a case
5695 */
5696 sas.sas_maxrto = asoc->stats.max_obs_rto;
5697 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
5698 sizeof(struct sockaddr_storage));
5699
5700 /* Mark beginning of a new observation period */
5701 asoc->stats.max_obs_rto = asoc->rto_min;
5702
5703 if (put_user(len, optlen))
5704 return -EFAULT;
5705
5706 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
5707
5708 if (copy_to_user(optval, &sas, len))
5709 return -EFAULT;
5710
5711 return 0;
5712 }
5713
5714 static int sctp_getsockopt(struct sock *sk, int level, int optname,
5715 char __user *optval, int __user *optlen)
5716 {
5717 int retval = 0;
5718 int len;
5719
5720 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
5721
5722 /* I can hardly begin to describe how wrong this is. This is
5723 * so broken as to be worse than useless. The API draft
5724 * REALLY is NOT helpful here... I am not convinced that the
5725 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5726 * are at all well-founded.
5727 */
5728 if (level != SOL_SCTP) {
5729 struct sctp_af *af = sctp_sk(sk)->pf->af;
5730
5731 retval = af->getsockopt(sk, level, optname, optval, optlen);
5732 return retval;
5733 }
5734
5735 if (get_user(len, optlen))
5736 return -EFAULT;
5737
5738 sctp_lock_sock(sk);
5739
5740 switch (optname) {
5741 case SCTP_STATUS:
5742 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5743 break;
5744 case SCTP_DISABLE_FRAGMENTS:
5745 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5746 optlen);
5747 break;
5748 case SCTP_EVENTS:
5749 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5750 break;
5751 case SCTP_AUTOCLOSE:
5752 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5753 break;
5754 case SCTP_SOCKOPT_PEELOFF:
5755 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5756 break;
5757 case SCTP_PEER_ADDR_PARAMS:
5758 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5759 optlen);
5760 break;
5761 case SCTP_DELAYED_SACK:
5762 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5763 optlen);
5764 break;
5765 case SCTP_INITMSG:
5766 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5767 break;
5768 case SCTP_GET_PEER_ADDRS:
5769 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5770 optlen);
5771 break;
5772 case SCTP_GET_LOCAL_ADDRS:
5773 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5774 optlen);
5775 break;
5776 case SCTP_SOCKOPT_CONNECTX3:
5777 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5778 break;
5779 case SCTP_DEFAULT_SEND_PARAM:
5780 retval = sctp_getsockopt_default_send_param(sk, len,
5781 optval, optlen);
5782 break;
5783 case SCTP_PRIMARY_ADDR:
5784 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5785 break;
5786 case SCTP_NODELAY:
5787 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5788 break;
5789 case SCTP_RTOINFO:
5790 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5791 break;
5792 case SCTP_ASSOCINFO:
5793 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5794 break;
5795 case SCTP_I_WANT_MAPPED_V4_ADDR:
5796 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5797 break;
5798 case SCTP_MAXSEG:
5799 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5800 break;
5801 case SCTP_GET_PEER_ADDR_INFO:
5802 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5803 optlen);
5804 break;
5805 case SCTP_ADAPTATION_LAYER:
5806 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5807 optlen);
5808 break;
5809 case SCTP_CONTEXT:
5810 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5811 break;
5812 case SCTP_FRAGMENT_INTERLEAVE:
5813 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5814 optlen);
5815 break;
5816 case SCTP_PARTIAL_DELIVERY_POINT:
5817 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5818 optlen);
5819 break;
5820 case SCTP_MAX_BURST:
5821 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5822 break;
5823 case SCTP_AUTH_KEY:
5824 case SCTP_AUTH_CHUNK:
5825 case SCTP_AUTH_DELETE_KEY:
5826 retval = -EOPNOTSUPP;
5827 break;
5828 case SCTP_HMAC_IDENT:
5829 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5830 break;
5831 case SCTP_AUTH_ACTIVE_KEY:
5832 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5833 break;
5834 case SCTP_PEER_AUTH_CHUNKS:
5835 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5836 optlen);
5837 break;
5838 case SCTP_LOCAL_AUTH_CHUNKS:
5839 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5840 optlen);
5841 break;
5842 case SCTP_GET_ASSOC_NUMBER:
5843 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5844 break;
5845 case SCTP_GET_ASSOC_ID_LIST:
5846 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5847 break;
5848 case SCTP_AUTO_ASCONF:
5849 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5850 break;
5851 case SCTP_PEER_ADDR_THLDS:
5852 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
5853 break;
5854 case SCTP_GET_ASSOC_STATS:
5855 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
5856 break;
5857 default:
5858 retval = -ENOPROTOOPT;
5859 break;
5860 }
5861
5862 sctp_release_sock(sk);
5863 return retval;
5864 }
5865
5866 static void sctp_hash(struct sock *sk)
5867 {
5868 /* STUB */
5869 }
5870
5871 static void sctp_unhash(struct sock *sk)
5872 {
5873 /* STUB */
5874 }
5875
5876 /* Check if port is acceptable. Possibly find first available port.
5877 *
5878 * The port hash table (contained in the 'global' SCTP protocol storage
5879 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5880 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5881 * list (the list number is the port number hashed out, so as you
5882 * would expect from a hash function, all the ports in a given list have
5883 * such a number that hashes out to the same list number; you were
5884 * expecting that, right?); so each list has a set of ports, with a
5885 * link to the socket (struct sock) that uses it, the port number and
5886 * a fastreuse flag (FIXME: NPI ipg).
5887 */
5888 static struct sctp_bind_bucket *sctp_bucket_create(
5889 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
5890
5891 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5892 {
5893 struct sctp_bind_hashbucket *head; /* hash list */
5894 struct sctp_bind_bucket *pp;
5895 unsigned short snum;
5896 int ret;
5897
5898 snum = ntohs(addr->v4.sin_port);
5899
5900 pr_debug("%s: begins, snum:%d\n", __func__, snum);
5901
5902 sctp_local_bh_disable();
5903
5904 if (snum == 0) {
5905 /* Search for an available port. */
5906 int low, high, remaining, index;
5907 unsigned int rover;
5908
5909 inet_get_local_port_range(sock_net(sk), &low, &high);
5910 remaining = (high - low) + 1;
5911 rover = net_random() % remaining + low;
5912
5913 do {
5914 rover++;
5915 if ((rover < low) || (rover > high))
5916 rover = low;
5917 if (inet_is_reserved_local_port(rover))
5918 continue;
5919 index = sctp_phashfn(sock_net(sk), rover);
5920 head = &sctp_port_hashtable[index];
5921 sctp_spin_lock(&head->lock);
5922 sctp_for_each_hentry(pp, &head->chain)
5923 if ((pp->port == rover) &&
5924 net_eq(sock_net(sk), pp->net))
5925 goto next;
5926 break;
5927 next:
5928 sctp_spin_unlock(&head->lock);
5929 } while (--remaining > 0);
5930
5931 /* Exhausted local port range during search? */
5932 ret = 1;
5933 if (remaining <= 0)
5934 goto fail;
5935
5936 /* OK, here is the one we will use. HEAD (the port
5937 * hash table list entry) is non-NULL and we hold it's
5938 * mutex.
5939 */
5940 snum = rover;
5941 } else {
5942 /* We are given an specific port number; we verify
5943 * that it is not being used. If it is used, we will
5944 * exahust the search in the hash list corresponding
5945 * to the port number (snum) - we detect that with the
5946 * port iterator, pp being NULL.
5947 */
5948 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
5949 sctp_spin_lock(&head->lock);
5950 sctp_for_each_hentry(pp, &head->chain) {
5951 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
5952 goto pp_found;
5953 }
5954 }
5955 pp = NULL;
5956 goto pp_not_found;
5957 pp_found:
5958 if (!hlist_empty(&pp->owner)) {
5959 /* We had a port hash table hit - there is an
5960 * available port (pp != NULL) and it is being
5961 * used by other socket (pp->owner not empty); that other
5962 * socket is going to be sk2.
5963 */
5964 int reuse = sk->sk_reuse;
5965 struct sock *sk2;
5966
5967 pr_debug("%s: found a possible match\n", __func__);
5968
5969 if (pp->fastreuse && sk->sk_reuse &&
5970 sk->sk_state != SCTP_SS_LISTENING)
5971 goto success;
5972
5973 /* Run through the list of sockets bound to the port
5974 * (pp->port) [via the pointers bind_next and
5975 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5976 * we get the endpoint they describe and run through
5977 * the endpoint's list of IP (v4 or v6) addresses,
5978 * comparing each of the addresses with the address of
5979 * the socket sk. If we find a match, then that means
5980 * that this port/socket (sk) combination are already
5981 * in an endpoint.
5982 */
5983 sk_for_each_bound(sk2, &pp->owner) {
5984 struct sctp_endpoint *ep2;
5985 ep2 = sctp_sk(sk2)->ep;
5986
5987 if (sk == sk2 ||
5988 (reuse && sk2->sk_reuse &&
5989 sk2->sk_state != SCTP_SS_LISTENING))
5990 continue;
5991
5992 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5993 sctp_sk(sk2), sctp_sk(sk))) {
5994 ret = (long)sk2;
5995 goto fail_unlock;
5996 }
5997 }
5998
5999 pr_debug("%s: found a match\n", __func__);
6000 }
6001 pp_not_found:
6002 /* If there was a hash table miss, create a new port. */
6003 ret = 1;
6004 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
6005 goto fail_unlock;
6006
6007 /* In either case (hit or miss), make sure fastreuse is 1 only
6008 * if sk->sk_reuse is too (that is, if the caller requested
6009 * SO_REUSEADDR on this socket -sk-).
6010 */
6011 if (hlist_empty(&pp->owner)) {
6012 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
6013 pp->fastreuse = 1;
6014 else
6015 pp->fastreuse = 0;
6016 } else if (pp->fastreuse &&
6017 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6018 pp->fastreuse = 0;
6019
6020 /* We are set, so fill up all the data in the hash table
6021 * entry, tie the socket list information with the rest of the
6022 * sockets FIXME: Blurry, NPI (ipg).
6023 */
6024 success:
6025 if (!sctp_sk(sk)->bind_hash) {
6026 inet_sk(sk)->inet_num = snum;
6027 sk_add_bind_node(sk, &pp->owner);
6028 sctp_sk(sk)->bind_hash = pp;
6029 }
6030 ret = 0;
6031
6032 fail_unlock:
6033 sctp_spin_unlock(&head->lock);
6034
6035 fail:
6036 sctp_local_bh_enable();
6037 return ret;
6038 }
6039
6040 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6041 * port is requested.
6042 */
6043 static int sctp_get_port(struct sock *sk, unsigned short snum)
6044 {
6045 union sctp_addr addr;
6046 struct sctp_af *af = sctp_sk(sk)->pf->af;
6047
6048 /* Set up a dummy address struct from the sk. */
6049 af->from_sk(&addr, sk);
6050 addr.v4.sin_port = htons(snum);
6051
6052 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6053 return !!sctp_get_port_local(sk, &addr);
6054 }
6055
6056 /*
6057 * Move a socket to LISTENING state.
6058 */
6059 static int sctp_listen_start(struct sock *sk, int backlog)
6060 {
6061 struct sctp_sock *sp = sctp_sk(sk);
6062 struct sctp_endpoint *ep = sp->ep;
6063 struct crypto_hash *tfm = NULL;
6064 char alg[32];
6065
6066 /* Allocate HMAC for generating cookie. */
6067 if (!sp->hmac && sp->sctp_hmac_alg) {
6068 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6069 tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC);
6070 if (IS_ERR(tfm)) {
6071 net_info_ratelimited("failed to load transform for %s: %ld\n",
6072 sp->sctp_hmac_alg, PTR_ERR(tfm));
6073 return -ENOSYS;
6074 }
6075 sctp_sk(sk)->hmac = tfm;
6076 }
6077
6078 /*
6079 * If a bind() or sctp_bindx() is not called prior to a listen()
6080 * call that allows new associations to be accepted, the system
6081 * picks an ephemeral port and will choose an address set equivalent
6082 * to binding with a wildcard address.
6083 *
6084 * This is not currently spelled out in the SCTP sockets
6085 * extensions draft, but follows the practice as seen in TCP
6086 * sockets.
6087 *
6088 */
6089 sk->sk_state = SCTP_SS_LISTENING;
6090 if (!ep->base.bind_addr.port) {
6091 if (sctp_autobind(sk))
6092 return -EAGAIN;
6093 } else {
6094 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6095 sk->sk_state = SCTP_SS_CLOSED;
6096 return -EADDRINUSE;
6097 }
6098 }
6099
6100 sk->sk_max_ack_backlog = backlog;
6101 sctp_hash_endpoint(ep);
6102 return 0;
6103 }
6104
6105 /*
6106 * 4.1.3 / 5.1.3 listen()
6107 *
6108 * By default, new associations are not accepted for UDP style sockets.
6109 * An application uses listen() to mark a socket as being able to
6110 * accept new associations.
6111 *
6112 * On TCP style sockets, applications use listen() to ready the SCTP
6113 * endpoint for accepting inbound associations.
6114 *
6115 * On both types of endpoints a backlog of '0' disables listening.
6116 *
6117 * Move a socket to LISTENING state.
6118 */
6119 int sctp_inet_listen(struct socket *sock, int backlog)
6120 {
6121 struct sock *sk = sock->sk;
6122 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6123 int err = -EINVAL;
6124
6125 if (unlikely(backlog < 0))
6126 return err;
6127
6128 sctp_lock_sock(sk);
6129
6130 /* Peeled-off sockets are not allowed to listen(). */
6131 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6132 goto out;
6133
6134 if (sock->state != SS_UNCONNECTED)
6135 goto out;
6136
6137 /* If backlog is zero, disable listening. */
6138 if (!backlog) {
6139 if (sctp_sstate(sk, CLOSED))
6140 goto out;
6141
6142 err = 0;
6143 sctp_unhash_endpoint(ep);
6144 sk->sk_state = SCTP_SS_CLOSED;
6145 if (sk->sk_reuse)
6146 sctp_sk(sk)->bind_hash->fastreuse = 1;
6147 goto out;
6148 }
6149
6150 /* If we are already listening, just update the backlog */
6151 if (sctp_sstate(sk, LISTENING))
6152 sk->sk_max_ack_backlog = backlog;
6153 else {
6154 err = sctp_listen_start(sk, backlog);
6155 if (err)
6156 goto out;
6157 }
6158
6159 err = 0;
6160 out:
6161 sctp_release_sock(sk);
6162 return err;
6163 }
6164
6165 /*
6166 * This function is done by modeling the current datagram_poll() and the
6167 * tcp_poll(). Note that, based on these implementations, we don't
6168 * lock the socket in this function, even though it seems that,
6169 * ideally, locking or some other mechanisms can be used to ensure
6170 * the integrity of the counters (sndbuf and wmem_alloc) used
6171 * in this place. We assume that we don't need locks either until proven
6172 * otherwise.
6173 *
6174 * Another thing to note is that we include the Async I/O support
6175 * here, again, by modeling the current TCP/UDP code. We don't have
6176 * a good way to test with it yet.
6177 */
6178 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6179 {
6180 struct sock *sk = sock->sk;
6181 struct sctp_sock *sp = sctp_sk(sk);
6182 unsigned int mask;
6183
6184 poll_wait(file, sk_sleep(sk), wait);
6185
6186 /* A TCP-style listening socket becomes readable when the accept queue
6187 * is not empty.
6188 */
6189 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6190 return (!list_empty(&sp->ep->asocs)) ?
6191 (POLLIN | POLLRDNORM) : 0;
6192
6193 mask = 0;
6194
6195 /* Is there any exceptional events? */
6196 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6197 mask |= POLLERR |
6198 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
6199 if (sk->sk_shutdown & RCV_SHUTDOWN)
6200 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6201 if (sk->sk_shutdown == SHUTDOWN_MASK)
6202 mask |= POLLHUP;
6203
6204 /* Is it readable? Reconsider this code with TCP-style support. */
6205 if (!skb_queue_empty(&sk->sk_receive_queue))
6206 mask |= POLLIN | POLLRDNORM;
6207
6208 /* The association is either gone or not ready. */
6209 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6210 return mask;
6211
6212 /* Is it writable? */
6213 if (sctp_writeable(sk)) {
6214 mask |= POLLOUT | POLLWRNORM;
6215 } else {
6216 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6217 /*
6218 * Since the socket is not locked, the buffer
6219 * might be made available after the writeable check and
6220 * before the bit is set. This could cause a lost I/O
6221 * signal. tcp_poll() has a race breaker for this race
6222 * condition. Based on their implementation, we put
6223 * in the following code to cover it as well.
6224 */
6225 if (sctp_writeable(sk))
6226 mask |= POLLOUT | POLLWRNORM;
6227 }
6228 return mask;
6229 }
6230
6231 /********************************************************************
6232 * 2nd Level Abstractions
6233 ********************************************************************/
6234
6235 static struct sctp_bind_bucket *sctp_bucket_create(
6236 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6237 {
6238 struct sctp_bind_bucket *pp;
6239
6240 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6241 if (pp) {
6242 SCTP_DBG_OBJCNT_INC(bind_bucket);
6243 pp->port = snum;
6244 pp->fastreuse = 0;
6245 INIT_HLIST_HEAD(&pp->owner);
6246 pp->net = net;
6247 hlist_add_head(&pp->node, &head->chain);
6248 }
6249 return pp;
6250 }
6251
6252 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6253 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6254 {
6255 if (pp && hlist_empty(&pp->owner)) {
6256 __hlist_del(&pp->node);
6257 kmem_cache_free(sctp_bucket_cachep, pp);
6258 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6259 }
6260 }
6261
6262 /* Release this socket's reference to a local port. */
6263 static inline void __sctp_put_port(struct sock *sk)
6264 {
6265 struct sctp_bind_hashbucket *head =
6266 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6267 inet_sk(sk)->inet_num)];
6268 struct sctp_bind_bucket *pp;
6269
6270 sctp_spin_lock(&head->lock);
6271 pp = sctp_sk(sk)->bind_hash;
6272 __sk_del_bind_node(sk);
6273 sctp_sk(sk)->bind_hash = NULL;
6274 inet_sk(sk)->inet_num = 0;
6275 sctp_bucket_destroy(pp);
6276 sctp_spin_unlock(&head->lock);
6277 }
6278
6279 void sctp_put_port(struct sock *sk)
6280 {
6281 sctp_local_bh_disable();
6282 __sctp_put_port(sk);
6283 sctp_local_bh_enable();
6284 }
6285
6286 /*
6287 * The system picks an ephemeral port and choose an address set equivalent
6288 * to binding with a wildcard address.
6289 * One of those addresses will be the primary address for the association.
6290 * This automatically enables the multihoming capability of SCTP.
6291 */
6292 static int sctp_autobind(struct sock *sk)
6293 {
6294 union sctp_addr autoaddr;
6295 struct sctp_af *af;
6296 __be16 port;
6297
6298 /* Initialize a local sockaddr structure to INADDR_ANY. */
6299 af = sctp_sk(sk)->pf->af;
6300
6301 port = htons(inet_sk(sk)->inet_num);
6302 af->inaddr_any(&autoaddr, port);
6303
6304 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6305 }
6306
6307 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6308 *
6309 * From RFC 2292
6310 * 4.2 The cmsghdr Structure *
6311 *
6312 * When ancillary data is sent or received, any number of ancillary data
6313 * objects can be specified by the msg_control and msg_controllen members of
6314 * the msghdr structure, because each object is preceded by
6315 * a cmsghdr structure defining the object's length (the cmsg_len member).
6316 * Historically Berkeley-derived implementations have passed only one object
6317 * at a time, but this API allows multiple objects to be
6318 * passed in a single call to sendmsg() or recvmsg(). The following example
6319 * shows two ancillary data objects in a control buffer.
6320 *
6321 * |<--------------------------- msg_controllen -------------------------->|
6322 * | |
6323 *
6324 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6325 *
6326 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6327 * | | |
6328 *
6329 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6330 *
6331 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6332 * | | | | |
6333 *
6334 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6335 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6336 *
6337 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6338 *
6339 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6340 * ^
6341 * |
6342 *
6343 * msg_control
6344 * points here
6345 */
6346 static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
6347 {
6348 struct cmsghdr *cmsg;
6349 struct msghdr *my_msg = (struct msghdr *)msg;
6350
6351 for (cmsg = CMSG_FIRSTHDR(msg);
6352 cmsg != NULL;
6353 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6354 if (!CMSG_OK(my_msg, cmsg))
6355 return -EINVAL;
6356
6357 /* Should we parse this header or ignore? */
6358 if (cmsg->cmsg_level != IPPROTO_SCTP)
6359 continue;
6360
6361 /* Strictly check lengths following example in SCM code. */
6362 switch (cmsg->cmsg_type) {
6363 case SCTP_INIT:
6364 /* SCTP Socket API Extension
6365 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6366 *
6367 * This cmsghdr structure provides information for
6368 * initializing new SCTP associations with sendmsg().
6369 * The SCTP_INITMSG socket option uses this same data
6370 * structure. This structure is not used for
6371 * recvmsg().
6372 *
6373 * cmsg_level cmsg_type cmsg_data[]
6374 * ------------ ------------ ----------------------
6375 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6376 */
6377 if (cmsg->cmsg_len !=
6378 CMSG_LEN(sizeof(struct sctp_initmsg)))
6379 return -EINVAL;
6380 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6381 break;
6382
6383 case SCTP_SNDRCV:
6384 /* SCTP Socket API Extension
6385 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6386 *
6387 * This cmsghdr structure specifies SCTP options for
6388 * sendmsg() and describes SCTP header information
6389 * about a received message through recvmsg().
6390 *
6391 * cmsg_level cmsg_type cmsg_data[]
6392 * ------------ ------------ ----------------------
6393 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6394 */
6395 if (cmsg->cmsg_len !=
6396 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6397 return -EINVAL;
6398
6399 cmsgs->info =
6400 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6401
6402 /* Minimally, validate the sinfo_flags. */
6403 if (cmsgs->info->sinfo_flags &
6404 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6405 SCTP_ABORT | SCTP_EOF))
6406 return -EINVAL;
6407 break;
6408
6409 default:
6410 return -EINVAL;
6411 }
6412 }
6413 return 0;
6414 }
6415
6416 /*
6417 * Wait for a packet..
6418 * Note: This function is the same function as in core/datagram.c
6419 * with a few modifications to make lksctp work.
6420 */
6421 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
6422 {
6423 int error;
6424 DEFINE_WAIT(wait);
6425
6426 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6427
6428 /* Socket errors? */
6429 error = sock_error(sk);
6430 if (error)
6431 goto out;
6432
6433 if (!skb_queue_empty(&sk->sk_receive_queue))
6434 goto ready;
6435
6436 /* Socket shut down? */
6437 if (sk->sk_shutdown & RCV_SHUTDOWN)
6438 goto out;
6439
6440 /* Sequenced packets can come disconnected. If so we report the
6441 * problem.
6442 */
6443 error = -ENOTCONN;
6444
6445 /* Is there a good reason to think that we may receive some data? */
6446 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6447 goto out;
6448
6449 /* Handle signals. */
6450 if (signal_pending(current))
6451 goto interrupted;
6452
6453 /* Let another process have a go. Since we are going to sleep
6454 * anyway. Note: This may cause odd behaviors if the message
6455 * does not fit in the user's buffer, but this seems to be the
6456 * only way to honor MSG_DONTWAIT realistically.
6457 */
6458 sctp_release_sock(sk);
6459 *timeo_p = schedule_timeout(*timeo_p);
6460 sctp_lock_sock(sk);
6461
6462 ready:
6463 finish_wait(sk_sleep(sk), &wait);
6464 return 0;
6465
6466 interrupted:
6467 error = sock_intr_errno(*timeo_p);
6468
6469 out:
6470 finish_wait(sk_sleep(sk), &wait);
6471 *err = error;
6472 return error;
6473 }
6474
6475 /* Receive a datagram.
6476 * Note: This is pretty much the same routine as in core/datagram.c
6477 * with a few changes to make lksctp work.
6478 */
6479 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6480 int noblock, int *err)
6481 {
6482 int error;
6483 struct sk_buff *skb;
6484 long timeo;
6485
6486 timeo = sock_rcvtimeo(sk, noblock);
6487
6488 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
6489 MAX_SCHEDULE_TIMEOUT);
6490
6491 do {
6492 /* Again only user level code calls this function,
6493 * so nothing interrupt level
6494 * will suddenly eat the receive_queue.
6495 *
6496 * Look at current nfs client by the way...
6497 * However, this function was correct in any case. 8)
6498 */
6499 if (flags & MSG_PEEK) {
6500 spin_lock_bh(&sk->sk_receive_queue.lock);
6501 skb = skb_peek(&sk->sk_receive_queue);
6502 if (skb)
6503 atomic_inc(&skb->users);
6504 spin_unlock_bh(&sk->sk_receive_queue.lock);
6505 } else {
6506 skb = skb_dequeue(&sk->sk_receive_queue);
6507 }
6508
6509 if (skb)
6510 return skb;
6511
6512 /* Caller is allowed not to check sk->sk_err before calling. */
6513 error = sock_error(sk);
6514 if (error)
6515 goto no_packet;
6516
6517 if (sk->sk_shutdown & RCV_SHUTDOWN)
6518 break;
6519
6520 /* User doesn't want to wait. */
6521 error = -EAGAIN;
6522 if (!timeo)
6523 goto no_packet;
6524 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6525
6526 return NULL;
6527
6528 no_packet:
6529 *err = error;
6530 return NULL;
6531 }
6532
6533 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6534 static void __sctp_write_space(struct sctp_association *asoc)
6535 {
6536 struct sock *sk = asoc->base.sk;
6537 struct socket *sock = sk->sk_socket;
6538
6539 if ((sctp_wspace(asoc) > 0) && sock) {
6540 if (waitqueue_active(&asoc->wait))
6541 wake_up_interruptible(&asoc->wait);
6542
6543 if (sctp_writeable(sk)) {
6544 wait_queue_head_t *wq = sk_sleep(sk);
6545
6546 if (wq && waitqueue_active(wq))
6547 wake_up_interruptible(wq);
6548
6549 /* Note that we try to include the Async I/O support
6550 * here by modeling from the current TCP/UDP code.
6551 * We have not tested with it yet.
6552 */
6553 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6554 sock_wake_async(sock,
6555 SOCK_WAKE_SPACE, POLL_OUT);
6556 }
6557 }
6558 }
6559
6560 /* Do accounting for the sndbuf space.
6561 * Decrement the used sndbuf space of the corresponding association by the
6562 * data size which was just transmitted(freed).
6563 */
6564 static void sctp_wfree(struct sk_buff *skb)
6565 {
6566 struct sctp_association *asoc;
6567 struct sctp_chunk *chunk;
6568 struct sock *sk;
6569
6570 /* Get the saved chunk pointer. */
6571 chunk = *((struct sctp_chunk **)(skb->cb));
6572 asoc = chunk->asoc;
6573 sk = asoc->base.sk;
6574 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6575 sizeof(struct sk_buff) +
6576 sizeof(struct sctp_chunk);
6577
6578 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6579
6580 /*
6581 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6582 */
6583 sk->sk_wmem_queued -= skb->truesize;
6584 sk_mem_uncharge(sk, skb->truesize);
6585
6586 sock_wfree(skb);
6587 __sctp_write_space(asoc);
6588
6589 sctp_association_put(asoc);
6590 }
6591
6592 /* Do accounting for the receive space on the socket.
6593 * Accounting for the association is done in ulpevent.c
6594 * We set this as a destructor for the cloned data skbs so that
6595 * accounting is done at the correct time.
6596 */
6597 void sctp_sock_rfree(struct sk_buff *skb)
6598 {
6599 struct sock *sk = skb->sk;
6600 struct sctp_ulpevent *event = sctp_skb2event(skb);
6601
6602 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6603
6604 /*
6605 * Mimic the behavior of sock_rfree
6606 */
6607 sk_mem_uncharge(sk, event->rmem_len);
6608 }
6609
6610
6611 /* Helper function to wait for space in the sndbuf. */
6612 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6613 size_t msg_len)
6614 {
6615 struct sock *sk = asoc->base.sk;
6616 int err = 0;
6617 long current_timeo = *timeo_p;
6618 DEFINE_WAIT(wait);
6619
6620 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
6621 *timeo_p, msg_len);
6622
6623 /* Increment the association's refcnt. */
6624 sctp_association_hold(asoc);
6625
6626 /* Wait on the association specific sndbuf space. */
6627 for (;;) {
6628 prepare_to_wait_exclusive(&asoc->wait, &wait,
6629 TASK_INTERRUPTIBLE);
6630 if (!*timeo_p)
6631 goto do_nonblock;
6632 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6633 asoc->base.dead)
6634 goto do_error;
6635 if (signal_pending(current))
6636 goto do_interrupted;
6637 if (msg_len <= sctp_wspace(asoc))
6638 break;
6639
6640 /* Let another process have a go. Since we are going
6641 * to sleep anyway.
6642 */
6643 sctp_release_sock(sk);
6644 current_timeo = schedule_timeout(current_timeo);
6645 BUG_ON(sk != asoc->base.sk);
6646 sctp_lock_sock(sk);
6647
6648 *timeo_p = current_timeo;
6649 }
6650
6651 out:
6652 finish_wait(&asoc->wait, &wait);
6653
6654 /* Release the association's refcnt. */
6655 sctp_association_put(asoc);
6656
6657 return err;
6658
6659 do_error:
6660 err = -EPIPE;
6661 goto out;
6662
6663 do_interrupted:
6664 err = sock_intr_errno(*timeo_p);
6665 goto out;
6666
6667 do_nonblock:
6668 err = -EAGAIN;
6669 goto out;
6670 }
6671
6672 void sctp_data_ready(struct sock *sk, int len)
6673 {
6674 struct socket_wq *wq;
6675
6676 rcu_read_lock();
6677 wq = rcu_dereference(sk->sk_wq);
6678 if (wq_has_sleeper(wq))
6679 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6680 POLLRDNORM | POLLRDBAND);
6681 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6682 rcu_read_unlock();
6683 }
6684
6685 /* If socket sndbuf has changed, wake up all per association waiters. */
6686 void sctp_write_space(struct sock *sk)
6687 {
6688 struct sctp_association *asoc;
6689
6690 /* Wake up the tasks in each wait queue. */
6691 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6692 __sctp_write_space(asoc);
6693 }
6694 }
6695
6696 /* Is there any sndbuf space available on the socket?
6697 *
6698 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6699 * associations on the same socket. For a UDP-style socket with
6700 * multiple associations, it is possible for it to be "unwriteable"
6701 * prematurely. I assume that this is acceptable because
6702 * a premature "unwriteable" is better than an accidental "writeable" which
6703 * would cause an unwanted block under certain circumstances. For the 1-1
6704 * UDP-style sockets or TCP-style sockets, this code should work.
6705 * - Daisy
6706 */
6707 static int sctp_writeable(struct sock *sk)
6708 {
6709 int amt = 0;
6710
6711 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6712 if (amt < 0)
6713 amt = 0;
6714 return amt;
6715 }
6716
6717 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6718 * returns immediately with EINPROGRESS.
6719 */
6720 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6721 {
6722 struct sock *sk = asoc->base.sk;
6723 int err = 0;
6724 long current_timeo = *timeo_p;
6725 DEFINE_WAIT(wait);
6726
6727 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
6728
6729 /* Increment the association's refcnt. */
6730 sctp_association_hold(asoc);
6731
6732 for (;;) {
6733 prepare_to_wait_exclusive(&asoc->wait, &wait,
6734 TASK_INTERRUPTIBLE);
6735 if (!*timeo_p)
6736 goto do_nonblock;
6737 if (sk->sk_shutdown & RCV_SHUTDOWN)
6738 break;
6739 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6740 asoc->base.dead)
6741 goto do_error;
6742 if (signal_pending(current))
6743 goto do_interrupted;
6744
6745 if (sctp_state(asoc, ESTABLISHED))
6746 break;
6747
6748 /* Let another process have a go. Since we are going
6749 * to sleep anyway.
6750 */
6751 sctp_release_sock(sk);
6752 current_timeo = schedule_timeout(current_timeo);
6753 sctp_lock_sock(sk);
6754
6755 *timeo_p = current_timeo;
6756 }
6757
6758 out:
6759 finish_wait(&asoc->wait, &wait);
6760
6761 /* Release the association's refcnt. */
6762 sctp_association_put(asoc);
6763
6764 return err;
6765
6766 do_error:
6767 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6768 err = -ETIMEDOUT;
6769 else
6770 err = -ECONNREFUSED;
6771 goto out;
6772
6773 do_interrupted:
6774 err = sock_intr_errno(*timeo_p);
6775 goto out;
6776
6777 do_nonblock:
6778 err = -EINPROGRESS;
6779 goto out;
6780 }
6781
6782 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6783 {
6784 struct sctp_endpoint *ep;
6785 int err = 0;
6786 DEFINE_WAIT(wait);
6787
6788 ep = sctp_sk(sk)->ep;
6789
6790
6791 for (;;) {
6792 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6793 TASK_INTERRUPTIBLE);
6794
6795 if (list_empty(&ep->asocs)) {
6796 sctp_release_sock(sk);
6797 timeo = schedule_timeout(timeo);
6798 sctp_lock_sock(sk);
6799 }
6800
6801 err = -EINVAL;
6802 if (!sctp_sstate(sk, LISTENING))
6803 break;
6804
6805 err = 0;
6806 if (!list_empty(&ep->asocs))
6807 break;
6808
6809 err = sock_intr_errno(timeo);
6810 if (signal_pending(current))
6811 break;
6812
6813 err = -EAGAIN;
6814 if (!timeo)
6815 break;
6816 }
6817
6818 finish_wait(sk_sleep(sk), &wait);
6819
6820 return err;
6821 }
6822
6823 static void sctp_wait_for_close(struct sock *sk, long timeout)
6824 {
6825 DEFINE_WAIT(wait);
6826
6827 do {
6828 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6829 if (list_empty(&sctp_sk(sk)->ep->asocs))
6830 break;
6831 sctp_release_sock(sk);
6832 timeout = schedule_timeout(timeout);
6833 sctp_lock_sock(sk);
6834 } while (!signal_pending(current) && timeout);
6835
6836 finish_wait(sk_sleep(sk), &wait);
6837 }
6838
6839 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6840 {
6841 struct sk_buff *frag;
6842
6843 if (!skb->data_len)
6844 goto done;
6845
6846 /* Don't forget the fragments. */
6847 skb_walk_frags(skb, frag)
6848 sctp_skb_set_owner_r_frag(frag, sk);
6849
6850 done:
6851 sctp_skb_set_owner_r(skb, sk);
6852 }
6853
6854 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6855 struct sctp_association *asoc)
6856 {
6857 struct inet_sock *inet = inet_sk(sk);
6858 struct inet_sock *newinet;
6859
6860 newsk->sk_type = sk->sk_type;
6861 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6862 newsk->sk_flags = sk->sk_flags;
6863 newsk->sk_no_check = sk->sk_no_check;
6864 newsk->sk_reuse = sk->sk_reuse;
6865
6866 newsk->sk_shutdown = sk->sk_shutdown;
6867 newsk->sk_destruct = sctp_destruct_sock;
6868 newsk->sk_family = sk->sk_family;
6869 newsk->sk_protocol = IPPROTO_SCTP;
6870 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6871 newsk->sk_sndbuf = sk->sk_sndbuf;
6872 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6873 newsk->sk_lingertime = sk->sk_lingertime;
6874 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6875 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6876
6877 newinet = inet_sk(newsk);
6878
6879 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6880 * getsockname() and getpeername()
6881 */
6882 newinet->inet_sport = inet->inet_sport;
6883 newinet->inet_saddr = inet->inet_saddr;
6884 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6885 newinet->inet_dport = htons(asoc->peer.port);
6886 newinet->pmtudisc = inet->pmtudisc;
6887 newinet->inet_id = asoc->next_tsn ^ jiffies;
6888
6889 newinet->uc_ttl = inet->uc_ttl;
6890 newinet->mc_loop = 1;
6891 newinet->mc_ttl = 1;
6892 newinet->mc_index = 0;
6893 newinet->mc_list = NULL;
6894 }
6895
6896 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6897 * and its messages to the newsk.
6898 */
6899 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6900 struct sctp_association *assoc,
6901 sctp_socket_type_t type)
6902 {
6903 struct sctp_sock *oldsp = sctp_sk(oldsk);
6904 struct sctp_sock *newsp = sctp_sk(newsk);
6905 struct sctp_bind_bucket *pp; /* hash list port iterator */
6906 struct sctp_endpoint *newep = newsp->ep;
6907 struct sk_buff *skb, *tmp;
6908 struct sctp_ulpevent *event;
6909 struct sctp_bind_hashbucket *head;
6910 struct list_head tmplist;
6911
6912 /* Migrate socket buffer sizes and all the socket level options to the
6913 * new socket.
6914 */
6915 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6916 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6917 /* Brute force copy old sctp opt. */
6918 if (oldsp->do_auto_asconf) {
6919 memcpy(&tmplist, &newsp->auto_asconf_list, sizeof(tmplist));
6920 inet_sk_copy_descendant(newsk, oldsk);
6921 memcpy(&newsp->auto_asconf_list, &tmplist, sizeof(tmplist));
6922 } else
6923 inet_sk_copy_descendant(newsk, oldsk);
6924
6925 /* Restore the ep value that was overwritten with the above structure
6926 * copy.
6927 */
6928 newsp->ep = newep;
6929 newsp->hmac = NULL;
6930
6931 /* Hook this new socket in to the bind_hash list. */
6932 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
6933 inet_sk(oldsk)->inet_num)];
6934 sctp_local_bh_disable();
6935 sctp_spin_lock(&head->lock);
6936 pp = sctp_sk(oldsk)->bind_hash;
6937 sk_add_bind_node(newsk, &pp->owner);
6938 sctp_sk(newsk)->bind_hash = pp;
6939 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6940 sctp_spin_unlock(&head->lock);
6941 sctp_local_bh_enable();
6942
6943 /* Copy the bind_addr list from the original endpoint to the new
6944 * endpoint so that we can handle restarts properly
6945 */
6946 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6947 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6948
6949 /* Move any messages in the old socket's receive queue that are for the
6950 * peeled off association to the new socket's receive queue.
6951 */
6952 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6953 event = sctp_skb2event(skb);
6954 if (event->asoc == assoc) {
6955 __skb_unlink(skb, &oldsk->sk_receive_queue);
6956 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6957 sctp_skb_set_owner_r_frag(skb, newsk);
6958 }
6959 }
6960
6961 /* Clean up any messages pending delivery due to partial
6962 * delivery. Three cases:
6963 * 1) No partial deliver; no work.
6964 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6965 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6966 */
6967 skb_queue_head_init(&newsp->pd_lobby);
6968 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6969
6970 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6971 struct sk_buff_head *queue;
6972
6973 /* Decide which queue to move pd_lobby skbs to. */
6974 if (assoc->ulpq.pd_mode) {
6975 queue = &newsp->pd_lobby;
6976 } else
6977 queue = &newsk->sk_receive_queue;
6978
6979 /* Walk through the pd_lobby, looking for skbs that
6980 * need moved to the new socket.
6981 */
6982 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6983 event = sctp_skb2event(skb);
6984 if (event->asoc == assoc) {
6985 __skb_unlink(skb, &oldsp->pd_lobby);
6986 __skb_queue_tail(queue, skb);
6987 sctp_skb_set_owner_r_frag(skb, newsk);
6988 }
6989 }
6990
6991 /* Clear up any skbs waiting for the partial
6992 * delivery to finish.
6993 */
6994 if (assoc->ulpq.pd_mode)
6995 sctp_clear_pd(oldsk, NULL);
6996
6997 }
6998
6999 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
7000 sctp_skb_set_owner_r_frag(skb, newsk);
7001
7002 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
7003 sctp_skb_set_owner_r_frag(skb, newsk);
7004
7005 /* Set the type of socket to indicate that it is peeled off from the
7006 * original UDP-style socket or created with the accept() call on a
7007 * TCP-style socket..
7008 */
7009 newsp->type = type;
7010
7011 /* Mark the new socket "in-use" by the user so that any packets
7012 * that may arrive on the association after we've moved it are
7013 * queued to the backlog. This prevents a potential race between
7014 * backlog processing on the old socket and new-packet processing
7015 * on the new socket.
7016 *
7017 * The caller has just allocated newsk so we can guarantee that other
7018 * paths won't try to lock it and then oldsk.
7019 */
7020 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7021 sctp_assoc_migrate(assoc, newsk);
7022
7023 /* If the association on the newsk is already closed before accept()
7024 * is called, set RCV_SHUTDOWN flag.
7025 */
7026 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
7027 newsk->sk_shutdown |= RCV_SHUTDOWN;
7028
7029 newsk->sk_state = SCTP_SS_ESTABLISHED;
7030 sctp_release_sock(newsk);
7031 }
7032
7033
7034 /* This proto struct describes the ULP interface for SCTP. */
7035 struct proto sctp_prot = {
7036 .name = "SCTP",
7037 .owner = THIS_MODULE,
7038 .close = sctp_close,
7039 .connect = sctp_connect,
7040 .disconnect = sctp_disconnect,
7041 .accept = sctp_accept,
7042 .ioctl = sctp_ioctl,
7043 .init = sctp_init_sock,
7044 .destroy = sctp_destroy_sock,
7045 .shutdown = sctp_shutdown,
7046 .setsockopt = sctp_setsockopt,
7047 .getsockopt = sctp_getsockopt,
7048 .sendmsg = sctp_sendmsg,
7049 .recvmsg = sctp_recvmsg,
7050 .bind = sctp_bind,
7051 .backlog_rcv = sctp_backlog_rcv,
7052 .hash = sctp_hash,
7053 .unhash = sctp_unhash,
7054 .get_port = sctp_get_port,
7055 .obj_size = sizeof(struct sctp_sock),
7056 .sysctl_mem = sysctl_sctp_mem,
7057 .sysctl_rmem = sysctl_sctp_rmem,
7058 .sysctl_wmem = sysctl_sctp_wmem,
7059 .memory_pressure = &sctp_memory_pressure,
7060 .enter_memory_pressure = sctp_enter_memory_pressure,
7061 .memory_allocated = &sctp_memory_allocated,
7062 .sockets_allocated = &sctp_sockets_allocated,
7063 };
7064
7065 #if IS_ENABLED(CONFIG_IPV6)
7066
7067 struct proto sctpv6_prot = {
7068 .name = "SCTPv6",
7069 .owner = THIS_MODULE,
7070 .close = sctp_close,
7071 .connect = sctp_connect,
7072 .disconnect = sctp_disconnect,
7073 .accept = sctp_accept,
7074 .ioctl = sctp_ioctl,
7075 .init = sctp_init_sock,
7076 .destroy = sctp_destroy_sock,
7077 .shutdown = sctp_shutdown,
7078 .setsockopt = sctp_setsockopt,
7079 .getsockopt = sctp_getsockopt,
7080 .sendmsg = sctp_sendmsg,
7081 .recvmsg = sctp_recvmsg,
7082 .bind = sctp_bind,
7083 .backlog_rcv = sctp_backlog_rcv,
7084 .hash = sctp_hash,
7085 .unhash = sctp_unhash,
7086 .get_port = sctp_get_port,
7087 .obj_size = sizeof(struct sctp6_sock),
7088 .sysctl_mem = sysctl_sctp_mem,
7089 .sysctl_rmem = sysctl_sctp_rmem,
7090 .sysctl_wmem = sysctl_sctp_wmem,
7091 .memory_pressure = &sctp_memory_pressure,
7092 .enter_memory_pressure = sctp_enter_memory_pressure,
7093 .memory_allocated = &sctp_memory_allocated,
7094 .sockets_allocated = &sctp_sockets_allocated,
7095 };
7096 #endif /* IS_ENABLED(CONFIG_IPV6) */
This page took 0.478915 seconds and 6 git commands to generate.