Merge tag 'imx-fixes-3.13-2' of git://git.linaro.org/people/shawnguo/linux-2.6 into...
[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, write to
32 * the Free Software Foundation, 59 Temple Place - Suite 330,
33 * Boston, MA 02111-1307, USA.
34 *
35 * Please send any bug reports or fixes you make to the
36 * email address(es):
37 * lksctp developers <linux-sctp@vger.kernel.org>
38 *
39 * Written or modified by:
40 * La Monte H.P. Yarroll <piggy@acm.org>
41 * Narasimha Budihal <narsi@refcode.org>
42 * Karl Knutson <karl@athena.chicago.il.us>
43 * Jon Grimm <jgrimm@us.ibm.com>
44 * Xingang Guo <xingang.guo@intel.com>
45 * Daisy Chang <daisyc@us.ibm.com>
46 * Sridhar Samudrala <samudrala@us.ibm.com>
47 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
48 * Ardelle Fan <ardelle.fan@intel.com>
49 * Ryan Layer <rmlayer@us.ibm.com>
50 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
51 * Kevin Gao <kevin.gao@intel.com>
52 */
53
54 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
55
56 #include <linux/types.h>
57 #include <linux/kernel.h>
58 #include <linux/wait.h>
59 #include <linux/time.h>
60 #include <linux/ip.h>
61 #include <linux/capability.h>
62 #include <linux/fcntl.h>
63 #include <linux/poll.h>
64 #include <linux/init.h>
65 #include <linux/crypto.h>
66 #include <linux/slab.h>
67 #include <linux/file.h>
68
69 #include <net/ip.h>
70 #include <net/icmp.h>
71 #include <net/route.h>
72 #include <net/ipv6.h>
73 #include <net/inet_common.h>
74
75 #include <linux/socket.h> /* for sa_family_t */
76 #include <linux/export.h>
77 #include <net/sock.h>
78 #include <net/sctp/sctp.h>
79 #include <net/sctp/sm.h>
80
81 /* Forward declarations for internal helper functions. */
82 static int sctp_writeable(struct sock *sk);
83 static void sctp_wfree(struct sk_buff *skb);
84 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
85 size_t msg_len);
86 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
87 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
88 static int sctp_wait_for_accept(struct sock *sk, long timeo);
89 static void sctp_wait_for_close(struct sock *sk, long timeo);
90 static void sctp_destruct_sock(struct sock *sk);
91 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
92 union sctp_addr *addr, int len);
93 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
94 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
95 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
96 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
97 static int sctp_send_asconf(struct sctp_association *asoc,
98 struct sctp_chunk *chunk);
99 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
100 static int sctp_autobind(struct sock *sk);
101 static void sctp_sock_migrate(struct sock *, struct sock *,
102 struct sctp_association *, sctp_socket_type_t);
103
104 extern struct kmem_cache *sctp_bucket_cachep;
105 extern long sysctl_sctp_mem[3];
106 extern int sysctl_sctp_rmem[3];
107 extern int sysctl_sctp_wmem[3];
108
109 static int sctp_memory_pressure;
110 static atomic_long_t sctp_memory_allocated;
111 struct percpu_counter sctp_sockets_allocated;
112
113 static void sctp_enter_memory_pressure(struct sock *sk)
114 {
115 sctp_memory_pressure = 1;
116 }
117
118
119 /* Get the sndbuf space available at the time on the association. */
120 static inline int sctp_wspace(struct sctp_association *asoc)
121 {
122 int amt;
123
124 if (asoc->ep->sndbuf_policy)
125 amt = asoc->sndbuf_used;
126 else
127 amt = sk_wmem_alloc_get(asoc->base.sk);
128
129 if (amt >= asoc->base.sk->sk_sndbuf) {
130 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
131 amt = 0;
132 else {
133 amt = sk_stream_wspace(asoc->base.sk);
134 if (amt < 0)
135 amt = 0;
136 }
137 } else {
138 amt = asoc->base.sk->sk_sndbuf - amt;
139 }
140 return amt;
141 }
142
143 /* Increment the used sndbuf space count of the corresponding association by
144 * the size of the outgoing data chunk.
145 * Also, set the skb destructor for sndbuf accounting later.
146 *
147 * Since it is always 1-1 between chunk and skb, and also a new skb is always
148 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
149 * destructor in the data chunk skb for the purpose of the sndbuf space
150 * tracking.
151 */
152 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
153 {
154 struct sctp_association *asoc = chunk->asoc;
155 struct sock *sk = asoc->base.sk;
156
157 /* The sndbuf space is tracked per association. */
158 sctp_association_hold(asoc);
159
160 skb_set_owner_w(chunk->skb, sk);
161
162 chunk->skb->destructor = sctp_wfree;
163 /* Save the chunk pointer in skb for sctp_wfree to use later. */
164 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
165
166 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
167 sizeof(struct sk_buff) +
168 sizeof(struct sctp_chunk);
169
170 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
171 sk->sk_wmem_queued += chunk->skb->truesize;
172 sk_mem_charge(sk, chunk->skb->truesize);
173 }
174
175 /* Verify that this is a valid address. */
176 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
177 int len)
178 {
179 struct sctp_af *af;
180
181 /* Verify basic sockaddr. */
182 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
183 if (!af)
184 return -EINVAL;
185
186 /* Is this a valid SCTP address? */
187 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
188 return -EINVAL;
189
190 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
191 return -EINVAL;
192
193 return 0;
194 }
195
196 /* Look up the association by its id. If this is not a UDP-style
197 * socket, the ID field is always ignored.
198 */
199 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
200 {
201 struct sctp_association *asoc = NULL;
202
203 /* If this is not a UDP-style socket, assoc id should be ignored. */
204 if (!sctp_style(sk, UDP)) {
205 /* Return NULL if the socket state is not ESTABLISHED. It
206 * could be a TCP-style listening socket or a socket which
207 * hasn't yet called connect() to establish an association.
208 */
209 if (!sctp_sstate(sk, ESTABLISHED))
210 return NULL;
211
212 /* Get the first and the only association from the list. */
213 if (!list_empty(&sctp_sk(sk)->ep->asocs))
214 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
215 struct sctp_association, asocs);
216 return asoc;
217 }
218
219 /* Otherwise this is a UDP-style socket. */
220 if (!id || (id == (sctp_assoc_t)-1))
221 return NULL;
222
223 spin_lock_bh(&sctp_assocs_id_lock);
224 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
225 spin_unlock_bh(&sctp_assocs_id_lock);
226
227 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
228 return NULL;
229
230 return asoc;
231 }
232
233 /* Look up the transport from an address and an assoc id. If both address and
234 * id are specified, the associations matching the address and the id should be
235 * the same.
236 */
237 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
238 struct sockaddr_storage *addr,
239 sctp_assoc_t id)
240 {
241 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
242 struct sctp_transport *transport;
243 union sctp_addr *laddr = (union sctp_addr *)addr;
244
245 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
246 laddr,
247 &transport);
248
249 if (!addr_asoc)
250 return NULL;
251
252 id_asoc = sctp_id2assoc(sk, id);
253 if (id_asoc && (id_asoc != addr_asoc))
254 return NULL;
255
256 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
257 (union sctp_addr *)addr);
258
259 return transport;
260 }
261
262 /* API 3.1.2 bind() - UDP Style Syntax
263 * The syntax of bind() is,
264 *
265 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
266 *
267 * sd - the socket descriptor returned by socket().
268 * addr - the address structure (struct sockaddr_in or struct
269 * sockaddr_in6 [RFC 2553]),
270 * addr_len - the size of the address structure.
271 */
272 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
273 {
274 int retval = 0;
275
276 sctp_lock_sock(sk);
277
278 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
279 addr, addr_len);
280
281 /* Disallow binding twice. */
282 if (!sctp_sk(sk)->ep->base.bind_addr.port)
283 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
284 addr_len);
285 else
286 retval = -EINVAL;
287
288 sctp_release_sock(sk);
289
290 return retval;
291 }
292
293 static long sctp_get_port_local(struct sock *, union sctp_addr *);
294
295 /* Verify this is a valid sockaddr. */
296 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
297 union sctp_addr *addr, int len)
298 {
299 struct sctp_af *af;
300
301 /* Check minimum size. */
302 if (len < sizeof (struct sockaddr))
303 return NULL;
304
305 /* V4 mapped address are really of AF_INET family */
306 if (addr->sa.sa_family == AF_INET6 &&
307 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
308 if (!opt->pf->af_supported(AF_INET, opt))
309 return NULL;
310 } else {
311 /* Does this PF support this AF? */
312 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
313 return NULL;
314 }
315
316 /* If we get this far, af is valid. */
317 af = sctp_get_af_specific(addr->sa.sa_family);
318
319 if (len < af->sockaddr_len)
320 return NULL;
321
322 return af;
323 }
324
325 /* Bind a local address either to an endpoint or to an association. */
326 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
327 {
328 struct net *net = sock_net(sk);
329 struct sctp_sock *sp = sctp_sk(sk);
330 struct sctp_endpoint *ep = sp->ep;
331 struct sctp_bind_addr *bp = &ep->base.bind_addr;
332 struct sctp_af *af;
333 unsigned short snum;
334 int ret = 0;
335
336 /* Common sockaddr verification. */
337 af = sctp_sockaddr_af(sp, addr, len);
338 if (!af) {
339 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
340 __func__, sk, addr, len);
341 return -EINVAL;
342 }
343
344 snum = ntohs(addr->v4.sin_port);
345
346 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
347 __func__, sk, &addr->sa, bp->port, snum, len);
348
349 /* PF specific bind() address verification. */
350 if (!sp->pf->bind_verify(sp, addr))
351 return -EADDRNOTAVAIL;
352
353 /* We must either be unbound, or bind to the same port.
354 * It's OK to allow 0 ports if we are already bound.
355 * We'll just inhert an already bound port in this case
356 */
357 if (bp->port) {
358 if (!snum)
359 snum = bp->port;
360 else if (snum != bp->port) {
361 pr_debug("%s: new port %d doesn't match existing port "
362 "%d\n", __func__, snum, bp->port);
363 return -EINVAL;
364 }
365 }
366
367 if (snum && snum < PROT_SOCK &&
368 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
369 return -EACCES;
370
371 /* See if the address matches any of the addresses we may have
372 * already bound before checking against other endpoints.
373 */
374 if (sctp_bind_addr_match(bp, addr, sp))
375 return -EINVAL;
376
377 /* Make sure we are allowed to bind here.
378 * The function sctp_get_port_local() does duplicate address
379 * detection.
380 */
381 addr->v4.sin_port = htons(snum);
382 if ((ret = sctp_get_port_local(sk, addr))) {
383 return -EADDRINUSE;
384 }
385
386 /* Refresh ephemeral port. */
387 if (!bp->port)
388 bp->port = inet_sk(sk)->inet_num;
389
390 /* Add the address to the bind address list.
391 * Use GFP_ATOMIC since BHs will be disabled.
392 */
393 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
394
395 /* Copy back into socket for getsockname() use. */
396 if (!ret) {
397 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
398 af->to_sk_saddr(addr, sk);
399 }
400
401 return ret;
402 }
403
404 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
405 *
406 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
407 * at any one time. If a sender, after sending an ASCONF chunk, decides
408 * it needs to transfer another ASCONF Chunk, it MUST wait until the
409 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
410 * subsequent ASCONF. Note this restriction binds each side, so at any
411 * time two ASCONF may be in-transit on any given association (one sent
412 * from each endpoint).
413 */
414 static int sctp_send_asconf(struct sctp_association *asoc,
415 struct sctp_chunk *chunk)
416 {
417 struct net *net = sock_net(asoc->base.sk);
418 int retval = 0;
419
420 /* If there is an outstanding ASCONF chunk, queue it for later
421 * transmission.
422 */
423 if (asoc->addip_last_asconf) {
424 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
425 goto out;
426 }
427
428 /* Hold the chunk until an ASCONF_ACK is received. */
429 sctp_chunk_hold(chunk);
430 retval = sctp_primitive_ASCONF(net, asoc, chunk);
431 if (retval)
432 sctp_chunk_free(chunk);
433 else
434 asoc->addip_last_asconf = chunk;
435
436 out:
437 return retval;
438 }
439
440 /* Add a list of addresses as bind addresses to local endpoint or
441 * association.
442 *
443 * Basically run through each address specified in the addrs/addrcnt
444 * array/length pair, determine if it is IPv6 or IPv4 and call
445 * sctp_do_bind() on it.
446 *
447 * If any of them fails, then the operation will be reversed and the
448 * ones that were added will be removed.
449 *
450 * Only sctp_setsockopt_bindx() is supposed to call this function.
451 */
452 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
453 {
454 int cnt;
455 int retval = 0;
456 void *addr_buf;
457 struct sockaddr *sa_addr;
458 struct sctp_af *af;
459
460 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
461 addrs, addrcnt);
462
463 addr_buf = addrs;
464 for (cnt = 0; cnt < addrcnt; cnt++) {
465 /* The list may contain either IPv4 or IPv6 address;
466 * determine the address length for walking thru the list.
467 */
468 sa_addr = addr_buf;
469 af = sctp_get_af_specific(sa_addr->sa_family);
470 if (!af) {
471 retval = -EINVAL;
472 goto err_bindx_add;
473 }
474
475 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
476 af->sockaddr_len);
477
478 addr_buf += af->sockaddr_len;
479
480 err_bindx_add:
481 if (retval < 0) {
482 /* Failed. Cleanup the ones that have been added */
483 if (cnt > 0)
484 sctp_bindx_rem(sk, addrs, cnt);
485 return retval;
486 }
487 }
488
489 return retval;
490 }
491
492 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
493 * associations that are part of the endpoint indicating that a list of local
494 * addresses are added to the endpoint.
495 *
496 * If any of the addresses is already in the bind address list of the
497 * association, we do not send the chunk for that association. But it will not
498 * affect other associations.
499 *
500 * Only sctp_setsockopt_bindx() is supposed to call this function.
501 */
502 static int sctp_send_asconf_add_ip(struct sock *sk,
503 struct sockaddr *addrs,
504 int addrcnt)
505 {
506 struct net *net = sock_net(sk);
507 struct sctp_sock *sp;
508 struct sctp_endpoint *ep;
509 struct sctp_association *asoc;
510 struct sctp_bind_addr *bp;
511 struct sctp_chunk *chunk;
512 struct sctp_sockaddr_entry *laddr;
513 union sctp_addr *addr;
514 union sctp_addr saveaddr;
515 void *addr_buf;
516 struct sctp_af *af;
517 struct list_head *p;
518 int i;
519 int retval = 0;
520
521 if (!net->sctp.addip_enable)
522 return retval;
523
524 sp = sctp_sk(sk);
525 ep = sp->ep;
526
527 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
528 __func__, sk, addrs, addrcnt);
529
530 list_for_each_entry(asoc, &ep->asocs, asocs) {
531 if (!asoc->peer.asconf_capable)
532 continue;
533
534 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
535 continue;
536
537 if (!sctp_state(asoc, ESTABLISHED))
538 continue;
539
540 /* Check if any address in the packed array of addresses is
541 * in the bind address list of the association. If so,
542 * do not send the asconf chunk to its peer, but continue with
543 * other associations.
544 */
545 addr_buf = addrs;
546 for (i = 0; i < addrcnt; i++) {
547 addr = addr_buf;
548 af = sctp_get_af_specific(addr->v4.sin_family);
549 if (!af) {
550 retval = -EINVAL;
551 goto out;
552 }
553
554 if (sctp_assoc_lookup_laddr(asoc, addr))
555 break;
556
557 addr_buf += af->sockaddr_len;
558 }
559 if (i < addrcnt)
560 continue;
561
562 /* Use the first valid address in bind addr list of
563 * association as Address Parameter of ASCONF CHUNK.
564 */
565 bp = &asoc->base.bind_addr;
566 p = bp->address_list.next;
567 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
568 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
569 addrcnt, SCTP_PARAM_ADD_IP);
570 if (!chunk) {
571 retval = -ENOMEM;
572 goto out;
573 }
574
575 /* Add the new addresses to the bind address list with
576 * use_as_src set to 0.
577 */
578 addr_buf = addrs;
579 for (i = 0; i < addrcnt; i++) {
580 addr = addr_buf;
581 af = sctp_get_af_specific(addr->v4.sin_family);
582 memcpy(&saveaddr, addr, af->sockaddr_len);
583 retval = sctp_add_bind_addr(bp, &saveaddr,
584 SCTP_ADDR_NEW, GFP_ATOMIC);
585 addr_buf += af->sockaddr_len;
586 }
587 if (asoc->src_out_of_asoc_ok) {
588 struct sctp_transport *trans;
589
590 list_for_each_entry(trans,
591 &asoc->peer.transport_addr_list, transports) {
592 /* Clear the source and route cache */
593 dst_release(trans->dst);
594 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
595 2*asoc->pathmtu, 4380));
596 trans->ssthresh = asoc->peer.i.a_rwnd;
597 trans->rto = asoc->rto_initial;
598 sctp_max_rto(asoc, trans);
599 trans->rtt = trans->srtt = trans->rttvar = 0;
600 sctp_transport_route(trans, NULL,
601 sctp_sk(asoc->base.sk));
602 }
603 }
604 retval = sctp_send_asconf(asoc, chunk);
605 }
606
607 out:
608 return retval;
609 }
610
611 /* Remove a list of addresses from bind addresses list. Do not remove the
612 * last address.
613 *
614 * Basically run through each address specified in the addrs/addrcnt
615 * array/length pair, determine if it is IPv6 or IPv4 and call
616 * sctp_del_bind() on it.
617 *
618 * If any of them fails, then the operation will be reversed and the
619 * ones that were removed will be added back.
620 *
621 * At least one address has to be left; if only one address is
622 * available, the operation will return -EBUSY.
623 *
624 * Only sctp_setsockopt_bindx() is supposed to call this function.
625 */
626 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
627 {
628 struct sctp_sock *sp = sctp_sk(sk);
629 struct sctp_endpoint *ep = sp->ep;
630 int cnt;
631 struct sctp_bind_addr *bp = &ep->base.bind_addr;
632 int retval = 0;
633 void *addr_buf;
634 union sctp_addr *sa_addr;
635 struct sctp_af *af;
636
637 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
638 __func__, sk, addrs, addrcnt);
639
640 addr_buf = addrs;
641 for (cnt = 0; cnt < addrcnt; cnt++) {
642 /* If the bind address list is empty or if there is only one
643 * bind address, there is nothing more to be removed (we need
644 * at least one address here).
645 */
646 if (list_empty(&bp->address_list) ||
647 (sctp_list_single_entry(&bp->address_list))) {
648 retval = -EBUSY;
649 goto err_bindx_rem;
650 }
651
652 sa_addr = addr_buf;
653 af = sctp_get_af_specific(sa_addr->sa.sa_family);
654 if (!af) {
655 retval = -EINVAL;
656 goto err_bindx_rem;
657 }
658
659 if (!af->addr_valid(sa_addr, sp, NULL)) {
660 retval = -EADDRNOTAVAIL;
661 goto err_bindx_rem;
662 }
663
664 if (sa_addr->v4.sin_port &&
665 sa_addr->v4.sin_port != htons(bp->port)) {
666 retval = -EINVAL;
667 goto err_bindx_rem;
668 }
669
670 if (!sa_addr->v4.sin_port)
671 sa_addr->v4.sin_port = htons(bp->port);
672
673 /* FIXME - There is probably a need to check if sk->sk_saddr and
674 * sk->sk_rcv_addr are currently set to one of the addresses to
675 * be removed. This is something which needs to be looked into
676 * when we are fixing the outstanding issues with multi-homing
677 * socket routing and failover schemes. Refer to comments in
678 * sctp_do_bind(). -daisy
679 */
680 retval = sctp_del_bind_addr(bp, sa_addr);
681
682 addr_buf += af->sockaddr_len;
683 err_bindx_rem:
684 if (retval < 0) {
685 /* Failed. Add the ones that has been removed back */
686 if (cnt > 0)
687 sctp_bindx_add(sk, addrs, cnt);
688 return retval;
689 }
690 }
691
692 return retval;
693 }
694
695 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
696 * the associations that are part of the endpoint indicating that a list of
697 * local addresses are removed from the endpoint.
698 *
699 * If any of the addresses is already in the bind address list of the
700 * association, we do not send the chunk for that association. But it will not
701 * affect other associations.
702 *
703 * Only sctp_setsockopt_bindx() is supposed to call this function.
704 */
705 static int sctp_send_asconf_del_ip(struct sock *sk,
706 struct sockaddr *addrs,
707 int addrcnt)
708 {
709 struct net *net = sock_net(sk);
710 struct sctp_sock *sp;
711 struct sctp_endpoint *ep;
712 struct sctp_association *asoc;
713 struct sctp_transport *transport;
714 struct sctp_bind_addr *bp;
715 struct sctp_chunk *chunk;
716 union sctp_addr *laddr;
717 void *addr_buf;
718 struct sctp_af *af;
719 struct sctp_sockaddr_entry *saddr;
720 int i;
721 int retval = 0;
722 int stored = 0;
723
724 chunk = NULL;
725 if (!net->sctp.addip_enable)
726 return retval;
727
728 sp = sctp_sk(sk);
729 ep = sp->ep;
730
731 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
732 __func__, sk, addrs, addrcnt);
733
734 list_for_each_entry(asoc, &ep->asocs, asocs) {
735
736 if (!asoc->peer.asconf_capable)
737 continue;
738
739 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
740 continue;
741
742 if (!sctp_state(asoc, ESTABLISHED))
743 continue;
744
745 /* Check if any address in the packed array of addresses is
746 * not present in the bind address list of the association.
747 * If so, do not send the asconf chunk to its peer, but
748 * continue with other associations.
749 */
750 addr_buf = addrs;
751 for (i = 0; i < addrcnt; i++) {
752 laddr = addr_buf;
753 af = sctp_get_af_specific(laddr->v4.sin_family);
754 if (!af) {
755 retval = -EINVAL;
756 goto out;
757 }
758
759 if (!sctp_assoc_lookup_laddr(asoc, laddr))
760 break;
761
762 addr_buf += af->sockaddr_len;
763 }
764 if (i < addrcnt)
765 continue;
766
767 /* Find one address in the association's bind address list
768 * that is not in the packed array of addresses. This is to
769 * make sure that we do not delete all the addresses in the
770 * association.
771 */
772 bp = &asoc->base.bind_addr;
773 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
774 addrcnt, sp);
775 if ((laddr == NULL) && (addrcnt == 1)) {
776 if (asoc->asconf_addr_del_pending)
777 continue;
778 asoc->asconf_addr_del_pending =
779 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
780 if (asoc->asconf_addr_del_pending == NULL) {
781 retval = -ENOMEM;
782 goto out;
783 }
784 asoc->asconf_addr_del_pending->sa.sa_family =
785 addrs->sa_family;
786 asoc->asconf_addr_del_pending->v4.sin_port =
787 htons(bp->port);
788 if (addrs->sa_family == AF_INET) {
789 struct sockaddr_in *sin;
790
791 sin = (struct sockaddr_in *)addrs;
792 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
793 } else if (addrs->sa_family == AF_INET6) {
794 struct sockaddr_in6 *sin6;
795
796 sin6 = (struct sockaddr_in6 *)addrs;
797 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
798 }
799
800 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
801 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
802 asoc->asconf_addr_del_pending);
803
804 asoc->src_out_of_asoc_ok = 1;
805 stored = 1;
806 goto skip_mkasconf;
807 }
808
809 if (laddr == NULL)
810 return -EINVAL;
811
812 /* We do not need RCU protection throughout this loop
813 * because this is done under a socket lock from the
814 * setsockopt call.
815 */
816 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
817 SCTP_PARAM_DEL_IP);
818 if (!chunk) {
819 retval = -ENOMEM;
820 goto out;
821 }
822
823 skip_mkasconf:
824 /* Reset use_as_src flag for the addresses in the bind address
825 * list that are to be deleted.
826 */
827 addr_buf = addrs;
828 for (i = 0; i < addrcnt; i++) {
829 laddr = addr_buf;
830 af = sctp_get_af_specific(laddr->v4.sin_family);
831 list_for_each_entry(saddr, &bp->address_list, list) {
832 if (sctp_cmp_addr_exact(&saddr->a, laddr))
833 saddr->state = SCTP_ADDR_DEL;
834 }
835 addr_buf += af->sockaddr_len;
836 }
837
838 /* Update the route and saddr entries for all the transports
839 * as some of the addresses in the bind address list are
840 * about to be deleted and cannot be used as source addresses.
841 */
842 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
843 transports) {
844 dst_release(transport->dst);
845 sctp_transport_route(transport, NULL,
846 sctp_sk(asoc->base.sk));
847 }
848
849 if (stored)
850 /* We don't need to transmit ASCONF */
851 continue;
852 retval = sctp_send_asconf(asoc, chunk);
853 }
854 out:
855 return retval;
856 }
857
858 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
859 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
860 {
861 struct sock *sk = sctp_opt2sk(sp);
862 union sctp_addr *addr;
863 struct sctp_af *af;
864
865 /* It is safe to write port space in caller. */
866 addr = &addrw->a;
867 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
868 af = sctp_get_af_specific(addr->sa.sa_family);
869 if (!af)
870 return -EINVAL;
871 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
872 return -EINVAL;
873
874 if (addrw->state == SCTP_ADDR_NEW)
875 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
876 else
877 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
878 }
879
880 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
881 *
882 * API 8.1
883 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
884 * int flags);
885 *
886 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
887 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
888 * or IPv6 addresses.
889 *
890 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
891 * Section 3.1.2 for this usage.
892 *
893 * addrs is a pointer to an array of one or more socket addresses. Each
894 * address is contained in its appropriate structure (i.e. struct
895 * sockaddr_in or struct sockaddr_in6) the family of the address type
896 * must be used to distinguish the address length (note that this
897 * representation is termed a "packed array" of addresses). The caller
898 * specifies the number of addresses in the array with addrcnt.
899 *
900 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
901 * -1, and sets errno to the appropriate error code.
902 *
903 * For SCTP, the port given in each socket address must be the same, or
904 * sctp_bindx() will fail, setting errno to EINVAL.
905 *
906 * The flags parameter is formed from the bitwise OR of zero or more of
907 * the following currently defined flags:
908 *
909 * SCTP_BINDX_ADD_ADDR
910 *
911 * SCTP_BINDX_REM_ADDR
912 *
913 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
914 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
915 * addresses from the association. The two flags are mutually exclusive;
916 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
917 * not remove all addresses from an association; sctp_bindx() will
918 * reject such an attempt with EINVAL.
919 *
920 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
921 * additional addresses with an endpoint after calling bind(). Or use
922 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
923 * socket is associated with so that no new association accepted will be
924 * associated with those addresses. If the endpoint supports dynamic
925 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
926 * endpoint to send the appropriate message to the peer to change the
927 * peers address lists.
928 *
929 * Adding and removing addresses from a connected association is
930 * optional functionality. Implementations that do not support this
931 * functionality should return EOPNOTSUPP.
932 *
933 * Basically do nothing but copying the addresses from user to kernel
934 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
935 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
936 * from userspace.
937 *
938 * We don't use copy_from_user() for optimization: we first do the
939 * sanity checks (buffer size -fast- and access check-healthy
940 * pointer); if all of those succeed, then we can alloc the memory
941 * (expensive operation) needed to copy the data to kernel. Then we do
942 * the copying without checking the user space area
943 * (__copy_from_user()).
944 *
945 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
946 * it.
947 *
948 * sk The sk of the socket
949 * addrs The pointer to the addresses in user land
950 * addrssize Size of the addrs buffer
951 * op Operation to perform (add or remove, see the flags of
952 * sctp_bindx)
953 *
954 * Returns 0 if ok, <0 errno code on error.
955 */
956 static int sctp_setsockopt_bindx(struct sock* sk,
957 struct sockaddr __user *addrs,
958 int addrs_size, int op)
959 {
960 struct sockaddr *kaddrs;
961 int err;
962 int addrcnt = 0;
963 int walk_size = 0;
964 struct sockaddr *sa_addr;
965 void *addr_buf;
966 struct sctp_af *af;
967
968 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
969 __func__, sk, addrs, addrs_size, op);
970
971 if (unlikely(addrs_size <= 0))
972 return -EINVAL;
973
974 /* Check the user passed a healthy pointer. */
975 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
976 return -EFAULT;
977
978 /* Alloc space for the address array in kernel memory. */
979 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
980 if (unlikely(!kaddrs))
981 return -ENOMEM;
982
983 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
984 kfree(kaddrs);
985 return -EFAULT;
986 }
987
988 /* Walk through the addrs buffer and count the number of addresses. */
989 addr_buf = kaddrs;
990 while (walk_size < addrs_size) {
991 if (walk_size + sizeof(sa_family_t) > addrs_size) {
992 kfree(kaddrs);
993 return -EINVAL;
994 }
995
996 sa_addr = addr_buf;
997 af = sctp_get_af_specific(sa_addr->sa_family);
998
999 /* If the address family is not supported or if this address
1000 * causes the address buffer to overflow return EINVAL.
1001 */
1002 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1003 kfree(kaddrs);
1004 return -EINVAL;
1005 }
1006 addrcnt++;
1007 addr_buf += af->sockaddr_len;
1008 walk_size += af->sockaddr_len;
1009 }
1010
1011 /* Do the work. */
1012 switch (op) {
1013 case SCTP_BINDX_ADD_ADDR:
1014 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1015 if (err)
1016 goto out;
1017 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1018 break;
1019
1020 case SCTP_BINDX_REM_ADDR:
1021 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1022 if (err)
1023 goto out;
1024 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1025 break;
1026
1027 default:
1028 err = -EINVAL;
1029 break;
1030 }
1031
1032 out:
1033 kfree(kaddrs);
1034
1035 return err;
1036 }
1037
1038 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1039 *
1040 * Common routine for handling connect() and sctp_connectx().
1041 * Connect will come in with just a single address.
1042 */
1043 static int __sctp_connect(struct sock* sk,
1044 struct sockaddr *kaddrs,
1045 int addrs_size,
1046 sctp_assoc_t *assoc_id)
1047 {
1048 struct net *net = sock_net(sk);
1049 struct sctp_sock *sp;
1050 struct sctp_endpoint *ep;
1051 struct sctp_association *asoc = NULL;
1052 struct sctp_association *asoc2;
1053 struct sctp_transport *transport;
1054 union sctp_addr to;
1055 struct sctp_af *af;
1056 sctp_scope_t scope;
1057 long timeo;
1058 int err = 0;
1059 int addrcnt = 0;
1060 int walk_size = 0;
1061 union sctp_addr *sa_addr = NULL;
1062 void *addr_buf;
1063 unsigned short port;
1064 unsigned int f_flags = 0;
1065
1066 sp = sctp_sk(sk);
1067 ep = sp->ep;
1068
1069 /* connect() cannot be done on a socket that is already in ESTABLISHED
1070 * state - UDP-style peeled off socket or a TCP-style socket that
1071 * is already connected.
1072 * It cannot be done even on a TCP-style listening socket.
1073 */
1074 if (sctp_sstate(sk, ESTABLISHED) ||
1075 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1076 err = -EISCONN;
1077 goto out_free;
1078 }
1079
1080 /* Walk through the addrs buffer and count the number of addresses. */
1081 addr_buf = kaddrs;
1082 while (walk_size < addrs_size) {
1083 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1084 err = -EINVAL;
1085 goto out_free;
1086 }
1087
1088 sa_addr = addr_buf;
1089 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1090
1091 /* If the address family is not supported or if this address
1092 * causes the address buffer to overflow return EINVAL.
1093 */
1094 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1095 err = -EINVAL;
1096 goto out_free;
1097 }
1098
1099 port = ntohs(sa_addr->v4.sin_port);
1100
1101 /* Save current address so we can work with it */
1102 memcpy(&to, sa_addr, af->sockaddr_len);
1103
1104 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1105 if (err)
1106 goto out_free;
1107
1108 /* Make sure the destination port is correctly set
1109 * in all addresses.
1110 */
1111 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1112 err = -EINVAL;
1113 goto out_free;
1114 }
1115
1116 /* Check if there already is a matching association on the
1117 * endpoint (other than the one created here).
1118 */
1119 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1120 if (asoc2 && asoc2 != asoc) {
1121 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1122 err = -EISCONN;
1123 else
1124 err = -EALREADY;
1125 goto out_free;
1126 }
1127
1128 /* If we could not find a matching association on the endpoint,
1129 * make sure that there is no peeled-off association matching
1130 * the peer address even on another socket.
1131 */
1132 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1133 err = -EADDRNOTAVAIL;
1134 goto out_free;
1135 }
1136
1137 if (!asoc) {
1138 /* If a bind() or sctp_bindx() is not called prior to
1139 * an sctp_connectx() call, the system picks an
1140 * ephemeral port and will choose an address set
1141 * equivalent to binding with a wildcard address.
1142 */
1143 if (!ep->base.bind_addr.port) {
1144 if (sctp_autobind(sk)) {
1145 err = -EAGAIN;
1146 goto out_free;
1147 }
1148 } else {
1149 /*
1150 * If an unprivileged user inherits a 1-many
1151 * style socket with open associations on a
1152 * privileged port, it MAY be permitted to
1153 * accept new associations, but it SHOULD NOT
1154 * be permitted to open new associations.
1155 */
1156 if (ep->base.bind_addr.port < PROT_SOCK &&
1157 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1158 err = -EACCES;
1159 goto out_free;
1160 }
1161 }
1162
1163 scope = sctp_scope(&to);
1164 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1165 if (!asoc) {
1166 err = -ENOMEM;
1167 goto out_free;
1168 }
1169
1170 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1171 GFP_KERNEL);
1172 if (err < 0) {
1173 goto out_free;
1174 }
1175
1176 }
1177
1178 /* Prime the peer's transport structures. */
1179 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1180 SCTP_UNKNOWN);
1181 if (!transport) {
1182 err = -ENOMEM;
1183 goto out_free;
1184 }
1185
1186 addrcnt++;
1187 addr_buf += af->sockaddr_len;
1188 walk_size += af->sockaddr_len;
1189 }
1190
1191 /* In case the user of sctp_connectx() wants an association
1192 * id back, assign one now.
1193 */
1194 if (assoc_id) {
1195 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1196 if (err < 0)
1197 goto out_free;
1198 }
1199
1200 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1201 if (err < 0) {
1202 goto out_free;
1203 }
1204
1205 /* Initialize sk's dport and daddr for getpeername() */
1206 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1207 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1208 af->to_sk_daddr(sa_addr, sk);
1209 sk->sk_err = 0;
1210
1211 /* in-kernel sockets don't generally have a file allocated to them
1212 * if all they do is call sock_create_kern().
1213 */
1214 if (sk->sk_socket->file)
1215 f_flags = sk->sk_socket->file->f_flags;
1216
1217 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1218
1219 err = sctp_wait_for_connect(asoc, &timeo);
1220 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1221 *assoc_id = asoc->assoc_id;
1222
1223 /* Don't free association on exit. */
1224 asoc = NULL;
1225
1226 out_free:
1227 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1228 __func__, asoc, kaddrs, err);
1229
1230 if (asoc) {
1231 /* sctp_primitive_ASSOCIATE may have added this association
1232 * To the hash table, try to unhash it, just in case, its a noop
1233 * if it wasn't hashed so we're safe
1234 */
1235 sctp_unhash_established(asoc);
1236 sctp_association_free(asoc);
1237 }
1238 return err;
1239 }
1240
1241 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1242 *
1243 * API 8.9
1244 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1245 * sctp_assoc_t *asoc);
1246 *
1247 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1248 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1249 * or IPv6 addresses.
1250 *
1251 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1252 * Section 3.1.2 for this usage.
1253 *
1254 * addrs is a pointer to an array of one or more socket addresses. Each
1255 * address is contained in its appropriate structure (i.e. struct
1256 * sockaddr_in or struct sockaddr_in6) the family of the address type
1257 * must be used to distengish the address length (note that this
1258 * representation is termed a "packed array" of addresses). The caller
1259 * specifies the number of addresses in the array with addrcnt.
1260 *
1261 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1262 * the association id of the new association. On failure, sctp_connectx()
1263 * returns -1, and sets errno to the appropriate error code. The assoc_id
1264 * is not touched by the kernel.
1265 *
1266 * For SCTP, the port given in each socket address must be the same, or
1267 * sctp_connectx() will fail, setting errno to EINVAL.
1268 *
1269 * An application can use sctp_connectx to initiate an association with
1270 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1271 * allows a caller to specify multiple addresses at which a peer can be
1272 * reached. The way the SCTP stack uses the list of addresses to set up
1273 * the association is implementation dependent. This function only
1274 * specifies that the stack will try to make use of all the addresses in
1275 * the list when needed.
1276 *
1277 * Note that the list of addresses passed in is only used for setting up
1278 * the association. It does not necessarily equal the set of addresses
1279 * the peer uses for the resulting association. If the caller wants to
1280 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1281 * retrieve them after the association has been set up.
1282 *
1283 * Basically do nothing but copying the addresses from user to kernel
1284 * land and invoking either sctp_connectx(). This is used for tunneling
1285 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1286 *
1287 * We don't use copy_from_user() for optimization: we first do the
1288 * sanity checks (buffer size -fast- and access check-healthy
1289 * pointer); if all of those succeed, then we can alloc the memory
1290 * (expensive operation) needed to copy the data to kernel. Then we do
1291 * the copying without checking the user space area
1292 * (__copy_from_user()).
1293 *
1294 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1295 * it.
1296 *
1297 * sk The sk of the socket
1298 * addrs The pointer to the addresses in user land
1299 * addrssize Size of the addrs buffer
1300 *
1301 * Returns >=0 if ok, <0 errno code on error.
1302 */
1303 static int __sctp_setsockopt_connectx(struct sock* sk,
1304 struct sockaddr __user *addrs,
1305 int addrs_size,
1306 sctp_assoc_t *assoc_id)
1307 {
1308 int err = 0;
1309 struct sockaddr *kaddrs;
1310
1311 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1312 __func__, sk, addrs, addrs_size);
1313
1314 if (unlikely(addrs_size <= 0))
1315 return -EINVAL;
1316
1317 /* Check the user passed a healthy pointer. */
1318 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1319 return -EFAULT;
1320
1321 /* Alloc space for the address array in kernel memory. */
1322 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1323 if (unlikely(!kaddrs))
1324 return -ENOMEM;
1325
1326 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1327 err = -EFAULT;
1328 } else {
1329 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1330 }
1331
1332 kfree(kaddrs);
1333
1334 return err;
1335 }
1336
1337 /*
1338 * This is an older interface. It's kept for backward compatibility
1339 * to the option that doesn't provide association id.
1340 */
1341 static int sctp_setsockopt_connectx_old(struct sock* sk,
1342 struct sockaddr __user *addrs,
1343 int addrs_size)
1344 {
1345 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1346 }
1347
1348 /*
1349 * New interface for the API. The since the API is done with a socket
1350 * option, to make it simple we feed back the association id is as a return
1351 * indication to the call. Error is always negative and association id is
1352 * always positive.
1353 */
1354 static int sctp_setsockopt_connectx(struct sock* sk,
1355 struct sockaddr __user *addrs,
1356 int addrs_size)
1357 {
1358 sctp_assoc_t assoc_id = 0;
1359 int err = 0;
1360
1361 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1362
1363 if (err)
1364 return err;
1365 else
1366 return assoc_id;
1367 }
1368
1369 /*
1370 * New (hopefully final) interface for the API.
1371 * We use the sctp_getaddrs_old structure so that use-space library
1372 * can avoid any unnecessary allocations. The only defferent part
1373 * is that we store the actual length of the address buffer into the
1374 * addrs_num structure member. That way we can re-use the existing
1375 * code.
1376 */
1377 static int sctp_getsockopt_connectx3(struct sock* sk, int len,
1378 char __user *optval,
1379 int __user *optlen)
1380 {
1381 struct sctp_getaddrs_old param;
1382 sctp_assoc_t assoc_id = 0;
1383 int err = 0;
1384
1385 if (len < sizeof(param))
1386 return -EINVAL;
1387
1388 if (copy_from_user(&param, optval, sizeof(param)))
1389 return -EFAULT;
1390
1391 err = __sctp_setsockopt_connectx(sk,
1392 (struct sockaddr __user *)param.addrs,
1393 param.addr_num, &assoc_id);
1394
1395 if (err == 0 || err == -EINPROGRESS) {
1396 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1397 return -EFAULT;
1398 if (put_user(sizeof(assoc_id), optlen))
1399 return -EFAULT;
1400 }
1401
1402 return err;
1403 }
1404
1405 /* API 3.1.4 close() - UDP Style Syntax
1406 * Applications use close() to perform graceful shutdown (as described in
1407 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1408 * by a UDP-style socket.
1409 *
1410 * The syntax is
1411 *
1412 * ret = close(int sd);
1413 *
1414 * sd - the socket descriptor of the associations to be closed.
1415 *
1416 * To gracefully shutdown a specific association represented by the
1417 * UDP-style socket, an application should use the sendmsg() call,
1418 * passing no user data, but including the appropriate flag in the
1419 * ancillary data (see Section xxxx).
1420 *
1421 * If sd in the close() call is a branched-off socket representing only
1422 * one association, the shutdown is performed on that association only.
1423 *
1424 * 4.1.6 close() - TCP Style Syntax
1425 *
1426 * Applications use close() to gracefully close down an association.
1427 *
1428 * The syntax is:
1429 *
1430 * int close(int sd);
1431 *
1432 * sd - the socket descriptor of the association to be closed.
1433 *
1434 * After an application calls close() on a socket descriptor, no further
1435 * socket operations will succeed on that descriptor.
1436 *
1437 * API 7.1.4 SO_LINGER
1438 *
1439 * An application using the TCP-style socket can use this option to
1440 * perform the SCTP ABORT primitive. The linger option structure is:
1441 *
1442 * struct linger {
1443 * int l_onoff; // option on/off
1444 * int l_linger; // linger time
1445 * };
1446 *
1447 * To enable the option, set l_onoff to 1. If the l_linger value is set
1448 * to 0, calling close() is the same as the ABORT primitive. If the
1449 * value is set to a negative value, the setsockopt() call will return
1450 * an error. If the value is set to a positive value linger_time, the
1451 * close() can be blocked for at most linger_time ms. If the graceful
1452 * shutdown phase does not finish during this period, close() will
1453 * return but the graceful shutdown phase continues in the system.
1454 */
1455 static void sctp_close(struct sock *sk, long timeout)
1456 {
1457 struct net *net = sock_net(sk);
1458 struct sctp_endpoint *ep;
1459 struct sctp_association *asoc;
1460 struct list_head *pos, *temp;
1461 unsigned int data_was_unread;
1462
1463 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1464
1465 sctp_lock_sock(sk);
1466 sk->sk_shutdown = SHUTDOWN_MASK;
1467 sk->sk_state = SCTP_SS_CLOSING;
1468
1469 ep = sctp_sk(sk)->ep;
1470
1471 /* Clean up any skbs sitting on the receive queue. */
1472 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1473 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1474
1475 /* Walk all associations on an endpoint. */
1476 list_for_each_safe(pos, temp, &ep->asocs) {
1477 asoc = list_entry(pos, struct sctp_association, asocs);
1478
1479 if (sctp_style(sk, TCP)) {
1480 /* A closed association can still be in the list if
1481 * it belongs to a TCP-style listening socket that is
1482 * not yet accepted. If so, free it. If not, send an
1483 * ABORT or SHUTDOWN based on the linger options.
1484 */
1485 if (sctp_state(asoc, CLOSED)) {
1486 sctp_unhash_established(asoc);
1487 sctp_association_free(asoc);
1488 continue;
1489 }
1490 }
1491
1492 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1493 !skb_queue_empty(&asoc->ulpq.reasm) ||
1494 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1495 struct sctp_chunk *chunk;
1496
1497 chunk = sctp_make_abort_user(asoc, NULL, 0);
1498 if (chunk)
1499 sctp_primitive_ABORT(net, asoc, chunk);
1500 } else
1501 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1502 }
1503
1504 /* On a TCP-style socket, block for at most linger_time if set. */
1505 if (sctp_style(sk, TCP) && timeout)
1506 sctp_wait_for_close(sk, timeout);
1507
1508 /* This will run the backlog queue. */
1509 sctp_release_sock(sk);
1510
1511 /* Supposedly, no process has access to the socket, but
1512 * the net layers still may.
1513 */
1514 sctp_local_bh_disable();
1515 sctp_bh_lock_sock(sk);
1516
1517 /* Hold the sock, since sk_common_release() will put sock_put()
1518 * and we have just a little more cleanup.
1519 */
1520 sock_hold(sk);
1521 sk_common_release(sk);
1522
1523 sctp_bh_unlock_sock(sk);
1524 sctp_local_bh_enable();
1525
1526 sock_put(sk);
1527
1528 SCTP_DBG_OBJCNT_DEC(sock);
1529 }
1530
1531 /* Handle EPIPE error. */
1532 static int sctp_error(struct sock *sk, int flags, int err)
1533 {
1534 if (err == -EPIPE)
1535 err = sock_error(sk) ? : -EPIPE;
1536 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1537 send_sig(SIGPIPE, current, 0);
1538 return err;
1539 }
1540
1541 /* API 3.1.3 sendmsg() - UDP Style Syntax
1542 *
1543 * An application uses sendmsg() and recvmsg() calls to transmit data to
1544 * and receive data from its peer.
1545 *
1546 * ssize_t sendmsg(int socket, const struct msghdr *message,
1547 * int flags);
1548 *
1549 * socket - the socket descriptor of the endpoint.
1550 * message - pointer to the msghdr structure which contains a single
1551 * user message and possibly some ancillary data.
1552 *
1553 * See Section 5 for complete description of the data
1554 * structures.
1555 *
1556 * flags - flags sent or received with the user message, see Section
1557 * 5 for complete description of the flags.
1558 *
1559 * Note: This function could use a rewrite especially when explicit
1560 * connect support comes in.
1561 */
1562 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1563
1564 static int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1565
1566 static int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1567 struct msghdr *msg, size_t msg_len)
1568 {
1569 struct net *net = sock_net(sk);
1570 struct sctp_sock *sp;
1571 struct sctp_endpoint *ep;
1572 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1573 struct sctp_transport *transport, *chunk_tp;
1574 struct sctp_chunk *chunk;
1575 union sctp_addr to;
1576 struct sockaddr *msg_name = NULL;
1577 struct sctp_sndrcvinfo default_sinfo;
1578 struct sctp_sndrcvinfo *sinfo;
1579 struct sctp_initmsg *sinit;
1580 sctp_assoc_t associd = 0;
1581 sctp_cmsgs_t cmsgs = { NULL };
1582 int err;
1583 sctp_scope_t scope;
1584 long timeo;
1585 __u16 sinfo_flags = 0;
1586 struct sctp_datamsg *datamsg;
1587 int msg_flags = msg->msg_flags;
1588
1589 err = 0;
1590 sp = sctp_sk(sk);
1591 ep = sp->ep;
1592
1593 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1594 msg, msg_len, ep);
1595
1596 /* We cannot send a message over a TCP-style listening socket. */
1597 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1598 err = -EPIPE;
1599 goto out_nounlock;
1600 }
1601
1602 /* Parse out the SCTP CMSGs. */
1603 err = sctp_msghdr_parse(msg, &cmsgs);
1604 if (err) {
1605 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1606 goto out_nounlock;
1607 }
1608
1609 /* Fetch the destination address for this packet. This
1610 * address only selects the association--it is not necessarily
1611 * the address we will send to.
1612 * For a peeled-off socket, msg_name is ignored.
1613 */
1614 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1615 int msg_namelen = msg->msg_namelen;
1616
1617 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1618 msg_namelen);
1619 if (err)
1620 return err;
1621
1622 if (msg_namelen > sizeof(to))
1623 msg_namelen = sizeof(to);
1624 memcpy(&to, msg->msg_name, msg_namelen);
1625 msg_name = msg->msg_name;
1626 }
1627
1628 sinfo = cmsgs.info;
1629 sinit = cmsgs.init;
1630
1631 /* Did the user specify SNDRCVINFO? */
1632 if (sinfo) {
1633 sinfo_flags = sinfo->sinfo_flags;
1634 associd = sinfo->sinfo_assoc_id;
1635 }
1636
1637 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1638 msg_len, sinfo_flags);
1639
1640 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1641 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1642 err = -EINVAL;
1643 goto out_nounlock;
1644 }
1645
1646 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1647 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1648 * If SCTP_ABORT is set, the message length could be non zero with
1649 * the msg_iov set to the user abort reason.
1650 */
1651 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1652 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1653 err = -EINVAL;
1654 goto out_nounlock;
1655 }
1656
1657 /* If SCTP_ADDR_OVER is set, there must be an address
1658 * specified in msg_name.
1659 */
1660 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1661 err = -EINVAL;
1662 goto out_nounlock;
1663 }
1664
1665 transport = NULL;
1666
1667 pr_debug("%s: about to look up association\n", __func__);
1668
1669 sctp_lock_sock(sk);
1670
1671 /* If a msg_name has been specified, assume this is to be used. */
1672 if (msg_name) {
1673 /* Look for a matching association on the endpoint. */
1674 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1675 if (!asoc) {
1676 /* If we could not find a matching association on the
1677 * endpoint, make sure that it is not a TCP-style
1678 * socket that already has an association or there is
1679 * no peeled-off association on another socket.
1680 */
1681 if ((sctp_style(sk, TCP) &&
1682 sctp_sstate(sk, ESTABLISHED)) ||
1683 sctp_endpoint_is_peeled_off(ep, &to)) {
1684 err = -EADDRNOTAVAIL;
1685 goto out_unlock;
1686 }
1687 }
1688 } else {
1689 asoc = sctp_id2assoc(sk, associd);
1690 if (!asoc) {
1691 err = -EPIPE;
1692 goto out_unlock;
1693 }
1694 }
1695
1696 if (asoc) {
1697 pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1698
1699 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1700 * socket that has an association in CLOSED state. This can
1701 * happen when an accepted socket has an association that is
1702 * already CLOSED.
1703 */
1704 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1705 err = -EPIPE;
1706 goto out_unlock;
1707 }
1708
1709 if (sinfo_flags & SCTP_EOF) {
1710 pr_debug("%s: shutting down association:%p\n",
1711 __func__, asoc);
1712
1713 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1714 err = 0;
1715 goto out_unlock;
1716 }
1717 if (sinfo_flags & SCTP_ABORT) {
1718
1719 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1720 if (!chunk) {
1721 err = -ENOMEM;
1722 goto out_unlock;
1723 }
1724
1725 pr_debug("%s: aborting association:%p\n",
1726 __func__, asoc);
1727
1728 sctp_primitive_ABORT(net, asoc, chunk);
1729 err = 0;
1730 goto out_unlock;
1731 }
1732 }
1733
1734 /* Do we need to create the association? */
1735 if (!asoc) {
1736 pr_debug("%s: there is no association yet\n", __func__);
1737
1738 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1739 err = -EINVAL;
1740 goto out_unlock;
1741 }
1742
1743 /* Check for invalid stream against the stream counts,
1744 * either the default or the user specified stream counts.
1745 */
1746 if (sinfo) {
1747 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1748 /* Check against the defaults. */
1749 if (sinfo->sinfo_stream >=
1750 sp->initmsg.sinit_num_ostreams) {
1751 err = -EINVAL;
1752 goto out_unlock;
1753 }
1754 } else {
1755 /* Check against the requested. */
1756 if (sinfo->sinfo_stream >=
1757 sinit->sinit_num_ostreams) {
1758 err = -EINVAL;
1759 goto out_unlock;
1760 }
1761 }
1762 }
1763
1764 /*
1765 * API 3.1.2 bind() - UDP Style Syntax
1766 * If a bind() or sctp_bindx() is not called prior to a
1767 * sendmsg() call that initiates a new association, the
1768 * system picks an ephemeral port and will choose an address
1769 * set equivalent to binding with a wildcard address.
1770 */
1771 if (!ep->base.bind_addr.port) {
1772 if (sctp_autobind(sk)) {
1773 err = -EAGAIN;
1774 goto out_unlock;
1775 }
1776 } else {
1777 /*
1778 * If an unprivileged user inherits a one-to-many
1779 * style socket with open associations on a privileged
1780 * port, it MAY be permitted to accept new associations,
1781 * but it SHOULD NOT be permitted to open new
1782 * associations.
1783 */
1784 if (ep->base.bind_addr.port < PROT_SOCK &&
1785 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1786 err = -EACCES;
1787 goto out_unlock;
1788 }
1789 }
1790
1791 scope = sctp_scope(&to);
1792 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1793 if (!new_asoc) {
1794 err = -ENOMEM;
1795 goto out_unlock;
1796 }
1797 asoc = new_asoc;
1798 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1799 if (err < 0) {
1800 err = -ENOMEM;
1801 goto out_free;
1802 }
1803
1804 /* If the SCTP_INIT ancillary data is specified, set all
1805 * the association init values accordingly.
1806 */
1807 if (sinit) {
1808 if (sinit->sinit_num_ostreams) {
1809 asoc->c.sinit_num_ostreams =
1810 sinit->sinit_num_ostreams;
1811 }
1812 if (sinit->sinit_max_instreams) {
1813 asoc->c.sinit_max_instreams =
1814 sinit->sinit_max_instreams;
1815 }
1816 if (sinit->sinit_max_attempts) {
1817 asoc->max_init_attempts
1818 = sinit->sinit_max_attempts;
1819 }
1820 if (sinit->sinit_max_init_timeo) {
1821 asoc->max_init_timeo =
1822 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1823 }
1824 }
1825
1826 /* Prime the peer's transport structures. */
1827 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1828 if (!transport) {
1829 err = -ENOMEM;
1830 goto out_free;
1831 }
1832 }
1833
1834 /* ASSERT: we have a valid association at this point. */
1835 pr_debug("%s: we have a valid association\n", __func__);
1836
1837 if (!sinfo) {
1838 /* If the user didn't specify SNDRCVINFO, make up one with
1839 * some defaults.
1840 */
1841 memset(&default_sinfo, 0, sizeof(default_sinfo));
1842 default_sinfo.sinfo_stream = asoc->default_stream;
1843 default_sinfo.sinfo_flags = asoc->default_flags;
1844 default_sinfo.sinfo_ppid = asoc->default_ppid;
1845 default_sinfo.sinfo_context = asoc->default_context;
1846 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1847 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1848 sinfo = &default_sinfo;
1849 }
1850
1851 /* API 7.1.7, the sndbuf size per association bounds the
1852 * maximum size of data that can be sent in a single send call.
1853 */
1854 if (msg_len > sk->sk_sndbuf) {
1855 err = -EMSGSIZE;
1856 goto out_free;
1857 }
1858
1859 if (asoc->pmtu_pending)
1860 sctp_assoc_pending_pmtu(sk, asoc);
1861
1862 /* If fragmentation is disabled and the message length exceeds the
1863 * association fragmentation point, return EMSGSIZE. The I-D
1864 * does not specify what this error is, but this looks like
1865 * a great fit.
1866 */
1867 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1868 err = -EMSGSIZE;
1869 goto out_free;
1870 }
1871
1872 /* Check for invalid stream. */
1873 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1874 err = -EINVAL;
1875 goto out_free;
1876 }
1877
1878 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1879 if (!sctp_wspace(asoc)) {
1880 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1881 if (err)
1882 goto out_free;
1883 }
1884
1885 /* If an address is passed with the sendto/sendmsg call, it is used
1886 * to override the primary destination address in the TCP model, or
1887 * when SCTP_ADDR_OVER flag is set in the UDP model.
1888 */
1889 if ((sctp_style(sk, TCP) && msg_name) ||
1890 (sinfo_flags & SCTP_ADDR_OVER)) {
1891 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1892 if (!chunk_tp) {
1893 err = -EINVAL;
1894 goto out_free;
1895 }
1896 } else
1897 chunk_tp = NULL;
1898
1899 /* Auto-connect, if we aren't connected already. */
1900 if (sctp_state(asoc, CLOSED)) {
1901 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1902 if (err < 0)
1903 goto out_free;
1904
1905 pr_debug("%s: we associated primitively\n", __func__);
1906 }
1907
1908 /* Break the message into multiple chunks of maximum size. */
1909 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1910 if (IS_ERR(datamsg)) {
1911 err = PTR_ERR(datamsg);
1912 goto out_free;
1913 }
1914
1915 /* Now send the (possibly) fragmented message. */
1916 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1917 sctp_chunk_hold(chunk);
1918
1919 /* Do accounting for the write space. */
1920 sctp_set_owner_w(chunk);
1921
1922 chunk->transport = chunk_tp;
1923 }
1924
1925 /* Send it to the lower layers. Note: all chunks
1926 * must either fail or succeed. The lower layer
1927 * works that way today. Keep it that way or this
1928 * breaks.
1929 */
1930 err = sctp_primitive_SEND(net, asoc, datamsg);
1931 /* Did the lower layer accept the chunk? */
1932 if (err) {
1933 sctp_datamsg_free(datamsg);
1934 goto out_free;
1935 }
1936
1937 pr_debug("%s: we sent primitively\n", __func__);
1938
1939 sctp_datamsg_put(datamsg);
1940 err = msg_len;
1941
1942 /* If we are already past ASSOCIATE, the lower
1943 * layers are responsible for association cleanup.
1944 */
1945 goto out_unlock;
1946
1947 out_free:
1948 if (new_asoc) {
1949 sctp_unhash_established(asoc);
1950 sctp_association_free(asoc);
1951 }
1952 out_unlock:
1953 sctp_release_sock(sk);
1954
1955 out_nounlock:
1956 return sctp_error(sk, msg_flags, err);
1957
1958 #if 0
1959 do_sock_err:
1960 if (msg_len)
1961 err = msg_len;
1962 else
1963 err = sock_error(sk);
1964 goto out;
1965
1966 do_interrupted:
1967 if (msg_len)
1968 err = msg_len;
1969 goto out;
1970 #endif /* 0 */
1971 }
1972
1973 /* This is an extended version of skb_pull() that removes the data from the
1974 * start of a skb even when data is spread across the list of skb's in the
1975 * frag_list. len specifies the total amount of data that needs to be removed.
1976 * when 'len' bytes could be removed from the skb, it returns 0.
1977 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1978 * could not be removed.
1979 */
1980 static int sctp_skb_pull(struct sk_buff *skb, int len)
1981 {
1982 struct sk_buff *list;
1983 int skb_len = skb_headlen(skb);
1984 int rlen;
1985
1986 if (len <= skb_len) {
1987 __skb_pull(skb, len);
1988 return 0;
1989 }
1990 len -= skb_len;
1991 __skb_pull(skb, skb_len);
1992
1993 skb_walk_frags(skb, list) {
1994 rlen = sctp_skb_pull(list, len);
1995 skb->len -= (len-rlen);
1996 skb->data_len -= (len-rlen);
1997
1998 if (!rlen)
1999 return 0;
2000
2001 len = rlen;
2002 }
2003
2004 return len;
2005 }
2006
2007 /* API 3.1.3 recvmsg() - UDP Style Syntax
2008 *
2009 * ssize_t recvmsg(int socket, struct msghdr *message,
2010 * int flags);
2011 *
2012 * socket - the socket descriptor of the endpoint.
2013 * message - pointer to the msghdr structure which contains a single
2014 * user message and possibly some ancillary data.
2015 *
2016 * See Section 5 for complete description of the data
2017 * structures.
2018 *
2019 * flags - flags sent or received with the user message, see Section
2020 * 5 for complete description of the flags.
2021 */
2022 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2023
2024 static int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2025 struct msghdr *msg, size_t len, int noblock,
2026 int flags, int *addr_len)
2027 {
2028 struct sctp_ulpevent *event = NULL;
2029 struct sctp_sock *sp = sctp_sk(sk);
2030 struct sk_buff *skb;
2031 int copied;
2032 int err = 0;
2033 int skb_len;
2034
2035 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2036 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2037 addr_len);
2038
2039 sctp_lock_sock(sk);
2040
2041 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2042 err = -ENOTCONN;
2043 goto out;
2044 }
2045
2046 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2047 if (!skb)
2048 goto out;
2049
2050 /* Get the total length of the skb including any skb's in the
2051 * frag_list.
2052 */
2053 skb_len = skb->len;
2054
2055 copied = skb_len;
2056 if (copied > len)
2057 copied = len;
2058
2059 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2060
2061 event = sctp_skb2event(skb);
2062
2063 if (err)
2064 goto out_free;
2065
2066 sock_recv_ts_and_drops(msg, sk, skb);
2067 if (sctp_ulpevent_is_notification(event)) {
2068 msg->msg_flags |= MSG_NOTIFICATION;
2069 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2070 } else {
2071 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2072 }
2073
2074 /* Check if we allow SCTP_SNDRCVINFO. */
2075 if (sp->subscribe.sctp_data_io_event)
2076 sctp_ulpevent_read_sndrcvinfo(event, msg);
2077 #if 0
2078 /* FIXME: we should be calling IP/IPv6 layers. */
2079 if (sk->sk_protinfo.af_inet.cmsg_flags)
2080 ip_cmsg_recv(msg, skb);
2081 #endif
2082
2083 err = copied;
2084
2085 /* If skb's length exceeds the user's buffer, update the skb and
2086 * push it back to the receive_queue so that the next call to
2087 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2088 */
2089 if (skb_len > copied) {
2090 msg->msg_flags &= ~MSG_EOR;
2091 if (flags & MSG_PEEK)
2092 goto out_free;
2093 sctp_skb_pull(skb, copied);
2094 skb_queue_head(&sk->sk_receive_queue, skb);
2095
2096 /* When only partial message is copied to the user, increase
2097 * rwnd by that amount. If all the data in the skb is read,
2098 * rwnd is updated when the event is freed.
2099 */
2100 if (!sctp_ulpevent_is_notification(event))
2101 sctp_assoc_rwnd_increase(event->asoc, copied);
2102 goto out;
2103 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2104 (event->msg_flags & MSG_EOR))
2105 msg->msg_flags |= MSG_EOR;
2106 else
2107 msg->msg_flags &= ~MSG_EOR;
2108
2109 out_free:
2110 if (flags & MSG_PEEK) {
2111 /* Release the skb reference acquired after peeking the skb in
2112 * sctp_skb_recv_datagram().
2113 */
2114 kfree_skb(skb);
2115 } else {
2116 /* Free the event which includes releasing the reference to
2117 * the owner of the skb, freeing the skb and updating the
2118 * rwnd.
2119 */
2120 sctp_ulpevent_free(event);
2121 }
2122 out:
2123 sctp_release_sock(sk);
2124 return err;
2125 }
2126
2127 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2128 *
2129 * This option is a on/off flag. If enabled no SCTP message
2130 * fragmentation will be performed. Instead if a message being sent
2131 * exceeds the current PMTU size, the message will NOT be sent and
2132 * instead a error will be indicated to the user.
2133 */
2134 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2135 char __user *optval,
2136 unsigned int optlen)
2137 {
2138 int val;
2139
2140 if (optlen < sizeof(int))
2141 return -EINVAL;
2142
2143 if (get_user(val, (int __user *)optval))
2144 return -EFAULT;
2145
2146 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2147
2148 return 0;
2149 }
2150
2151 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2152 unsigned int optlen)
2153 {
2154 struct sctp_association *asoc;
2155 struct sctp_ulpevent *event;
2156
2157 if (optlen > sizeof(struct sctp_event_subscribe))
2158 return -EINVAL;
2159 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2160 return -EFAULT;
2161
2162 /*
2163 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2164 * if there is no data to be sent or retransmit, the stack will
2165 * immediately send up this notification.
2166 */
2167 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2168 &sctp_sk(sk)->subscribe)) {
2169 asoc = sctp_id2assoc(sk, 0);
2170
2171 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2172 event = sctp_ulpevent_make_sender_dry_event(asoc,
2173 GFP_ATOMIC);
2174 if (!event)
2175 return -ENOMEM;
2176
2177 sctp_ulpq_tail_event(&asoc->ulpq, event);
2178 }
2179 }
2180
2181 return 0;
2182 }
2183
2184 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2185 *
2186 * This socket option is applicable to the UDP-style socket only. When
2187 * set it will cause associations that are idle for more than the
2188 * specified number of seconds to automatically close. An association
2189 * being idle is defined an association that has NOT sent or received
2190 * user data. The special value of '0' indicates that no automatic
2191 * close of any associations should be performed. The option expects an
2192 * integer defining the number of seconds of idle time before an
2193 * association is closed.
2194 */
2195 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2196 unsigned int optlen)
2197 {
2198 struct sctp_sock *sp = sctp_sk(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 return 0;
2209 }
2210
2211 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2212 *
2213 * Applications can enable or disable heartbeats for any peer address of
2214 * an association, modify an address's heartbeat interval, force a
2215 * heartbeat to be sent immediately, and adjust the address's maximum
2216 * number of retransmissions sent before an address is considered
2217 * unreachable. The following structure is used to access and modify an
2218 * address's parameters:
2219 *
2220 * struct sctp_paddrparams {
2221 * sctp_assoc_t spp_assoc_id;
2222 * struct sockaddr_storage spp_address;
2223 * uint32_t spp_hbinterval;
2224 * uint16_t spp_pathmaxrxt;
2225 * uint32_t spp_pathmtu;
2226 * uint32_t spp_sackdelay;
2227 * uint32_t spp_flags;
2228 * };
2229 *
2230 * spp_assoc_id - (one-to-many style socket) This is filled in the
2231 * application, and identifies the association for
2232 * this query.
2233 * spp_address - This specifies which address is of interest.
2234 * spp_hbinterval - This contains the value of the heartbeat interval,
2235 * in milliseconds. If a value of zero
2236 * is present in this field then no changes are to
2237 * be made to this parameter.
2238 * spp_pathmaxrxt - This contains the maximum number of
2239 * retransmissions before this address shall be
2240 * considered unreachable. If a value of zero
2241 * is present in this field then no changes are to
2242 * be made to this parameter.
2243 * spp_pathmtu - When Path MTU discovery is disabled the value
2244 * specified here will be the "fixed" path mtu.
2245 * Note that if the spp_address field is empty
2246 * then all associations on this address will
2247 * have this fixed path mtu set upon them.
2248 *
2249 * spp_sackdelay - When delayed sack is enabled, this value specifies
2250 * the number of milliseconds that sacks will be delayed
2251 * for. This value will apply to all addresses of an
2252 * association if the spp_address field is empty. Note
2253 * also, that if delayed sack is enabled and this
2254 * value is set to 0, no change is made to the last
2255 * recorded delayed sack timer value.
2256 *
2257 * spp_flags - These flags are used to control various features
2258 * on an association. The flag field may contain
2259 * zero or more of the following options.
2260 *
2261 * SPP_HB_ENABLE - Enable heartbeats on the
2262 * specified address. Note that if the address
2263 * field is empty all addresses for the association
2264 * have heartbeats enabled upon them.
2265 *
2266 * SPP_HB_DISABLE - Disable heartbeats on the
2267 * speicifed address. Note that if the address
2268 * field is empty all addresses for the association
2269 * will have their heartbeats disabled. Note also
2270 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2271 * mutually exclusive, only one of these two should
2272 * be specified. Enabling both fields will have
2273 * undetermined results.
2274 *
2275 * SPP_HB_DEMAND - Request a user initiated heartbeat
2276 * to be made immediately.
2277 *
2278 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2279 * heartbeat delayis to be set to the value of 0
2280 * milliseconds.
2281 *
2282 * SPP_PMTUD_ENABLE - This field will enable PMTU
2283 * discovery upon the specified address. Note that
2284 * if the address feild is empty then all addresses
2285 * on the association are effected.
2286 *
2287 * SPP_PMTUD_DISABLE - This field will disable PMTU
2288 * discovery upon the specified address. Note that
2289 * if the address feild is empty then all addresses
2290 * on the association are effected. Not also that
2291 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2292 * exclusive. Enabling both will have undetermined
2293 * results.
2294 *
2295 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2296 * on delayed sack. The time specified in spp_sackdelay
2297 * is used to specify the sack delay for this address. Note
2298 * that if spp_address is empty then all addresses will
2299 * enable delayed sack and take on the sack delay
2300 * value specified in spp_sackdelay.
2301 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2302 * off delayed sack. If the spp_address field is blank then
2303 * delayed sack is disabled for the entire association. Note
2304 * also that this field is mutually exclusive to
2305 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2306 * results.
2307 */
2308 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2309 struct sctp_transport *trans,
2310 struct sctp_association *asoc,
2311 struct sctp_sock *sp,
2312 int hb_change,
2313 int pmtud_change,
2314 int sackdelay_change)
2315 {
2316 int error;
2317
2318 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2319 struct net *net = sock_net(trans->asoc->base.sk);
2320
2321 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2322 if (error)
2323 return error;
2324 }
2325
2326 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2327 * this field is ignored. Note also that a value of zero indicates
2328 * the current setting should be left unchanged.
2329 */
2330 if (params->spp_flags & SPP_HB_ENABLE) {
2331
2332 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2333 * set. This lets us use 0 value when this flag
2334 * is set.
2335 */
2336 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2337 params->spp_hbinterval = 0;
2338
2339 if (params->spp_hbinterval ||
2340 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2341 if (trans) {
2342 trans->hbinterval =
2343 msecs_to_jiffies(params->spp_hbinterval);
2344 } else if (asoc) {
2345 asoc->hbinterval =
2346 msecs_to_jiffies(params->spp_hbinterval);
2347 } else {
2348 sp->hbinterval = params->spp_hbinterval;
2349 }
2350 }
2351 }
2352
2353 if (hb_change) {
2354 if (trans) {
2355 trans->param_flags =
2356 (trans->param_flags & ~SPP_HB) | hb_change;
2357 } else if (asoc) {
2358 asoc->param_flags =
2359 (asoc->param_flags & ~SPP_HB) | hb_change;
2360 } else {
2361 sp->param_flags =
2362 (sp->param_flags & ~SPP_HB) | hb_change;
2363 }
2364 }
2365
2366 /* When Path MTU discovery is disabled the value specified here will
2367 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2368 * include the flag SPP_PMTUD_DISABLE for this field to have any
2369 * effect).
2370 */
2371 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2372 if (trans) {
2373 trans->pathmtu = params->spp_pathmtu;
2374 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2375 } else if (asoc) {
2376 asoc->pathmtu = params->spp_pathmtu;
2377 sctp_frag_point(asoc, params->spp_pathmtu);
2378 } else {
2379 sp->pathmtu = params->spp_pathmtu;
2380 }
2381 }
2382
2383 if (pmtud_change) {
2384 if (trans) {
2385 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2386 (params->spp_flags & SPP_PMTUD_ENABLE);
2387 trans->param_flags =
2388 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2389 if (update) {
2390 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2391 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2392 }
2393 } else if (asoc) {
2394 asoc->param_flags =
2395 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2396 } else {
2397 sp->param_flags =
2398 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2399 }
2400 }
2401
2402 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2403 * value of this field is ignored. Note also that a value of zero
2404 * indicates the current setting should be left unchanged.
2405 */
2406 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2407 if (trans) {
2408 trans->sackdelay =
2409 msecs_to_jiffies(params->spp_sackdelay);
2410 } else if (asoc) {
2411 asoc->sackdelay =
2412 msecs_to_jiffies(params->spp_sackdelay);
2413 } else {
2414 sp->sackdelay = params->spp_sackdelay;
2415 }
2416 }
2417
2418 if (sackdelay_change) {
2419 if (trans) {
2420 trans->param_flags =
2421 (trans->param_flags & ~SPP_SACKDELAY) |
2422 sackdelay_change;
2423 } else if (asoc) {
2424 asoc->param_flags =
2425 (asoc->param_flags & ~SPP_SACKDELAY) |
2426 sackdelay_change;
2427 } else {
2428 sp->param_flags =
2429 (sp->param_flags & ~SPP_SACKDELAY) |
2430 sackdelay_change;
2431 }
2432 }
2433
2434 /* Note that a value of zero indicates the current setting should be
2435 left unchanged.
2436 */
2437 if (params->spp_pathmaxrxt) {
2438 if (trans) {
2439 trans->pathmaxrxt = params->spp_pathmaxrxt;
2440 } else if (asoc) {
2441 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2442 } else {
2443 sp->pathmaxrxt = params->spp_pathmaxrxt;
2444 }
2445 }
2446
2447 return 0;
2448 }
2449
2450 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2451 char __user *optval,
2452 unsigned int optlen)
2453 {
2454 struct sctp_paddrparams params;
2455 struct sctp_transport *trans = NULL;
2456 struct sctp_association *asoc = NULL;
2457 struct sctp_sock *sp = sctp_sk(sk);
2458 int error;
2459 int hb_change, pmtud_change, sackdelay_change;
2460
2461 if (optlen != sizeof(struct sctp_paddrparams))
2462 return - EINVAL;
2463
2464 if (copy_from_user(&params, optval, optlen))
2465 return -EFAULT;
2466
2467 /* Validate flags and value parameters. */
2468 hb_change = params.spp_flags & SPP_HB;
2469 pmtud_change = params.spp_flags & SPP_PMTUD;
2470 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2471
2472 if (hb_change == SPP_HB ||
2473 pmtud_change == SPP_PMTUD ||
2474 sackdelay_change == SPP_SACKDELAY ||
2475 params.spp_sackdelay > 500 ||
2476 (params.spp_pathmtu &&
2477 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2478 return -EINVAL;
2479
2480 /* If an address other than INADDR_ANY is specified, and
2481 * no transport is found, then the request is invalid.
2482 */
2483 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
2484 trans = sctp_addr_id2transport(sk, &params.spp_address,
2485 params.spp_assoc_id);
2486 if (!trans)
2487 return -EINVAL;
2488 }
2489
2490 /* Get association, if assoc_id != 0 and the socket is a one
2491 * to many style socket, and an association was not found, then
2492 * the id was invalid.
2493 */
2494 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2495 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2496 return -EINVAL;
2497
2498 /* Heartbeat demand can only be sent on a transport or
2499 * association, but not a socket.
2500 */
2501 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2502 return -EINVAL;
2503
2504 /* Process parameters. */
2505 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2506 hb_change, pmtud_change,
2507 sackdelay_change);
2508
2509 if (error)
2510 return error;
2511
2512 /* If changes are for association, also apply parameters to each
2513 * transport.
2514 */
2515 if (!trans && asoc) {
2516 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2517 transports) {
2518 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2519 hb_change, pmtud_change,
2520 sackdelay_change);
2521 }
2522 }
2523
2524 return 0;
2525 }
2526
2527 /*
2528 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2529 *
2530 * This option will effect the way delayed acks are performed. This
2531 * option allows you to get or set the delayed ack time, in
2532 * milliseconds. It also allows changing the delayed ack frequency.
2533 * Changing the frequency to 1 disables the delayed sack algorithm. If
2534 * the assoc_id is 0, then this sets or gets the endpoints default
2535 * values. If the assoc_id field is non-zero, then the set or get
2536 * effects the specified association for the one to many model (the
2537 * assoc_id field is ignored by the one to one model). Note that if
2538 * sack_delay or sack_freq are 0 when setting this option, then the
2539 * current values will remain unchanged.
2540 *
2541 * struct sctp_sack_info {
2542 * sctp_assoc_t sack_assoc_id;
2543 * uint32_t sack_delay;
2544 * uint32_t sack_freq;
2545 * };
2546 *
2547 * sack_assoc_id - This parameter, indicates which association the user
2548 * is performing an action upon. Note that if this field's value is
2549 * zero then the endpoints default value is changed (effecting future
2550 * associations only).
2551 *
2552 * sack_delay - This parameter contains the number of milliseconds that
2553 * the user is requesting the delayed ACK timer be set to. Note that
2554 * this value is defined in the standard to be between 200 and 500
2555 * milliseconds.
2556 *
2557 * sack_freq - This parameter contains the number of packets that must
2558 * be received before a sack is sent without waiting for the delay
2559 * timer to expire. The default value for this is 2, setting this
2560 * value to 1 will disable the delayed sack algorithm.
2561 */
2562
2563 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2564 char __user *optval, unsigned int optlen)
2565 {
2566 struct sctp_sack_info params;
2567 struct sctp_transport *trans = NULL;
2568 struct sctp_association *asoc = NULL;
2569 struct sctp_sock *sp = sctp_sk(sk);
2570
2571 if (optlen == sizeof(struct sctp_sack_info)) {
2572 if (copy_from_user(&params, optval, optlen))
2573 return -EFAULT;
2574
2575 if (params.sack_delay == 0 && params.sack_freq == 0)
2576 return 0;
2577 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2578 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2579 pr_warn("Use struct sctp_sack_info instead\n");
2580 if (copy_from_user(&params, optval, optlen))
2581 return -EFAULT;
2582
2583 if (params.sack_delay == 0)
2584 params.sack_freq = 1;
2585 else
2586 params.sack_freq = 0;
2587 } else
2588 return - EINVAL;
2589
2590 /* Validate value parameter. */
2591 if (params.sack_delay > 500)
2592 return -EINVAL;
2593
2594 /* Get association, if sack_assoc_id != 0 and the socket is a one
2595 * to many style socket, and an association was not found, then
2596 * the id was invalid.
2597 */
2598 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2599 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2600 return -EINVAL;
2601
2602 if (params.sack_delay) {
2603 if (asoc) {
2604 asoc->sackdelay =
2605 msecs_to_jiffies(params.sack_delay);
2606 asoc->param_flags =
2607 (asoc->param_flags & ~SPP_SACKDELAY) |
2608 SPP_SACKDELAY_ENABLE;
2609 } else {
2610 sp->sackdelay = params.sack_delay;
2611 sp->param_flags =
2612 (sp->param_flags & ~SPP_SACKDELAY) |
2613 SPP_SACKDELAY_ENABLE;
2614 }
2615 }
2616
2617 if (params.sack_freq == 1) {
2618 if (asoc) {
2619 asoc->param_flags =
2620 (asoc->param_flags & ~SPP_SACKDELAY) |
2621 SPP_SACKDELAY_DISABLE;
2622 } else {
2623 sp->param_flags =
2624 (sp->param_flags & ~SPP_SACKDELAY) |
2625 SPP_SACKDELAY_DISABLE;
2626 }
2627 } else if (params.sack_freq > 1) {
2628 if (asoc) {
2629 asoc->sackfreq = params.sack_freq;
2630 asoc->param_flags =
2631 (asoc->param_flags & ~SPP_SACKDELAY) |
2632 SPP_SACKDELAY_ENABLE;
2633 } else {
2634 sp->sackfreq = params.sack_freq;
2635 sp->param_flags =
2636 (sp->param_flags & ~SPP_SACKDELAY) |
2637 SPP_SACKDELAY_ENABLE;
2638 }
2639 }
2640
2641 /* If change is for association, also apply to each transport. */
2642 if (asoc) {
2643 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2644 transports) {
2645 if (params.sack_delay) {
2646 trans->sackdelay =
2647 msecs_to_jiffies(params.sack_delay);
2648 trans->param_flags =
2649 (trans->param_flags & ~SPP_SACKDELAY) |
2650 SPP_SACKDELAY_ENABLE;
2651 }
2652 if (params.sack_freq == 1) {
2653 trans->param_flags =
2654 (trans->param_flags & ~SPP_SACKDELAY) |
2655 SPP_SACKDELAY_DISABLE;
2656 } else if (params.sack_freq > 1) {
2657 trans->sackfreq = params.sack_freq;
2658 trans->param_flags =
2659 (trans->param_flags & ~SPP_SACKDELAY) |
2660 SPP_SACKDELAY_ENABLE;
2661 }
2662 }
2663 }
2664
2665 return 0;
2666 }
2667
2668 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2669 *
2670 * Applications can specify protocol parameters for the default association
2671 * initialization. The option name argument to setsockopt() and getsockopt()
2672 * is SCTP_INITMSG.
2673 *
2674 * Setting initialization parameters is effective only on an unconnected
2675 * socket (for UDP-style sockets only future associations are effected
2676 * by the change). With TCP-style sockets, this option is inherited by
2677 * sockets derived from a listener socket.
2678 */
2679 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2680 {
2681 struct sctp_initmsg sinit;
2682 struct sctp_sock *sp = sctp_sk(sk);
2683
2684 if (optlen != sizeof(struct sctp_initmsg))
2685 return -EINVAL;
2686 if (copy_from_user(&sinit, optval, optlen))
2687 return -EFAULT;
2688
2689 if (sinit.sinit_num_ostreams)
2690 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2691 if (sinit.sinit_max_instreams)
2692 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2693 if (sinit.sinit_max_attempts)
2694 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2695 if (sinit.sinit_max_init_timeo)
2696 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2697
2698 return 0;
2699 }
2700
2701 /*
2702 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2703 *
2704 * Applications that wish to use the sendto() system call may wish to
2705 * specify a default set of parameters that would normally be supplied
2706 * through the inclusion of ancillary data. This socket option allows
2707 * such an application to set the default sctp_sndrcvinfo structure.
2708 * The application that wishes to use this socket option simply passes
2709 * in to this call the sctp_sndrcvinfo structure defined in Section
2710 * 5.2.2) The input parameters accepted by this call include
2711 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2712 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2713 * to this call if the caller is using the UDP model.
2714 */
2715 static int sctp_setsockopt_default_send_param(struct sock *sk,
2716 char __user *optval,
2717 unsigned int optlen)
2718 {
2719 struct sctp_sndrcvinfo info;
2720 struct sctp_association *asoc;
2721 struct sctp_sock *sp = sctp_sk(sk);
2722
2723 if (optlen != sizeof(struct sctp_sndrcvinfo))
2724 return -EINVAL;
2725 if (copy_from_user(&info, optval, optlen))
2726 return -EFAULT;
2727
2728 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2729 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2730 return -EINVAL;
2731
2732 if (asoc) {
2733 asoc->default_stream = info.sinfo_stream;
2734 asoc->default_flags = info.sinfo_flags;
2735 asoc->default_ppid = info.sinfo_ppid;
2736 asoc->default_context = info.sinfo_context;
2737 asoc->default_timetolive = info.sinfo_timetolive;
2738 } else {
2739 sp->default_stream = info.sinfo_stream;
2740 sp->default_flags = info.sinfo_flags;
2741 sp->default_ppid = info.sinfo_ppid;
2742 sp->default_context = info.sinfo_context;
2743 sp->default_timetolive = info.sinfo_timetolive;
2744 }
2745
2746 return 0;
2747 }
2748
2749 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2750 *
2751 * Requests that the local SCTP stack use the enclosed peer address as
2752 * the association primary. The enclosed address must be one of the
2753 * association peer's addresses.
2754 */
2755 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2756 unsigned int optlen)
2757 {
2758 struct sctp_prim prim;
2759 struct sctp_transport *trans;
2760
2761 if (optlen != sizeof(struct sctp_prim))
2762 return -EINVAL;
2763
2764 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2765 return -EFAULT;
2766
2767 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2768 if (!trans)
2769 return -EINVAL;
2770
2771 sctp_assoc_set_primary(trans->asoc, trans);
2772
2773 return 0;
2774 }
2775
2776 /*
2777 * 7.1.5 SCTP_NODELAY
2778 *
2779 * Turn on/off any Nagle-like algorithm. This means that packets are
2780 * generally sent as soon as possible and no unnecessary delays are
2781 * introduced, at the cost of more packets in the network. Expects an
2782 * integer boolean flag.
2783 */
2784 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2785 unsigned int optlen)
2786 {
2787 int val;
2788
2789 if (optlen < sizeof(int))
2790 return -EINVAL;
2791 if (get_user(val, (int __user *)optval))
2792 return -EFAULT;
2793
2794 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2795 return 0;
2796 }
2797
2798 /*
2799 *
2800 * 7.1.1 SCTP_RTOINFO
2801 *
2802 * The protocol parameters used to initialize and bound retransmission
2803 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2804 * and modify these parameters.
2805 * All parameters are time values, in milliseconds. A value of 0, when
2806 * modifying the parameters, indicates that the current value should not
2807 * be changed.
2808 *
2809 */
2810 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2811 {
2812 struct sctp_rtoinfo rtoinfo;
2813 struct sctp_association *asoc;
2814
2815 if (optlen != sizeof (struct sctp_rtoinfo))
2816 return -EINVAL;
2817
2818 if (copy_from_user(&rtoinfo, optval, optlen))
2819 return -EFAULT;
2820
2821 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2822
2823 /* Set the values to the specific association */
2824 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2825 return -EINVAL;
2826
2827 if (asoc) {
2828 if (rtoinfo.srto_initial != 0)
2829 asoc->rto_initial =
2830 msecs_to_jiffies(rtoinfo.srto_initial);
2831 if (rtoinfo.srto_max != 0)
2832 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2833 if (rtoinfo.srto_min != 0)
2834 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2835 } else {
2836 /* If there is no association or the association-id = 0
2837 * set the values to the endpoint.
2838 */
2839 struct sctp_sock *sp = sctp_sk(sk);
2840
2841 if (rtoinfo.srto_initial != 0)
2842 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2843 if (rtoinfo.srto_max != 0)
2844 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2845 if (rtoinfo.srto_min != 0)
2846 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2847 }
2848
2849 return 0;
2850 }
2851
2852 /*
2853 *
2854 * 7.1.2 SCTP_ASSOCINFO
2855 *
2856 * This option is used to tune the maximum retransmission attempts
2857 * of the association.
2858 * Returns an error if the new association retransmission value is
2859 * greater than the sum of the retransmission value of the peer.
2860 * See [SCTP] for more information.
2861 *
2862 */
2863 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2864 {
2865
2866 struct sctp_assocparams assocparams;
2867 struct sctp_association *asoc;
2868
2869 if (optlen != sizeof(struct sctp_assocparams))
2870 return -EINVAL;
2871 if (copy_from_user(&assocparams, optval, optlen))
2872 return -EFAULT;
2873
2874 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2875
2876 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2877 return -EINVAL;
2878
2879 /* Set the values to the specific association */
2880 if (asoc) {
2881 if (assocparams.sasoc_asocmaxrxt != 0) {
2882 __u32 path_sum = 0;
2883 int paths = 0;
2884 struct sctp_transport *peer_addr;
2885
2886 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2887 transports) {
2888 path_sum += peer_addr->pathmaxrxt;
2889 paths++;
2890 }
2891
2892 /* Only validate asocmaxrxt if we have more than
2893 * one path/transport. We do this because path
2894 * retransmissions are only counted when we have more
2895 * then one path.
2896 */
2897 if (paths > 1 &&
2898 assocparams.sasoc_asocmaxrxt > path_sum)
2899 return -EINVAL;
2900
2901 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2902 }
2903
2904 if (assocparams.sasoc_cookie_life != 0)
2905 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
2906 } else {
2907 /* Set the values to the endpoint */
2908 struct sctp_sock *sp = sctp_sk(sk);
2909
2910 if (assocparams.sasoc_asocmaxrxt != 0)
2911 sp->assocparams.sasoc_asocmaxrxt =
2912 assocparams.sasoc_asocmaxrxt;
2913 if (assocparams.sasoc_cookie_life != 0)
2914 sp->assocparams.sasoc_cookie_life =
2915 assocparams.sasoc_cookie_life;
2916 }
2917 return 0;
2918 }
2919
2920 /*
2921 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2922 *
2923 * This socket option is a boolean flag which turns on or off mapped V4
2924 * addresses. If this option is turned on and the socket is type
2925 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2926 * If this option is turned off, then no mapping will be done of V4
2927 * addresses and a user will receive both PF_INET6 and PF_INET type
2928 * addresses on the socket.
2929 */
2930 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2931 {
2932 int val;
2933 struct sctp_sock *sp = sctp_sk(sk);
2934
2935 if (optlen < sizeof(int))
2936 return -EINVAL;
2937 if (get_user(val, (int __user *)optval))
2938 return -EFAULT;
2939 if (val)
2940 sp->v4mapped = 1;
2941 else
2942 sp->v4mapped = 0;
2943
2944 return 0;
2945 }
2946
2947 /*
2948 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2949 * This option will get or set the maximum size to put in any outgoing
2950 * SCTP DATA chunk. If a message is larger than this size it will be
2951 * fragmented by SCTP into the specified size. Note that the underlying
2952 * SCTP implementation may fragment into smaller sized chunks when the
2953 * PMTU of the underlying association is smaller than the value set by
2954 * the user. The default value for this option is '0' which indicates
2955 * the user is NOT limiting fragmentation and only the PMTU will effect
2956 * SCTP's choice of DATA chunk size. Note also that values set larger
2957 * than the maximum size of an IP datagram will effectively let SCTP
2958 * control fragmentation (i.e. the same as setting this option to 0).
2959 *
2960 * The following structure is used to access and modify this parameter:
2961 *
2962 * struct sctp_assoc_value {
2963 * sctp_assoc_t assoc_id;
2964 * uint32_t assoc_value;
2965 * };
2966 *
2967 * assoc_id: This parameter is ignored for one-to-one style sockets.
2968 * For one-to-many style sockets this parameter indicates which
2969 * association the user is performing an action upon. Note that if
2970 * this field's value is zero then the endpoints default value is
2971 * changed (effecting future associations only).
2972 * assoc_value: This parameter specifies the maximum size in bytes.
2973 */
2974 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2975 {
2976 struct sctp_assoc_value params;
2977 struct sctp_association *asoc;
2978 struct sctp_sock *sp = sctp_sk(sk);
2979 int val;
2980
2981 if (optlen == sizeof(int)) {
2982 pr_warn("Use of int in maxseg socket option deprecated\n");
2983 pr_warn("Use struct sctp_assoc_value instead\n");
2984 if (copy_from_user(&val, optval, optlen))
2985 return -EFAULT;
2986 params.assoc_id = 0;
2987 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2988 if (copy_from_user(&params, optval, optlen))
2989 return -EFAULT;
2990 val = params.assoc_value;
2991 } else
2992 return -EINVAL;
2993
2994 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2995 return -EINVAL;
2996
2997 asoc = sctp_id2assoc(sk, params.assoc_id);
2998 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2999 return -EINVAL;
3000
3001 if (asoc) {
3002 if (val == 0) {
3003 val = asoc->pathmtu;
3004 val -= sp->pf->af->net_header_len;
3005 val -= sizeof(struct sctphdr) +
3006 sizeof(struct sctp_data_chunk);
3007 }
3008 asoc->user_frag = val;
3009 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3010 } else {
3011 sp->user_frag = val;
3012 }
3013
3014 return 0;
3015 }
3016
3017
3018 /*
3019 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3020 *
3021 * Requests that the peer mark the enclosed address as the association
3022 * primary. The enclosed address must be one of the association's
3023 * locally bound addresses. The following structure is used to make a
3024 * set primary request:
3025 */
3026 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3027 unsigned int optlen)
3028 {
3029 struct net *net = sock_net(sk);
3030 struct sctp_sock *sp;
3031 struct sctp_association *asoc = NULL;
3032 struct sctp_setpeerprim prim;
3033 struct sctp_chunk *chunk;
3034 struct sctp_af *af;
3035 int err;
3036
3037 sp = sctp_sk(sk);
3038
3039 if (!net->sctp.addip_enable)
3040 return -EPERM;
3041
3042 if (optlen != sizeof(struct sctp_setpeerprim))
3043 return -EINVAL;
3044
3045 if (copy_from_user(&prim, optval, optlen))
3046 return -EFAULT;
3047
3048 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3049 if (!asoc)
3050 return -EINVAL;
3051
3052 if (!asoc->peer.asconf_capable)
3053 return -EPERM;
3054
3055 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3056 return -EPERM;
3057
3058 if (!sctp_state(asoc, ESTABLISHED))
3059 return -ENOTCONN;
3060
3061 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3062 if (!af)
3063 return -EINVAL;
3064
3065 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3066 return -EADDRNOTAVAIL;
3067
3068 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3069 return -EADDRNOTAVAIL;
3070
3071 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3072 chunk = sctp_make_asconf_set_prim(asoc,
3073 (union sctp_addr *)&prim.sspp_addr);
3074 if (!chunk)
3075 return -ENOMEM;
3076
3077 err = sctp_send_asconf(asoc, chunk);
3078
3079 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3080
3081 return err;
3082 }
3083
3084 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3085 unsigned int optlen)
3086 {
3087 struct sctp_setadaptation adaptation;
3088
3089 if (optlen != sizeof(struct sctp_setadaptation))
3090 return -EINVAL;
3091 if (copy_from_user(&adaptation, optval, optlen))
3092 return -EFAULT;
3093
3094 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3095
3096 return 0;
3097 }
3098
3099 /*
3100 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3101 *
3102 * The context field in the sctp_sndrcvinfo structure is normally only
3103 * used when a failed message is retrieved holding the value that was
3104 * sent down on the actual send call. This option allows the setting of
3105 * a default context on an association basis that will be received on
3106 * reading messages from the peer. This is especially helpful in the
3107 * one-2-many model for an application to keep some reference to an
3108 * internal state machine that is processing messages on the
3109 * association. Note that the setting of this value only effects
3110 * received messages from the peer and does not effect the value that is
3111 * saved with outbound messages.
3112 */
3113 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3114 unsigned int optlen)
3115 {
3116 struct sctp_assoc_value params;
3117 struct sctp_sock *sp;
3118 struct sctp_association *asoc;
3119
3120 if (optlen != sizeof(struct sctp_assoc_value))
3121 return -EINVAL;
3122 if (copy_from_user(&params, optval, optlen))
3123 return -EFAULT;
3124
3125 sp = sctp_sk(sk);
3126
3127 if (params.assoc_id != 0) {
3128 asoc = sctp_id2assoc(sk, params.assoc_id);
3129 if (!asoc)
3130 return -EINVAL;
3131 asoc->default_rcv_context = params.assoc_value;
3132 } else {
3133 sp->default_rcv_context = params.assoc_value;
3134 }
3135
3136 return 0;
3137 }
3138
3139 /*
3140 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3141 *
3142 * This options will at a minimum specify if the implementation is doing
3143 * fragmented interleave. Fragmented interleave, for a one to many
3144 * socket, is when subsequent calls to receive a message may return
3145 * parts of messages from different associations. Some implementations
3146 * may allow you to turn this value on or off. If so, when turned off,
3147 * no fragment interleave will occur (which will cause a head of line
3148 * blocking amongst multiple associations sharing the same one to many
3149 * socket). When this option is turned on, then each receive call may
3150 * come from a different association (thus the user must receive data
3151 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3152 * association each receive belongs to.
3153 *
3154 * This option takes a boolean value. A non-zero value indicates that
3155 * fragmented interleave is on. A value of zero indicates that
3156 * fragmented interleave is off.
3157 *
3158 * Note that it is important that an implementation that allows this
3159 * option to be turned on, have it off by default. Otherwise an unaware
3160 * application using the one to many model may become confused and act
3161 * incorrectly.
3162 */
3163 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3164 char __user *optval,
3165 unsigned int optlen)
3166 {
3167 int val;
3168
3169 if (optlen != sizeof(int))
3170 return -EINVAL;
3171 if (get_user(val, (int __user *)optval))
3172 return -EFAULT;
3173
3174 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3175
3176 return 0;
3177 }
3178
3179 /*
3180 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3181 * (SCTP_PARTIAL_DELIVERY_POINT)
3182 *
3183 * This option will set or get the SCTP partial delivery point. This
3184 * point is the size of a message where the partial delivery API will be
3185 * invoked to help free up rwnd space for the peer. Setting this to a
3186 * lower value will cause partial deliveries to happen more often. The
3187 * calls argument is an integer that sets or gets the partial delivery
3188 * point. Note also that the call will fail if the user attempts to set
3189 * this value larger than the socket receive buffer size.
3190 *
3191 * Note that any single message having a length smaller than or equal to
3192 * the SCTP partial delivery point will be delivered in one single read
3193 * call as long as the user provided buffer is large enough to hold the
3194 * message.
3195 */
3196 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3197 char __user *optval,
3198 unsigned int optlen)
3199 {
3200 u32 val;
3201
3202 if (optlen != sizeof(u32))
3203 return -EINVAL;
3204 if (get_user(val, (int __user *)optval))
3205 return -EFAULT;
3206
3207 /* Note: We double the receive buffer from what the user sets
3208 * it to be, also initial rwnd is based on rcvbuf/2.
3209 */
3210 if (val > (sk->sk_rcvbuf >> 1))
3211 return -EINVAL;
3212
3213 sctp_sk(sk)->pd_point = val;
3214
3215 return 0; /* is this the right error code? */
3216 }
3217
3218 /*
3219 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3220 *
3221 * This option will allow a user to change the maximum burst of packets
3222 * that can be emitted by this association. Note that the default value
3223 * is 4, and some implementations may restrict this setting so that it
3224 * can only be lowered.
3225 *
3226 * NOTE: This text doesn't seem right. Do this on a socket basis with
3227 * future associations inheriting the socket value.
3228 */
3229 static int sctp_setsockopt_maxburst(struct sock *sk,
3230 char __user *optval,
3231 unsigned int optlen)
3232 {
3233 struct sctp_assoc_value params;
3234 struct sctp_sock *sp;
3235 struct sctp_association *asoc;
3236 int val;
3237 int assoc_id = 0;
3238
3239 if (optlen == sizeof(int)) {
3240 pr_warn("Use of int in max_burst socket option deprecated\n");
3241 pr_warn("Use struct sctp_assoc_value instead\n");
3242 if (copy_from_user(&val, optval, optlen))
3243 return -EFAULT;
3244 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3245 if (copy_from_user(&params, optval, optlen))
3246 return -EFAULT;
3247 val = params.assoc_value;
3248 assoc_id = params.assoc_id;
3249 } else
3250 return -EINVAL;
3251
3252 sp = sctp_sk(sk);
3253
3254 if (assoc_id != 0) {
3255 asoc = sctp_id2assoc(sk, assoc_id);
3256 if (!asoc)
3257 return -EINVAL;
3258 asoc->max_burst = val;
3259 } else
3260 sp->max_burst = val;
3261
3262 return 0;
3263 }
3264
3265 /*
3266 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3267 *
3268 * This set option adds a chunk type that the user is requesting to be
3269 * received only in an authenticated way. Changes to the list of chunks
3270 * will only effect future associations on the socket.
3271 */
3272 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3273 char __user *optval,
3274 unsigned int optlen)
3275 {
3276 struct net *net = sock_net(sk);
3277 struct sctp_authchunk val;
3278
3279 if (!net->sctp.auth_enable)
3280 return -EACCES;
3281
3282 if (optlen != sizeof(struct sctp_authchunk))
3283 return -EINVAL;
3284 if (copy_from_user(&val, optval, optlen))
3285 return -EFAULT;
3286
3287 switch (val.sauth_chunk) {
3288 case SCTP_CID_INIT:
3289 case SCTP_CID_INIT_ACK:
3290 case SCTP_CID_SHUTDOWN_COMPLETE:
3291 case SCTP_CID_AUTH:
3292 return -EINVAL;
3293 }
3294
3295 /* add this chunk id to the endpoint */
3296 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3297 }
3298
3299 /*
3300 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3301 *
3302 * This option gets or sets the list of HMAC algorithms that the local
3303 * endpoint requires the peer to use.
3304 */
3305 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3306 char __user *optval,
3307 unsigned int optlen)
3308 {
3309 struct net *net = sock_net(sk);
3310 struct sctp_hmacalgo *hmacs;
3311 u32 idents;
3312 int err;
3313
3314 if (!net->sctp.auth_enable)
3315 return -EACCES;
3316
3317 if (optlen < sizeof(struct sctp_hmacalgo))
3318 return -EINVAL;
3319
3320 hmacs= memdup_user(optval, optlen);
3321 if (IS_ERR(hmacs))
3322 return PTR_ERR(hmacs);
3323
3324 idents = hmacs->shmac_num_idents;
3325 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3326 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3327 err = -EINVAL;
3328 goto out;
3329 }
3330
3331 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3332 out:
3333 kfree(hmacs);
3334 return err;
3335 }
3336
3337 /*
3338 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3339 *
3340 * This option will set a shared secret key which is used to build an
3341 * association shared key.
3342 */
3343 static int sctp_setsockopt_auth_key(struct sock *sk,
3344 char __user *optval,
3345 unsigned int optlen)
3346 {
3347 struct net *net = sock_net(sk);
3348 struct sctp_authkey *authkey;
3349 struct sctp_association *asoc;
3350 int ret;
3351
3352 if (!net->sctp.auth_enable)
3353 return -EACCES;
3354
3355 if (optlen <= sizeof(struct sctp_authkey))
3356 return -EINVAL;
3357
3358 authkey= memdup_user(optval, optlen);
3359 if (IS_ERR(authkey))
3360 return PTR_ERR(authkey);
3361
3362 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3363 ret = -EINVAL;
3364 goto out;
3365 }
3366
3367 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3368 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3369 ret = -EINVAL;
3370 goto out;
3371 }
3372
3373 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3374 out:
3375 kzfree(authkey);
3376 return ret;
3377 }
3378
3379 /*
3380 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3381 *
3382 * This option will get or set the active shared key to be used to build
3383 * the association shared key.
3384 */
3385 static int sctp_setsockopt_active_key(struct sock *sk,
3386 char __user *optval,
3387 unsigned int optlen)
3388 {
3389 struct net *net = sock_net(sk);
3390 struct sctp_authkeyid val;
3391 struct sctp_association *asoc;
3392
3393 if (!net->sctp.auth_enable)
3394 return -EACCES;
3395
3396 if (optlen != sizeof(struct sctp_authkeyid))
3397 return -EINVAL;
3398 if (copy_from_user(&val, optval, optlen))
3399 return -EFAULT;
3400
3401 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3402 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3403 return -EINVAL;
3404
3405 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3406 val.scact_keynumber);
3407 }
3408
3409 /*
3410 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3411 *
3412 * This set option will delete a shared secret key from use.
3413 */
3414 static int sctp_setsockopt_del_key(struct sock *sk,
3415 char __user *optval,
3416 unsigned int optlen)
3417 {
3418 struct net *net = sock_net(sk);
3419 struct sctp_authkeyid val;
3420 struct sctp_association *asoc;
3421
3422 if (!net->sctp.auth_enable)
3423 return -EACCES;
3424
3425 if (optlen != sizeof(struct sctp_authkeyid))
3426 return -EINVAL;
3427 if (copy_from_user(&val, optval, optlen))
3428 return -EFAULT;
3429
3430 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3431 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3432 return -EINVAL;
3433
3434 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3435 val.scact_keynumber);
3436
3437 }
3438
3439 /*
3440 * 8.1.23 SCTP_AUTO_ASCONF
3441 *
3442 * This option will enable or disable the use of the automatic generation of
3443 * ASCONF chunks to add and delete addresses to an existing association. Note
3444 * that this option has two caveats namely: a) it only affects sockets that
3445 * are bound to all addresses available to the SCTP stack, and b) the system
3446 * administrator may have an overriding control that turns the ASCONF feature
3447 * off no matter what setting the socket option may have.
3448 * This option expects an integer boolean flag, where a non-zero value turns on
3449 * the option, and a zero value turns off the option.
3450 * Note. In this implementation, socket operation overrides default parameter
3451 * being set by sysctl as well as FreeBSD implementation
3452 */
3453 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3454 unsigned int optlen)
3455 {
3456 int val;
3457 struct sctp_sock *sp = sctp_sk(sk);
3458
3459 if (optlen < sizeof(int))
3460 return -EINVAL;
3461 if (get_user(val, (int __user *)optval))
3462 return -EFAULT;
3463 if (!sctp_is_ep_boundall(sk) && val)
3464 return -EINVAL;
3465 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3466 return 0;
3467
3468 if (val == 0 && sp->do_auto_asconf) {
3469 list_del(&sp->auto_asconf_list);
3470 sp->do_auto_asconf = 0;
3471 } else if (val && !sp->do_auto_asconf) {
3472 list_add_tail(&sp->auto_asconf_list,
3473 &sock_net(sk)->sctp.auto_asconf_splist);
3474 sp->do_auto_asconf = 1;
3475 }
3476 return 0;
3477 }
3478
3479
3480 /*
3481 * SCTP_PEER_ADDR_THLDS
3482 *
3483 * This option allows us to alter the partially failed threshold for one or all
3484 * transports in an association. See Section 6.1 of:
3485 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3486 */
3487 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3488 char __user *optval,
3489 unsigned int optlen)
3490 {
3491 struct sctp_paddrthlds val;
3492 struct sctp_transport *trans;
3493 struct sctp_association *asoc;
3494
3495 if (optlen < sizeof(struct sctp_paddrthlds))
3496 return -EINVAL;
3497 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3498 sizeof(struct sctp_paddrthlds)))
3499 return -EFAULT;
3500
3501
3502 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3503 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3504 if (!asoc)
3505 return -ENOENT;
3506 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3507 transports) {
3508 if (val.spt_pathmaxrxt)
3509 trans->pathmaxrxt = val.spt_pathmaxrxt;
3510 trans->pf_retrans = val.spt_pathpfthld;
3511 }
3512
3513 if (val.spt_pathmaxrxt)
3514 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3515 asoc->pf_retrans = val.spt_pathpfthld;
3516 } else {
3517 trans = sctp_addr_id2transport(sk, &val.spt_address,
3518 val.spt_assoc_id);
3519 if (!trans)
3520 return -ENOENT;
3521
3522 if (val.spt_pathmaxrxt)
3523 trans->pathmaxrxt = val.spt_pathmaxrxt;
3524 trans->pf_retrans = val.spt_pathpfthld;
3525 }
3526
3527 return 0;
3528 }
3529
3530 /* API 6.2 setsockopt(), getsockopt()
3531 *
3532 * Applications use setsockopt() and getsockopt() to set or retrieve
3533 * socket options. Socket options are used to change the default
3534 * behavior of sockets calls. They are described in Section 7.
3535 *
3536 * The syntax is:
3537 *
3538 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3539 * int __user *optlen);
3540 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3541 * int optlen);
3542 *
3543 * sd - the socket descript.
3544 * level - set to IPPROTO_SCTP for all SCTP options.
3545 * optname - the option name.
3546 * optval - the buffer to store the value of the option.
3547 * optlen - the size of the buffer.
3548 */
3549 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3550 char __user *optval, unsigned int optlen)
3551 {
3552 int retval = 0;
3553
3554 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3555
3556 /* I can hardly begin to describe how wrong this is. This is
3557 * so broken as to be worse than useless. The API draft
3558 * REALLY is NOT helpful here... I am not convinced that the
3559 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3560 * are at all well-founded.
3561 */
3562 if (level != SOL_SCTP) {
3563 struct sctp_af *af = sctp_sk(sk)->pf->af;
3564 retval = af->setsockopt(sk, level, optname, optval, optlen);
3565 goto out_nounlock;
3566 }
3567
3568 sctp_lock_sock(sk);
3569
3570 switch (optname) {
3571 case SCTP_SOCKOPT_BINDX_ADD:
3572 /* 'optlen' is the size of the addresses buffer. */
3573 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3574 optlen, SCTP_BINDX_ADD_ADDR);
3575 break;
3576
3577 case SCTP_SOCKOPT_BINDX_REM:
3578 /* 'optlen' is the size of the addresses buffer. */
3579 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3580 optlen, SCTP_BINDX_REM_ADDR);
3581 break;
3582
3583 case SCTP_SOCKOPT_CONNECTX_OLD:
3584 /* 'optlen' is the size of the addresses buffer. */
3585 retval = sctp_setsockopt_connectx_old(sk,
3586 (struct sockaddr __user *)optval,
3587 optlen);
3588 break;
3589
3590 case SCTP_SOCKOPT_CONNECTX:
3591 /* 'optlen' is the size of the addresses buffer. */
3592 retval = sctp_setsockopt_connectx(sk,
3593 (struct sockaddr __user *)optval,
3594 optlen);
3595 break;
3596
3597 case SCTP_DISABLE_FRAGMENTS:
3598 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3599 break;
3600
3601 case SCTP_EVENTS:
3602 retval = sctp_setsockopt_events(sk, optval, optlen);
3603 break;
3604
3605 case SCTP_AUTOCLOSE:
3606 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3607 break;
3608
3609 case SCTP_PEER_ADDR_PARAMS:
3610 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3611 break;
3612
3613 case SCTP_DELAYED_SACK:
3614 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3615 break;
3616 case SCTP_PARTIAL_DELIVERY_POINT:
3617 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3618 break;
3619
3620 case SCTP_INITMSG:
3621 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3622 break;
3623 case SCTP_DEFAULT_SEND_PARAM:
3624 retval = sctp_setsockopt_default_send_param(sk, optval,
3625 optlen);
3626 break;
3627 case SCTP_PRIMARY_ADDR:
3628 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3629 break;
3630 case SCTP_SET_PEER_PRIMARY_ADDR:
3631 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3632 break;
3633 case SCTP_NODELAY:
3634 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3635 break;
3636 case SCTP_RTOINFO:
3637 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3638 break;
3639 case SCTP_ASSOCINFO:
3640 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3641 break;
3642 case SCTP_I_WANT_MAPPED_V4_ADDR:
3643 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3644 break;
3645 case SCTP_MAXSEG:
3646 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3647 break;
3648 case SCTP_ADAPTATION_LAYER:
3649 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3650 break;
3651 case SCTP_CONTEXT:
3652 retval = sctp_setsockopt_context(sk, optval, optlen);
3653 break;
3654 case SCTP_FRAGMENT_INTERLEAVE:
3655 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3656 break;
3657 case SCTP_MAX_BURST:
3658 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3659 break;
3660 case SCTP_AUTH_CHUNK:
3661 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3662 break;
3663 case SCTP_HMAC_IDENT:
3664 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3665 break;
3666 case SCTP_AUTH_KEY:
3667 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3668 break;
3669 case SCTP_AUTH_ACTIVE_KEY:
3670 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3671 break;
3672 case SCTP_AUTH_DELETE_KEY:
3673 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3674 break;
3675 case SCTP_AUTO_ASCONF:
3676 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3677 break;
3678 case SCTP_PEER_ADDR_THLDS:
3679 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3680 break;
3681 default:
3682 retval = -ENOPROTOOPT;
3683 break;
3684 }
3685
3686 sctp_release_sock(sk);
3687
3688 out_nounlock:
3689 return retval;
3690 }
3691
3692 /* API 3.1.6 connect() - UDP Style Syntax
3693 *
3694 * An application may use the connect() call in the UDP model to initiate an
3695 * association without sending data.
3696 *
3697 * The syntax is:
3698 *
3699 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3700 *
3701 * sd: the socket descriptor to have a new association added to.
3702 *
3703 * nam: the address structure (either struct sockaddr_in or struct
3704 * sockaddr_in6 defined in RFC2553 [7]).
3705 *
3706 * len: the size of the address.
3707 */
3708 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
3709 int addr_len)
3710 {
3711 int err = 0;
3712 struct sctp_af *af;
3713
3714 sctp_lock_sock(sk);
3715
3716 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
3717 addr, addr_len);
3718
3719 /* Validate addr_len before calling common connect/connectx routine. */
3720 af = sctp_get_af_specific(addr->sa_family);
3721 if (!af || addr_len < af->sockaddr_len) {
3722 err = -EINVAL;
3723 } else {
3724 /* Pass correct addr len to common routine (so it knows there
3725 * is only one address being passed.
3726 */
3727 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3728 }
3729
3730 sctp_release_sock(sk);
3731 return err;
3732 }
3733
3734 /* FIXME: Write comments. */
3735 static int sctp_disconnect(struct sock *sk, int flags)
3736 {
3737 return -EOPNOTSUPP; /* STUB */
3738 }
3739
3740 /* 4.1.4 accept() - TCP Style Syntax
3741 *
3742 * Applications use accept() call to remove an established SCTP
3743 * association from the accept queue of the endpoint. A new socket
3744 * descriptor will be returned from accept() to represent the newly
3745 * formed association.
3746 */
3747 static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3748 {
3749 struct sctp_sock *sp;
3750 struct sctp_endpoint *ep;
3751 struct sock *newsk = NULL;
3752 struct sctp_association *asoc;
3753 long timeo;
3754 int error = 0;
3755
3756 sctp_lock_sock(sk);
3757
3758 sp = sctp_sk(sk);
3759 ep = sp->ep;
3760
3761 if (!sctp_style(sk, TCP)) {
3762 error = -EOPNOTSUPP;
3763 goto out;
3764 }
3765
3766 if (!sctp_sstate(sk, LISTENING)) {
3767 error = -EINVAL;
3768 goto out;
3769 }
3770
3771 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3772
3773 error = sctp_wait_for_accept(sk, timeo);
3774 if (error)
3775 goto out;
3776
3777 /* We treat the list of associations on the endpoint as the accept
3778 * queue and pick the first association on the list.
3779 */
3780 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3781
3782 newsk = sp->pf->create_accept_sk(sk, asoc);
3783 if (!newsk) {
3784 error = -ENOMEM;
3785 goto out;
3786 }
3787
3788 /* Populate the fields of the newsk from the oldsk and migrate the
3789 * asoc to the newsk.
3790 */
3791 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3792
3793 out:
3794 sctp_release_sock(sk);
3795 *err = error;
3796 return newsk;
3797 }
3798
3799 /* The SCTP ioctl handler. */
3800 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3801 {
3802 int rc = -ENOTCONN;
3803
3804 sctp_lock_sock(sk);
3805
3806 /*
3807 * SEQPACKET-style sockets in LISTENING state are valid, for
3808 * SCTP, so only discard TCP-style sockets in LISTENING state.
3809 */
3810 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3811 goto out;
3812
3813 switch (cmd) {
3814 case SIOCINQ: {
3815 struct sk_buff *skb;
3816 unsigned int amount = 0;
3817
3818 skb = skb_peek(&sk->sk_receive_queue);
3819 if (skb != NULL) {
3820 /*
3821 * We will only return the amount of this packet since
3822 * that is all that will be read.
3823 */
3824 amount = skb->len;
3825 }
3826 rc = put_user(amount, (int __user *)arg);
3827 break;
3828 }
3829 default:
3830 rc = -ENOIOCTLCMD;
3831 break;
3832 }
3833 out:
3834 sctp_release_sock(sk);
3835 return rc;
3836 }
3837
3838 /* This is the function which gets called during socket creation to
3839 * initialized the SCTP-specific portion of the sock.
3840 * The sock structure should already be zero-filled memory.
3841 */
3842 static int sctp_init_sock(struct sock *sk)
3843 {
3844 struct net *net = sock_net(sk);
3845 struct sctp_sock *sp;
3846
3847 pr_debug("%s: sk:%p\n", __func__, sk);
3848
3849 sp = sctp_sk(sk);
3850
3851 /* Initialize the SCTP per socket area. */
3852 switch (sk->sk_type) {
3853 case SOCK_SEQPACKET:
3854 sp->type = SCTP_SOCKET_UDP;
3855 break;
3856 case SOCK_STREAM:
3857 sp->type = SCTP_SOCKET_TCP;
3858 break;
3859 default:
3860 return -ESOCKTNOSUPPORT;
3861 }
3862
3863 /* Initialize default send parameters. These parameters can be
3864 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3865 */
3866 sp->default_stream = 0;
3867 sp->default_ppid = 0;
3868 sp->default_flags = 0;
3869 sp->default_context = 0;
3870 sp->default_timetolive = 0;
3871
3872 sp->default_rcv_context = 0;
3873 sp->max_burst = net->sctp.max_burst;
3874
3875 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
3876
3877 /* Initialize default setup parameters. These parameters
3878 * can be modified with the SCTP_INITMSG socket option or
3879 * overridden by the SCTP_INIT CMSG.
3880 */
3881 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3882 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3883 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
3884 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
3885
3886 /* Initialize default RTO related parameters. These parameters can
3887 * be modified for with the SCTP_RTOINFO socket option.
3888 */
3889 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
3890 sp->rtoinfo.srto_max = net->sctp.rto_max;
3891 sp->rtoinfo.srto_min = net->sctp.rto_min;
3892
3893 /* Initialize default association related parameters. These parameters
3894 * can be modified with the SCTP_ASSOCINFO socket option.
3895 */
3896 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
3897 sp->assocparams.sasoc_number_peer_destinations = 0;
3898 sp->assocparams.sasoc_peer_rwnd = 0;
3899 sp->assocparams.sasoc_local_rwnd = 0;
3900 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
3901
3902 /* Initialize default event subscriptions. By default, all the
3903 * options are off.
3904 */
3905 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3906
3907 /* Default Peer Address Parameters. These defaults can
3908 * be modified via SCTP_PEER_ADDR_PARAMS
3909 */
3910 sp->hbinterval = net->sctp.hb_interval;
3911 sp->pathmaxrxt = net->sctp.max_retrans_path;
3912 sp->pathmtu = 0; // allow default discovery
3913 sp->sackdelay = net->sctp.sack_timeout;
3914 sp->sackfreq = 2;
3915 sp->param_flags = SPP_HB_ENABLE |
3916 SPP_PMTUD_ENABLE |
3917 SPP_SACKDELAY_ENABLE;
3918
3919 /* If enabled no SCTP message fragmentation will be performed.
3920 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3921 */
3922 sp->disable_fragments = 0;
3923
3924 /* Enable Nagle algorithm by default. */
3925 sp->nodelay = 0;
3926
3927 /* Enable by default. */
3928 sp->v4mapped = 1;
3929
3930 /* Auto-close idle associations after the configured
3931 * number of seconds. A value of 0 disables this
3932 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3933 * for UDP-style sockets only.
3934 */
3935 sp->autoclose = 0;
3936
3937 /* User specified fragmentation limit. */
3938 sp->user_frag = 0;
3939
3940 sp->adaptation_ind = 0;
3941
3942 sp->pf = sctp_get_pf_specific(sk->sk_family);
3943
3944 /* Control variables for partial data delivery. */
3945 atomic_set(&sp->pd_mode, 0);
3946 skb_queue_head_init(&sp->pd_lobby);
3947 sp->frag_interleave = 0;
3948
3949 /* Create a per socket endpoint structure. Even if we
3950 * change the data structure relationships, this may still
3951 * be useful for storing pre-connect address information.
3952 */
3953 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
3954 if (!sp->ep)
3955 return -ENOMEM;
3956
3957 sp->hmac = NULL;
3958
3959 sk->sk_destruct = sctp_destruct_sock;
3960
3961 SCTP_DBG_OBJCNT_INC(sock);
3962
3963 local_bh_disable();
3964 percpu_counter_inc(&sctp_sockets_allocated);
3965 sock_prot_inuse_add(net, sk->sk_prot, 1);
3966 if (net->sctp.default_auto_asconf) {
3967 list_add_tail(&sp->auto_asconf_list,
3968 &net->sctp.auto_asconf_splist);
3969 sp->do_auto_asconf = 1;
3970 } else
3971 sp->do_auto_asconf = 0;
3972 local_bh_enable();
3973
3974 return 0;
3975 }
3976
3977 /* Cleanup any SCTP per socket resources. */
3978 static void sctp_destroy_sock(struct sock *sk)
3979 {
3980 struct sctp_sock *sp;
3981
3982 pr_debug("%s: sk:%p\n", __func__, sk);
3983
3984 /* Release our hold on the endpoint. */
3985 sp = sctp_sk(sk);
3986 /* This could happen during socket init, thus we bail out
3987 * early, since the rest of the below is not setup either.
3988 */
3989 if (sp->ep == NULL)
3990 return;
3991
3992 if (sp->do_auto_asconf) {
3993 sp->do_auto_asconf = 0;
3994 list_del(&sp->auto_asconf_list);
3995 }
3996 sctp_endpoint_free(sp->ep);
3997 local_bh_disable();
3998 percpu_counter_dec(&sctp_sockets_allocated);
3999 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4000 local_bh_enable();
4001 }
4002
4003 /* Triggered when there are no references on the socket anymore */
4004 static void sctp_destruct_sock(struct sock *sk)
4005 {
4006 struct sctp_sock *sp = sctp_sk(sk);
4007
4008 /* Free up the HMAC transform. */
4009 crypto_free_hash(sp->hmac);
4010
4011 inet_sock_destruct(sk);
4012 }
4013
4014 /* API 4.1.7 shutdown() - TCP Style Syntax
4015 * int shutdown(int socket, int how);
4016 *
4017 * sd - the socket descriptor of the association to be closed.
4018 * how - Specifies the type of shutdown. The values are
4019 * as follows:
4020 * SHUT_RD
4021 * Disables further receive operations. No SCTP
4022 * protocol action is taken.
4023 * SHUT_WR
4024 * Disables further send operations, and initiates
4025 * the SCTP shutdown sequence.
4026 * SHUT_RDWR
4027 * Disables further send and receive operations
4028 * and initiates the SCTP shutdown sequence.
4029 */
4030 static void sctp_shutdown(struct sock *sk, int how)
4031 {
4032 struct net *net = sock_net(sk);
4033 struct sctp_endpoint *ep;
4034 struct sctp_association *asoc;
4035
4036 if (!sctp_style(sk, TCP))
4037 return;
4038
4039 if (how & SEND_SHUTDOWN) {
4040 ep = sctp_sk(sk)->ep;
4041 if (!list_empty(&ep->asocs)) {
4042 asoc = list_entry(ep->asocs.next,
4043 struct sctp_association, asocs);
4044 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4045 }
4046 }
4047 }
4048
4049 /* 7.2.1 Association Status (SCTP_STATUS)
4050
4051 * Applications can retrieve current status information about an
4052 * association, including association state, peer receiver window size,
4053 * number of unacked data chunks, and number of data chunks pending
4054 * receipt. This information is read-only.
4055 */
4056 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4057 char __user *optval,
4058 int __user *optlen)
4059 {
4060 struct sctp_status status;
4061 struct sctp_association *asoc = NULL;
4062 struct sctp_transport *transport;
4063 sctp_assoc_t associd;
4064 int retval = 0;
4065
4066 if (len < sizeof(status)) {
4067 retval = -EINVAL;
4068 goto out;
4069 }
4070
4071 len = sizeof(status);
4072 if (copy_from_user(&status, optval, len)) {
4073 retval = -EFAULT;
4074 goto out;
4075 }
4076
4077 associd = status.sstat_assoc_id;
4078 asoc = sctp_id2assoc(sk, associd);
4079 if (!asoc) {
4080 retval = -EINVAL;
4081 goto out;
4082 }
4083
4084 transport = asoc->peer.primary_path;
4085
4086 status.sstat_assoc_id = sctp_assoc2id(asoc);
4087 status.sstat_state = asoc->state;
4088 status.sstat_rwnd = asoc->peer.rwnd;
4089 status.sstat_unackdata = asoc->unack_data;
4090
4091 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4092 status.sstat_instrms = asoc->c.sinit_max_instreams;
4093 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4094 status.sstat_fragmentation_point = asoc->frag_point;
4095 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4096 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4097 transport->af_specific->sockaddr_len);
4098 /* Map ipv4 address into v4-mapped-on-v6 address. */
4099 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4100 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4101 status.sstat_primary.spinfo_state = transport->state;
4102 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4103 status.sstat_primary.spinfo_srtt = transport->srtt;
4104 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4105 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4106
4107 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4108 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4109
4110 if (put_user(len, optlen)) {
4111 retval = -EFAULT;
4112 goto out;
4113 }
4114
4115 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4116 __func__, len, status.sstat_state, status.sstat_rwnd,
4117 status.sstat_assoc_id);
4118
4119 if (copy_to_user(optval, &status, len)) {
4120 retval = -EFAULT;
4121 goto out;
4122 }
4123
4124 out:
4125 return retval;
4126 }
4127
4128
4129 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4130 *
4131 * Applications can retrieve information about a specific peer address
4132 * of an association, including its reachability state, congestion
4133 * window, and retransmission timer values. This information is
4134 * read-only.
4135 */
4136 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4137 char __user *optval,
4138 int __user *optlen)
4139 {
4140 struct sctp_paddrinfo pinfo;
4141 struct sctp_transport *transport;
4142 int retval = 0;
4143
4144 if (len < sizeof(pinfo)) {
4145 retval = -EINVAL;
4146 goto out;
4147 }
4148
4149 len = sizeof(pinfo);
4150 if (copy_from_user(&pinfo, optval, len)) {
4151 retval = -EFAULT;
4152 goto out;
4153 }
4154
4155 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4156 pinfo.spinfo_assoc_id);
4157 if (!transport)
4158 return -EINVAL;
4159
4160 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4161 pinfo.spinfo_state = transport->state;
4162 pinfo.spinfo_cwnd = transport->cwnd;
4163 pinfo.spinfo_srtt = transport->srtt;
4164 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4165 pinfo.spinfo_mtu = transport->pathmtu;
4166
4167 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4168 pinfo.spinfo_state = SCTP_ACTIVE;
4169
4170 if (put_user(len, optlen)) {
4171 retval = -EFAULT;
4172 goto out;
4173 }
4174
4175 if (copy_to_user(optval, &pinfo, len)) {
4176 retval = -EFAULT;
4177 goto out;
4178 }
4179
4180 out:
4181 return retval;
4182 }
4183
4184 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4185 *
4186 * This option is a on/off flag. If enabled no SCTP message
4187 * fragmentation will be performed. Instead if a message being sent
4188 * exceeds the current PMTU size, the message will NOT be sent and
4189 * instead a error will be indicated to the user.
4190 */
4191 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4192 char __user *optval, int __user *optlen)
4193 {
4194 int val;
4195
4196 if (len < sizeof(int))
4197 return -EINVAL;
4198
4199 len = sizeof(int);
4200 val = (sctp_sk(sk)->disable_fragments == 1);
4201 if (put_user(len, optlen))
4202 return -EFAULT;
4203 if (copy_to_user(optval, &val, len))
4204 return -EFAULT;
4205 return 0;
4206 }
4207
4208 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4209 *
4210 * This socket option is used to specify various notifications and
4211 * ancillary data the user wishes to receive.
4212 */
4213 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4214 int __user *optlen)
4215 {
4216 if (len <= 0)
4217 return -EINVAL;
4218 if (len > sizeof(struct sctp_event_subscribe))
4219 len = sizeof(struct sctp_event_subscribe);
4220 if (put_user(len, optlen))
4221 return -EFAULT;
4222 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4223 return -EFAULT;
4224 return 0;
4225 }
4226
4227 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4228 *
4229 * This socket option is applicable to the UDP-style socket only. When
4230 * set it will cause associations that are idle for more than the
4231 * specified number of seconds to automatically close. An association
4232 * being idle is defined an association that has NOT sent or received
4233 * user data. The special value of '0' indicates that no automatic
4234 * close of any associations should be performed. The option expects an
4235 * integer defining the number of seconds of idle time before an
4236 * association is closed.
4237 */
4238 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4239 {
4240 /* Applicable to UDP-style socket only */
4241 if (sctp_style(sk, TCP))
4242 return -EOPNOTSUPP;
4243 if (len < sizeof(int))
4244 return -EINVAL;
4245 len = sizeof(int);
4246 if (put_user(len, optlen))
4247 return -EFAULT;
4248 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4249 return -EFAULT;
4250 return 0;
4251 }
4252
4253 /* Helper routine to branch off an association to a new socket. */
4254 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4255 {
4256 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4257 struct socket *sock;
4258 struct sctp_af *af;
4259 int err = 0;
4260
4261 if (!asoc)
4262 return -EINVAL;
4263
4264 /* An association cannot be branched off from an already peeled-off
4265 * socket, nor is this supported for tcp style sockets.
4266 */
4267 if (!sctp_style(sk, UDP))
4268 return -EINVAL;
4269
4270 /* Create a new socket. */
4271 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4272 if (err < 0)
4273 return err;
4274
4275 sctp_copy_sock(sock->sk, sk, asoc);
4276
4277 /* Make peeled-off sockets more like 1-1 accepted sockets.
4278 * Set the daddr and initialize id to something more random
4279 */
4280 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4281 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4282
4283 /* Populate the fields of the newsk from the oldsk and migrate the
4284 * asoc to the newsk.
4285 */
4286 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4287
4288 *sockp = sock;
4289
4290 return err;
4291 }
4292 EXPORT_SYMBOL(sctp_do_peeloff);
4293
4294 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4295 {
4296 sctp_peeloff_arg_t peeloff;
4297 struct socket *newsock;
4298 struct file *newfile;
4299 int retval = 0;
4300
4301 if (len < sizeof(sctp_peeloff_arg_t))
4302 return -EINVAL;
4303 len = sizeof(sctp_peeloff_arg_t);
4304 if (copy_from_user(&peeloff, optval, len))
4305 return -EFAULT;
4306
4307 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4308 if (retval < 0)
4309 goto out;
4310
4311 /* Map the socket to an unused fd that can be returned to the user. */
4312 retval = get_unused_fd_flags(0);
4313 if (retval < 0) {
4314 sock_release(newsock);
4315 goto out;
4316 }
4317
4318 newfile = sock_alloc_file(newsock, 0, NULL);
4319 if (unlikely(IS_ERR(newfile))) {
4320 put_unused_fd(retval);
4321 sock_release(newsock);
4322 return PTR_ERR(newfile);
4323 }
4324
4325 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4326 retval);
4327
4328 /* Return the fd mapped to the new socket. */
4329 if (put_user(len, optlen)) {
4330 fput(newfile);
4331 put_unused_fd(retval);
4332 return -EFAULT;
4333 }
4334 peeloff.sd = retval;
4335 if (copy_to_user(optval, &peeloff, len)) {
4336 fput(newfile);
4337 put_unused_fd(retval);
4338 return -EFAULT;
4339 }
4340 fd_install(retval, newfile);
4341 out:
4342 return retval;
4343 }
4344
4345 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4346 *
4347 * Applications can enable or disable heartbeats for any peer address of
4348 * an association, modify an address's heartbeat interval, force a
4349 * heartbeat to be sent immediately, and adjust the address's maximum
4350 * number of retransmissions sent before an address is considered
4351 * unreachable. The following structure is used to access and modify an
4352 * address's parameters:
4353 *
4354 * struct sctp_paddrparams {
4355 * sctp_assoc_t spp_assoc_id;
4356 * struct sockaddr_storage spp_address;
4357 * uint32_t spp_hbinterval;
4358 * uint16_t spp_pathmaxrxt;
4359 * uint32_t spp_pathmtu;
4360 * uint32_t spp_sackdelay;
4361 * uint32_t spp_flags;
4362 * };
4363 *
4364 * spp_assoc_id - (one-to-many style socket) This is filled in the
4365 * application, and identifies the association for
4366 * this query.
4367 * spp_address - This specifies which address is of interest.
4368 * spp_hbinterval - This contains the value of the heartbeat interval,
4369 * in milliseconds. If a value of zero
4370 * is present in this field then no changes are to
4371 * be made to this parameter.
4372 * spp_pathmaxrxt - This contains the maximum number of
4373 * retransmissions before this address shall be
4374 * considered unreachable. If a value of zero
4375 * is present in this field then no changes are to
4376 * be made to this parameter.
4377 * spp_pathmtu - When Path MTU discovery is disabled the value
4378 * specified here will be the "fixed" path mtu.
4379 * Note that if the spp_address field is empty
4380 * then all associations on this address will
4381 * have this fixed path mtu set upon them.
4382 *
4383 * spp_sackdelay - When delayed sack is enabled, this value specifies
4384 * the number of milliseconds that sacks will be delayed
4385 * for. This value will apply to all addresses of an
4386 * association if the spp_address field is empty. Note
4387 * also, that if delayed sack is enabled and this
4388 * value is set to 0, no change is made to the last
4389 * recorded delayed sack timer value.
4390 *
4391 * spp_flags - These flags are used to control various features
4392 * on an association. The flag field may contain
4393 * zero or more of the following options.
4394 *
4395 * SPP_HB_ENABLE - Enable heartbeats on the
4396 * specified address. Note that if the address
4397 * field is empty all addresses for the association
4398 * have heartbeats enabled upon them.
4399 *
4400 * SPP_HB_DISABLE - Disable heartbeats on the
4401 * speicifed address. Note that if the address
4402 * field is empty all addresses for the association
4403 * will have their heartbeats disabled. Note also
4404 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4405 * mutually exclusive, only one of these two should
4406 * be specified. Enabling both fields will have
4407 * undetermined results.
4408 *
4409 * SPP_HB_DEMAND - Request a user initiated heartbeat
4410 * to be made immediately.
4411 *
4412 * SPP_PMTUD_ENABLE - This field will enable PMTU
4413 * discovery upon the specified address. Note that
4414 * if the address feild is empty then all addresses
4415 * on the association are effected.
4416 *
4417 * SPP_PMTUD_DISABLE - This field will disable PMTU
4418 * discovery upon the specified address. Note that
4419 * if the address feild is empty then all addresses
4420 * on the association are effected. Not also that
4421 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4422 * exclusive. Enabling both will have undetermined
4423 * results.
4424 *
4425 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4426 * on delayed sack. The time specified in spp_sackdelay
4427 * is used to specify the sack delay for this address. Note
4428 * that if spp_address is empty then all addresses will
4429 * enable delayed sack and take on the sack delay
4430 * value specified in spp_sackdelay.
4431 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4432 * off delayed sack. If the spp_address field is blank then
4433 * delayed sack is disabled for the entire association. Note
4434 * also that this field is mutually exclusive to
4435 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4436 * results.
4437 */
4438 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4439 char __user *optval, int __user *optlen)
4440 {
4441 struct sctp_paddrparams params;
4442 struct sctp_transport *trans = NULL;
4443 struct sctp_association *asoc = NULL;
4444 struct sctp_sock *sp = sctp_sk(sk);
4445
4446 if (len < sizeof(struct sctp_paddrparams))
4447 return -EINVAL;
4448 len = sizeof(struct sctp_paddrparams);
4449 if (copy_from_user(&params, optval, len))
4450 return -EFAULT;
4451
4452 /* If an address other than INADDR_ANY is specified, and
4453 * no transport is found, then the request is invalid.
4454 */
4455 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
4456 trans = sctp_addr_id2transport(sk, &params.spp_address,
4457 params.spp_assoc_id);
4458 if (!trans) {
4459 pr_debug("%s: failed no transport\n", __func__);
4460 return -EINVAL;
4461 }
4462 }
4463
4464 /* Get association, if assoc_id != 0 and the socket is a one
4465 * to many style socket, and an association was not found, then
4466 * the id was invalid.
4467 */
4468 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4469 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4470 pr_debug("%s: failed no association\n", __func__);
4471 return -EINVAL;
4472 }
4473
4474 if (trans) {
4475 /* Fetch transport values. */
4476 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4477 params.spp_pathmtu = trans->pathmtu;
4478 params.spp_pathmaxrxt = trans->pathmaxrxt;
4479 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4480
4481 /*draft-11 doesn't say what to return in spp_flags*/
4482 params.spp_flags = trans->param_flags;
4483 } else if (asoc) {
4484 /* Fetch association values. */
4485 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4486 params.spp_pathmtu = asoc->pathmtu;
4487 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4488 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4489
4490 /*draft-11 doesn't say what to return in spp_flags*/
4491 params.spp_flags = asoc->param_flags;
4492 } else {
4493 /* Fetch socket values. */
4494 params.spp_hbinterval = sp->hbinterval;
4495 params.spp_pathmtu = sp->pathmtu;
4496 params.spp_sackdelay = sp->sackdelay;
4497 params.spp_pathmaxrxt = sp->pathmaxrxt;
4498
4499 /*draft-11 doesn't say what to return in spp_flags*/
4500 params.spp_flags = sp->param_flags;
4501 }
4502
4503 if (copy_to_user(optval, &params, len))
4504 return -EFAULT;
4505
4506 if (put_user(len, optlen))
4507 return -EFAULT;
4508
4509 return 0;
4510 }
4511
4512 /*
4513 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4514 *
4515 * This option will effect the way delayed acks are performed. This
4516 * option allows you to get or set the delayed ack time, in
4517 * milliseconds. It also allows changing the delayed ack frequency.
4518 * Changing the frequency to 1 disables the delayed sack algorithm. If
4519 * the assoc_id is 0, then this sets or gets the endpoints default
4520 * values. If the assoc_id field is non-zero, then the set or get
4521 * effects the specified association for the one to many model (the
4522 * assoc_id field is ignored by the one to one model). Note that if
4523 * sack_delay or sack_freq are 0 when setting this option, then the
4524 * current values will remain unchanged.
4525 *
4526 * struct sctp_sack_info {
4527 * sctp_assoc_t sack_assoc_id;
4528 * uint32_t sack_delay;
4529 * uint32_t sack_freq;
4530 * };
4531 *
4532 * sack_assoc_id - This parameter, indicates which association the user
4533 * is performing an action upon. Note that if this field's value is
4534 * zero then the endpoints default value is changed (effecting future
4535 * associations only).
4536 *
4537 * sack_delay - This parameter contains the number of milliseconds that
4538 * the user is requesting the delayed ACK timer be set to. Note that
4539 * this value is defined in the standard to be between 200 and 500
4540 * milliseconds.
4541 *
4542 * sack_freq - This parameter contains the number of packets that must
4543 * be received before a sack is sent without waiting for the delay
4544 * timer to expire. The default value for this is 2, setting this
4545 * value to 1 will disable the delayed sack algorithm.
4546 */
4547 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4548 char __user *optval,
4549 int __user *optlen)
4550 {
4551 struct sctp_sack_info params;
4552 struct sctp_association *asoc = NULL;
4553 struct sctp_sock *sp = sctp_sk(sk);
4554
4555 if (len >= sizeof(struct sctp_sack_info)) {
4556 len = sizeof(struct sctp_sack_info);
4557
4558 if (copy_from_user(&params, optval, len))
4559 return -EFAULT;
4560 } else if (len == sizeof(struct sctp_assoc_value)) {
4561 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4562 pr_warn("Use struct sctp_sack_info instead\n");
4563 if (copy_from_user(&params, optval, len))
4564 return -EFAULT;
4565 } else
4566 return - EINVAL;
4567
4568 /* Get association, if sack_assoc_id != 0 and the socket is a one
4569 * to many style socket, and an association was not found, then
4570 * the id was invalid.
4571 */
4572 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4573 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4574 return -EINVAL;
4575
4576 if (asoc) {
4577 /* Fetch association values. */
4578 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4579 params.sack_delay = jiffies_to_msecs(
4580 asoc->sackdelay);
4581 params.sack_freq = asoc->sackfreq;
4582
4583 } else {
4584 params.sack_delay = 0;
4585 params.sack_freq = 1;
4586 }
4587 } else {
4588 /* Fetch socket values. */
4589 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4590 params.sack_delay = sp->sackdelay;
4591 params.sack_freq = sp->sackfreq;
4592 } else {
4593 params.sack_delay = 0;
4594 params.sack_freq = 1;
4595 }
4596 }
4597
4598 if (copy_to_user(optval, &params, len))
4599 return -EFAULT;
4600
4601 if (put_user(len, optlen))
4602 return -EFAULT;
4603
4604 return 0;
4605 }
4606
4607 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4608 *
4609 * Applications can specify protocol parameters for the default association
4610 * initialization. The option name argument to setsockopt() and getsockopt()
4611 * is SCTP_INITMSG.
4612 *
4613 * Setting initialization parameters is effective only on an unconnected
4614 * socket (for UDP-style sockets only future associations are effected
4615 * by the change). With TCP-style sockets, this option is inherited by
4616 * sockets derived from a listener socket.
4617 */
4618 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4619 {
4620 if (len < sizeof(struct sctp_initmsg))
4621 return -EINVAL;
4622 len = sizeof(struct sctp_initmsg);
4623 if (put_user(len, optlen))
4624 return -EFAULT;
4625 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4626 return -EFAULT;
4627 return 0;
4628 }
4629
4630
4631 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4632 char __user *optval, int __user *optlen)
4633 {
4634 struct sctp_association *asoc;
4635 int cnt = 0;
4636 struct sctp_getaddrs getaddrs;
4637 struct sctp_transport *from;
4638 void __user *to;
4639 union sctp_addr temp;
4640 struct sctp_sock *sp = sctp_sk(sk);
4641 int addrlen;
4642 size_t space_left;
4643 int bytes_copied;
4644
4645 if (len < sizeof(struct sctp_getaddrs))
4646 return -EINVAL;
4647
4648 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4649 return -EFAULT;
4650
4651 /* For UDP-style sockets, id specifies the association to query. */
4652 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4653 if (!asoc)
4654 return -EINVAL;
4655
4656 to = optval + offsetof(struct sctp_getaddrs,addrs);
4657 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4658
4659 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4660 transports) {
4661 memcpy(&temp, &from->ipaddr, sizeof(temp));
4662 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4663 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4664 if (space_left < addrlen)
4665 return -ENOMEM;
4666 if (copy_to_user(to, &temp, addrlen))
4667 return -EFAULT;
4668 to += addrlen;
4669 cnt++;
4670 space_left -= addrlen;
4671 }
4672
4673 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4674 return -EFAULT;
4675 bytes_copied = ((char __user *)to) - optval;
4676 if (put_user(bytes_copied, optlen))
4677 return -EFAULT;
4678
4679 return 0;
4680 }
4681
4682 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4683 size_t space_left, int *bytes_copied)
4684 {
4685 struct sctp_sockaddr_entry *addr;
4686 union sctp_addr temp;
4687 int cnt = 0;
4688 int addrlen;
4689 struct net *net = sock_net(sk);
4690
4691 rcu_read_lock();
4692 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
4693 if (!addr->valid)
4694 continue;
4695
4696 if ((PF_INET == sk->sk_family) &&
4697 (AF_INET6 == addr->a.sa.sa_family))
4698 continue;
4699 if ((PF_INET6 == sk->sk_family) &&
4700 inet_v6_ipv6only(sk) &&
4701 (AF_INET == addr->a.sa.sa_family))
4702 continue;
4703 memcpy(&temp, &addr->a, sizeof(temp));
4704 if (!temp.v4.sin_port)
4705 temp.v4.sin_port = htons(port);
4706
4707 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4708 &temp);
4709 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4710 if (space_left < addrlen) {
4711 cnt = -ENOMEM;
4712 break;
4713 }
4714 memcpy(to, &temp, addrlen);
4715
4716 to += addrlen;
4717 cnt ++;
4718 space_left -= addrlen;
4719 *bytes_copied += addrlen;
4720 }
4721 rcu_read_unlock();
4722
4723 return cnt;
4724 }
4725
4726
4727 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4728 char __user *optval, int __user *optlen)
4729 {
4730 struct sctp_bind_addr *bp;
4731 struct sctp_association *asoc;
4732 int cnt = 0;
4733 struct sctp_getaddrs getaddrs;
4734 struct sctp_sockaddr_entry *addr;
4735 void __user *to;
4736 union sctp_addr temp;
4737 struct sctp_sock *sp = sctp_sk(sk);
4738 int addrlen;
4739 int err = 0;
4740 size_t space_left;
4741 int bytes_copied = 0;
4742 void *addrs;
4743 void *buf;
4744
4745 if (len < sizeof(struct sctp_getaddrs))
4746 return -EINVAL;
4747
4748 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4749 return -EFAULT;
4750
4751 /*
4752 * For UDP-style sockets, id specifies the association to query.
4753 * If the id field is set to the value '0' then the locally bound
4754 * addresses are returned without regard to any particular
4755 * association.
4756 */
4757 if (0 == getaddrs.assoc_id) {
4758 bp = &sctp_sk(sk)->ep->base.bind_addr;
4759 } else {
4760 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4761 if (!asoc)
4762 return -EINVAL;
4763 bp = &asoc->base.bind_addr;
4764 }
4765
4766 to = optval + offsetof(struct sctp_getaddrs,addrs);
4767 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4768
4769 addrs = kmalloc(space_left, GFP_KERNEL);
4770 if (!addrs)
4771 return -ENOMEM;
4772
4773 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4774 * addresses from the global local address list.
4775 */
4776 if (sctp_list_single_entry(&bp->address_list)) {
4777 addr = list_entry(bp->address_list.next,
4778 struct sctp_sockaddr_entry, list);
4779 if (sctp_is_any(sk, &addr->a)) {
4780 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4781 space_left, &bytes_copied);
4782 if (cnt < 0) {
4783 err = cnt;
4784 goto out;
4785 }
4786 goto copy_getaddrs;
4787 }
4788 }
4789
4790 buf = addrs;
4791 /* Protection on the bound address list is not needed since
4792 * in the socket option context we hold a socket lock and
4793 * thus the bound address list can't change.
4794 */
4795 list_for_each_entry(addr, &bp->address_list, list) {
4796 memcpy(&temp, &addr->a, sizeof(temp));
4797 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4798 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4799 if (space_left < addrlen) {
4800 err = -ENOMEM; /*fixme: right error?*/
4801 goto out;
4802 }
4803 memcpy(buf, &temp, addrlen);
4804 buf += addrlen;
4805 bytes_copied += addrlen;
4806 cnt ++;
4807 space_left -= addrlen;
4808 }
4809
4810 copy_getaddrs:
4811 if (copy_to_user(to, addrs, bytes_copied)) {
4812 err = -EFAULT;
4813 goto out;
4814 }
4815 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4816 err = -EFAULT;
4817 goto out;
4818 }
4819 if (put_user(bytes_copied, optlen))
4820 err = -EFAULT;
4821 out:
4822 kfree(addrs);
4823 return err;
4824 }
4825
4826 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4827 *
4828 * Requests that the local SCTP stack use the enclosed peer address as
4829 * the association primary. The enclosed address must be one of the
4830 * association peer's addresses.
4831 */
4832 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4833 char __user *optval, int __user *optlen)
4834 {
4835 struct sctp_prim prim;
4836 struct sctp_association *asoc;
4837 struct sctp_sock *sp = sctp_sk(sk);
4838
4839 if (len < sizeof(struct sctp_prim))
4840 return -EINVAL;
4841
4842 len = sizeof(struct sctp_prim);
4843
4844 if (copy_from_user(&prim, optval, len))
4845 return -EFAULT;
4846
4847 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4848 if (!asoc)
4849 return -EINVAL;
4850
4851 if (!asoc->peer.primary_path)
4852 return -ENOTCONN;
4853
4854 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4855 asoc->peer.primary_path->af_specific->sockaddr_len);
4856
4857 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4858 (union sctp_addr *)&prim.ssp_addr);
4859
4860 if (put_user(len, optlen))
4861 return -EFAULT;
4862 if (copy_to_user(optval, &prim, len))
4863 return -EFAULT;
4864
4865 return 0;
4866 }
4867
4868 /*
4869 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4870 *
4871 * Requests that the local endpoint set the specified Adaptation Layer
4872 * Indication parameter for all future INIT and INIT-ACK exchanges.
4873 */
4874 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4875 char __user *optval, int __user *optlen)
4876 {
4877 struct sctp_setadaptation adaptation;
4878
4879 if (len < sizeof(struct sctp_setadaptation))
4880 return -EINVAL;
4881
4882 len = sizeof(struct sctp_setadaptation);
4883
4884 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4885
4886 if (put_user(len, optlen))
4887 return -EFAULT;
4888 if (copy_to_user(optval, &adaptation, len))
4889 return -EFAULT;
4890
4891 return 0;
4892 }
4893
4894 /*
4895 *
4896 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4897 *
4898 * Applications that wish to use the sendto() system call may wish to
4899 * specify a default set of parameters that would normally be supplied
4900 * through the inclusion of ancillary data. This socket option allows
4901 * such an application to set the default sctp_sndrcvinfo structure.
4902
4903
4904 * The application that wishes to use this socket option simply passes
4905 * in to this call the sctp_sndrcvinfo structure defined in Section
4906 * 5.2.2) The input parameters accepted by this call include
4907 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4908 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4909 * to this call if the caller is using the UDP model.
4910 *
4911 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4912 */
4913 static int sctp_getsockopt_default_send_param(struct sock *sk,
4914 int len, char __user *optval,
4915 int __user *optlen)
4916 {
4917 struct sctp_sndrcvinfo info;
4918 struct sctp_association *asoc;
4919 struct sctp_sock *sp = sctp_sk(sk);
4920
4921 if (len < sizeof(struct sctp_sndrcvinfo))
4922 return -EINVAL;
4923
4924 len = sizeof(struct sctp_sndrcvinfo);
4925
4926 if (copy_from_user(&info, optval, len))
4927 return -EFAULT;
4928
4929 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4930 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4931 return -EINVAL;
4932
4933 if (asoc) {
4934 info.sinfo_stream = asoc->default_stream;
4935 info.sinfo_flags = asoc->default_flags;
4936 info.sinfo_ppid = asoc->default_ppid;
4937 info.sinfo_context = asoc->default_context;
4938 info.sinfo_timetolive = asoc->default_timetolive;
4939 } else {
4940 info.sinfo_stream = sp->default_stream;
4941 info.sinfo_flags = sp->default_flags;
4942 info.sinfo_ppid = sp->default_ppid;
4943 info.sinfo_context = sp->default_context;
4944 info.sinfo_timetolive = sp->default_timetolive;
4945 }
4946
4947 if (put_user(len, optlen))
4948 return -EFAULT;
4949 if (copy_to_user(optval, &info, len))
4950 return -EFAULT;
4951
4952 return 0;
4953 }
4954
4955 /*
4956 *
4957 * 7.1.5 SCTP_NODELAY
4958 *
4959 * Turn on/off any Nagle-like algorithm. This means that packets are
4960 * generally sent as soon as possible and no unnecessary delays are
4961 * introduced, at the cost of more packets in the network. Expects an
4962 * integer boolean flag.
4963 */
4964
4965 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4966 char __user *optval, int __user *optlen)
4967 {
4968 int val;
4969
4970 if (len < sizeof(int))
4971 return -EINVAL;
4972
4973 len = sizeof(int);
4974 val = (sctp_sk(sk)->nodelay == 1);
4975 if (put_user(len, optlen))
4976 return -EFAULT;
4977 if (copy_to_user(optval, &val, len))
4978 return -EFAULT;
4979 return 0;
4980 }
4981
4982 /*
4983 *
4984 * 7.1.1 SCTP_RTOINFO
4985 *
4986 * The protocol parameters used to initialize and bound retransmission
4987 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4988 * and modify these parameters.
4989 * All parameters are time values, in milliseconds. A value of 0, when
4990 * modifying the parameters, indicates that the current value should not
4991 * be changed.
4992 *
4993 */
4994 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4995 char __user *optval,
4996 int __user *optlen) {
4997 struct sctp_rtoinfo rtoinfo;
4998 struct sctp_association *asoc;
4999
5000 if (len < sizeof (struct sctp_rtoinfo))
5001 return -EINVAL;
5002
5003 len = sizeof(struct sctp_rtoinfo);
5004
5005 if (copy_from_user(&rtoinfo, optval, len))
5006 return -EFAULT;
5007
5008 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5009
5010 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5011 return -EINVAL;
5012
5013 /* Values corresponding to the specific association. */
5014 if (asoc) {
5015 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5016 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5017 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5018 } else {
5019 /* Values corresponding to the endpoint. */
5020 struct sctp_sock *sp = sctp_sk(sk);
5021
5022 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5023 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5024 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5025 }
5026
5027 if (put_user(len, optlen))
5028 return -EFAULT;
5029
5030 if (copy_to_user(optval, &rtoinfo, len))
5031 return -EFAULT;
5032
5033 return 0;
5034 }
5035
5036 /*
5037 *
5038 * 7.1.2 SCTP_ASSOCINFO
5039 *
5040 * This option is used to tune the maximum retransmission attempts
5041 * of the association.
5042 * Returns an error if the new association retransmission value is
5043 * greater than the sum of the retransmission value of the peer.
5044 * See [SCTP] for more information.
5045 *
5046 */
5047 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5048 char __user *optval,
5049 int __user *optlen)
5050 {
5051
5052 struct sctp_assocparams assocparams;
5053 struct sctp_association *asoc;
5054 struct list_head *pos;
5055 int cnt = 0;
5056
5057 if (len < sizeof (struct sctp_assocparams))
5058 return -EINVAL;
5059
5060 len = sizeof(struct sctp_assocparams);
5061
5062 if (copy_from_user(&assocparams, optval, len))
5063 return -EFAULT;
5064
5065 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5066
5067 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5068 return -EINVAL;
5069
5070 /* Values correspoinding to the specific association */
5071 if (asoc) {
5072 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5073 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5074 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5075 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5076
5077 list_for_each(pos, &asoc->peer.transport_addr_list) {
5078 cnt ++;
5079 }
5080
5081 assocparams.sasoc_number_peer_destinations = cnt;
5082 } else {
5083 /* Values corresponding to the endpoint */
5084 struct sctp_sock *sp = sctp_sk(sk);
5085
5086 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5087 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5088 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5089 assocparams.sasoc_cookie_life =
5090 sp->assocparams.sasoc_cookie_life;
5091 assocparams.sasoc_number_peer_destinations =
5092 sp->assocparams.
5093 sasoc_number_peer_destinations;
5094 }
5095
5096 if (put_user(len, optlen))
5097 return -EFAULT;
5098
5099 if (copy_to_user(optval, &assocparams, len))
5100 return -EFAULT;
5101
5102 return 0;
5103 }
5104
5105 /*
5106 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5107 *
5108 * This socket option is a boolean flag which turns on or off mapped V4
5109 * addresses. If this option is turned on and the socket is type
5110 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5111 * If this option is turned off, then no mapping will be done of V4
5112 * addresses and a user will receive both PF_INET6 and PF_INET type
5113 * addresses on the socket.
5114 */
5115 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5116 char __user *optval, int __user *optlen)
5117 {
5118 int val;
5119 struct sctp_sock *sp = sctp_sk(sk);
5120
5121 if (len < sizeof(int))
5122 return -EINVAL;
5123
5124 len = sizeof(int);
5125 val = sp->v4mapped;
5126 if (put_user(len, optlen))
5127 return -EFAULT;
5128 if (copy_to_user(optval, &val, len))
5129 return -EFAULT;
5130
5131 return 0;
5132 }
5133
5134 /*
5135 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5136 * (chapter and verse is quoted at sctp_setsockopt_context())
5137 */
5138 static int sctp_getsockopt_context(struct sock *sk, int len,
5139 char __user *optval, int __user *optlen)
5140 {
5141 struct sctp_assoc_value params;
5142 struct sctp_sock *sp;
5143 struct sctp_association *asoc;
5144
5145 if (len < sizeof(struct sctp_assoc_value))
5146 return -EINVAL;
5147
5148 len = sizeof(struct sctp_assoc_value);
5149
5150 if (copy_from_user(&params, optval, len))
5151 return -EFAULT;
5152
5153 sp = sctp_sk(sk);
5154
5155 if (params.assoc_id != 0) {
5156 asoc = sctp_id2assoc(sk, params.assoc_id);
5157 if (!asoc)
5158 return -EINVAL;
5159 params.assoc_value = asoc->default_rcv_context;
5160 } else {
5161 params.assoc_value = sp->default_rcv_context;
5162 }
5163
5164 if (put_user(len, optlen))
5165 return -EFAULT;
5166 if (copy_to_user(optval, &params, len))
5167 return -EFAULT;
5168
5169 return 0;
5170 }
5171
5172 /*
5173 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5174 * This option will get or set the maximum size to put in any outgoing
5175 * SCTP DATA chunk. If a message is larger than this size it will be
5176 * fragmented by SCTP into the specified size. Note that the underlying
5177 * SCTP implementation may fragment into smaller sized chunks when the
5178 * PMTU of the underlying association is smaller than the value set by
5179 * the user. The default value for this option is '0' which indicates
5180 * the user is NOT limiting fragmentation and only the PMTU will effect
5181 * SCTP's choice of DATA chunk size. Note also that values set larger
5182 * than the maximum size of an IP datagram will effectively let SCTP
5183 * control fragmentation (i.e. the same as setting this option to 0).
5184 *
5185 * The following structure is used to access and modify this parameter:
5186 *
5187 * struct sctp_assoc_value {
5188 * sctp_assoc_t assoc_id;
5189 * uint32_t assoc_value;
5190 * };
5191 *
5192 * assoc_id: This parameter is ignored for one-to-one style sockets.
5193 * For one-to-many style sockets this parameter indicates which
5194 * association the user is performing an action upon. Note that if
5195 * this field's value is zero then the endpoints default value is
5196 * changed (effecting future associations only).
5197 * assoc_value: This parameter specifies the maximum size in bytes.
5198 */
5199 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5200 char __user *optval, int __user *optlen)
5201 {
5202 struct sctp_assoc_value params;
5203 struct sctp_association *asoc;
5204
5205 if (len == sizeof(int)) {
5206 pr_warn("Use of int in maxseg socket option deprecated\n");
5207 pr_warn("Use struct sctp_assoc_value instead\n");
5208 params.assoc_id = 0;
5209 } else if (len >= sizeof(struct sctp_assoc_value)) {
5210 len = sizeof(struct sctp_assoc_value);
5211 if (copy_from_user(&params, optval, sizeof(params)))
5212 return -EFAULT;
5213 } else
5214 return -EINVAL;
5215
5216 asoc = sctp_id2assoc(sk, params.assoc_id);
5217 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5218 return -EINVAL;
5219
5220 if (asoc)
5221 params.assoc_value = asoc->frag_point;
5222 else
5223 params.assoc_value = sctp_sk(sk)->user_frag;
5224
5225 if (put_user(len, optlen))
5226 return -EFAULT;
5227 if (len == sizeof(int)) {
5228 if (copy_to_user(optval, &params.assoc_value, len))
5229 return -EFAULT;
5230 } else {
5231 if (copy_to_user(optval, &params, len))
5232 return -EFAULT;
5233 }
5234
5235 return 0;
5236 }
5237
5238 /*
5239 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5240 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5241 */
5242 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5243 char __user *optval, int __user *optlen)
5244 {
5245 int val;
5246
5247 if (len < sizeof(int))
5248 return -EINVAL;
5249
5250 len = sizeof(int);
5251
5252 val = sctp_sk(sk)->frag_interleave;
5253 if (put_user(len, optlen))
5254 return -EFAULT;
5255 if (copy_to_user(optval, &val, len))
5256 return -EFAULT;
5257
5258 return 0;
5259 }
5260
5261 /*
5262 * 7.1.25. Set or Get the sctp partial delivery point
5263 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5264 */
5265 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5266 char __user *optval,
5267 int __user *optlen)
5268 {
5269 u32 val;
5270
5271 if (len < sizeof(u32))
5272 return -EINVAL;
5273
5274 len = sizeof(u32);
5275
5276 val = sctp_sk(sk)->pd_point;
5277 if (put_user(len, optlen))
5278 return -EFAULT;
5279 if (copy_to_user(optval, &val, len))
5280 return -EFAULT;
5281
5282 return 0;
5283 }
5284
5285 /*
5286 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5287 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5288 */
5289 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5290 char __user *optval,
5291 int __user *optlen)
5292 {
5293 struct sctp_assoc_value params;
5294 struct sctp_sock *sp;
5295 struct sctp_association *asoc;
5296
5297 if (len == sizeof(int)) {
5298 pr_warn("Use of int in max_burst socket option deprecated\n");
5299 pr_warn("Use struct sctp_assoc_value instead\n");
5300 params.assoc_id = 0;
5301 } else if (len >= sizeof(struct sctp_assoc_value)) {
5302 len = sizeof(struct sctp_assoc_value);
5303 if (copy_from_user(&params, optval, len))
5304 return -EFAULT;
5305 } else
5306 return -EINVAL;
5307
5308 sp = sctp_sk(sk);
5309
5310 if (params.assoc_id != 0) {
5311 asoc = sctp_id2assoc(sk, params.assoc_id);
5312 if (!asoc)
5313 return -EINVAL;
5314 params.assoc_value = asoc->max_burst;
5315 } else
5316 params.assoc_value = sp->max_burst;
5317
5318 if (len == sizeof(int)) {
5319 if (copy_to_user(optval, &params.assoc_value, len))
5320 return -EFAULT;
5321 } else {
5322 if (copy_to_user(optval, &params, len))
5323 return -EFAULT;
5324 }
5325
5326 return 0;
5327
5328 }
5329
5330 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5331 char __user *optval, int __user *optlen)
5332 {
5333 struct net *net = sock_net(sk);
5334 struct sctp_hmacalgo __user *p = (void __user *)optval;
5335 struct sctp_hmac_algo_param *hmacs;
5336 __u16 data_len = 0;
5337 u32 num_idents;
5338
5339 if (!net->sctp.auth_enable)
5340 return -EACCES;
5341
5342 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5343 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5344
5345 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5346 return -EINVAL;
5347
5348 len = sizeof(struct sctp_hmacalgo) + data_len;
5349 num_idents = data_len / sizeof(u16);
5350
5351 if (put_user(len, optlen))
5352 return -EFAULT;
5353 if (put_user(num_idents, &p->shmac_num_idents))
5354 return -EFAULT;
5355 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5356 return -EFAULT;
5357 return 0;
5358 }
5359
5360 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5361 char __user *optval, int __user *optlen)
5362 {
5363 struct net *net = sock_net(sk);
5364 struct sctp_authkeyid val;
5365 struct sctp_association *asoc;
5366
5367 if (!net->sctp.auth_enable)
5368 return -EACCES;
5369
5370 if (len < sizeof(struct sctp_authkeyid))
5371 return -EINVAL;
5372 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5373 return -EFAULT;
5374
5375 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5376 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5377 return -EINVAL;
5378
5379 if (asoc)
5380 val.scact_keynumber = asoc->active_key_id;
5381 else
5382 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5383
5384 len = sizeof(struct sctp_authkeyid);
5385 if (put_user(len, optlen))
5386 return -EFAULT;
5387 if (copy_to_user(optval, &val, len))
5388 return -EFAULT;
5389
5390 return 0;
5391 }
5392
5393 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5394 char __user *optval, int __user *optlen)
5395 {
5396 struct net *net = sock_net(sk);
5397 struct sctp_authchunks __user *p = (void __user *)optval;
5398 struct sctp_authchunks val;
5399 struct sctp_association *asoc;
5400 struct sctp_chunks_param *ch;
5401 u32 num_chunks = 0;
5402 char __user *to;
5403
5404 if (!net->sctp.auth_enable)
5405 return -EACCES;
5406
5407 if (len < sizeof(struct sctp_authchunks))
5408 return -EINVAL;
5409
5410 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5411 return -EFAULT;
5412
5413 to = p->gauth_chunks;
5414 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5415 if (!asoc)
5416 return -EINVAL;
5417
5418 ch = asoc->peer.peer_chunks;
5419 if (!ch)
5420 goto num;
5421
5422 /* See if the user provided enough room for all the data */
5423 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5424 if (len < num_chunks)
5425 return -EINVAL;
5426
5427 if (copy_to_user(to, ch->chunks, num_chunks))
5428 return -EFAULT;
5429 num:
5430 len = sizeof(struct sctp_authchunks) + num_chunks;
5431 if (put_user(len, optlen)) return -EFAULT;
5432 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5433 return -EFAULT;
5434 return 0;
5435 }
5436
5437 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5438 char __user *optval, int __user *optlen)
5439 {
5440 struct net *net = sock_net(sk);
5441 struct sctp_authchunks __user *p = (void __user *)optval;
5442 struct sctp_authchunks val;
5443 struct sctp_association *asoc;
5444 struct sctp_chunks_param *ch;
5445 u32 num_chunks = 0;
5446 char __user *to;
5447
5448 if (!net->sctp.auth_enable)
5449 return -EACCES;
5450
5451 if (len < sizeof(struct sctp_authchunks))
5452 return -EINVAL;
5453
5454 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5455 return -EFAULT;
5456
5457 to = p->gauth_chunks;
5458 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5459 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5460 return -EINVAL;
5461
5462 if (asoc)
5463 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5464 else
5465 ch = sctp_sk(sk)->ep->auth_chunk_list;
5466
5467 if (!ch)
5468 goto num;
5469
5470 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5471 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5472 return -EINVAL;
5473
5474 if (copy_to_user(to, ch->chunks, num_chunks))
5475 return -EFAULT;
5476 num:
5477 len = sizeof(struct sctp_authchunks) + num_chunks;
5478 if (put_user(len, optlen))
5479 return -EFAULT;
5480 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5481 return -EFAULT;
5482
5483 return 0;
5484 }
5485
5486 /*
5487 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5488 * This option gets the current number of associations that are attached
5489 * to a one-to-many style socket. The option value is an uint32_t.
5490 */
5491 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5492 char __user *optval, int __user *optlen)
5493 {
5494 struct sctp_sock *sp = sctp_sk(sk);
5495 struct sctp_association *asoc;
5496 u32 val = 0;
5497
5498 if (sctp_style(sk, TCP))
5499 return -EOPNOTSUPP;
5500
5501 if (len < sizeof(u32))
5502 return -EINVAL;
5503
5504 len = sizeof(u32);
5505
5506 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5507 val++;
5508 }
5509
5510 if (put_user(len, optlen))
5511 return -EFAULT;
5512 if (copy_to_user(optval, &val, len))
5513 return -EFAULT;
5514
5515 return 0;
5516 }
5517
5518 /*
5519 * 8.1.23 SCTP_AUTO_ASCONF
5520 * See the corresponding setsockopt entry as description
5521 */
5522 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5523 char __user *optval, int __user *optlen)
5524 {
5525 int val = 0;
5526
5527 if (len < sizeof(int))
5528 return -EINVAL;
5529
5530 len = sizeof(int);
5531 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5532 val = 1;
5533 if (put_user(len, optlen))
5534 return -EFAULT;
5535 if (copy_to_user(optval, &val, len))
5536 return -EFAULT;
5537 return 0;
5538 }
5539
5540 /*
5541 * 8.2.6. Get the Current Identifiers of Associations
5542 * (SCTP_GET_ASSOC_ID_LIST)
5543 *
5544 * This option gets the current list of SCTP association identifiers of
5545 * the SCTP associations handled by a one-to-many style socket.
5546 */
5547 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5548 char __user *optval, int __user *optlen)
5549 {
5550 struct sctp_sock *sp = sctp_sk(sk);
5551 struct sctp_association *asoc;
5552 struct sctp_assoc_ids *ids;
5553 u32 num = 0;
5554
5555 if (sctp_style(sk, TCP))
5556 return -EOPNOTSUPP;
5557
5558 if (len < sizeof(struct sctp_assoc_ids))
5559 return -EINVAL;
5560
5561 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5562 num++;
5563 }
5564
5565 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5566 return -EINVAL;
5567
5568 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5569
5570 ids = kmalloc(len, GFP_KERNEL);
5571 if (unlikely(!ids))
5572 return -ENOMEM;
5573
5574 ids->gaids_number_of_ids = num;
5575 num = 0;
5576 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5577 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5578 }
5579
5580 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5581 kfree(ids);
5582 return -EFAULT;
5583 }
5584
5585 kfree(ids);
5586 return 0;
5587 }
5588
5589 /*
5590 * SCTP_PEER_ADDR_THLDS
5591 *
5592 * This option allows us to fetch the partially failed threshold for one or all
5593 * transports in an association. See Section 6.1 of:
5594 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5595 */
5596 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
5597 char __user *optval,
5598 int len,
5599 int __user *optlen)
5600 {
5601 struct sctp_paddrthlds val;
5602 struct sctp_transport *trans;
5603 struct sctp_association *asoc;
5604
5605 if (len < sizeof(struct sctp_paddrthlds))
5606 return -EINVAL;
5607 len = sizeof(struct sctp_paddrthlds);
5608 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
5609 return -EFAULT;
5610
5611 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
5612 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
5613 if (!asoc)
5614 return -ENOENT;
5615
5616 val.spt_pathpfthld = asoc->pf_retrans;
5617 val.spt_pathmaxrxt = asoc->pathmaxrxt;
5618 } else {
5619 trans = sctp_addr_id2transport(sk, &val.spt_address,
5620 val.spt_assoc_id);
5621 if (!trans)
5622 return -ENOENT;
5623
5624 val.spt_pathmaxrxt = trans->pathmaxrxt;
5625 val.spt_pathpfthld = trans->pf_retrans;
5626 }
5627
5628 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
5629 return -EFAULT;
5630
5631 return 0;
5632 }
5633
5634 /*
5635 * SCTP_GET_ASSOC_STATS
5636 *
5637 * This option retrieves local per endpoint statistics. It is modeled
5638 * after OpenSolaris' implementation
5639 */
5640 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
5641 char __user *optval,
5642 int __user *optlen)
5643 {
5644 struct sctp_assoc_stats sas;
5645 struct sctp_association *asoc = NULL;
5646
5647 /* User must provide at least the assoc id */
5648 if (len < sizeof(sctp_assoc_t))
5649 return -EINVAL;
5650
5651 /* Allow the struct to grow and fill in as much as possible */
5652 len = min_t(size_t, len, sizeof(sas));
5653
5654 if (copy_from_user(&sas, optval, len))
5655 return -EFAULT;
5656
5657 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
5658 if (!asoc)
5659 return -EINVAL;
5660
5661 sas.sas_rtxchunks = asoc->stats.rtxchunks;
5662 sas.sas_gapcnt = asoc->stats.gapcnt;
5663 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
5664 sas.sas_osacks = asoc->stats.osacks;
5665 sas.sas_isacks = asoc->stats.isacks;
5666 sas.sas_octrlchunks = asoc->stats.octrlchunks;
5667 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
5668 sas.sas_oodchunks = asoc->stats.oodchunks;
5669 sas.sas_iodchunks = asoc->stats.iodchunks;
5670 sas.sas_ouodchunks = asoc->stats.ouodchunks;
5671 sas.sas_iuodchunks = asoc->stats.iuodchunks;
5672 sas.sas_idupchunks = asoc->stats.idupchunks;
5673 sas.sas_opackets = asoc->stats.opackets;
5674 sas.sas_ipackets = asoc->stats.ipackets;
5675
5676 /* New high max rto observed, will return 0 if not a single
5677 * RTO update took place. obs_rto_ipaddr will be bogus
5678 * in such a case
5679 */
5680 sas.sas_maxrto = asoc->stats.max_obs_rto;
5681 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
5682 sizeof(struct sockaddr_storage));
5683
5684 /* Mark beginning of a new observation period */
5685 asoc->stats.max_obs_rto = asoc->rto_min;
5686
5687 if (put_user(len, optlen))
5688 return -EFAULT;
5689
5690 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
5691
5692 if (copy_to_user(optval, &sas, len))
5693 return -EFAULT;
5694
5695 return 0;
5696 }
5697
5698 static int sctp_getsockopt(struct sock *sk, int level, int optname,
5699 char __user *optval, int __user *optlen)
5700 {
5701 int retval = 0;
5702 int len;
5703
5704 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
5705
5706 /* I can hardly begin to describe how wrong this is. This is
5707 * so broken as to be worse than useless. The API draft
5708 * REALLY is NOT helpful here... I am not convinced that the
5709 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5710 * are at all well-founded.
5711 */
5712 if (level != SOL_SCTP) {
5713 struct sctp_af *af = sctp_sk(sk)->pf->af;
5714
5715 retval = af->getsockopt(sk, level, optname, optval, optlen);
5716 return retval;
5717 }
5718
5719 if (get_user(len, optlen))
5720 return -EFAULT;
5721
5722 sctp_lock_sock(sk);
5723
5724 switch (optname) {
5725 case SCTP_STATUS:
5726 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5727 break;
5728 case SCTP_DISABLE_FRAGMENTS:
5729 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5730 optlen);
5731 break;
5732 case SCTP_EVENTS:
5733 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5734 break;
5735 case SCTP_AUTOCLOSE:
5736 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5737 break;
5738 case SCTP_SOCKOPT_PEELOFF:
5739 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5740 break;
5741 case SCTP_PEER_ADDR_PARAMS:
5742 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5743 optlen);
5744 break;
5745 case SCTP_DELAYED_SACK:
5746 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5747 optlen);
5748 break;
5749 case SCTP_INITMSG:
5750 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5751 break;
5752 case SCTP_GET_PEER_ADDRS:
5753 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5754 optlen);
5755 break;
5756 case SCTP_GET_LOCAL_ADDRS:
5757 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5758 optlen);
5759 break;
5760 case SCTP_SOCKOPT_CONNECTX3:
5761 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5762 break;
5763 case SCTP_DEFAULT_SEND_PARAM:
5764 retval = sctp_getsockopt_default_send_param(sk, len,
5765 optval, optlen);
5766 break;
5767 case SCTP_PRIMARY_ADDR:
5768 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5769 break;
5770 case SCTP_NODELAY:
5771 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5772 break;
5773 case SCTP_RTOINFO:
5774 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5775 break;
5776 case SCTP_ASSOCINFO:
5777 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5778 break;
5779 case SCTP_I_WANT_MAPPED_V4_ADDR:
5780 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5781 break;
5782 case SCTP_MAXSEG:
5783 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5784 break;
5785 case SCTP_GET_PEER_ADDR_INFO:
5786 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5787 optlen);
5788 break;
5789 case SCTP_ADAPTATION_LAYER:
5790 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5791 optlen);
5792 break;
5793 case SCTP_CONTEXT:
5794 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5795 break;
5796 case SCTP_FRAGMENT_INTERLEAVE:
5797 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5798 optlen);
5799 break;
5800 case SCTP_PARTIAL_DELIVERY_POINT:
5801 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5802 optlen);
5803 break;
5804 case SCTP_MAX_BURST:
5805 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5806 break;
5807 case SCTP_AUTH_KEY:
5808 case SCTP_AUTH_CHUNK:
5809 case SCTP_AUTH_DELETE_KEY:
5810 retval = -EOPNOTSUPP;
5811 break;
5812 case SCTP_HMAC_IDENT:
5813 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5814 break;
5815 case SCTP_AUTH_ACTIVE_KEY:
5816 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5817 break;
5818 case SCTP_PEER_AUTH_CHUNKS:
5819 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5820 optlen);
5821 break;
5822 case SCTP_LOCAL_AUTH_CHUNKS:
5823 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5824 optlen);
5825 break;
5826 case SCTP_GET_ASSOC_NUMBER:
5827 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5828 break;
5829 case SCTP_GET_ASSOC_ID_LIST:
5830 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5831 break;
5832 case SCTP_AUTO_ASCONF:
5833 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5834 break;
5835 case SCTP_PEER_ADDR_THLDS:
5836 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
5837 break;
5838 case SCTP_GET_ASSOC_STATS:
5839 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
5840 break;
5841 default:
5842 retval = -ENOPROTOOPT;
5843 break;
5844 }
5845
5846 sctp_release_sock(sk);
5847 return retval;
5848 }
5849
5850 static void sctp_hash(struct sock *sk)
5851 {
5852 /* STUB */
5853 }
5854
5855 static void sctp_unhash(struct sock *sk)
5856 {
5857 /* STUB */
5858 }
5859
5860 /* Check if port is acceptable. Possibly find first available port.
5861 *
5862 * The port hash table (contained in the 'global' SCTP protocol storage
5863 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5864 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5865 * list (the list number is the port number hashed out, so as you
5866 * would expect from a hash function, all the ports in a given list have
5867 * such a number that hashes out to the same list number; you were
5868 * expecting that, right?); so each list has a set of ports, with a
5869 * link to the socket (struct sock) that uses it, the port number and
5870 * a fastreuse flag (FIXME: NPI ipg).
5871 */
5872 static struct sctp_bind_bucket *sctp_bucket_create(
5873 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
5874
5875 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5876 {
5877 struct sctp_bind_hashbucket *head; /* hash list */
5878 struct sctp_bind_bucket *pp;
5879 unsigned short snum;
5880 int ret;
5881
5882 snum = ntohs(addr->v4.sin_port);
5883
5884 pr_debug("%s: begins, snum:%d\n", __func__, snum);
5885
5886 sctp_local_bh_disable();
5887
5888 if (snum == 0) {
5889 /* Search for an available port. */
5890 int low, high, remaining, index;
5891 unsigned int rover;
5892
5893 inet_get_local_port_range(sock_net(sk), &low, &high);
5894 remaining = (high - low) + 1;
5895 rover = net_random() % remaining + low;
5896
5897 do {
5898 rover++;
5899 if ((rover < low) || (rover > high))
5900 rover = low;
5901 if (inet_is_reserved_local_port(rover))
5902 continue;
5903 index = sctp_phashfn(sock_net(sk), rover);
5904 head = &sctp_port_hashtable[index];
5905 sctp_spin_lock(&head->lock);
5906 sctp_for_each_hentry(pp, &head->chain)
5907 if ((pp->port == rover) &&
5908 net_eq(sock_net(sk), pp->net))
5909 goto next;
5910 break;
5911 next:
5912 sctp_spin_unlock(&head->lock);
5913 } while (--remaining > 0);
5914
5915 /* Exhausted local port range during search? */
5916 ret = 1;
5917 if (remaining <= 0)
5918 goto fail;
5919
5920 /* OK, here is the one we will use. HEAD (the port
5921 * hash table list entry) is non-NULL and we hold it's
5922 * mutex.
5923 */
5924 snum = rover;
5925 } else {
5926 /* We are given an specific port number; we verify
5927 * that it is not being used. If it is used, we will
5928 * exahust the search in the hash list corresponding
5929 * to the port number (snum) - we detect that with the
5930 * port iterator, pp being NULL.
5931 */
5932 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
5933 sctp_spin_lock(&head->lock);
5934 sctp_for_each_hentry(pp, &head->chain) {
5935 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
5936 goto pp_found;
5937 }
5938 }
5939 pp = NULL;
5940 goto pp_not_found;
5941 pp_found:
5942 if (!hlist_empty(&pp->owner)) {
5943 /* We had a port hash table hit - there is an
5944 * available port (pp != NULL) and it is being
5945 * used by other socket (pp->owner not empty); that other
5946 * socket is going to be sk2.
5947 */
5948 int reuse = sk->sk_reuse;
5949 struct sock *sk2;
5950
5951 pr_debug("%s: found a possible match\n", __func__);
5952
5953 if (pp->fastreuse && sk->sk_reuse &&
5954 sk->sk_state != SCTP_SS_LISTENING)
5955 goto success;
5956
5957 /* Run through the list of sockets bound to the port
5958 * (pp->port) [via the pointers bind_next and
5959 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5960 * we get the endpoint they describe and run through
5961 * the endpoint's list of IP (v4 or v6) addresses,
5962 * comparing each of the addresses with the address of
5963 * the socket sk. If we find a match, then that means
5964 * that this port/socket (sk) combination are already
5965 * in an endpoint.
5966 */
5967 sk_for_each_bound(sk2, &pp->owner) {
5968 struct sctp_endpoint *ep2;
5969 ep2 = sctp_sk(sk2)->ep;
5970
5971 if (sk == sk2 ||
5972 (reuse && sk2->sk_reuse &&
5973 sk2->sk_state != SCTP_SS_LISTENING))
5974 continue;
5975
5976 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5977 sctp_sk(sk2), sctp_sk(sk))) {
5978 ret = (long)sk2;
5979 goto fail_unlock;
5980 }
5981 }
5982
5983 pr_debug("%s: found a match\n", __func__);
5984 }
5985 pp_not_found:
5986 /* If there was a hash table miss, create a new port. */
5987 ret = 1;
5988 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
5989 goto fail_unlock;
5990
5991 /* In either case (hit or miss), make sure fastreuse is 1 only
5992 * if sk->sk_reuse is too (that is, if the caller requested
5993 * SO_REUSEADDR on this socket -sk-).
5994 */
5995 if (hlist_empty(&pp->owner)) {
5996 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5997 pp->fastreuse = 1;
5998 else
5999 pp->fastreuse = 0;
6000 } else if (pp->fastreuse &&
6001 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6002 pp->fastreuse = 0;
6003
6004 /* We are set, so fill up all the data in the hash table
6005 * entry, tie the socket list information with the rest of the
6006 * sockets FIXME: Blurry, NPI (ipg).
6007 */
6008 success:
6009 if (!sctp_sk(sk)->bind_hash) {
6010 inet_sk(sk)->inet_num = snum;
6011 sk_add_bind_node(sk, &pp->owner);
6012 sctp_sk(sk)->bind_hash = pp;
6013 }
6014 ret = 0;
6015
6016 fail_unlock:
6017 sctp_spin_unlock(&head->lock);
6018
6019 fail:
6020 sctp_local_bh_enable();
6021 return ret;
6022 }
6023
6024 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6025 * port is requested.
6026 */
6027 static int sctp_get_port(struct sock *sk, unsigned short snum)
6028 {
6029 union sctp_addr addr;
6030 struct sctp_af *af = sctp_sk(sk)->pf->af;
6031
6032 /* Set up a dummy address struct from the sk. */
6033 af->from_sk(&addr, sk);
6034 addr.v4.sin_port = htons(snum);
6035
6036 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6037 return !!sctp_get_port_local(sk, &addr);
6038 }
6039
6040 /*
6041 * Move a socket to LISTENING state.
6042 */
6043 static int sctp_listen_start(struct sock *sk, int backlog)
6044 {
6045 struct sctp_sock *sp = sctp_sk(sk);
6046 struct sctp_endpoint *ep = sp->ep;
6047 struct crypto_hash *tfm = NULL;
6048 char alg[32];
6049
6050 /* Allocate HMAC for generating cookie. */
6051 if (!sp->hmac && sp->sctp_hmac_alg) {
6052 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6053 tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC);
6054 if (IS_ERR(tfm)) {
6055 net_info_ratelimited("failed to load transform for %s: %ld\n",
6056 sp->sctp_hmac_alg, PTR_ERR(tfm));
6057 return -ENOSYS;
6058 }
6059 sctp_sk(sk)->hmac = tfm;
6060 }
6061
6062 /*
6063 * If a bind() or sctp_bindx() is not called prior to a listen()
6064 * call that allows new associations to be accepted, the system
6065 * picks an ephemeral port and will choose an address set equivalent
6066 * to binding with a wildcard address.
6067 *
6068 * This is not currently spelled out in the SCTP sockets
6069 * extensions draft, but follows the practice as seen in TCP
6070 * sockets.
6071 *
6072 */
6073 sk->sk_state = SCTP_SS_LISTENING;
6074 if (!ep->base.bind_addr.port) {
6075 if (sctp_autobind(sk))
6076 return -EAGAIN;
6077 } else {
6078 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6079 sk->sk_state = SCTP_SS_CLOSED;
6080 return -EADDRINUSE;
6081 }
6082 }
6083
6084 sk->sk_max_ack_backlog = backlog;
6085 sctp_hash_endpoint(ep);
6086 return 0;
6087 }
6088
6089 /*
6090 * 4.1.3 / 5.1.3 listen()
6091 *
6092 * By default, new associations are not accepted for UDP style sockets.
6093 * An application uses listen() to mark a socket as being able to
6094 * accept new associations.
6095 *
6096 * On TCP style sockets, applications use listen() to ready the SCTP
6097 * endpoint for accepting inbound associations.
6098 *
6099 * On both types of endpoints a backlog of '0' disables listening.
6100 *
6101 * Move a socket to LISTENING state.
6102 */
6103 int sctp_inet_listen(struct socket *sock, int backlog)
6104 {
6105 struct sock *sk = sock->sk;
6106 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6107 int err = -EINVAL;
6108
6109 if (unlikely(backlog < 0))
6110 return err;
6111
6112 sctp_lock_sock(sk);
6113
6114 /* Peeled-off sockets are not allowed to listen(). */
6115 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6116 goto out;
6117
6118 if (sock->state != SS_UNCONNECTED)
6119 goto out;
6120
6121 /* If backlog is zero, disable listening. */
6122 if (!backlog) {
6123 if (sctp_sstate(sk, CLOSED))
6124 goto out;
6125
6126 err = 0;
6127 sctp_unhash_endpoint(ep);
6128 sk->sk_state = SCTP_SS_CLOSED;
6129 if (sk->sk_reuse)
6130 sctp_sk(sk)->bind_hash->fastreuse = 1;
6131 goto out;
6132 }
6133
6134 /* If we are already listening, just update the backlog */
6135 if (sctp_sstate(sk, LISTENING))
6136 sk->sk_max_ack_backlog = backlog;
6137 else {
6138 err = sctp_listen_start(sk, backlog);
6139 if (err)
6140 goto out;
6141 }
6142
6143 err = 0;
6144 out:
6145 sctp_release_sock(sk);
6146 return err;
6147 }
6148
6149 /*
6150 * This function is done by modeling the current datagram_poll() and the
6151 * tcp_poll(). Note that, based on these implementations, we don't
6152 * lock the socket in this function, even though it seems that,
6153 * ideally, locking or some other mechanisms can be used to ensure
6154 * the integrity of the counters (sndbuf and wmem_alloc) used
6155 * in this place. We assume that we don't need locks either until proven
6156 * otherwise.
6157 *
6158 * Another thing to note is that we include the Async I/O support
6159 * here, again, by modeling the current TCP/UDP code. We don't have
6160 * a good way to test with it yet.
6161 */
6162 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6163 {
6164 struct sock *sk = sock->sk;
6165 struct sctp_sock *sp = sctp_sk(sk);
6166 unsigned int mask;
6167
6168 poll_wait(file, sk_sleep(sk), wait);
6169
6170 /* A TCP-style listening socket becomes readable when the accept queue
6171 * is not empty.
6172 */
6173 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6174 return (!list_empty(&sp->ep->asocs)) ?
6175 (POLLIN | POLLRDNORM) : 0;
6176
6177 mask = 0;
6178
6179 /* Is there any exceptional events? */
6180 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6181 mask |= POLLERR |
6182 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
6183 if (sk->sk_shutdown & RCV_SHUTDOWN)
6184 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6185 if (sk->sk_shutdown == SHUTDOWN_MASK)
6186 mask |= POLLHUP;
6187
6188 /* Is it readable? Reconsider this code with TCP-style support. */
6189 if (!skb_queue_empty(&sk->sk_receive_queue))
6190 mask |= POLLIN | POLLRDNORM;
6191
6192 /* The association is either gone or not ready. */
6193 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6194 return mask;
6195
6196 /* Is it writable? */
6197 if (sctp_writeable(sk)) {
6198 mask |= POLLOUT | POLLWRNORM;
6199 } else {
6200 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6201 /*
6202 * Since the socket is not locked, the buffer
6203 * might be made available after the writeable check and
6204 * before the bit is set. This could cause a lost I/O
6205 * signal. tcp_poll() has a race breaker for this race
6206 * condition. Based on their implementation, we put
6207 * in the following code to cover it as well.
6208 */
6209 if (sctp_writeable(sk))
6210 mask |= POLLOUT | POLLWRNORM;
6211 }
6212 return mask;
6213 }
6214
6215 /********************************************************************
6216 * 2nd Level Abstractions
6217 ********************************************************************/
6218
6219 static struct sctp_bind_bucket *sctp_bucket_create(
6220 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6221 {
6222 struct sctp_bind_bucket *pp;
6223
6224 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6225 if (pp) {
6226 SCTP_DBG_OBJCNT_INC(bind_bucket);
6227 pp->port = snum;
6228 pp->fastreuse = 0;
6229 INIT_HLIST_HEAD(&pp->owner);
6230 pp->net = net;
6231 hlist_add_head(&pp->node, &head->chain);
6232 }
6233 return pp;
6234 }
6235
6236 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6237 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6238 {
6239 if (pp && hlist_empty(&pp->owner)) {
6240 __hlist_del(&pp->node);
6241 kmem_cache_free(sctp_bucket_cachep, pp);
6242 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6243 }
6244 }
6245
6246 /* Release this socket's reference to a local port. */
6247 static inline void __sctp_put_port(struct sock *sk)
6248 {
6249 struct sctp_bind_hashbucket *head =
6250 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6251 inet_sk(sk)->inet_num)];
6252 struct sctp_bind_bucket *pp;
6253
6254 sctp_spin_lock(&head->lock);
6255 pp = sctp_sk(sk)->bind_hash;
6256 __sk_del_bind_node(sk);
6257 sctp_sk(sk)->bind_hash = NULL;
6258 inet_sk(sk)->inet_num = 0;
6259 sctp_bucket_destroy(pp);
6260 sctp_spin_unlock(&head->lock);
6261 }
6262
6263 void sctp_put_port(struct sock *sk)
6264 {
6265 sctp_local_bh_disable();
6266 __sctp_put_port(sk);
6267 sctp_local_bh_enable();
6268 }
6269
6270 /*
6271 * The system picks an ephemeral port and choose an address set equivalent
6272 * to binding with a wildcard address.
6273 * One of those addresses will be the primary address for the association.
6274 * This automatically enables the multihoming capability of SCTP.
6275 */
6276 static int sctp_autobind(struct sock *sk)
6277 {
6278 union sctp_addr autoaddr;
6279 struct sctp_af *af;
6280 __be16 port;
6281
6282 /* Initialize a local sockaddr structure to INADDR_ANY. */
6283 af = sctp_sk(sk)->pf->af;
6284
6285 port = htons(inet_sk(sk)->inet_num);
6286 af->inaddr_any(&autoaddr, port);
6287
6288 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6289 }
6290
6291 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6292 *
6293 * From RFC 2292
6294 * 4.2 The cmsghdr Structure *
6295 *
6296 * When ancillary data is sent or received, any number of ancillary data
6297 * objects can be specified by the msg_control and msg_controllen members of
6298 * the msghdr structure, because each object is preceded by
6299 * a cmsghdr structure defining the object's length (the cmsg_len member).
6300 * Historically Berkeley-derived implementations have passed only one object
6301 * at a time, but this API allows multiple objects to be
6302 * passed in a single call to sendmsg() or recvmsg(). The following example
6303 * shows two ancillary data objects in a control buffer.
6304 *
6305 * |<--------------------------- msg_controllen -------------------------->|
6306 * | |
6307 *
6308 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6309 *
6310 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6311 * | | |
6312 *
6313 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6314 *
6315 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6316 * | | | | |
6317 *
6318 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6319 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6320 *
6321 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6322 *
6323 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6324 * ^
6325 * |
6326 *
6327 * msg_control
6328 * points here
6329 */
6330 static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
6331 {
6332 struct cmsghdr *cmsg;
6333 struct msghdr *my_msg = (struct msghdr *)msg;
6334
6335 for (cmsg = CMSG_FIRSTHDR(msg);
6336 cmsg != NULL;
6337 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6338 if (!CMSG_OK(my_msg, cmsg))
6339 return -EINVAL;
6340
6341 /* Should we parse this header or ignore? */
6342 if (cmsg->cmsg_level != IPPROTO_SCTP)
6343 continue;
6344
6345 /* Strictly check lengths following example in SCM code. */
6346 switch (cmsg->cmsg_type) {
6347 case SCTP_INIT:
6348 /* SCTP Socket API Extension
6349 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6350 *
6351 * This cmsghdr structure provides information for
6352 * initializing new SCTP associations with sendmsg().
6353 * The SCTP_INITMSG socket option uses this same data
6354 * structure. This structure is not used for
6355 * recvmsg().
6356 *
6357 * cmsg_level cmsg_type cmsg_data[]
6358 * ------------ ------------ ----------------------
6359 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6360 */
6361 if (cmsg->cmsg_len !=
6362 CMSG_LEN(sizeof(struct sctp_initmsg)))
6363 return -EINVAL;
6364 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6365 break;
6366
6367 case SCTP_SNDRCV:
6368 /* SCTP Socket API Extension
6369 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6370 *
6371 * This cmsghdr structure specifies SCTP options for
6372 * sendmsg() and describes SCTP header information
6373 * about a received message through recvmsg().
6374 *
6375 * cmsg_level cmsg_type cmsg_data[]
6376 * ------------ ------------ ----------------------
6377 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6378 */
6379 if (cmsg->cmsg_len !=
6380 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6381 return -EINVAL;
6382
6383 cmsgs->info =
6384 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6385
6386 /* Minimally, validate the sinfo_flags. */
6387 if (cmsgs->info->sinfo_flags &
6388 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6389 SCTP_ABORT | SCTP_EOF))
6390 return -EINVAL;
6391 break;
6392
6393 default:
6394 return -EINVAL;
6395 }
6396 }
6397 return 0;
6398 }
6399
6400 /*
6401 * Wait for a packet..
6402 * Note: This function is the same function as in core/datagram.c
6403 * with a few modifications to make lksctp work.
6404 */
6405 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6406 {
6407 int error;
6408 DEFINE_WAIT(wait);
6409
6410 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6411
6412 /* Socket errors? */
6413 error = sock_error(sk);
6414 if (error)
6415 goto out;
6416
6417 if (!skb_queue_empty(&sk->sk_receive_queue))
6418 goto ready;
6419
6420 /* Socket shut down? */
6421 if (sk->sk_shutdown & RCV_SHUTDOWN)
6422 goto out;
6423
6424 /* Sequenced packets can come disconnected. If so we report the
6425 * problem.
6426 */
6427 error = -ENOTCONN;
6428
6429 /* Is there a good reason to think that we may receive some data? */
6430 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6431 goto out;
6432
6433 /* Handle signals. */
6434 if (signal_pending(current))
6435 goto interrupted;
6436
6437 /* Let another process have a go. Since we are going to sleep
6438 * anyway. Note: This may cause odd behaviors if the message
6439 * does not fit in the user's buffer, but this seems to be the
6440 * only way to honor MSG_DONTWAIT realistically.
6441 */
6442 sctp_release_sock(sk);
6443 *timeo_p = schedule_timeout(*timeo_p);
6444 sctp_lock_sock(sk);
6445
6446 ready:
6447 finish_wait(sk_sleep(sk), &wait);
6448 return 0;
6449
6450 interrupted:
6451 error = sock_intr_errno(*timeo_p);
6452
6453 out:
6454 finish_wait(sk_sleep(sk), &wait);
6455 *err = error;
6456 return error;
6457 }
6458
6459 /* Receive a datagram.
6460 * Note: This is pretty much the same routine as in core/datagram.c
6461 * with a few changes to make lksctp work.
6462 */
6463 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6464 int noblock, int *err)
6465 {
6466 int error;
6467 struct sk_buff *skb;
6468 long timeo;
6469
6470 timeo = sock_rcvtimeo(sk, noblock);
6471
6472 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
6473 MAX_SCHEDULE_TIMEOUT);
6474
6475 do {
6476 /* Again only user level code calls this function,
6477 * so nothing interrupt level
6478 * will suddenly eat the receive_queue.
6479 *
6480 * Look at current nfs client by the way...
6481 * However, this function was correct in any case. 8)
6482 */
6483 if (flags & MSG_PEEK) {
6484 spin_lock_bh(&sk->sk_receive_queue.lock);
6485 skb = skb_peek(&sk->sk_receive_queue);
6486 if (skb)
6487 atomic_inc(&skb->users);
6488 spin_unlock_bh(&sk->sk_receive_queue.lock);
6489 } else {
6490 skb = skb_dequeue(&sk->sk_receive_queue);
6491 }
6492
6493 if (skb)
6494 return skb;
6495
6496 /* Caller is allowed not to check sk->sk_err before calling. */
6497 error = sock_error(sk);
6498 if (error)
6499 goto no_packet;
6500
6501 if (sk->sk_shutdown & RCV_SHUTDOWN)
6502 break;
6503
6504 /* User doesn't want to wait. */
6505 error = -EAGAIN;
6506 if (!timeo)
6507 goto no_packet;
6508 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6509
6510 return NULL;
6511
6512 no_packet:
6513 *err = error;
6514 return NULL;
6515 }
6516
6517 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6518 static void __sctp_write_space(struct sctp_association *asoc)
6519 {
6520 struct sock *sk = asoc->base.sk;
6521 struct socket *sock = sk->sk_socket;
6522
6523 if ((sctp_wspace(asoc) > 0) && sock) {
6524 if (waitqueue_active(&asoc->wait))
6525 wake_up_interruptible(&asoc->wait);
6526
6527 if (sctp_writeable(sk)) {
6528 wait_queue_head_t *wq = sk_sleep(sk);
6529
6530 if (wq && waitqueue_active(wq))
6531 wake_up_interruptible(wq);
6532
6533 /* Note that we try to include the Async I/O support
6534 * here by modeling from the current TCP/UDP code.
6535 * We have not tested with it yet.
6536 */
6537 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6538 sock_wake_async(sock,
6539 SOCK_WAKE_SPACE, POLL_OUT);
6540 }
6541 }
6542 }
6543
6544 /* Do accounting for the sndbuf space.
6545 * Decrement the used sndbuf space of the corresponding association by the
6546 * data size which was just transmitted(freed).
6547 */
6548 static void sctp_wfree(struct sk_buff *skb)
6549 {
6550 struct sctp_association *asoc;
6551 struct sctp_chunk *chunk;
6552 struct sock *sk;
6553
6554 /* Get the saved chunk pointer. */
6555 chunk = *((struct sctp_chunk **)(skb->cb));
6556 asoc = chunk->asoc;
6557 sk = asoc->base.sk;
6558 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6559 sizeof(struct sk_buff) +
6560 sizeof(struct sctp_chunk);
6561
6562 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6563
6564 /*
6565 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6566 */
6567 sk->sk_wmem_queued -= skb->truesize;
6568 sk_mem_uncharge(sk, skb->truesize);
6569
6570 sock_wfree(skb);
6571 __sctp_write_space(asoc);
6572
6573 sctp_association_put(asoc);
6574 }
6575
6576 /* Do accounting for the receive space on the socket.
6577 * Accounting for the association is done in ulpevent.c
6578 * We set this as a destructor for the cloned data skbs so that
6579 * accounting is done at the correct time.
6580 */
6581 void sctp_sock_rfree(struct sk_buff *skb)
6582 {
6583 struct sock *sk = skb->sk;
6584 struct sctp_ulpevent *event = sctp_skb2event(skb);
6585
6586 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6587
6588 /*
6589 * Mimic the behavior of sock_rfree
6590 */
6591 sk_mem_uncharge(sk, event->rmem_len);
6592 }
6593
6594
6595 /* Helper function to wait for space in the sndbuf. */
6596 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6597 size_t msg_len)
6598 {
6599 struct sock *sk = asoc->base.sk;
6600 int err = 0;
6601 long current_timeo = *timeo_p;
6602 DEFINE_WAIT(wait);
6603
6604 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
6605 *timeo_p, msg_len);
6606
6607 /* Increment the association's refcnt. */
6608 sctp_association_hold(asoc);
6609
6610 /* Wait on the association specific sndbuf space. */
6611 for (;;) {
6612 prepare_to_wait_exclusive(&asoc->wait, &wait,
6613 TASK_INTERRUPTIBLE);
6614 if (!*timeo_p)
6615 goto do_nonblock;
6616 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6617 asoc->base.dead)
6618 goto do_error;
6619 if (signal_pending(current))
6620 goto do_interrupted;
6621 if (msg_len <= sctp_wspace(asoc))
6622 break;
6623
6624 /* Let another process have a go. Since we are going
6625 * to sleep anyway.
6626 */
6627 sctp_release_sock(sk);
6628 current_timeo = schedule_timeout(current_timeo);
6629 BUG_ON(sk != asoc->base.sk);
6630 sctp_lock_sock(sk);
6631
6632 *timeo_p = current_timeo;
6633 }
6634
6635 out:
6636 finish_wait(&asoc->wait, &wait);
6637
6638 /* Release the association's refcnt. */
6639 sctp_association_put(asoc);
6640
6641 return err;
6642
6643 do_error:
6644 err = -EPIPE;
6645 goto out;
6646
6647 do_interrupted:
6648 err = sock_intr_errno(*timeo_p);
6649 goto out;
6650
6651 do_nonblock:
6652 err = -EAGAIN;
6653 goto out;
6654 }
6655
6656 void sctp_data_ready(struct sock *sk, int len)
6657 {
6658 struct socket_wq *wq;
6659
6660 rcu_read_lock();
6661 wq = rcu_dereference(sk->sk_wq);
6662 if (wq_has_sleeper(wq))
6663 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6664 POLLRDNORM | POLLRDBAND);
6665 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6666 rcu_read_unlock();
6667 }
6668
6669 /* If socket sndbuf has changed, wake up all per association waiters. */
6670 void sctp_write_space(struct sock *sk)
6671 {
6672 struct sctp_association *asoc;
6673
6674 /* Wake up the tasks in each wait queue. */
6675 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6676 __sctp_write_space(asoc);
6677 }
6678 }
6679
6680 /* Is there any sndbuf space available on the socket?
6681 *
6682 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6683 * associations on the same socket. For a UDP-style socket with
6684 * multiple associations, it is possible for it to be "unwriteable"
6685 * prematurely. I assume that this is acceptable because
6686 * a premature "unwriteable" is better than an accidental "writeable" which
6687 * would cause an unwanted block under certain circumstances. For the 1-1
6688 * UDP-style sockets or TCP-style sockets, this code should work.
6689 * - Daisy
6690 */
6691 static int sctp_writeable(struct sock *sk)
6692 {
6693 int amt = 0;
6694
6695 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6696 if (amt < 0)
6697 amt = 0;
6698 return amt;
6699 }
6700
6701 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6702 * returns immediately with EINPROGRESS.
6703 */
6704 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6705 {
6706 struct sock *sk = asoc->base.sk;
6707 int err = 0;
6708 long current_timeo = *timeo_p;
6709 DEFINE_WAIT(wait);
6710
6711 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
6712
6713 /* Increment the association's refcnt. */
6714 sctp_association_hold(asoc);
6715
6716 for (;;) {
6717 prepare_to_wait_exclusive(&asoc->wait, &wait,
6718 TASK_INTERRUPTIBLE);
6719 if (!*timeo_p)
6720 goto do_nonblock;
6721 if (sk->sk_shutdown & RCV_SHUTDOWN)
6722 break;
6723 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6724 asoc->base.dead)
6725 goto do_error;
6726 if (signal_pending(current))
6727 goto do_interrupted;
6728
6729 if (sctp_state(asoc, ESTABLISHED))
6730 break;
6731
6732 /* Let another process have a go. Since we are going
6733 * to sleep anyway.
6734 */
6735 sctp_release_sock(sk);
6736 current_timeo = schedule_timeout(current_timeo);
6737 sctp_lock_sock(sk);
6738
6739 *timeo_p = current_timeo;
6740 }
6741
6742 out:
6743 finish_wait(&asoc->wait, &wait);
6744
6745 /* Release the association's refcnt. */
6746 sctp_association_put(asoc);
6747
6748 return err;
6749
6750 do_error:
6751 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6752 err = -ETIMEDOUT;
6753 else
6754 err = -ECONNREFUSED;
6755 goto out;
6756
6757 do_interrupted:
6758 err = sock_intr_errno(*timeo_p);
6759 goto out;
6760
6761 do_nonblock:
6762 err = -EINPROGRESS;
6763 goto out;
6764 }
6765
6766 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6767 {
6768 struct sctp_endpoint *ep;
6769 int err = 0;
6770 DEFINE_WAIT(wait);
6771
6772 ep = sctp_sk(sk)->ep;
6773
6774
6775 for (;;) {
6776 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6777 TASK_INTERRUPTIBLE);
6778
6779 if (list_empty(&ep->asocs)) {
6780 sctp_release_sock(sk);
6781 timeo = schedule_timeout(timeo);
6782 sctp_lock_sock(sk);
6783 }
6784
6785 err = -EINVAL;
6786 if (!sctp_sstate(sk, LISTENING))
6787 break;
6788
6789 err = 0;
6790 if (!list_empty(&ep->asocs))
6791 break;
6792
6793 err = sock_intr_errno(timeo);
6794 if (signal_pending(current))
6795 break;
6796
6797 err = -EAGAIN;
6798 if (!timeo)
6799 break;
6800 }
6801
6802 finish_wait(sk_sleep(sk), &wait);
6803
6804 return err;
6805 }
6806
6807 static void sctp_wait_for_close(struct sock *sk, long timeout)
6808 {
6809 DEFINE_WAIT(wait);
6810
6811 do {
6812 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6813 if (list_empty(&sctp_sk(sk)->ep->asocs))
6814 break;
6815 sctp_release_sock(sk);
6816 timeout = schedule_timeout(timeout);
6817 sctp_lock_sock(sk);
6818 } while (!signal_pending(current) && timeout);
6819
6820 finish_wait(sk_sleep(sk), &wait);
6821 }
6822
6823 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6824 {
6825 struct sk_buff *frag;
6826
6827 if (!skb->data_len)
6828 goto done;
6829
6830 /* Don't forget the fragments. */
6831 skb_walk_frags(skb, frag)
6832 sctp_skb_set_owner_r_frag(frag, sk);
6833
6834 done:
6835 sctp_skb_set_owner_r(skb, sk);
6836 }
6837
6838 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6839 struct sctp_association *asoc)
6840 {
6841 struct inet_sock *inet = inet_sk(sk);
6842 struct inet_sock *newinet;
6843
6844 newsk->sk_type = sk->sk_type;
6845 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6846 newsk->sk_flags = sk->sk_flags;
6847 newsk->sk_no_check = sk->sk_no_check;
6848 newsk->sk_reuse = sk->sk_reuse;
6849
6850 newsk->sk_shutdown = sk->sk_shutdown;
6851 newsk->sk_destruct = sctp_destruct_sock;
6852 newsk->sk_family = sk->sk_family;
6853 newsk->sk_protocol = IPPROTO_SCTP;
6854 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6855 newsk->sk_sndbuf = sk->sk_sndbuf;
6856 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6857 newsk->sk_lingertime = sk->sk_lingertime;
6858 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6859 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6860
6861 newinet = inet_sk(newsk);
6862
6863 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6864 * getsockname() and getpeername()
6865 */
6866 newinet->inet_sport = inet->inet_sport;
6867 newinet->inet_saddr = inet->inet_saddr;
6868 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6869 newinet->inet_dport = htons(asoc->peer.port);
6870 newinet->pmtudisc = inet->pmtudisc;
6871 newinet->inet_id = asoc->next_tsn ^ jiffies;
6872
6873 newinet->uc_ttl = inet->uc_ttl;
6874 newinet->mc_loop = 1;
6875 newinet->mc_ttl = 1;
6876 newinet->mc_index = 0;
6877 newinet->mc_list = NULL;
6878 }
6879
6880 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6881 * and its messages to the newsk.
6882 */
6883 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6884 struct sctp_association *assoc,
6885 sctp_socket_type_t type)
6886 {
6887 struct sctp_sock *oldsp = sctp_sk(oldsk);
6888 struct sctp_sock *newsp = sctp_sk(newsk);
6889 struct sctp_bind_bucket *pp; /* hash list port iterator */
6890 struct sctp_endpoint *newep = newsp->ep;
6891 struct sk_buff *skb, *tmp;
6892 struct sctp_ulpevent *event;
6893 struct sctp_bind_hashbucket *head;
6894 struct list_head tmplist;
6895
6896 /* Migrate socket buffer sizes and all the socket level options to the
6897 * new socket.
6898 */
6899 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6900 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6901 /* Brute force copy old sctp opt. */
6902 if (oldsp->do_auto_asconf) {
6903 memcpy(&tmplist, &newsp->auto_asconf_list, sizeof(tmplist));
6904 inet_sk_copy_descendant(newsk, oldsk);
6905 memcpy(&newsp->auto_asconf_list, &tmplist, sizeof(tmplist));
6906 } else
6907 inet_sk_copy_descendant(newsk, oldsk);
6908
6909 /* Restore the ep value that was overwritten with the above structure
6910 * copy.
6911 */
6912 newsp->ep = newep;
6913 newsp->hmac = NULL;
6914
6915 /* Hook this new socket in to the bind_hash list. */
6916 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
6917 inet_sk(oldsk)->inet_num)];
6918 sctp_local_bh_disable();
6919 sctp_spin_lock(&head->lock);
6920 pp = sctp_sk(oldsk)->bind_hash;
6921 sk_add_bind_node(newsk, &pp->owner);
6922 sctp_sk(newsk)->bind_hash = pp;
6923 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6924 sctp_spin_unlock(&head->lock);
6925 sctp_local_bh_enable();
6926
6927 /* Copy the bind_addr list from the original endpoint to the new
6928 * endpoint so that we can handle restarts properly
6929 */
6930 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6931 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6932
6933 /* Move any messages in the old socket's receive queue that are for the
6934 * peeled off association to the new socket's receive queue.
6935 */
6936 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6937 event = sctp_skb2event(skb);
6938 if (event->asoc == assoc) {
6939 __skb_unlink(skb, &oldsk->sk_receive_queue);
6940 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6941 sctp_skb_set_owner_r_frag(skb, newsk);
6942 }
6943 }
6944
6945 /* Clean up any messages pending delivery due to partial
6946 * delivery. Three cases:
6947 * 1) No partial deliver; no work.
6948 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6949 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6950 */
6951 skb_queue_head_init(&newsp->pd_lobby);
6952 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6953
6954 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6955 struct sk_buff_head *queue;
6956
6957 /* Decide which queue to move pd_lobby skbs to. */
6958 if (assoc->ulpq.pd_mode) {
6959 queue = &newsp->pd_lobby;
6960 } else
6961 queue = &newsk->sk_receive_queue;
6962
6963 /* Walk through the pd_lobby, looking for skbs that
6964 * need moved to the new socket.
6965 */
6966 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6967 event = sctp_skb2event(skb);
6968 if (event->asoc == assoc) {
6969 __skb_unlink(skb, &oldsp->pd_lobby);
6970 __skb_queue_tail(queue, skb);
6971 sctp_skb_set_owner_r_frag(skb, newsk);
6972 }
6973 }
6974
6975 /* Clear up any skbs waiting for the partial
6976 * delivery to finish.
6977 */
6978 if (assoc->ulpq.pd_mode)
6979 sctp_clear_pd(oldsk, NULL);
6980
6981 }
6982
6983 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6984 sctp_skb_set_owner_r_frag(skb, newsk);
6985
6986 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6987 sctp_skb_set_owner_r_frag(skb, newsk);
6988
6989 /* Set the type of socket to indicate that it is peeled off from the
6990 * original UDP-style socket or created with the accept() call on a
6991 * TCP-style socket..
6992 */
6993 newsp->type = type;
6994
6995 /* Mark the new socket "in-use" by the user so that any packets
6996 * that may arrive on the association after we've moved it are
6997 * queued to the backlog. This prevents a potential race between
6998 * backlog processing on the old socket and new-packet processing
6999 * on the new socket.
7000 *
7001 * The caller has just allocated newsk so we can guarantee that other
7002 * paths won't try to lock it and then oldsk.
7003 */
7004 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7005 sctp_assoc_migrate(assoc, newsk);
7006
7007 /* If the association on the newsk is already closed before accept()
7008 * is called, set RCV_SHUTDOWN flag.
7009 */
7010 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
7011 newsk->sk_shutdown |= RCV_SHUTDOWN;
7012
7013 newsk->sk_state = SCTP_SS_ESTABLISHED;
7014 sctp_release_sock(newsk);
7015 }
7016
7017
7018 /* This proto struct describes the ULP interface for SCTP. */
7019 struct proto sctp_prot = {
7020 .name = "SCTP",
7021 .owner = THIS_MODULE,
7022 .close = sctp_close,
7023 .connect = sctp_connect,
7024 .disconnect = sctp_disconnect,
7025 .accept = sctp_accept,
7026 .ioctl = sctp_ioctl,
7027 .init = sctp_init_sock,
7028 .destroy = sctp_destroy_sock,
7029 .shutdown = sctp_shutdown,
7030 .setsockopt = sctp_setsockopt,
7031 .getsockopt = sctp_getsockopt,
7032 .sendmsg = sctp_sendmsg,
7033 .recvmsg = sctp_recvmsg,
7034 .bind = sctp_bind,
7035 .backlog_rcv = sctp_backlog_rcv,
7036 .hash = sctp_hash,
7037 .unhash = sctp_unhash,
7038 .get_port = sctp_get_port,
7039 .obj_size = sizeof(struct sctp_sock),
7040 .sysctl_mem = sysctl_sctp_mem,
7041 .sysctl_rmem = sysctl_sctp_rmem,
7042 .sysctl_wmem = sysctl_sctp_wmem,
7043 .memory_pressure = &sctp_memory_pressure,
7044 .enter_memory_pressure = sctp_enter_memory_pressure,
7045 .memory_allocated = &sctp_memory_allocated,
7046 .sockets_allocated = &sctp_sockets_allocated,
7047 };
7048
7049 #if IS_ENABLED(CONFIG_IPV6)
7050
7051 struct proto sctpv6_prot = {
7052 .name = "SCTPv6",
7053 .owner = THIS_MODULE,
7054 .close = sctp_close,
7055 .connect = sctp_connect,
7056 .disconnect = sctp_disconnect,
7057 .accept = sctp_accept,
7058 .ioctl = sctp_ioctl,
7059 .init = sctp_init_sock,
7060 .destroy = sctp_destroy_sock,
7061 .shutdown = sctp_shutdown,
7062 .setsockopt = sctp_setsockopt,
7063 .getsockopt = sctp_getsockopt,
7064 .sendmsg = sctp_sendmsg,
7065 .recvmsg = sctp_recvmsg,
7066 .bind = sctp_bind,
7067 .backlog_rcv = sctp_backlog_rcv,
7068 .hash = sctp_hash,
7069 .unhash = sctp_unhash,
7070 .get_port = sctp_get_port,
7071 .obj_size = sizeof(struct sctp6_sock),
7072 .sysctl_mem = sysctl_sctp_mem,
7073 .sysctl_rmem = sysctl_sctp_rmem,
7074 .sysctl_wmem = sysctl_sctp_wmem,
7075 .memory_pressure = &sctp_memory_pressure,
7076 .enter_memory_pressure = sctp_enter_memory_pressure,
7077 .memory_allocated = &sctp_memory_allocated,
7078 .sockets_allocated = &sctp_sockets_allocated,
7079 };
7080 #endif /* IS_ENABLED(CONFIG_IPV6) */
This page took 0.20036 seconds and 6 git commands to generate.