sctp: add SCTP_PR_SUPPORTED on sctp sockopt
[deliverable/linux.git] / net / sctp / socket.c
1 /* SCTP kernel implementation
2 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001-2003 Intel Corp.
6 * Copyright (c) 2001-2002 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
8 *
9 * This file is part of the SCTP kernel implementation
10 *
11 * These functions interface with the sockets layer to implement the
12 * SCTP Extensions for the Sockets API.
13 *
14 * Note that the descriptions from the specification are USER level
15 * functions--this file is the functions which populate the struct proto
16 * for SCTP which is the BOTTOM of the sockets interface.
17 *
18 * This SCTP implementation is free software;
19 * you can redistribute it and/or modify it under the terms of
20 * the GNU General Public License as published by
21 * the Free Software Foundation; either version 2, or (at your option)
22 * any later version.
23 *
24 * This SCTP implementation is distributed in the hope that it
25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 * ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
29 *
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, see
32 * <http://www.gnu.org/licenses/>.
33 *
34 * Please send any bug reports or fixes you make to the
35 * email address(es):
36 * lksctp developers <linux-sctp@vger.kernel.org>
37 *
38 * Written or modified by:
39 * La Monte H.P. Yarroll <piggy@acm.org>
40 * Narasimha Budihal <narsi@refcode.org>
41 * Karl Knutson <karl@athena.chicago.il.us>
42 * Jon Grimm <jgrimm@us.ibm.com>
43 * Xingang Guo <xingang.guo@intel.com>
44 * Daisy Chang <daisyc@us.ibm.com>
45 * Sridhar Samudrala <samudrala@us.ibm.com>
46 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
47 * Ardelle Fan <ardelle.fan@intel.com>
48 * Ryan Layer <rmlayer@us.ibm.com>
49 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
50 * Kevin Gao <kevin.gao@intel.com>
51 */
52
53 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
54
55 #include <crypto/hash.h>
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/slab.h>
66 #include <linux/file.h>
67 #include <linux/compat.h>
68
69 #include <net/ip.h>
70 #include <net/icmp.h>
71 #include <net/route.h>
72 #include <net/ipv6.h>
73 #include <net/inet_common.h>
74 #include <net/busy_poll.h>
75
76 #include <linux/socket.h> /* for sa_family_t */
77 #include <linux/export.h>
78 #include <net/sock.h>
79 #include <net/sctp/sctp.h>
80 #include <net/sctp/sm.h>
81
82 /* Forward declarations for internal helper functions. */
83 static int sctp_writeable(struct sock *sk);
84 static void sctp_wfree(struct sk_buff *skb);
85 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
86 size_t msg_len);
87 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p);
88 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
89 static int sctp_wait_for_accept(struct sock *sk, long timeo);
90 static void sctp_wait_for_close(struct sock *sk, long timeo);
91 static void sctp_destruct_sock(struct sock *sk);
92 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
93 union sctp_addr *addr, int len);
94 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
95 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
96 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
97 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
98 static int sctp_send_asconf(struct sctp_association *asoc,
99 struct sctp_chunk *chunk);
100 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
101 static int sctp_autobind(struct sock *sk);
102 static void sctp_sock_migrate(struct sock *, struct sock *,
103 struct sctp_association *, sctp_socket_type_t);
104
105 static int sctp_memory_pressure;
106 static atomic_long_t sctp_memory_allocated;
107 struct percpu_counter sctp_sockets_allocated;
108
109 static void sctp_enter_memory_pressure(struct sock *sk)
110 {
111 sctp_memory_pressure = 1;
112 }
113
114
115 /* Get the sndbuf space available at the time on the association. */
116 static inline int sctp_wspace(struct sctp_association *asoc)
117 {
118 int amt;
119
120 if (asoc->ep->sndbuf_policy)
121 amt = asoc->sndbuf_used;
122 else
123 amt = sk_wmem_alloc_get(asoc->base.sk);
124
125 if (amt >= asoc->base.sk->sk_sndbuf) {
126 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
127 amt = 0;
128 else {
129 amt = sk_stream_wspace(asoc->base.sk);
130 if (amt < 0)
131 amt = 0;
132 }
133 } else {
134 amt = asoc->base.sk->sk_sndbuf - amt;
135 }
136 return amt;
137 }
138
139 /* Increment the used sndbuf space count of the corresponding association by
140 * the size of the outgoing data chunk.
141 * Also, set the skb destructor for sndbuf accounting later.
142 *
143 * Since it is always 1-1 between chunk and skb, and also a new skb is always
144 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
145 * destructor in the data chunk skb for the purpose of the sndbuf space
146 * tracking.
147 */
148 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
149 {
150 struct sctp_association *asoc = chunk->asoc;
151 struct sock *sk = asoc->base.sk;
152
153 /* The sndbuf space is tracked per association. */
154 sctp_association_hold(asoc);
155
156 skb_set_owner_w(chunk->skb, sk);
157
158 chunk->skb->destructor = sctp_wfree;
159 /* Save the chunk pointer in skb for sctp_wfree to use later. */
160 skb_shinfo(chunk->skb)->destructor_arg = chunk;
161
162 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
163 sizeof(struct sk_buff) +
164 sizeof(struct sctp_chunk);
165
166 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
167 sk->sk_wmem_queued += chunk->skb->truesize;
168 sk_mem_charge(sk, chunk->skb->truesize);
169 }
170
171 /* Verify that this is a valid address. */
172 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
173 int len)
174 {
175 struct sctp_af *af;
176
177 /* Verify basic sockaddr. */
178 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
179 if (!af)
180 return -EINVAL;
181
182 /* Is this a valid SCTP address? */
183 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
184 return -EINVAL;
185
186 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
187 return -EINVAL;
188
189 return 0;
190 }
191
192 /* Look up the association by its id. If this is not a UDP-style
193 * socket, the ID field is always ignored.
194 */
195 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
196 {
197 struct sctp_association *asoc = NULL;
198
199 /* If this is not a UDP-style socket, assoc id should be ignored. */
200 if (!sctp_style(sk, UDP)) {
201 /* Return NULL if the socket state is not ESTABLISHED. It
202 * could be a TCP-style listening socket or a socket which
203 * hasn't yet called connect() to establish an association.
204 */
205 if (!sctp_sstate(sk, ESTABLISHED))
206 return NULL;
207
208 /* Get the first and the only association from the list. */
209 if (!list_empty(&sctp_sk(sk)->ep->asocs))
210 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
211 struct sctp_association, asocs);
212 return asoc;
213 }
214
215 /* Otherwise this is a UDP-style socket. */
216 if (!id || (id == (sctp_assoc_t)-1))
217 return NULL;
218
219 spin_lock_bh(&sctp_assocs_id_lock);
220 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
221 spin_unlock_bh(&sctp_assocs_id_lock);
222
223 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
224 return NULL;
225
226 return asoc;
227 }
228
229 /* Look up the transport from an address and an assoc id. If both address and
230 * id are specified, the associations matching the address and the id should be
231 * the same.
232 */
233 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
234 struct sockaddr_storage *addr,
235 sctp_assoc_t id)
236 {
237 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
238 struct sctp_transport *transport;
239 union sctp_addr *laddr = (union sctp_addr *)addr;
240
241 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
242 laddr,
243 &transport);
244
245 if (!addr_asoc)
246 return NULL;
247
248 id_asoc = sctp_id2assoc(sk, id);
249 if (id_asoc && (id_asoc != addr_asoc))
250 return NULL;
251
252 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
253 (union sctp_addr *)addr);
254
255 return transport;
256 }
257
258 /* API 3.1.2 bind() - UDP Style Syntax
259 * The syntax of bind() is,
260 *
261 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
262 *
263 * sd - the socket descriptor returned by socket().
264 * addr - the address structure (struct sockaddr_in or struct
265 * sockaddr_in6 [RFC 2553]),
266 * addr_len - the size of the address structure.
267 */
268 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
269 {
270 int retval = 0;
271
272 lock_sock(sk);
273
274 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
275 addr, addr_len);
276
277 /* Disallow binding twice. */
278 if (!sctp_sk(sk)->ep->base.bind_addr.port)
279 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
280 addr_len);
281 else
282 retval = -EINVAL;
283
284 release_sock(sk);
285
286 return retval;
287 }
288
289 static long sctp_get_port_local(struct sock *, union sctp_addr *);
290
291 /* Verify this is a valid sockaddr. */
292 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
293 union sctp_addr *addr, int len)
294 {
295 struct sctp_af *af;
296
297 /* Check minimum size. */
298 if (len < sizeof (struct sockaddr))
299 return NULL;
300
301 /* V4 mapped address are really of AF_INET family */
302 if (addr->sa.sa_family == AF_INET6 &&
303 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
304 if (!opt->pf->af_supported(AF_INET, opt))
305 return NULL;
306 } else {
307 /* Does this PF support this AF? */
308 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
309 return NULL;
310 }
311
312 /* If we get this far, af is valid. */
313 af = sctp_get_af_specific(addr->sa.sa_family);
314
315 if (len < af->sockaddr_len)
316 return NULL;
317
318 return af;
319 }
320
321 /* Bind a local address either to an endpoint or to an association. */
322 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
323 {
324 struct net *net = sock_net(sk);
325 struct sctp_sock *sp = sctp_sk(sk);
326 struct sctp_endpoint *ep = sp->ep;
327 struct sctp_bind_addr *bp = &ep->base.bind_addr;
328 struct sctp_af *af;
329 unsigned short snum;
330 int ret = 0;
331
332 /* Common sockaddr verification. */
333 af = sctp_sockaddr_af(sp, addr, len);
334 if (!af) {
335 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
336 __func__, sk, addr, len);
337 return -EINVAL;
338 }
339
340 snum = ntohs(addr->v4.sin_port);
341
342 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
343 __func__, sk, &addr->sa, bp->port, snum, len);
344
345 /* PF specific bind() address verification. */
346 if (!sp->pf->bind_verify(sp, addr))
347 return -EADDRNOTAVAIL;
348
349 /* We must either be unbound, or bind to the same port.
350 * It's OK to allow 0 ports if we are already bound.
351 * We'll just inhert an already bound port in this case
352 */
353 if (bp->port) {
354 if (!snum)
355 snum = bp->port;
356 else if (snum != bp->port) {
357 pr_debug("%s: new port %d doesn't match existing port "
358 "%d\n", __func__, snum, bp->port);
359 return -EINVAL;
360 }
361 }
362
363 if (snum && snum < PROT_SOCK &&
364 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
365 return -EACCES;
366
367 /* See if the address matches any of the addresses we may have
368 * already bound before checking against other endpoints.
369 */
370 if (sctp_bind_addr_match(bp, addr, sp))
371 return -EINVAL;
372
373 /* Make sure we are allowed to bind here.
374 * The function sctp_get_port_local() does duplicate address
375 * detection.
376 */
377 addr->v4.sin_port = htons(snum);
378 if ((ret = sctp_get_port_local(sk, addr))) {
379 return -EADDRINUSE;
380 }
381
382 /* Refresh ephemeral port. */
383 if (!bp->port)
384 bp->port = inet_sk(sk)->inet_num;
385
386 /* Add the address to the bind address list.
387 * Use GFP_ATOMIC since BHs will be disabled.
388 */
389 ret = sctp_add_bind_addr(bp, addr, af->sockaddr_len,
390 SCTP_ADDR_SRC, GFP_ATOMIC);
391
392 /* Copy back into socket for getsockname() use. */
393 if (!ret) {
394 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
395 sp->pf->to_sk_saddr(addr, sk);
396 }
397
398 return ret;
399 }
400
401 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
402 *
403 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
404 * at any one time. If a sender, after sending an ASCONF chunk, decides
405 * it needs to transfer another ASCONF Chunk, it MUST wait until the
406 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
407 * subsequent ASCONF. Note this restriction binds each side, so at any
408 * time two ASCONF may be in-transit on any given association (one sent
409 * from each endpoint).
410 */
411 static int sctp_send_asconf(struct sctp_association *asoc,
412 struct sctp_chunk *chunk)
413 {
414 struct net *net = sock_net(asoc->base.sk);
415 int retval = 0;
416
417 /* If there is an outstanding ASCONF chunk, queue it for later
418 * transmission.
419 */
420 if (asoc->addip_last_asconf) {
421 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
422 goto out;
423 }
424
425 /* Hold the chunk until an ASCONF_ACK is received. */
426 sctp_chunk_hold(chunk);
427 retval = sctp_primitive_ASCONF(net, asoc, chunk);
428 if (retval)
429 sctp_chunk_free(chunk);
430 else
431 asoc->addip_last_asconf = chunk;
432
433 out:
434 return retval;
435 }
436
437 /* Add a list of addresses as bind addresses to local endpoint or
438 * association.
439 *
440 * Basically run through each address specified in the addrs/addrcnt
441 * array/length pair, determine if it is IPv6 or IPv4 and call
442 * sctp_do_bind() on it.
443 *
444 * If any of them fails, then the operation will be reversed and the
445 * ones that were added will be removed.
446 *
447 * Only sctp_setsockopt_bindx() is supposed to call this function.
448 */
449 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
450 {
451 int cnt;
452 int retval = 0;
453 void *addr_buf;
454 struct sockaddr *sa_addr;
455 struct sctp_af *af;
456
457 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
458 addrs, addrcnt);
459
460 addr_buf = addrs;
461 for (cnt = 0; cnt < addrcnt; cnt++) {
462 /* The list may contain either IPv4 or IPv6 address;
463 * determine the address length for walking thru the list.
464 */
465 sa_addr = addr_buf;
466 af = sctp_get_af_specific(sa_addr->sa_family);
467 if (!af) {
468 retval = -EINVAL;
469 goto err_bindx_add;
470 }
471
472 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
473 af->sockaddr_len);
474
475 addr_buf += af->sockaddr_len;
476
477 err_bindx_add:
478 if (retval < 0) {
479 /* Failed. Cleanup the ones that have been added */
480 if (cnt > 0)
481 sctp_bindx_rem(sk, addrs, cnt);
482 return retval;
483 }
484 }
485
486 return retval;
487 }
488
489 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
490 * associations that are part of the endpoint indicating that a list of local
491 * addresses are added to the endpoint.
492 *
493 * If any of the addresses is already in the bind address list of the
494 * association, we do not send the chunk for that association. But it will not
495 * affect other associations.
496 *
497 * Only sctp_setsockopt_bindx() is supposed to call this function.
498 */
499 static int sctp_send_asconf_add_ip(struct sock *sk,
500 struct sockaddr *addrs,
501 int addrcnt)
502 {
503 struct net *net = sock_net(sk);
504 struct sctp_sock *sp;
505 struct sctp_endpoint *ep;
506 struct sctp_association *asoc;
507 struct sctp_bind_addr *bp;
508 struct sctp_chunk *chunk;
509 struct sctp_sockaddr_entry *laddr;
510 union sctp_addr *addr;
511 union sctp_addr saveaddr;
512 void *addr_buf;
513 struct sctp_af *af;
514 struct list_head *p;
515 int i;
516 int retval = 0;
517
518 if (!net->sctp.addip_enable)
519 return retval;
520
521 sp = sctp_sk(sk);
522 ep = sp->ep;
523
524 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
525 __func__, sk, addrs, addrcnt);
526
527 list_for_each_entry(asoc, &ep->asocs, asocs) {
528 if (!asoc->peer.asconf_capable)
529 continue;
530
531 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
532 continue;
533
534 if (!sctp_state(asoc, ESTABLISHED))
535 continue;
536
537 /* Check if any address in the packed array of addresses is
538 * in the bind address list of the association. If so,
539 * do not send the asconf chunk to its peer, but continue with
540 * other associations.
541 */
542 addr_buf = addrs;
543 for (i = 0; i < addrcnt; i++) {
544 addr = addr_buf;
545 af = sctp_get_af_specific(addr->v4.sin_family);
546 if (!af) {
547 retval = -EINVAL;
548 goto out;
549 }
550
551 if (sctp_assoc_lookup_laddr(asoc, addr))
552 break;
553
554 addr_buf += af->sockaddr_len;
555 }
556 if (i < addrcnt)
557 continue;
558
559 /* Use the first valid address in bind addr list of
560 * association as Address Parameter of ASCONF CHUNK.
561 */
562 bp = &asoc->base.bind_addr;
563 p = bp->address_list.next;
564 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
565 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
566 addrcnt, SCTP_PARAM_ADD_IP);
567 if (!chunk) {
568 retval = -ENOMEM;
569 goto out;
570 }
571
572 /* Add the new addresses to the bind address list with
573 * use_as_src set to 0.
574 */
575 addr_buf = addrs;
576 for (i = 0; i < addrcnt; i++) {
577 addr = addr_buf;
578 af = sctp_get_af_specific(addr->v4.sin_family);
579 memcpy(&saveaddr, addr, af->sockaddr_len);
580 retval = sctp_add_bind_addr(bp, &saveaddr,
581 sizeof(saveaddr),
582 SCTP_ADDR_NEW, GFP_ATOMIC);
583 addr_buf += af->sockaddr_len;
584 }
585 if (asoc->src_out_of_asoc_ok) {
586 struct sctp_transport *trans;
587
588 list_for_each_entry(trans,
589 &asoc->peer.transport_addr_list, transports) {
590 /* Clear the source and route cache */
591 dst_release(trans->dst);
592 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
593 2*asoc->pathmtu, 4380));
594 trans->ssthresh = asoc->peer.i.a_rwnd;
595 trans->rto = asoc->rto_initial;
596 sctp_max_rto(asoc, trans);
597 trans->rtt = trans->srtt = trans->rttvar = 0;
598 sctp_transport_route(trans, NULL,
599 sctp_sk(asoc->base.sk));
600 }
601 }
602 retval = sctp_send_asconf(asoc, chunk);
603 }
604
605 out:
606 return retval;
607 }
608
609 /* Remove a list of addresses from bind addresses list. Do not remove the
610 * last address.
611 *
612 * Basically run through each address specified in the addrs/addrcnt
613 * array/length pair, determine if it is IPv6 or IPv4 and call
614 * sctp_del_bind() on it.
615 *
616 * If any of them fails, then the operation will be reversed and the
617 * ones that were removed will be added back.
618 *
619 * At least one address has to be left; if only one address is
620 * available, the operation will return -EBUSY.
621 *
622 * Only sctp_setsockopt_bindx() is supposed to call this function.
623 */
624 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
625 {
626 struct sctp_sock *sp = sctp_sk(sk);
627 struct sctp_endpoint *ep = sp->ep;
628 int cnt;
629 struct sctp_bind_addr *bp = &ep->base.bind_addr;
630 int retval = 0;
631 void *addr_buf;
632 union sctp_addr *sa_addr;
633 struct sctp_af *af;
634
635 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
636 __func__, sk, addrs, addrcnt);
637
638 addr_buf = addrs;
639 for (cnt = 0; cnt < addrcnt; cnt++) {
640 /* If the bind address list is empty or if there is only one
641 * bind address, there is nothing more to be removed (we need
642 * at least one address here).
643 */
644 if (list_empty(&bp->address_list) ||
645 (sctp_list_single_entry(&bp->address_list))) {
646 retval = -EBUSY;
647 goto err_bindx_rem;
648 }
649
650 sa_addr = addr_buf;
651 af = sctp_get_af_specific(sa_addr->sa.sa_family);
652 if (!af) {
653 retval = -EINVAL;
654 goto err_bindx_rem;
655 }
656
657 if (!af->addr_valid(sa_addr, sp, NULL)) {
658 retval = -EADDRNOTAVAIL;
659 goto err_bindx_rem;
660 }
661
662 if (sa_addr->v4.sin_port &&
663 sa_addr->v4.sin_port != htons(bp->port)) {
664 retval = -EINVAL;
665 goto err_bindx_rem;
666 }
667
668 if (!sa_addr->v4.sin_port)
669 sa_addr->v4.sin_port = htons(bp->port);
670
671 /* FIXME - There is probably a need to check if sk->sk_saddr and
672 * sk->sk_rcv_addr are currently set to one of the addresses to
673 * be removed. This is something which needs to be looked into
674 * when we are fixing the outstanding issues with multi-homing
675 * socket routing and failover schemes. Refer to comments in
676 * sctp_do_bind(). -daisy
677 */
678 retval = sctp_del_bind_addr(bp, sa_addr);
679
680 addr_buf += af->sockaddr_len;
681 err_bindx_rem:
682 if (retval < 0) {
683 /* Failed. Add the ones that has been removed back */
684 if (cnt > 0)
685 sctp_bindx_add(sk, addrs, cnt);
686 return retval;
687 }
688 }
689
690 return retval;
691 }
692
693 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
694 * the associations that are part of the endpoint indicating that a list of
695 * local addresses are removed from the endpoint.
696 *
697 * If any of the addresses is already in the bind address list of the
698 * association, we do not send the chunk for that association. But it will not
699 * affect other associations.
700 *
701 * Only sctp_setsockopt_bindx() is supposed to call this function.
702 */
703 static int sctp_send_asconf_del_ip(struct sock *sk,
704 struct sockaddr *addrs,
705 int addrcnt)
706 {
707 struct net *net = sock_net(sk);
708 struct sctp_sock *sp;
709 struct sctp_endpoint *ep;
710 struct sctp_association *asoc;
711 struct sctp_transport *transport;
712 struct sctp_bind_addr *bp;
713 struct sctp_chunk *chunk;
714 union sctp_addr *laddr;
715 void *addr_buf;
716 struct sctp_af *af;
717 struct sctp_sockaddr_entry *saddr;
718 int i;
719 int retval = 0;
720 int stored = 0;
721
722 chunk = NULL;
723 if (!net->sctp.addip_enable)
724 return retval;
725
726 sp = sctp_sk(sk);
727 ep = sp->ep;
728
729 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
730 __func__, sk, addrs, addrcnt);
731
732 list_for_each_entry(asoc, &ep->asocs, asocs) {
733
734 if (!asoc->peer.asconf_capable)
735 continue;
736
737 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
738 continue;
739
740 if (!sctp_state(asoc, ESTABLISHED))
741 continue;
742
743 /* Check if any address in the packed array of addresses is
744 * not present in the bind address list of the association.
745 * If so, do not send the asconf chunk to its peer, but
746 * continue with other associations.
747 */
748 addr_buf = addrs;
749 for (i = 0; i < addrcnt; i++) {
750 laddr = addr_buf;
751 af = sctp_get_af_specific(laddr->v4.sin_family);
752 if (!af) {
753 retval = -EINVAL;
754 goto out;
755 }
756
757 if (!sctp_assoc_lookup_laddr(asoc, laddr))
758 break;
759
760 addr_buf += af->sockaddr_len;
761 }
762 if (i < addrcnt)
763 continue;
764
765 /* Find one address in the association's bind address list
766 * that is not in the packed array of addresses. This is to
767 * make sure that we do not delete all the addresses in the
768 * association.
769 */
770 bp = &asoc->base.bind_addr;
771 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
772 addrcnt, sp);
773 if ((laddr == NULL) && (addrcnt == 1)) {
774 if (asoc->asconf_addr_del_pending)
775 continue;
776 asoc->asconf_addr_del_pending =
777 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
778 if (asoc->asconf_addr_del_pending == NULL) {
779 retval = -ENOMEM;
780 goto out;
781 }
782 asoc->asconf_addr_del_pending->sa.sa_family =
783 addrs->sa_family;
784 asoc->asconf_addr_del_pending->v4.sin_port =
785 htons(bp->port);
786 if (addrs->sa_family == AF_INET) {
787 struct sockaddr_in *sin;
788
789 sin = (struct sockaddr_in *)addrs;
790 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
791 } else if (addrs->sa_family == AF_INET6) {
792 struct sockaddr_in6 *sin6;
793
794 sin6 = (struct sockaddr_in6 *)addrs;
795 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
796 }
797
798 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
799 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
800 asoc->asconf_addr_del_pending);
801
802 asoc->src_out_of_asoc_ok = 1;
803 stored = 1;
804 goto skip_mkasconf;
805 }
806
807 if (laddr == NULL)
808 return -EINVAL;
809
810 /* We do not need RCU protection throughout this loop
811 * because this is done under a socket lock from the
812 * setsockopt call.
813 */
814 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
815 SCTP_PARAM_DEL_IP);
816 if (!chunk) {
817 retval = -ENOMEM;
818 goto out;
819 }
820
821 skip_mkasconf:
822 /* Reset use_as_src flag for the addresses in the bind address
823 * list that are to be deleted.
824 */
825 addr_buf = addrs;
826 for (i = 0; i < addrcnt; i++) {
827 laddr = addr_buf;
828 af = sctp_get_af_specific(laddr->v4.sin_family);
829 list_for_each_entry(saddr, &bp->address_list, list) {
830 if (sctp_cmp_addr_exact(&saddr->a, laddr))
831 saddr->state = SCTP_ADDR_DEL;
832 }
833 addr_buf += af->sockaddr_len;
834 }
835
836 /* Update the route and saddr entries for all the transports
837 * as some of the addresses in the bind address list are
838 * about to be deleted and cannot be used as source addresses.
839 */
840 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
841 transports) {
842 dst_release(transport->dst);
843 sctp_transport_route(transport, NULL,
844 sctp_sk(asoc->base.sk));
845 }
846
847 if (stored)
848 /* We don't need to transmit ASCONF */
849 continue;
850 retval = sctp_send_asconf(asoc, chunk);
851 }
852 out:
853 return retval;
854 }
855
856 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
857 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
858 {
859 struct sock *sk = sctp_opt2sk(sp);
860 union sctp_addr *addr;
861 struct sctp_af *af;
862
863 /* It is safe to write port space in caller. */
864 addr = &addrw->a;
865 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
866 af = sctp_get_af_specific(addr->sa.sa_family);
867 if (!af)
868 return -EINVAL;
869 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
870 return -EINVAL;
871
872 if (addrw->state == SCTP_ADDR_NEW)
873 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
874 else
875 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
876 }
877
878 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
879 *
880 * API 8.1
881 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
882 * int flags);
883 *
884 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
885 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
886 * or IPv6 addresses.
887 *
888 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
889 * Section 3.1.2 for this usage.
890 *
891 * addrs is a pointer to an array of one or more socket addresses. Each
892 * address is contained in its appropriate structure (i.e. struct
893 * sockaddr_in or struct sockaddr_in6) the family of the address type
894 * must be used to distinguish the address length (note that this
895 * representation is termed a "packed array" of addresses). The caller
896 * specifies the number of addresses in the array with addrcnt.
897 *
898 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
899 * -1, and sets errno to the appropriate error code.
900 *
901 * For SCTP, the port given in each socket address must be the same, or
902 * sctp_bindx() will fail, setting errno to EINVAL.
903 *
904 * The flags parameter is formed from the bitwise OR of zero or more of
905 * the following currently defined flags:
906 *
907 * SCTP_BINDX_ADD_ADDR
908 *
909 * SCTP_BINDX_REM_ADDR
910 *
911 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
912 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
913 * addresses from the association. The two flags are mutually exclusive;
914 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
915 * not remove all addresses from an association; sctp_bindx() will
916 * reject such an attempt with EINVAL.
917 *
918 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
919 * additional addresses with an endpoint after calling bind(). Or use
920 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
921 * socket is associated with so that no new association accepted will be
922 * associated with those addresses. If the endpoint supports dynamic
923 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
924 * endpoint to send the appropriate message to the peer to change the
925 * peers address lists.
926 *
927 * Adding and removing addresses from a connected association is
928 * optional functionality. Implementations that do not support this
929 * functionality should return EOPNOTSUPP.
930 *
931 * Basically do nothing but copying the addresses from user to kernel
932 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
933 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
934 * from userspace.
935 *
936 * We don't use copy_from_user() for optimization: we first do the
937 * sanity checks (buffer size -fast- and access check-healthy
938 * pointer); if all of those succeed, then we can alloc the memory
939 * (expensive operation) needed to copy the data to kernel. Then we do
940 * the copying without checking the user space area
941 * (__copy_from_user()).
942 *
943 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
944 * it.
945 *
946 * sk The sk of the socket
947 * addrs The pointer to the addresses in user land
948 * addrssize Size of the addrs buffer
949 * op Operation to perform (add or remove, see the flags of
950 * sctp_bindx)
951 *
952 * Returns 0 if ok, <0 errno code on error.
953 */
954 static int sctp_setsockopt_bindx(struct sock *sk,
955 struct sockaddr __user *addrs,
956 int addrs_size, int op)
957 {
958 struct sockaddr *kaddrs;
959 int err;
960 int addrcnt = 0;
961 int walk_size = 0;
962 struct sockaddr *sa_addr;
963 void *addr_buf;
964 struct sctp_af *af;
965
966 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
967 __func__, sk, addrs, addrs_size, op);
968
969 if (unlikely(addrs_size <= 0))
970 return -EINVAL;
971
972 /* Check the user passed a healthy pointer. */
973 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
974 return -EFAULT;
975
976 /* Alloc space for the address array in kernel memory. */
977 kaddrs = kmalloc(addrs_size, GFP_USER | __GFP_NOWARN);
978 if (unlikely(!kaddrs))
979 return -ENOMEM;
980
981 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
982 kfree(kaddrs);
983 return -EFAULT;
984 }
985
986 /* Walk through the addrs buffer and count the number of addresses. */
987 addr_buf = kaddrs;
988 while (walk_size < addrs_size) {
989 if (walk_size + sizeof(sa_family_t) > addrs_size) {
990 kfree(kaddrs);
991 return -EINVAL;
992 }
993
994 sa_addr = addr_buf;
995 af = sctp_get_af_specific(sa_addr->sa_family);
996
997 /* If the address family is not supported or if this address
998 * causes the address buffer to overflow return EINVAL.
999 */
1000 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1001 kfree(kaddrs);
1002 return -EINVAL;
1003 }
1004 addrcnt++;
1005 addr_buf += af->sockaddr_len;
1006 walk_size += af->sockaddr_len;
1007 }
1008
1009 /* Do the work. */
1010 switch (op) {
1011 case SCTP_BINDX_ADD_ADDR:
1012 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1013 if (err)
1014 goto out;
1015 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1016 break;
1017
1018 case SCTP_BINDX_REM_ADDR:
1019 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1020 if (err)
1021 goto out;
1022 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1023 break;
1024
1025 default:
1026 err = -EINVAL;
1027 break;
1028 }
1029
1030 out:
1031 kfree(kaddrs);
1032
1033 return err;
1034 }
1035
1036 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1037 *
1038 * Common routine for handling connect() and sctp_connectx().
1039 * Connect will come in with just a single address.
1040 */
1041 static int __sctp_connect(struct sock *sk,
1042 struct sockaddr *kaddrs,
1043 int addrs_size,
1044 sctp_assoc_t *assoc_id)
1045 {
1046 struct net *net = sock_net(sk);
1047 struct sctp_sock *sp;
1048 struct sctp_endpoint *ep;
1049 struct sctp_association *asoc = NULL;
1050 struct sctp_association *asoc2;
1051 struct sctp_transport *transport;
1052 union sctp_addr to;
1053 sctp_scope_t scope;
1054 long timeo;
1055 int err = 0;
1056 int addrcnt = 0;
1057 int walk_size = 0;
1058 union sctp_addr *sa_addr = NULL;
1059 void *addr_buf;
1060 unsigned short port;
1061 unsigned int f_flags = 0;
1062
1063 sp = sctp_sk(sk);
1064 ep = sp->ep;
1065
1066 /* connect() cannot be done on a socket that is already in ESTABLISHED
1067 * state - UDP-style peeled off socket or a TCP-style socket that
1068 * is already connected.
1069 * It cannot be done even on a TCP-style listening socket.
1070 */
1071 if (sctp_sstate(sk, ESTABLISHED) ||
1072 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1073 err = -EISCONN;
1074 goto out_free;
1075 }
1076
1077 /* Walk through the addrs buffer and count the number of addresses. */
1078 addr_buf = kaddrs;
1079 while (walk_size < addrs_size) {
1080 struct sctp_af *af;
1081
1082 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1083 err = -EINVAL;
1084 goto out_free;
1085 }
1086
1087 sa_addr = addr_buf;
1088 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1089
1090 /* If the address family is not supported or if this address
1091 * causes the address buffer to overflow return EINVAL.
1092 */
1093 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1094 err = -EINVAL;
1095 goto out_free;
1096 }
1097
1098 port = ntohs(sa_addr->v4.sin_port);
1099
1100 /* Save current address so we can work with it */
1101 memcpy(&to, sa_addr, af->sockaddr_len);
1102
1103 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1104 if (err)
1105 goto out_free;
1106
1107 /* Make sure the destination port is correctly set
1108 * in all addresses.
1109 */
1110 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1111 err = -EINVAL;
1112 goto out_free;
1113 }
1114
1115 /* Check if there already is a matching association on the
1116 * endpoint (other than the one created here).
1117 */
1118 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1119 if (asoc2 && asoc2 != asoc) {
1120 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1121 err = -EISCONN;
1122 else
1123 err = -EALREADY;
1124 goto out_free;
1125 }
1126
1127 /* If we could not find a matching association on the endpoint,
1128 * make sure that there is no peeled-off association matching
1129 * the peer address even on another socket.
1130 */
1131 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1132 err = -EADDRNOTAVAIL;
1133 goto out_free;
1134 }
1135
1136 if (!asoc) {
1137 /* If a bind() or sctp_bindx() is not called prior to
1138 * an sctp_connectx() call, the system picks an
1139 * ephemeral port and will choose an address set
1140 * equivalent to binding with a wildcard address.
1141 */
1142 if (!ep->base.bind_addr.port) {
1143 if (sctp_autobind(sk)) {
1144 err = -EAGAIN;
1145 goto out_free;
1146 }
1147 } else {
1148 /*
1149 * If an unprivileged user inherits a 1-many
1150 * style socket with open associations on a
1151 * privileged port, it MAY be permitted to
1152 * accept new associations, but it SHOULD NOT
1153 * be permitted to open new associations.
1154 */
1155 if (ep->base.bind_addr.port < PROT_SOCK &&
1156 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1157 err = -EACCES;
1158 goto out_free;
1159 }
1160 }
1161
1162 scope = sctp_scope(&to);
1163 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1164 if (!asoc) {
1165 err = -ENOMEM;
1166 goto out_free;
1167 }
1168
1169 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1170 GFP_KERNEL);
1171 if (err < 0) {
1172 goto out_free;
1173 }
1174
1175 }
1176
1177 /* Prime the peer's transport structures. */
1178 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1179 SCTP_UNKNOWN);
1180 if (!transport) {
1181 err = -ENOMEM;
1182 goto out_free;
1183 }
1184
1185 addrcnt++;
1186 addr_buf += af->sockaddr_len;
1187 walk_size += af->sockaddr_len;
1188 }
1189
1190 /* In case the user of sctp_connectx() wants an association
1191 * id back, assign one now.
1192 */
1193 if (assoc_id) {
1194 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1195 if (err < 0)
1196 goto out_free;
1197 }
1198
1199 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1200 if (err < 0) {
1201 goto out_free;
1202 }
1203
1204 /* Initialize sk's dport and daddr for getpeername() */
1205 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1206 sp->pf->to_sk_daddr(sa_addr, sk);
1207 sk->sk_err = 0;
1208
1209 /* in-kernel sockets don't generally have a file allocated to them
1210 * if all they do is call sock_create_kern().
1211 */
1212 if (sk->sk_socket->file)
1213 f_flags = sk->sk_socket->file->f_flags;
1214
1215 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1216
1217 err = sctp_wait_for_connect(asoc, &timeo);
1218 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1219 *assoc_id = asoc->assoc_id;
1220
1221 /* Don't free association on exit. */
1222 asoc = NULL;
1223
1224 out_free:
1225 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1226 __func__, asoc, kaddrs, err);
1227
1228 if (asoc) {
1229 /* sctp_primitive_ASSOCIATE may have added this association
1230 * To the hash table, try to unhash it, just in case, its a noop
1231 * if it wasn't hashed so we're safe
1232 */
1233 sctp_association_free(asoc);
1234 }
1235 return err;
1236 }
1237
1238 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1239 *
1240 * API 8.9
1241 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1242 * sctp_assoc_t *asoc);
1243 *
1244 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1245 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1246 * or IPv6 addresses.
1247 *
1248 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1249 * Section 3.1.2 for this usage.
1250 *
1251 * addrs is a pointer to an array of one or more socket addresses. Each
1252 * address is contained in its appropriate structure (i.e. struct
1253 * sockaddr_in or struct sockaddr_in6) the family of the address type
1254 * must be used to distengish the address length (note that this
1255 * representation is termed a "packed array" of addresses). The caller
1256 * specifies the number of addresses in the array with addrcnt.
1257 *
1258 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1259 * the association id of the new association. On failure, sctp_connectx()
1260 * returns -1, and sets errno to the appropriate error code. The assoc_id
1261 * is not touched by the kernel.
1262 *
1263 * For SCTP, the port given in each socket address must be the same, or
1264 * sctp_connectx() will fail, setting errno to EINVAL.
1265 *
1266 * An application can use sctp_connectx to initiate an association with
1267 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1268 * allows a caller to specify multiple addresses at which a peer can be
1269 * reached. The way the SCTP stack uses the list of addresses to set up
1270 * the association is implementation dependent. This function only
1271 * specifies that the stack will try to make use of all the addresses in
1272 * the list when needed.
1273 *
1274 * Note that the list of addresses passed in is only used for setting up
1275 * the association. It does not necessarily equal the set of addresses
1276 * the peer uses for the resulting association. If the caller wants to
1277 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1278 * retrieve them after the association has been set up.
1279 *
1280 * Basically do nothing but copying the addresses from user to kernel
1281 * land and invoking either sctp_connectx(). This is used for tunneling
1282 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1283 *
1284 * We don't use copy_from_user() for optimization: we first do the
1285 * sanity checks (buffer size -fast- and access check-healthy
1286 * pointer); if all of those succeed, then we can alloc the memory
1287 * (expensive operation) needed to copy the data to kernel. Then we do
1288 * the copying without checking the user space area
1289 * (__copy_from_user()).
1290 *
1291 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1292 * it.
1293 *
1294 * sk The sk of the socket
1295 * addrs The pointer to the addresses in user land
1296 * addrssize Size of the addrs buffer
1297 *
1298 * Returns >=0 if ok, <0 errno code on error.
1299 */
1300 static int __sctp_setsockopt_connectx(struct sock *sk,
1301 struct sockaddr __user *addrs,
1302 int addrs_size,
1303 sctp_assoc_t *assoc_id)
1304 {
1305 struct sockaddr *kaddrs;
1306 gfp_t gfp = GFP_KERNEL;
1307 int err = 0;
1308
1309 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1310 __func__, sk, addrs, addrs_size);
1311
1312 if (unlikely(addrs_size <= 0))
1313 return -EINVAL;
1314
1315 /* Check the user passed a healthy pointer. */
1316 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1317 return -EFAULT;
1318
1319 /* Alloc space for the address array in kernel memory. */
1320 if (sk->sk_socket->file)
1321 gfp = GFP_USER | __GFP_NOWARN;
1322 kaddrs = kmalloc(addrs_size, gfp);
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 different part
1373 * is that we store the actual length of the address buffer into the
1374 * addrs_num structure member. That way we can re-use the existing
1375 * code.
1376 */
1377 #ifdef CONFIG_COMPAT
1378 struct compat_sctp_getaddrs_old {
1379 sctp_assoc_t assoc_id;
1380 s32 addr_num;
1381 compat_uptr_t addrs; /* struct sockaddr * */
1382 };
1383 #endif
1384
1385 static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1386 char __user *optval,
1387 int __user *optlen)
1388 {
1389 struct sctp_getaddrs_old param;
1390 sctp_assoc_t assoc_id = 0;
1391 int err = 0;
1392
1393 #ifdef CONFIG_COMPAT
1394 if (in_compat_syscall()) {
1395 struct compat_sctp_getaddrs_old param32;
1396
1397 if (len < sizeof(param32))
1398 return -EINVAL;
1399 if (copy_from_user(&param32, optval, sizeof(param32)))
1400 return -EFAULT;
1401
1402 param.assoc_id = param32.assoc_id;
1403 param.addr_num = param32.addr_num;
1404 param.addrs = compat_ptr(param32.addrs);
1405 } else
1406 #endif
1407 {
1408 if (len < sizeof(param))
1409 return -EINVAL;
1410 if (copy_from_user(&param, optval, sizeof(param)))
1411 return -EFAULT;
1412 }
1413
1414 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1415 param.addrs, param.addr_num,
1416 &assoc_id);
1417 if (err == 0 || err == -EINPROGRESS) {
1418 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1419 return -EFAULT;
1420 if (put_user(sizeof(assoc_id), optlen))
1421 return -EFAULT;
1422 }
1423
1424 return err;
1425 }
1426
1427 /* API 3.1.4 close() - UDP Style Syntax
1428 * Applications use close() to perform graceful shutdown (as described in
1429 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1430 * by a UDP-style socket.
1431 *
1432 * The syntax is
1433 *
1434 * ret = close(int sd);
1435 *
1436 * sd - the socket descriptor of the associations to be closed.
1437 *
1438 * To gracefully shutdown a specific association represented by the
1439 * UDP-style socket, an application should use the sendmsg() call,
1440 * passing no user data, but including the appropriate flag in the
1441 * ancillary data (see Section xxxx).
1442 *
1443 * If sd in the close() call is a branched-off socket representing only
1444 * one association, the shutdown is performed on that association only.
1445 *
1446 * 4.1.6 close() - TCP Style Syntax
1447 *
1448 * Applications use close() to gracefully close down an association.
1449 *
1450 * The syntax is:
1451 *
1452 * int close(int sd);
1453 *
1454 * sd - the socket descriptor of the association to be closed.
1455 *
1456 * After an application calls close() on a socket descriptor, no further
1457 * socket operations will succeed on that descriptor.
1458 *
1459 * API 7.1.4 SO_LINGER
1460 *
1461 * An application using the TCP-style socket can use this option to
1462 * perform the SCTP ABORT primitive. The linger option structure is:
1463 *
1464 * struct linger {
1465 * int l_onoff; // option on/off
1466 * int l_linger; // linger time
1467 * };
1468 *
1469 * To enable the option, set l_onoff to 1. If the l_linger value is set
1470 * to 0, calling close() is the same as the ABORT primitive. If the
1471 * value is set to a negative value, the setsockopt() call will return
1472 * an error. If the value is set to a positive value linger_time, the
1473 * close() can be blocked for at most linger_time ms. If the graceful
1474 * shutdown phase does not finish during this period, close() will
1475 * return but the graceful shutdown phase continues in the system.
1476 */
1477 static void sctp_close(struct sock *sk, long timeout)
1478 {
1479 struct net *net = sock_net(sk);
1480 struct sctp_endpoint *ep;
1481 struct sctp_association *asoc;
1482 struct list_head *pos, *temp;
1483 unsigned int data_was_unread;
1484
1485 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1486
1487 lock_sock(sk);
1488 sk->sk_shutdown = SHUTDOWN_MASK;
1489 sk->sk_state = SCTP_SS_CLOSING;
1490
1491 ep = sctp_sk(sk)->ep;
1492
1493 /* Clean up any skbs sitting on the receive queue. */
1494 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1495 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1496
1497 /* Walk all associations on an endpoint. */
1498 list_for_each_safe(pos, temp, &ep->asocs) {
1499 asoc = list_entry(pos, struct sctp_association, asocs);
1500
1501 if (sctp_style(sk, TCP)) {
1502 /* A closed association can still be in the list if
1503 * it belongs to a TCP-style listening socket that is
1504 * not yet accepted. If so, free it. If not, send an
1505 * ABORT or SHUTDOWN based on the linger options.
1506 */
1507 if (sctp_state(asoc, CLOSED)) {
1508 sctp_association_free(asoc);
1509 continue;
1510 }
1511 }
1512
1513 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1514 !skb_queue_empty(&asoc->ulpq.reasm) ||
1515 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1516 struct sctp_chunk *chunk;
1517
1518 chunk = sctp_make_abort_user(asoc, NULL, 0);
1519 sctp_primitive_ABORT(net, asoc, chunk);
1520 } else
1521 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1522 }
1523
1524 /* On a TCP-style socket, block for at most linger_time if set. */
1525 if (sctp_style(sk, TCP) && timeout)
1526 sctp_wait_for_close(sk, timeout);
1527
1528 /* This will run the backlog queue. */
1529 release_sock(sk);
1530
1531 /* Supposedly, no process has access to the socket, but
1532 * the net layers still may.
1533 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1534 * held and that should be grabbed before socket lock.
1535 */
1536 spin_lock_bh(&net->sctp.addr_wq_lock);
1537 bh_lock_sock(sk);
1538
1539 /* Hold the sock, since sk_common_release() will put sock_put()
1540 * and we have just a little more cleanup.
1541 */
1542 sock_hold(sk);
1543 sk_common_release(sk);
1544
1545 bh_unlock_sock(sk);
1546 spin_unlock_bh(&net->sctp.addr_wq_lock);
1547
1548 sock_put(sk);
1549
1550 SCTP_DBG_OBJCNT_DEC(sock);
1551 }
1552
1553 /* Handle EPIPE error. */
1554 static int sctp_error(struct sock *sk, int flags, int err)
1555 {
1556 if (err == -EPIPE)
1557 err = sock_error(sk) ? : -EPIPE;
1558 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1559 send_sig(SIGPIPE, current, 0);
1560 return err;
1561 }
1562
1563 /* API 3.1.3 sendmsg() - UDP Style Syntax
1564 *
1565 * An application uses sendmsg() and recvmsg() calls to transmit data to
1566 * and receive data from its peer.
1567 *
1568 * ssize_t sendmsg(int socket, const struct msghdr *message,
1569 * int flags);
1570 *
1571 * socket - the socket descriptor of the endpoint.
1572 * message - pointer to the msghdr structure which contains a single
1573 * user message and possibly some ancillary data.
1574 *
1575 * See Section 5 for complete description of the data
1576 * structures.
1577 *
1578 * flags - flags sent or received with the user message, see Section
1579 * 5 for complete description of the flags.
1580 *
1581 * Note: This function could use a rewrite especially when explicit
1582 * connect support comes in.
1583 */
1584 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1585
1586 static int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1587
1588 static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len)
1589 {
1590 struct net *net = sock_net(sk);
1591 struct sctp_sock *sp;
1592 struct sctp_endpoint *ep;
1593 struct sctp_association *new_asoc = NULL, *asoc = NULL;
1594 struct sctp_transport *transport, *chunk_tp;
1595 struct sctp_chunk *chunk;
1596 union sctp_addr to;
1597 struct sockaddr *msg_name = NULL;
1598 struct sctp_sndrcvinfo default_sinfo;
1599 struct sctp_sndrcvinfo *sinfo;
1600 struct sctp_initmsg *sinit;
1601 sctp_assoc_t associd = 0;
1602 sctp_cmsgs_t cmsgs = { NULL };
1603 sctp_scope_t scope;
1604 bool fill_sinfo_ttl = false, wait_connect = false;
1605 struct sctp_datamsg *datamsg;
1606 int msg_flags = msg->msg_flags;
1607 __u16 sinfo_flags = 0;
1608 long timeo;
1609 int err;
1610
1611 err = 0;
1612 sp = sctp_sk(sk);
1613 ep = sp->ep;
1614
1615 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1616 msg, msg_len, ep);
1617
1618 /* We cannot send a message over a TCP-style listening socket. */
1619 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1620 err = -EPIPE;
1621 goto out_nounlock;
1622 }
1623
1624 /* Parse out the SCTP CMSGs. */
1625 err = sctp_msghdr_parse(msg, &cmsgs);
1626 if (err) {
1627 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1628 goto out_nounlock;
1629 }
1630
1631 /* Fetch the destination address for this packet. This
1632 * address only selects the association--it is not necessarily
1633 * the address we will send to.
1634 * For a peeled-off socket, msg_name is ignored.
1635 */
1636 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1637 int msg_namelen = msg->msg_namelen;
1638
1639 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1640 msg_namelen);
1641 if (err)
1642 return err;
1643
1644 if (msg_namelen > sizeof(to))
1645 msg_namelen = sizeof(to);
1646 memcpy(&to, msg->msg_name, msg_namelen);
1647 msg_name = msg->msg_name;
1648 }
1649
1650 sinit = cmsgs.init;
1651 if (cmsgs.sinfo != NULL) {
1652 memset(&default_sinfo, 0, sizeof(default_sinfo));
1653 default_sinfo.sinfo_stream = cmsgs.sinfo->snd_sid;
1654 default_sinfo.sinfo_flags = cmsgs.sinfo->snd_flags;
1655 default_sinfo.sinfo_ppid = cmsgs.sinfo->snd_ppid;
1656 default_sinfo.sinfo_context = cmsgs.sinfo->snd_context;
1657 default_sinfo.sinfo_assoc_id = cmsgs.sinfo->snd_assoc_id;
1658
1659 sinfo = &default_sinfo;
1660 fill_sinfo_ttl = true;
1661 } else {
1662 sinfo = cmsgs.srinfo;
1663 }
1664 /* Did the user specify SNDINFO/SNDRCVINFO? */
1665 if (sinfo) {
1666 sinfo_flags = sinfo->sinfo_flags;
1667 associd = sinfo->sinfo_assoc_id;
1668 }
1669
1670 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1671 msg_len, sinfo_flags);
1672
1673 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1674 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1675 err = -EINVAL;
1676 goto out_nounlock;
1677 }
1678
1679 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1680 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1681 * If SCTP_ABORT is set, the message length could be non zero with
1682 * the msg_iov set to the user abort reason.
1683 */
1684 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1685 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1686 err = -EINVAL;
1687 goto out_nounlock;
1688 }
1689
1690 /* If SCTP_ADDR_OVER is set, there must be an address
1691 * specified in msg_name.
1692 */
1693 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1694 err = -EINVAL;
1695 goto out_nounlock;
1696 }
1697
1698 transport = NULL;
1699
1700 pr_debug("%s: about to look up association\n", __func__);
1701
1702 lock_sock(sk);
1703
1704 /* If a msg_name has been specified, assume this is to be used. */
1705 if (msg_name) {
1706 /* Look for a matching association on the endpoint. */
1707 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1708 if (!asoc) {
1709 /* If we could not find a matching association on the
1710 * endpoint, make sure that it is not a TCP-style
1711 * socket that already has an association or there is
1712 * no peeled-off association on another socket.
1713 */
1714 if ((sctp_style(sk, TCP) &&
1715 sctp_sstate(sk, ESTABLISHED)) ||
1716 sctp_endpoint_is_peeled_off(ep, &to)) {
1717 err = -EADDRNOTAVAIL;
1718 goto out_unlock;
1719 }
1720 }
1721 } else {
1722 asoc = sctp_id2assoc(sk, associd);
1723 if (!asoc) {
1724 err = -EPIPE;
1725 goto out_unlock;
1726 }
1727 }
1728
1729 if (asoc) {
1730 pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1731
1732 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1733 * socket that has an association in CLOSED state. This can
1734 * happen when an accepted socket has an association that is
1735 * already CLOSED.
1736 */
1737 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1738 err = -EPIPE;
1739 goto out_unlock;
1740 }
1741
1742 if (sinfo_flags & SCTP_EOF) {
1743 pr_debug("%s: shutting down association:%p\n",
1744 __func__, asoc);
1745
1746 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1747 err = 0;
1748 goto out_unlock;
1749 }
1750 if (sinfo_flags & SCTP_ABORT) {
1751
1752 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1753 if (!chunk) {
1754 err = -ENOMEM;
1755 goto out_unlock;
1756 }
1757
1758 pr_debug("%s: aborting association:%p\n",
1759 __func__, asoc);
1760
1761 sctp_primitive_ABORT(net, asoc, chunk);
1762 err = 0;
1763 goto out_unlock;
1764 }
1765 }
1766
1767 /* Do we need to create the association? */
1768 if (!asoc) {
1769 pr_debug("%s: there is no association yet\n", __func__);
1770
1771 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1772 err = -EINVAL;
1773 goto out_unlock;
1774 }
1775
1776 /* Check for invalid stream against the stream counts,
1777 * either the default or the user specified stream counts.
1778 */
1779 if (sinfo) {
1780 if (!sinit || !sinit->sinit_num_ostreams) {
1781 /* Check against the defaults. */
1782 if (sinfo->sinfo_stream >=
1783 sp->initmsg.sinit_num_ostreams) {
1784 err = -EINVAL;
1785 goto out_unlock;
1786 }
1787 } else {
1788 /* Check against the requested. */
1789 if (sinfo->sinfo_stream >=
1790 sinit->sinit_num_ostreams) {
1791 err = -EINVAL;
1792 goto out_unlock;
1793 }
1794 }
1795 }
1796
1797 /*
1798 * API 3.1.2 bind() - UDP Style Syntax
1799 * If a bind() or sctp_bindx() is not called prior to a
1800 * sendmsg() call that initiates a new association, the
1801 * system picks an ephemeral port and will choose an address
1802 * set equivalent to binding with a wildcard address.
1803 */
1804 if (!ep->base.bind_addr.port) {
1805 if (sctp_autobind(sk)) {
1806 err = -EAGAIN;
1807 goto out_unlock;
1808 }
1809 } else {
1810 /*
1811 * If an unprivileged user inherits a one-to-many
1812 * style socket with open associations on a privileged
1813 * port, it MAY be permitted to accept new associations,
1814 * but it SHOULD NOT be permitted to open new
1815 * associations.
1816 */
1817 if (ep->base.bind_addr.port < PROT_SOCK &&
1818 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1819 err = -EACCES;
1820 goto out_unlock;
1821 }
1822 }
1823
1824 scope = sctp_scope(&to);
1825 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1826 if (!new_asoc) {
1827 err = -ENOMEM;
1828 goto out_unlock;
1829 }
1830 asoc = new_asoc;
1831 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1832 if (err < 0) {
1833 err = -ENOMEM;
1834 goto out_free;
1835 }
1836
1837 /* If the SCTP_INIT ancillary data is specified, set all
1838 * the association init values accordingly.
1839 */
1840 if (sinit) {
1841 if (sinit->sinit_num_ostreams) {
1842 asoc->c.sinit_num_ostreams =
1843 sinit->sinit_num_ostreams;
1844 }
1845 if (sinit->sinit_max_instreams) {
1846 asoc->c.sinit_max_instreams =
1847 sinit->sinit_max_instreams;
1848 }
1849 if (sinit->sinit_max_attempts) {
1850 asoc->max_init_attempts
1851 = sinit->sinit_max_attempts;
1852 }
1853 if (sinit->sinit_max_init_timeo) {
1854 asoc->max_init_timeo =
1855 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1856 }
1857 }
1858
1859 /* Prime the peer's transport structures. */
1860 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1861 if (!transport) {
1862 err = -ENOMEM;
1863 goto out_free;
1864 }
1865 }
1866
1867 /* ASSERT: we have a valid association at this point. */
1868 pr_debug("%s: we have a valid association\n", __func__);
1869
1870 if (!sinfo) {
1871 /* If the user didn't specify SNDINFO/SNDRCVINFO, make up
1872 * one with some defaults.
1873 */
1874 memset(&default_sinfo, 0, sizeof(default_sinfo));
1875 default_sinfo.sinfo_stream = asoc->default_stream;
1876 default_sinfo.sinfo_flags = asoc->default_flags;
1877 default_sinfo.sinfo_ppid = asoc->default_ppid;
1878 default_sinfo.sinfo_context = asoc->default_context;
1879 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1880 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1881
1882 sinfo = &default_sinfo;
1883 } else if (fill_sinfo_ttl) {
1884 /* In case SNDINFO was specified, we still need to fill
1885 * it with a default ttl from the assoc here.
1886 */
1887 sinfo->sinfo_timetolive = asoc->default_timetolive;
1888 }
1889
1890 /* API 7.1.7, the sndbuf size per association bounds the
1891 * maximum size of data that can be sent in a single send call.
1892 */
1893 if (msg_len > sk->sk_sndbuf) {
1894 err = -EMSGSIZE;
1895 goto out_free;
1896 }
1897
1898 if (asoc->pmtu_pending)
1899 sctp_assoc_pending_pmtu(sk, asoc);
1900
1901 /* If fragmentation is disabled and the message length exceeds the
1902 * association fragmentation point, return EMSGSIZE. The I-D
1903 * does not specify what this error is, but this looks like
1904 * a great fit.
1905 */
1906 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1907 err = -EMSGSIZE;
1908 goto out_free;
1909 }
1910
1911 /* Check for invalid stream. */
1912 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1913 err = -EINVAL;
1914 goto out_free;
1915 }
1916
1917 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1918 if (!sctp_wspace(asoc)) {
1919 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1920 if (err)
1921 goto out_free;
1922 }
1923
1924 /* If an address is passed with the sendto/sendmsg call, it is used
1925 * to override the primary destination address in the TCP model, or
1926 * when SCTP_ADDR_OVER flag is set in the UDP model.
1927 */
1928 if ((sctp_style(sk, TCP) && msg_name) ||
1929 (sinfo_flags & SCTP_ADDR_OVER)) {
1930 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1931 if (!chunk_tp) {
1932 err = -EINVAL;
1933 goto out_free;
1934 }
1935 } else
1936 chunk_tp = NULL;
1937
1938 /* Auto-connect, if we aren't connected already. */
1939 if (sctp_state(asoc, CLOSED)) {
1940 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1941 if (err < 0)
1942 goto out_free;
1943
1944 wait_connect = true;
1945 pr_debug("%s: we associated primitively\n", __func__);
1946 }
1947
1948 /* Break the message into multiple chunks of maximum size. */
1949 datamsg = sctp_datamsg_from_user(asoc, sinfo, &msg->msg_iter);
1950 if (IS_ERR(datamsg)) {
1951 err = PTR_ERR(datamsg);
1952 goto out_free;
1953 }
1954
1955 /* Now send the (possibly) fragmented message. */
1956 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1957 /* Do accounting for the write space. */
1958 sctp_set_owner_w(chunk);
1959
1960 chunk->transport = chunk_tp;
1961 }
1962
1963 /* Send it to the lower layers. Note: all chunks
1964 * must either fail or succeed. The lower layer
1965 * works that way today. Keep it that way or this
1966 * breaks.
1967 */
1968 err = sctp_primitive_SEND(net, asoc, datamsg);
1969 sctp_datamsg_put(datamsg);
1970 /* Did the lower layer accept the chunk? */
1971 if (err)
1972 goto out_free;
1973
1974 pr_debug("%s: we sent primitively\n", __func__);
1975
1976 err = msg_len;
1977
1978 if (unlikely(wait_connect)) {
1979 timeo = sock_sndtimeo(sk, msg_flags & MSG_DONTWAIT);
1980 sctp_wait_for_connect(asoc, &timeo);
1981 }
1982
1983 /* If we are already past ASSOCIATE, the lower
1984 * layers are responsible for association cleanup.
1985 */
1986 goto out_unlock;
1987
1988 out_free:
1989 if (new_asoc)
1990 sctp_association_free(asoc);
1991 out_unlock:
1992 release_sock(sk);
1993
1994 out_nounlock:
1995 return sctp_error(sk, msg_flags, err);
1996
1997 #if 0
1998 do_sock_err:
1999 if (msg_len)
2000 err = msg_len;
2001 else
2002 err = sock_error(sk);
2003 goto out;
2004
2005 do_interrupted:
2006 if (msg_len)
2007 err = msg_len;
2008 goto out;
2009 #endif /* 0 */
2010 }
2011
2012 /* This is an extended version of skb_pull() that removes the data from the
2013 * start of a skb even when data is spread across the list of skb's in the
2014 * frag_list. len specifies the total amount of data that needs to be removed.
2015 * when 'len' bytes could be removed from the skb, it returns 0.
2016 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2017 * could not be removed.
2018 */
2019 static int sctp_skb_pull(struct sk_buff *skb, int len)
2020 {
2021 struct sk_buff *list;
2022 int skb_len = skb_headlen(skb);
2023 int rlen;
2024
2025 if (len <= skb_len) {
2026 __skb_pull(skb, len);
2027 return 0;
2028 }
2029 len -= skb_len;
2030 __skb_pull(skb, skb_len);
2031
2032 skb_walk_frags(skb, list) {
2033 rlen = sctp_skb_pull(list, len);
2034 skb->len -= (len-rlen);
2035 skb->data_len -= (len-rlen);
2036
2037 if (!rlen)
2038 return 0;
2039
2040 len = rlen;
2041 }
2042
2043 return len;
2044 }
2045
2046 /* API 3.1.3 recvmsg() - UDP Style Syntax
2047 *
2048 * ssize_t recvmsg(int socket, struct msghdr *message,
2049 * int flags);
2050 *
2051 * socket - the socket descriptor of the endpoint.
2052 * message - pointer to the msghdr structure which contains a single
2053 * user message and possibly some ancillary data.
2054 *
2055 * See Section 5 for complete description of the data
2056 * structures.
2057 *
2058 * flags - flags sent or received with the user message, see Section
2059 * 5 for complete description of the flags.
2060 */
2061 static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
2062 int noblock, int flags, int *addr_len)
2063 {
2064 struct sctp_ulpevent *event = NULL;
2065 struct sctp_sock *sp = sctp_sk(sk);
2066 struct sk_buff *skb;
2067 int copied;
2068 int err = 0;
2069 int skb_len;
2070
2071 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2072 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2073 addr_len);
2074
2075 lock_sock(sk);
2076
2077 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2078 err = -ENOTCONN;
2079 goto out;
2080 }
2081
2082 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2083 if (!skb)
2084 goto out;
2085
2086 /* Get the total length of the skb including any skb's in the
2087 * frag_list.
2088 */
2089 skb_len = skb->len;
2090
2091 copied = skb_len;
2092 if (copied > len)
2093 copied = len;
2094
2095 err = skb_copy_datagram_msg(skb, 0, msg, copied);
2096
2097 event = sctp_skb2event(skb);
2098
2099 if (err)
2100 goto out_free;
2101
2102 sock_recv_ts_and_drops(msg, sk, skb);
2103 if (sctp_ulpevent_is_notification(event)) {
2104 msg->msg_flags |= MSG_NOTIFICATION;
2105 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2106 } else {
2107 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2108 }
2109
2110 /* Check if we allow SCTP_NXTINFO. */
2111 if (sp->recvnxtinfo)
2112 sctp_ulpevent_read_nxtinfo(event, msg, sk);
2113 /* Check if we allow SCTP_RCVINFO. */
2114 if (sp->recvrcvinfo)
2115 sctp_ulpevent_read_rcvinfo(event, msg);
2116 /* Check if we allow SCTP_SNDRCVINFO. */
2117 if (sp->subscribe.sctp_data_io_event)
2118 sctp_ulpevent_read_sndrcvinfo(event, msg);
2119
2120 err = copied;
2121
2122 /* If skb's length exceeds the user's buffer, update the skb and
2123 * push it back to the receive_queue so that the next call to
2124 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2125 */
2126 if (skb_len > copied) {
2127 msg->msg_flags &= ~MSG_EOR;
2128 if (flags & MSG_PEEK)
2129 goto out_free;
2130 sctp_skb_pull(skb, copied);
2131 skb_queue_head(&sk->sk_receive_queue, skb);
2132
2133 /* When only partial message is copied to the user, increase
2134 * rwnd by that amount. If all the data in the skb is read,
2135 * rwnd is updated when the event is freed.
2136 */
2137 if (!sctp_ulpevent_is_notification(event))
2138 sctp_assoc_rwnd_increase(event->asoc, copied);
2139 goto out;
2140 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2141 (event->msg_flags & MSG_EOR))
2142 msg->msg_flags |= MSG_EOR;
2143 else
2144 msg->msg_flags &= ~MSG_EOR;
2145
2146 out_free:
2147 if (flags & MSG_PEEK) {
2148 /* Release the skb reference acquired after peeking the skb in
2149 * sctp_skb_recv_datagram().
2150 */
2151 kfree_skb(skb);
2152 } else {
2153 /* Free the event which includes releasing the reference to
2154 * the owner of the skb, freeing the skb and updating the
2155 * rwnd.
2156 */
2157 sctp_ulpevent_free(event);
2158 }
2159 out:
2160 release_sock(sk);
2161 return err;
2162 }
2163
2164 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2165 *
2166 * This option is a on/off flag. If enabled no SCTP message
2167 * fragmentation will be performed. Instead if a message being sent
2168 * exceeds the current PMTU size, the message will NOT be sent and
2169 * instead a error will be indicated to the user.
2170 */
2171 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2172 char __user *optval,
2173 unsigned int optlen)
2174 {
2175 int val;
2176
2177 if (optlen < sizeof(int))
2178 return -EINVAL;
2179
2180 if (get_user(val, (int __user *)optval))
2181 return -EFAULT;
2182
2183 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2184
2185 return 0;
2186 }
2187
2188 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2189 unsigned int optlen)
2190 {
2191 struct sctp_association *asoc;
2192 struct sctp_ulpevent *event;
2193
2194 if (optlen > sizeof(struct sctp_event_subscribe))
2195 return -EINVAL;
2196 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2197 return -EFAULT;
2198
2199 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2200 * if there is no data to be sent or retransmit, the stack will
2201 * immediately send up this notification.
2202 */
2203 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2204 &sctp_sk(sk)->subscribe)) {
2205 asoc = sctp_id2assoc(sk, 0);
2206
2207 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2208 event = sctp_ulpevent_make_sender_dry_event(asoc,
2209 GFP_ATOMIC);
2210 if (!event)
2211 return -ENOMEM;
2212
2213 sctp_ulpq_tail_event(&asoc->ulpq, event);
2214 }
2215 }
2216
2217 return 0;
2218 }
2219
2220 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2221 *
2222 * This socket option is applicable to the UDP-style socket only. When
2223 * set it will cause associations that are idle for more than the
2224 * specified number of seconds to automatically close. An association
2225 * being idle is defined an association that has NOT sent or received
2226 * user data. The special value of '0' indicates that no automatic
2227 * close of any associations should be performed. The option expects an
2228 * integer defining the number of seconds of idle time before an
2229 * association is closed.
2230 */
2231 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2232 unsigned int optlen)
2233 {
2234 struct sctp_sock *sp = sctp_sk(sk);
2235 struct net *net = sock_net(sk);
2236
2237 /* Applicable to UDP-style socket only */
2238 if (sctp_style(sk, TCP))
2239 return -EOPNOTSUPP;
2240 if (optlen != sizeof(int))
2241 return -EINVAL;
2242 if (copy_from_user(&sp->autoclose, optval, optlen))
2243 return -EFAULT;
2244
2245 if (sp->autoclose > net->sctp.max_autoclose)
2246 sp->autoclose = net->sctp.max_autoclose;
2247
2248 return 0;
2249 }
2250
2251 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2252 *
2253 * Applications can enable or disable heartbeats for any peer address of
2254 * an association, modify an address's heartbeat interval, force a
2255 * heartbeat to be sent immediately, and adjust the address's maximum
2256 * number of retransmissions sent before an address is considered
2257 * unreachable. The following structure is used to access and modify an
2258 * address's parameters:
2259 *
2260 * struct sctp_paddrparams {
2261 * sctp_assoc_t spp_assoc_id;
2262 * struct sockaddr_storage spp_address;
2263 * uint32_t spp_hbinterval;
2264 * uint16_t spp_pathmaxrxt;
2265 * uint32_t spp_pathmtu;
2266 * uint32_t spp_sackdelay;
2267 * uint32_t spp_flags;
2268 * };
2269 *
2270 * spp_assoc_id - (one-to-many style socket) This is filled in the
2271 * application, and identifies the association for
2272 * this query.
2273 * spp_address - This specifies which address is of interest.
2274 * spp_hbinterval - This contains the value of the heartbeat interval,
2275 * in milliseconds. If a value of zero
2276 * is present in this field then no changes are to
2277 * be made to this parameter.
2278 * spp_pathmaxrxt - This contains the maximum number of
2279 * retransmissions before this address shall be
2280 * considered unreachable. If a value of zero
2281 * is present in this field then no changes are to
2282 * be made to this parameter.
2283 * spp_pathmtu - When Path MTU discovery is disabled the value
2284 * specified here will be the "fixed" path mtu.
2285 * Note that if the spp_address field is empty
2286 * then all associations on this address will
2287 * have this fixed path mtu set upon them.
2288 *
2289 * spp_sackdelay - When delayed sack is enabled, this value specifies
2290 * the number of milliseconds that sacks will be delayed
2291 * for. This value will apply to all addresses of an
2292 * association if the spp_address field is empty. Note
2293 * also, that if delayed sack is enabled and this
2294 * value is set to 0, no change is made to the last
2295 * recorded delayed sack timer value.
2296 *
2297 * spp_flags - These flags are used to control various features
2298 * on an association. The flag field may contain
2299 * zero or more of the following options.
2300 *
2301 * SPP_HB_ENABLE - Enable heartbeats on the
2302 * specified address. Note that if the address
2303 * field is empty all addresses for the association
2304 * have heartbeats enabled upon them.
2305 *
2306 * SPP_HB_DISABLE - Disable heartbeats on the
2307 * speicifed address. Note that if the address
2308 * field is empty all addresses for the association
2309 * will have their heartbeats disabled. Note also
2310 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2311 * mutually exclusive, only one of these two should
2312 * be specified. Enabling both fields will have
2313 * undetermined results.
2314 *
2315 * SPP_HB_DEMAND - Request a user initiated heartbeat
2316 * to be made immediately.
2317 *
2318 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2319 * heartbeat delayis to be set to the value of 0
2320 * milliseconds.
2321 *
2322 * SPP_PMTUD_ENABLE - This field will enable PMTU
2323 * discovery upon the specified address. Note that
2324 * if the address feild is empty then all addresses
2325 * on the association are effected.
2326 *
2327 * SPP_PMTUD_DISABLE - This field will disable PMTU
2328 * discovery upon the specified address. Note that
2329 * if the address feild is empty then all addresses
2330 * on the association are effected. Not also that
2331 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2332 * exclusive. Enabling both will have undetermined
2333 * results.
2334 *
2335 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2336 * on delayed sack. The time specified in spp_sackdelay
2337 * is used to specify the sack delay for this address. Note
2338 * that if spp_address is empty then all addresses will
2339 * enable delayed sack and take on the sack delay
2340 * value specified in spp_sackdelay.
2341 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2342 * off delayed sack. If the spp_address field is blank then
2343 * delayed sack is disabled for the entire association. Note
2344 * also that this field is mutually exclusive to
2345 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2346 * results.
2347 */
2348 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2349 struct sctp_transport *trans,
2350 struct sctp_association *asoc,
2351 struct sctp_sock *sp,
2352 int hb_change,
2353 int pmtud_change,
2354 int sackdelay_change)
2355 {
2356 int error;
2357
2358 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2359 struct net *net = sock_net(trans->asoc->base.sk);
2360
2361 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2362 if (error)
2363 return error;
2364 }
2365
2366 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2367 * this field is ignored. Note also that a value of zero indicates
2368 * the current setting should be left unchanged.
2369 */
2370 if (params->spp_flags & SPP_HB_ENABLE) {
2371
2372 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2373 * set. This lets us use 0 value when this flag
2374 * is set.
2375 */
2376 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2377 params->spp_hbinterval = 0;
2378
2379 if (params->spp_hbinterval ||
2380 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2381 if (trans) {
2382 trans->hbinterval =
2383 msecs_to_jiffies(params->spp_hbinterval);
2384 } else if (asoc) {
2385 asoc->hbinterval =
2386 msecs_to_jiffies(params->spp_hbinterval);
2387 } else {
2388 sp->hbinterval = params->spp_hbinterval;
2389 }
2390 }
2391 }
2392
2393 if (hb_change) {
2394 if (trans) {
2395 trans->param_flags =
2396 (trans->param_flags & ~SPP_HB) | hb_change;
2397 } else if (asoc) {
2398 asoc->param_flags =
2399 (asoc->param_flags & ~SPP_HB) | hb_change;
2400 } else {
2401 sp->param_flags =
2402 (sp->param_flags & ~SPP_HB) | hb_change;
2403 }
2404 }
2405
2406 /* When Path MTU discovery is disabled the value specified here will
2407 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2408 * include the flag SPP_PMTUD_DISABLE for this field to have any
2409 * effect).
2410 */
2411 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2412 if (trans) {
2413 trans->pathmtu = params->spp_pathmtu;
2414 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2415 } else if (asoc) {
2416 asoc->pathmtu = params->spp_pathmtu;
2417 sctp_frag_point(asoc, params->spp_pathmtu);
2418 } else {
2419 sp->pathmtu = params->spp_pathmtu;
2420 }
2421 }
2422
2423 if (pmtud_change) {
2424 if (trans) {
2425 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2426 (params->spp_flags & SPP_PMTUD_ENABLE);
2427 trans->param_flags =
2428 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2429 if (update) {
2430 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2431 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2432 }
2433 } else if (asoc) {
2434 asoc->param_flags =
2435 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2436 } else {
2437 sp->param_flags =
2438 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2439 }
2440 }
2441
2442 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2443 * value of this field is ignored. Note also that a value of zero
2444 * indicates the current setting should be left unchanged.
2445 */
2446 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2447 if (trans) {
2448 trans->sackdelay =
2449 msecs_to_jiffies(params->spp_sackdelay);
2450 } else if (asoc) {
2451 asoc->sackdelay =
2452 msecs_to_jiffies(params->spp_sackdelay);
2453 } else {
2454 sp->sackdelay = params->spp_sackdelay;
2455 }
2456 }
2457
2458 if (sackdelay_change) {
2459 if (trans) {
2460 trans->param_flags =
2461 (trans->param_flags & ~SPP_SACKDELAY) |
2462 sackdelay_change;
2463 } else if (asoc) {
2464 asoc->param_flags =
2465 (asoc->param_flags & ~SPP_SACKDELAY) |
2466 sackdelay_change;
2467 } else {
2468 sp->param_flags =
2469 (sp->param_flags & ~SPP_SACKDELAY) |
2470 sackdelay_change;
2471 }
2472 }
2473
2474 /* Note that a value of zero indicates the current setting should be
2475 left unchanged.
2476 */
2477 if (params->spp_pathmaxrxt) {
2478 if (trans) {
2479 trans->pathmaxrxt = params->spp_pathmaxrxt;
2480 } else if (asoc) {
2481 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2482 } else {
2483 sp->pathmaxrxt = params->spp_pathmaxrxt;
2484 }
2485 }
2486
2487 return 0;
2488 }
2489
2490 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2491 char __user *optval,
2492 unsigned int optlen)
2493 {
2494 struct sctp_paddrparams params;
2495 struct sctp_transport *trans = NULL;
2496 struct sctp_association *asoc = NULL;
2497 struct sctp_sock *sp = sctp_sk(sk);
2498 int error;
2499 int hb_change, pmtud_change, sackdelay_change;
2500
2501 if (optlen != sizeof(struct sctp_paddrparams))
2502 return -EINVAL;
2503
2504 if (copy_from_user(&params, optval, optlen))
2505 return -EFAULT;
2506
2507 /* Validate flags and value parameters. */
2508 hb_change = params.spp_flags & SPP_HB;
2509 pmtud_change = params.spp_flags & SPP_PMTUD;
2510 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2511
2512 if (hb_change == SPP_HB ||
2513 pmtud_change == SPP_PMTUD ||
2514 sackdelay_change == SPP_SACKDELAY ||
2515 params.spp_sackdelay > 500 ||
2516 (params.spp_pathmtu &&
2517 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2518 return -EINVAL;
2519
2520 /* If an address other than INADDR_ANY is specified, and
2521 * no transport is found, then the request is invalid.
2522 */
2523 if (!sctp_is_any(sk, (union sctp_addr *)&params.spp_address)) {
2524 trans = sctp_addr_id2transport(sk, &params.spp_address,
2525 params.spp_assoc_id);
2526 if (!trans)
2527 return -EINVAL;
2528 }
2529
2530 /* Get association, if assoc_id != 0 and the socket is a one
2531 * to many style socket, and an association was not found, then
2532 * the id was invalid.
2533 */
2534 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2535 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2536 return -EINVAL;
2537
2538 /* Heartbeat demand can only be sent on a transport or
2539 * association, but not a socket.
2540 */
2541 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2542 return -EINVAL;
2543
2544 /* Process parameters. */
2545 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2546 hb_change, pmtud_change,
2547 sackdelay_change);
2548
2549 if (error)
2550 return error;
2551
2552 /* If changes are for association, also apply parameters to each
2553 * transport.
2554 */
2555 if (!trans && asoc) {
2556 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2557 transports) {
2558 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2559 hb_change, pmtud_change,
2560 sackdelay_change);
2561 }
2562 }
2563
2564 return 0;
2565 }
2566
2567 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2568 {
2569 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2570 }
2571
2572 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2573 {
2574 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2575 }
2576
2577 /*
2578 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2579 *
2580 * This option will effect the way delayed acks are performed. This
2581 * option allows you to get or set the delayed ack time, in
2582 * milliseconds. It also allows changing the delayed ack frequency.
2583 * Changing the frequency to 1 disables the delayed sack algorithm. If
2584 * the assoc_id is 0, then this sets or gets the endpoints default
2585 * values. If the assoc_id field is non-zero, then the set or get
2586 * effects the specified association for the one to many model (the
2587 * assoc_id field is ignored by the one to one model). Note that if
2588 * sack_delay or sack_freq are 0 when setting this option, then the
2589 * current values will remain unchanged.
2590 *
2591 * struct sctp_sack_info {
2592 * sctp_assoc_t sack_assoc_id;
2593 * uint32_t sack_delay;
2594 * uint32_t sack_freq;
2595 * };
2596 *
2597 * sack_assoc_id - This parameter, indicates which association the user
2598 * is performing an action upon. Note that if this field's value is
2599 * zero then the endpoints default value is changed (effecting future
2600 * associations only).
2601 *
2602 * sack_delay - This parameter contains the number of milliseconds that
2603 * the user is requesting the delayed ACK timer be set to. Note that
2604 * this value is defined in the standard to be between 200 and 500
2605 * milliseconds.
2606 *
2607 * sack_freq - This parameter contains the number of packets that must
2608 * be received before a sack is sent without waiting for the delay
2609 * timer to expire. The default value for this is 2, setting this
2610 * value to 1 will disable the delayed sack algorithm.
2611 */
2612
2613 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2614 char __user *optval, unsigned int optlen)
2615 {
2616 struct sctp_sack_info params;
2617 struct sctp_transport *trans = NULL;
2618 struct sctp_association *asoc = NULL;
2619 struct sctp_sock *sp = sctp_sk(sk);
2620
2621 if (optlen == sizeof(struct sctp_sack_info)) {
2622 if (copy_from_user(&params, optval, optlen))
2623 return -EFAULT;
2624
2625 if (params.sack_delay == 0 && params.sack_freq == 0)
2626 return 0;
2627 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2628 pr_warn_ratelimited(DEPRECATED
2629 "%s (pid %d) "
2630 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2631 "Use struct sctp_sack_info instead\n",
2632 current->comm, task_pid_nr(current));
2633 if (copy_from_user(&params, optval, optlen))
2634 return -EFAULT;
2635
2636 if (params.sack_delay == 0)
2637 params.sack_freq = 1;
2638 else
2639 params.sack_freq = 0;
2640 } else
2641 return -EINVAL;
2642
2643 /* Validate value parameter. */
2644 if (params.sack_delay > 500)
2645 return -EINVAL;
2646
2647 /* Get association, if sack_assoc_id != 0 and the socket is a one
2648 * to many style socket, and an association was not found, then
2649 * the id was invalid.
2650 */
2651 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2652 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2653 return -EINVAL;
2654
2655 if (params.sack_delay) {
2656 if (asoc) {
2657 asoc->sackdelay =
2658 msecs_to_jiffies(params.sack_delay);
2659 asoc->param_flags =
2660 sctp_spp_sackdelay_enable(asoc->param_flags);
2661 } else {
2662 sp->sackdelay = params.sack_delay;
2663 sp->param_flags =
2664 sctp_spp_sackdelay_enable(sp->param_flags);
2665 }
2666 }
2667
2668 if (params.sack_freq == 1) {
2669 if (asoc) {
2670 asoc->param_flags =
2671 sctp_spp_sackdelay_disable(asoc->param_flags);
2672 } else {
2673 sp->param_flags =
2674 sctp_spp_sackdelay_disable(sp->param_flags);
2675 }
2676 } else if (params.sack_freq > 1) {
2677 if (asoc) {
2678 asoc->sackfreq = params.sack_freq;
2679 asoc->param_flags =
2680 sctp_spp_sackdelay_enable(asoc->param_flags);
2681 } else {
2682 sp->sackfreq = params.sack_freq;
2683 sp->param_flags =
2684 sctp_spp_sackdelay_enable(sp->param_flags);
2685 }
2686 }
2687
2688 /* If change is for association, also apply to each transport. */
2689 if (asoc) {
2690 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2691 transports) {
2692 if (params.sack_delay) {
2693 trans->sackdelay =
2694 msecs_to_jiffies(params.sack_delay);
2695 trans->param_flags =
2696 sctp_spp_sackdelay_enable(trans->param_flags);
2697 }
2698 if (params.sack_freq == 1) {
2699 trans->param_flags =
2700 sctp_spp_sackdelay_disable(trans->param_flags);
2701 } else if (params.sack_freq > 1) {
2702 trans->sackfreq = params.sack_freq;
2703 trans->param_flags =
2704 sctp_spp_sackdelay_enable(trans->param_flags);
2705 }
2706 }
2707 }
2708
2709 return 0;
2710 }
2711
2712 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2713 *
2714 * Applications can specify protocol parameters for the default association
2715 * initialization. The option name argument to setsockopt() and getsockopt()
2716 * is SCTP_INITMSG.
2717 *
2718 * Setting initialization parameters is effective only on an unconnected
2719 * socket (for UDP-style sockets only future associations are effected
2720 * by the change). With TCP-style sockets, this option is inherited by
2721 * sockets derived from a listener socket.
2722 */
2723 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2724 {
2725 struct sctp_initmsg sinit;
2726 struct sctp_sock *sp = sctp_sk(sk);
2727
2728 if (optlen != sizeof(struct sctp_initmsg))
2729 return -EINVAL;
2730 if (copy_from_user(&sinit, optval, optlen))
2731 return -EFAULT;
2732
2733 if (sinit.sinit_num_ostreams)
2734 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2735 if (sinit.sinit_max_instreams)
2736 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2737 if (sinit.sinit_max_attempts)
2738 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2739 if (sinit.sinit_max_init_timeo)
2740 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2741
2742 return 0;
2743 }
2744
2745 /*
2746 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2747 *
2748 * Applications that wish to use the sendto() system call may wish to
2749 * specify a default set of parameters that would normally be supplied
2750 * through the inclusion of ancillary data. This socket option allows
2751 * such an application to set the default sctp_sndrcvinfo structure.
2752 * The application that wishes to use this socket option simply passes
2753 * in to this call the sctp_sndrcvinfo structure defined in Section
2754 * 5.2.2) The input parameters accepted by this call include
2755 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2756 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2757 * to this call if the caller is using the UDP model.
2758 */
2759 static int sctp_setsockopt_default_send_param(struct sock *sk,
2760 char __user *optval,
2761 unsigned int optlen)
2762 {
2763 struct sctp_sock *sp = sctp_sk(sk);
2764 struct sctp_association *asoc;
2765 struct sctp_sndrcvinfo info;
2766
2767 if (optlen != sizeof(info))
2768 return -EINVAL;
2769 if (copy_from_user(&info, optval, optlen))
2770 return -EFAULT;
2771 if (info.sinfo_flags &
2772 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2773 SCTP_ABORT | SCTP_EOF))
2774 return -EINVAL;
2775
2776 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2777 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2778 return -EINVAL;
2779 if (asoc) {
2780 asoc->default_stream = info.sinfo_stream;
2781 asoc->default_flags = info.sinfo_flags;
2782 asoc->default_ppid = info.sinfo_ppid;
2783 asoc->default_context = info.sinfo_context;
2784 asoc->default_timetolive = info.sinfo_timetolive;
2785 } else {
2786 sp->default_stream = info.sinfo_stream;
2787 sp->default_flags = info.sinfo_flags;
2788 sp->default_ppid = info.sinfo_ppid;
2789 sp->default_context = info.sinfo_context;
2790 sp->default_timetolive = info.sinfo_timetolive;
2791 }
2792
2793 return 0;
2794 }
2795
2796 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
2797 * (SCTP_DEFAULT_SNDINFO)
2798 */
2799 static int sctp_setsockopt_default_sndinfo(struct sock *sk,
2800 char __user *optval,
2801 unsigned int optlen)
2802 {
2803 struct sctp_sock *sp = sctp_sk(sk);
2804 struct sctp_association *asoc;
2805 struct sctp_sndinfo info;
2806
2807 if (optlen != sizeof(info))
2808 return -EINVAL;
2809 if (copy_from_user(&info, optval, optlen))
2810 return -EFAULT;
2811 if (info.snd_flags &
2812 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2813 SCTP_ABORT | SCTP_EOF))
2814 return -EINVAL;
2815
2816 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
2817 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
2818 return -EINVAL;
2819 if (asoc) {
2820 asoc->default_stream = info.snd_sid;
2821 asoc->default_flags = info.snd_flags;
2822 asoc->default_ppid = info.snd_ppid;
2823 asoc->default_context = info.snd_context;
2824 } else {
2825 sp->default_stream = info.snd_sid;
2826 sp->default_flags = info.snd_flags;
2827 sp->default_ppid = info.snd_ppid;
2828 sp->default_context = info.snd_context;
2829 }
2830
2831 return 0;
2832 }
2833
2834 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2835 *
2836 * Requests that the local SCTP stack use the enclosed peer address as
2837 * the association primary. The enclosed address must be one of the
2838 * association peer's addresses.
2839 */
2840 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2841 unsigned int optlen)
2842 {
2843 struct sctp_prim prim;
2844 struct sctp_transport *trans;
2845
2846 if (optlen != sizeof(struct sctp_prim))
2847 return -EINVAL;
2848
2849 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2850 return -EFAULT;
2851
2852 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2853 if (!trans)
2854 return -EINVAL;
2855
2856 sctp_assoc_set_primary(trans->asoc, trans);
2857
2858 return 0;
2859 }
2860
2861 /*
2862 * 7.1.5 SCTP_NODELAY
2863 *
2864 * Turn on/off any Nagle-like algorithm. This means that packets are
2865 * generally sent as soon as possible and no unnecessary delays are
2866 * introduced, at the cost of more packets in the network. Expects an
2867 * integer boolean flag.
2868 */
2869 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2870 unsigned int optlen)
2871 {
2872 int val;
2873
2874 if (optlen < sizeof(int))
2875 return -EINVAL;
2876 if (get_user(val, (int __user *)optval))
2877 return -EFAULT;
2878
2879 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2880 return 0;
2881 }
2882
2883 /*
2884 *
2885 * 7.1.1 SCTP_RTOINFO
2886 *
2887 * The protocol parameters used to initialize and bound retransmission
2888 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2889 * and modify these parameters.
2890 * All parameters are time values, in milliseconds. A value of 0, when
2891 * modifying the parameters, indicates that the current value should not
2892 * be changed.
2893 *
2894 */
2895 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2896 {
2897 struct sctp_rtoinfo rtoinfo;
2898 struct sctp_association *asoc;
2899 unsigned long rto_min, rto_max;
2900 struct sctp_sock *sp = sctp_sk(sk);
2901
2902 if (optlen != sizeof (struct sctp_rtoinfo))
2903 return -EINVAL;
2904
2905 if (copy_from_user(&rtoinfo, optval, optlen))
2906 return -EFAULT;
2907
2908 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2909
2910 /* Set the values to the specific association */
2911 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2912 return -EINVAL;
2913
2914 rto_max = rtoinfo.srto_max;
2915 rto_min = rtoinfo.srto_min;
2916
2917 if (rto_max)
2918 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
2919 else
2920 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
2921
2922 if (rto_min)
2923 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
2924 else
2925 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
2926
2927 if (rto_min > rto_max)
2928 return -EINVAL;
2929
2930 if (asoc) {
2931 if (rtoinfo.srto_initial != 0)
2932 asoc->rto_initial =
2933 msecs_to_jiffies(rtoinfo.srto_initial);
2934 asoc->rto_max = rto_max;
2935 asoc->rto_min = rto_min;
2936 } else {
2937 /* If there is no association or the association-id = 0
2938 * set the values to the endpoint.
2939 */
2940 if (rtoinfo.srto_initial != 0)
2941 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2942 sp->rtoinfo.srto_max = rto_max;
2943 sp->rtoinfo.srto_min = rto_min;
2944 }
2945
2946 return 0;
2947 }
2948
2949 /*
2950 *
2951 * 7.1.2 SCTP_ASSOCINFO
2952 *
2953 * This option is used to tune the maximum retransmission attempts
2954 * of the association.
2955 * Returns an error if the new association retransmission value is
2956 * greater than the sum of the retransmission value of the peer.
2957 * See [SCTP] for more information.
2958 *
2959 */
2960 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2961 {
2962
2963 struct sctp_assocparams assocparams;
2964 struct sctp_association *asoc;
2965
2966 if (optlen != sizeof(struct sctp_assocparams))
2967 return -EINVAL;
2968 if (copy_from_user(&assocparams, optval, optlen))
2969 return -EFAULT;
2970
2971 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2972
2973 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2974 return -EINVAL;
2975
2976 /* Set the values to the specific association */
2977 if (asoc) {
2978 if (assocparams.sasoc_asocmaxrxt != 0) {
2979 __u32 path_sum = 0;
2980 int paths = 0;
2981 struct sctp_transport *peer_addr;
2982
2983 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2984 transports) {
2985 path_sum += peer_addr->pathmaxrxt;
2986 paths++;
2987 }
2988
2989 /* Only validate asocmaxrxt if we have more than
2990 * one path/transport. We do this because path
2991 * retransmissions are only counted when we have more
2992 * then one path.
2993 */
2994 if (paths > 1 &&
2995 assocparams.sasoc_asocmaxrxt > path_sum)
2996 return -EINVAL;
2997
2998 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2999 }
3000
3001 if (assocparams.sasoc_cookie_life != 0)
3002 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
3003 } else {
3004 /* Set the values to the endpoint */
3005 struct sctp_sock *sp = sctp_sk(sk);
3006
3007 if (assocparams.sasoc_asocmaxrxt != 0)
3008 sp->assocparams.sasoc_asocmaxrxt =
3009 assocparams.sasoc_asocmaxrxt;
3010 if (assocparams.sasoc_cookie_life != 0)
3011 sp->assocparams.sasoc_cookie_life =
3012 assocparams.sasoc_cookie_life;
3013 }
3014 return 0;
3015 }
3016
3017 /*
3018 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3019 *
3020 * This socket option is a boolean flag which turns on or off mapped V4
3021 * addresses. If this option is turned on and the socket is type
3022 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3023 * If this option is turned off, then no mapping will be done of V4
3024 * addresses and a user will receive both PF_INET6 and PF_INET type
3025 * addresses on the socket.
3026 */
3027 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
3028 {
3029 int val;
3030 struct sctp_sock *sp = sctp_sk(sk);
3031
3032 if (optlen < sizeof(int))
3033 return -EINVAL;
3034 if (get_user(val, (int __user *)optval))
3035 return -EFAULT;
3036 if (val)
3037 sp->v4mapped = 1;
3038 else
3039 sp->v4mapped = 0;
3040
3041 return 0;
3042 }
3043
3044 /*
3045 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3046 * This option will get or set the maximum size to put in any outgoing
3047 * SCTP DATA chunk. If a message is larger than this size it will be
3048 * fragmented by SCTP into the specified size. Note that the underlying
3049 * SCTP implementation may fragment into smaller sized chunks when the
3050 * PMTU of the underlying association is smaller than the value set by
3051 * the user. The default value for this option is '0' which indicates
3052 * the user is NOT limiting fragmentation and only the PMTU will effect
3053 * SCTP's choice of DATA chunk size. Note also that values set larger
3054 * than the maximum size of an IP datagram will effectively let SCTP
3055 * control fragmentation (i.e. the same as setting this option to 0).
3056 *
3057 * The following structure is used to access and modify this parameter:
3058 *
3059 * struct sctp_assoc_value {
3060 * sctp_assoc_t assoc_id;
3061 * uint32_t assoc_value;
3062 * };
3063 *
3064 * assoc_id: This parameter is ignored for one-to-one style sockets.
3065 * For one-to-many style sockets this parameter indicates which
3066 * association the user is performing an action upon. Note that if
3067 * this field's value is zero then the endpoints default value is
3068 * changed (effecting future associations only).
3069 * assoc_value: This parameter specifies the maximum size in bytes.
3070 */
3071 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3072 {
3073 struct sctp_assoc_value params;
3074 struct sctp_association *asoc;
3075 struct sctp_sock *sp = sctp_sk(sk);
3076 int val;
3077
3078 if (optlen == sizeof(int)) {
3079 pr_warn_ratelimited(DEPRECATED
3080 "%s (pid %d) "
3081 "Use of int in maxseg socket option.\n"
3082 "Use struct sctp_assoc_value instead\n",
3083 current->comm, task_pid_nr(current));
3084 if (copy_from_user(&val, optval, optlen))
3085 return -EFAULT;
3086 params.assoc_id = 0;
3087 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3088 if (copy_from_user(&params, optval, optlen))
3089 return -EFAULT;
3090 val = params.assoc_value;
3091 } else
3092 return -EINVAL;
3093
3094 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3095 return -EINVAL;
3096
3097 asoc = sctp_id2assoc(sk, params.assoc_id);
3098 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3099 return -EINVAL;
3100
3101 if (asoc) {
3102 if (val == 0) {
3103 val = asoc->pathmtu;
3104 val -= sp->pf->af->net_header_len;
3105 val -= sizeof(struct sctphdr) +
3106 sizeof(struct sctp_data_chunk);
3107 }
3108 asoc->user_frag = val;
3109 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3110 } else {
3111 sp->user_frag = val;
3112 }
3113
3114 return 0;
3115 }
3116
3117
3118 /*
3119 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3120 *
3121 * Requests that the peer mark the enclosed address as the association
3122 * primary. The enclosed address must be one of the association's
3123 * locally bound addresses. The following structure is used to make a
3124 * set primary request:
3125 */
3126 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3127 unsigned int optlen)
3128 {
3129 struct net *net = sock_net(sk);
3130 struct sctp_sock *sp;
3131 struct sctp_association *asoc = NULL;
3132 struct sctp_setpeerprim prim;
3133 struct sctp_chunk *chunk;
3134 struct sctp_af *af;
3135 int err;
3136
3137 sp = sctp_sk(sk);
3138
3139 if (!net->sctp.addip_enable)
3140 return -EPERM;
3141
3142 if (optlen != sizeof(struct sctp_setpeerprim))
3143 return -EINVAL;
3144
3145 if (copy_from_user(&prim, optval, optlen))
3146 return -EFAULT;
3147
3148 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3149 if (!asoc)
3150 return -EINVAL;
3151
3152 if (!asoc->peer.asconf_capable)
3153 return -EPERM;
3154
3155 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3156 return -EPERM;
3157
3158 if (!sctp_state(asoc, ESTABLISHED))
3159 return -ENOTCONN;
3160
3161 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3162 if (!af)
3163 return -EINVAL;
3164
3165 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3166 return -EADDRNOTAVAIL;
3167
3168 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3169 return -EADDRNOTAVAIL;
3170
3171 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3172 chunk = sctp_make_asconf_set_prim(asoc,
3173 (union sctp_addr *)&prim.sspp_addr);
3174 if (!chunk)
3175 return -ENOMEM;
3176
3177 err = sctp_send_asconf(asoc, chunk);
3178
3179 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3180
3181 return err;
3182 }
3183
3184 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3185 unsigned int optlen)
3186 {
3187 struct sctp_setadaptation adaptation;
3188
3189 if (optlen != sizeof(struct sctp_setadaptation))
3190 return -EINVAL;
3191 if (copy_from_user(&adaptation, optval, optlen))
3192 return -EFAULT;
3193
3194 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3195
3196 return 0;
3197 }
3198
3199 /*
3200 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3201 *
3202 * The context field in the sctp_sndrcvinfo structure is normally only
3203 * used when a failed message is retrieved holding the value that was
3204 * sent down on the actual send call. This option allows the setting of
3205 * a default context on an association basis that will be received on
3206 * reading messages from the peer. This is especially helpful in the
3207 * one-2-many model for an application to keep some reference to an
3208 * internal state machine that is processing messages on the
3209 * association. Note that the setting of this value only effects
3210 * received messages from the peer and does not effect the value that is
3211 * saved with outbound messages.
3212 */
3213 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3214 unsigned int optlen)
3215 {
3216 struct sctp_assoc_value params;
3217 struct sctp_sock *sp;
3218 struct sctp_association *asoc;
3219
3220 if (optlen != sizeof(struct sctp_assoc_value))
3221 return -EINVAL;
3222 if (copy_from_user(&params, optval, optlen))
3223 return -EFAULT;
3224
3225 sp = sctp_sk(sk);
3226
3227 if (params.assoc_id != 0) {
3228 asoc = sctp_id2assoc(sk, params.assoc_id);
3229 if (!asoc)
3230 return -EINVAL;
3231 asoc->default_rcv_context = params.assoc_value;
3232 } else {
3233 sp->default_rcv_context = params.assoc_value;
3234 }
3235
3236 return 0;
3237 }
3238
3239 /*
3240 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3241 *
3242 * This options will at a minimum specify if the implementation is doing
3243 * fragmented interleave. Fragmented interleave, for a one to many
3244 * socket, is when subsequent calls to receive a message may return
3245 * parts of messages from different associations. Some implementations
3246 * may allow you to turn this value on or off. If so, when turned off,
3247 * no fragment interleave will occur (which will cause a head of line
3248 * blocking amongst multiple associations sharing the same one to many
3249 * socket). When this option is turned on, then each receive call may
3250 * come from a different association (thus the user must receive data
3251 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3252 * association each receive belongs to.
3253 *
3254 * This option takes a boolean value. A non-zero value indicates that
3255 * fragmented interleave is on. A value of zero indicates that
3256 * fragmented interleave is off.
3257 *
3258 * Note that it is important that an implementation that allows this
3259 * option to be turned on, have it off by default. Otherwise an unaware
3260 * application using the one to many model may become confused and act
3261 * incorrectly.
3262 */
3263 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3264 char __user *optval,
3265 unsigned int optlen)
3266 {
3267 int val;
3268
3269 if (optlen != sizeof(int))
3270 return -EINVAL;
3271 if (get_user(val, (int __user *)optval))
3272 return -EFAULT;
3273
3274 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3275
3276 return 0;
3277 }
3278
3279 /*
3280 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3281 * (SCTP_PARTIAL_DELIVERY_POINT)
3282 *
3283 * This option will set or get the SCTP partial delivery point. This
3284 * point is the size of a message where the partial delivery API will be
3285 * invoked to help free up rwnd space for the peer. Setting this to a
3286 * lower value will cause partial deliveries to happen more often. The
3287 * calls argument is an integer that sets or gets the partial delivery
3288 * point. Note also that the call will fail if the user attempts to set
3289 * this value larger than the socket receive buffer size.
3290 *
3291 * Note that any single message having a length smaller than or equal to
3292 * the SCTP partial delivery point will be delivered in one single read
3293 * call as long as the user provided buffer is large enough to hold the
3294 * message.
3295 */
3296 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3297 char __user *optval,
3298 unsigned int optlen)
3299 {
3300 u32 val;
3301
3302 if (optlen != sizeof(u32))
3303 return -EINVAL;
3304 if (get_user(val, (int __user *)optval))
3305 return -EFAULT;
3306
3307 /* Note: We double the receive buffer from what the user sets
3308 * it to be, also initial rwnd is based on rcvbuf/2.
3309 */
3310 if (val > (sk->sk_rcvbuf >> 1))
3311 return -EINVAL;
3312
3313 sctp_sk(sk)->pd_point = val;
3314
3315 return 0; /* is this the right error code? */
3316 }
3317
3318 /*
3319 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3320 *
3321 * This option will allow a user to change the maximum burst of packets
3322 * that can be emitted by this association. Note that the default value
3323 * is 4, and some implementations may restrict this setting so that it
3324 * can only be lowered.
3325 *
3326 * NOTE: This text doesn't seem right. Do this on a socket basis with
3327 * future associations inheriting the socket value.
3328 */
3329 static int sctp_setsockopt_maxburst(struct sock *sk,
3330 char __user *optval,
3331 unsigned int optlen)
3332 {
3333 struct sctp_assoc_value params;
3334 struct sctp_sock *sp;
3335 struct sctp_association *asoc;
3336 int val;
3337 int assoc_id = 0;
3338
3339 if (optlen == sizeof(int)) {
3340 pr_warn_ratelimited(DEPRECATED
3341 "%s (pid %d) "
3342 "Use of int in max_burst socket option deprecated.\n"
3343 "Use struct sctp_assoc_value instead\n",
3344 current->comm, task_pid_nr(current));
3345 if (copy_from_user(&val, optval, optlen))
3346 return -EFAULT;
3347 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3348 if (copy_from_user(&params, optval, optlen))
3349 return -EFAULT;
3350 val = params.assoc_value;
3351 assoc_id = params.assoc_id;
3352 } else
3353 return -EINVAL;
3354
3355 sp = sctp_sk(sk);
3356
3357 if (assoc_id != 0) {
3358 asoc = sctp_id2assoc(sk, assoc_id);
3359 if (!asoc)
3360 return -EINVAL;
3361 asoc->max_burst = val;
3362 } else
3363 sp->max_burst = val;
3364
3365 return 0;
3366 }
3367
3368 /*
3369 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3370 *
3371 * This set option adds a chunk type that the user is requesting to be
3372 * received only in an authenticated way. Changes to the list of chunks
3373 * will only effect future associations on the socket.
3374 */
3375 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3376 char __user *optval,
3377 unsigned int optlen)
3378 {
3379 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3380 struct sctp_authchunk val;
3381
3382 if (!ep->auth_enable)
3383 return -EACCES;
3384
3385 if (optlen != sizeof(struct sctp_authchunk))
3386 return -EINVAL;
3387 if (copy_from_user(&val, optval, optlen))
3388 return -EFAULT;
3389
3390 switch (val.sauth_chunk) {
3391 case SCTP_CID_INIT:
3392 case SCTP_CID_INIT_ACK:
3393 case SCTP_CID_SHUTDOWN_COMPLETE:
3394 case SCTP_CID_AUTH:
3395 return -EINVAL;
3396 }
3397
3398 /* add this chunk id to the endpoint */
3399 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3400 }
3401
3402 /*
3403 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3404 *
3405 * This option gets or sets the list of HMAC algorithms that the local
3406 * endpoint requires the peer to use.
3407 */
3408 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3409 char __user *optval,
3410 unsigned int optlen)
3411 {
3412 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3413 struct sctp_hmacalgo *hmacs;
3414 u32 idents;
3415 int err;
3416
3417 if (!ep->auth_enable)
3418 return -EACCES;
3419
3420 if (optlen < sizeof(struct sctp_hmacalgo))
3421 return -EINVAL;
3422
3423 hmacs = memdup_user(optval, optlen);
3424 if (IS_ERR(hmacs))
3425 return PTR_ERR(hmacs);
3426
3427 idents = hmacs->shmac_num_idents;
3428 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3429 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3430 err = -EINVAL;
3431 goto out;
3432 }
3433
3434 err = sctp_auth_ep_set_hmacs(ep, hmacs);
3435 out:
3436 kfree(hmacs);
3437 return err;
3438 }
3439
3440 /*
3441 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3442 *
3443 * This option will set a shared secret key which is used to build an
3444 * association shared key.
3445 */
3446 static int sctp_setsockopt_auth_key(struct sock *sk,
3447 char __user *optval,
3448 unsigned int optlen)
3449 {
3450 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3451 struct sctp_authkey *authkey;
3452 struct sctp_association *asoc;
3453 int ret;
3454
3455 if (!ep->auth_enable)
3456 return -EACCES;
3457
3458 if (optlen <= sizeof(struct sctp_authkey))
3459 return -EINVAL;
3460
3461 authkey = memdup_user(optval, optlen);
3462 if (IS_ERR(authkey))
3463 return PTR_ERR(authkey);
3464
3465 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3466 ret = -EINVAL;
3467 goto out;
3468 }
3469
3470 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3471 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3472 ret = -EINVAL;
3473 goto out;
3474 }
3475
3476 ret = sctp_auth_set_key(ep, asoc, authkey);
3477 out:
3478 kzfree(authkey);
3479 return ret;
3480 }
3481
3482 /*
3483 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3484 *
3485 * This option will get or set the active shared key to be used to build
3486 * the association shared key.
3487 */
3488 static int sctp_setsockopt_active_key(struct sock *sk,
3489 char __user *optval,
3490 unsigned int optlen)
3491 {
3492 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3493 struct sctp_authkeyid val;
3494 struct sctp_association *asoc;
3495
3496 if (!ep->auth_enable)
3497 return -EACCES;
3498
3499 if (optlen != sizeof(struct sctp_authkeyid))
3500 return -EINVAL;
3501 if (copy_from_user(&val, optval, optlen))
3502 return -EFAULT;
3503
3504 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3505 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3506 return -EINVAL;
3507
3508 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3509 }
3510
3511 /*
3512 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3513 *
3514 * This set option will delete a shared secret key from use.
3515 */
3516 static int sctp_setsockopt_del_key(struct sock *sk,
3517 char __user *optval,
3518 unsigned int optlen)
3519 {
3520 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3521 struct sctp_authkeyid val;
3522 struct sctp_association *asoc;
3523
3524 if (!ep->auth_enable)
3525 return -EACCES;
3526
3527 if (optlen != sizeof(struct sctp_authkeyid))
3528 return -EINVAL;
3529 if (copy_from_user(&val, optval, optlen))
3530 return -EFAULT;
3531
3532 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3533 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3534 return -EINVAL;
3535
3536 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3537
3538 }
3539
3540 /*
3541 * 8.1.23 SCTP_AUTO_ASCONF
3542 *
3543 * This option will enable or disable the use of the automatic generation of
3544 * ASCONF chunks to add and delete addresses to an existing association. Note
3545 * that this option has two caveats namely: a) it only affects sockets that
3546 * are bound to all addresses available to the SCTP stack, and b) the system
3547 * administrator may have an overriding control that turns the ASCONF feature
3548 * off no matter what setting the socket option may have.
3549 * This option expects an integer boolean flag, where a non-zero value turns on
3550 * the option, and a zero value turns off the option.
3551 * Note. In this implementation, socket operation overrides default parameter
3552 * being set by sysctl as well as FreeBSD implementation
3553 */
3554 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3555 unsigned int optlen)
3556 {
3557 int val;
3558 struct sctp_sock *sp = sctp_sk(sk);
3559
3560 if (optlen < sizeof(int))
3561 return -EINVAL;
3562 if (get_user(val, (int __user *)optval))
3563 return -EFAULT;
3564 if (!sctp_is_ep_boundall(sk) && val)
3565 return -EINVAL;
3566 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3567 return 0;
3568
3569 spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3570 if (val == 0 && sp->do_auto_asconf) {
3571 list_del(&sp->auto_asconf_list);
3572 sp->do_auto_asconf = 0;
3573 } else if (val && !sp->do_auto_asconf) {
3574 list_add_tail(&sp->auto_asconf_list,
3575 &sock_net(sk)->sctp.auto_asconf_splist);
3576 sp->do_auto_asconf = 1;
3577 }
3578 spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3579 return 0;
3580 }
3581
3582 /*
3583 * SCTP_PEER_ADDR_THLDS
3584 *
3585 * This option allows us to alter the partially failed threshold for one or all
3586 * transports in an association. See Section 6.1 of:
3587 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3588 */
3589 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3590 char __user *optval,
3591 unsigned int optlen)
3592 {
3593 struct sctp_paddrthlds val;
3594 struct sctp_transport *trans;
3595 struct sctp_association *asoc;
3596
3597 if (optlen < sizeof(struct sctp_paddrthlds))
3598 return -EINVAL;
3599 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3600 sizeof(struct sctp_paddrthlds)))
3601 return -EFAULT;
3602
3603
3604 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3605 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3606 if (!asoc)
3607 return -ENOENT;
3608 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3609 transports) {
3610 if (val.spt_pathmaxrxt)
3611 trans->pathmaxrxt = val.spt_pathmaxrxt;
3612 trans->pf_retrans = val.spt_pathpfthld;
3613 }
3614
3615 if (val.spt_pathmaxrxt)
3616 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3617 asoc->pf_retrans = val.spt_pathpfthld;
3618 } else {
3619 trans = sctp_addr_id2transport(sk, &val.spt_address,
3620 val.spt_assoc_id);
3621 if (!trans)
3622 return -ENOENT;
3623
3624 if (val.spt_pathmaxrxt)
3625 trans->pathmaxrxt = val.spt_pathmaxrxt;
3626 trans->pf_retrans = val.spt_pathpfthld;
3627 }
3628
3629 return 0;
3630 }
3631
3632 static int sctp_setsockopt_recvrcvinfo(struct sock *sk,
3633 char __user *optval,
3634 unsigned int optlen)
3635 {
3636 int val;
3637
3638 if (optlen < sizeof(int))
3639 return -EINVAL;
3640 if (get_user(val, (int __user *) optval))
3641 return -EFAULT;
3642
3643 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1;
3644
3645 return 0;
3646 }
3647
3648 static int sctp_setsockopt_recvnxtinfo(struct sock *sk,
3649 char __user *optval,
3650 unsigned int optlen)
3651 {
3652 int val;
3653
3654 if (optlen < sizeof(int))
3655 return -EINVAL;
3656 if (get_user(val, (int __user *) optval))
3657 return -EFAULT;
3658
3659 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1;
3660
3661 return 0;
3662 }
3663
3664 static int sctp_setsockopt_pr_supported(struct sock *sk,
3665 char __user *optval,
3666 unsigned int optlen)
3667 {
3668 struct sctp_assoc_value params;
3669 struct sctp_association *asoc;
3670 int retval = -EINVAL;
3671
3672 if (optlen != sizeof(params))
3673 goto out;
3674
3675 if (copy_from_user(&params, optval, optlen)) {
3676 retval = -EFAULT;
3677 goto out;
3678 }
3679
3680 asoc = sctp_id2assoc(sk, params.assoc_id);
3681 if (asoc) {
3682 asoc->prsctp_enable = !!params.assoc_value;
3683 } else if (!params.assoc_id) {
3684 struct sctp_sock *sp = sctp_sk(sk);
3685
3686 sp->ep->prsctp_enable = !!params.assoc_value;
3687 } else {
3688 goto out;
3689 }
3690
3691 retval = 0;
3692
3693 out:
3694 return retval;
3695 }
3696
3697 /* API 6.2 setsockopt(), getsockopt()
3698 *
3699 * Applications use setsockopt() and getsockopt() to set or retrieve
3700 * socket options. Socket options are used to change the default
3701 * behavior of sockets calls. They are described in Section 7.
3702 *
3703 * The syntax is:
3704 *
3705 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3706 * int __user *optlen);
3707 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3708 * int optlen);
3709 *
3710 * sd - the socket descript.
3711 * level - set to IPPROTO_SCTP for all SCTP options.
3712 * optname - the option name.
3713 * optval - the buffer to store the value of the option.
3714 * optlen - the size of the buffer.
3715 */
3716 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3717 char __user *optval, unsigned int optlen)
3718 {
3719 int retval = 0;
3720
3721 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3722
3723 /* I can hardly begin to describe how wrong this is. This is
3724 * so broken as to be worse than useless. The API draft
3725 * REALLY is NOT helpful here... I am not convinced that the
3726 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3727 * are at all well-founded.
3728 */
3729 if (level != SOL_SCTP) {
3730 struct sctp_af *af = sctp_sk(sk)->pf->af;
3731 retval = af->setsockopt(sk, level, optname, optval, optlen);
3732 goto out_nounlock;
3733 }
3734
3735 lock_sock(sk);
3736
3737 switch (optname) {
3738 case SCTP_SOCKOPT_BINDX_ADD:
3739 /* 'optlen' is the size of the addresses buffer. */
3740 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3741 optlen, SCTP_BINDX_ADD_ADDR);
3742 break;
3743
3744 case SCTP_SOCKOPT_BINDX_REM:
3745 /* 'optlen' is the size of the addresses buffer. */
3746 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3747 optlen, SCTP_BINDX_REM_ADDR);
3748 break;
3749
3750 case SCTP_SOCKOPT_CONNECTX_OLD:
3751 /* 'optlen' is the size of the addresses buffer. */
3752 retval = sctp_setsockopt_connectx_old(sk,
3753 (struct sockaddr __user *)optval,
3754 optlen);
3755 break;
3756
3757 case SCTP_SOCKOPT_CONNECTX:
3758 /* 'optlen' is the size of the addresses buffer. */
3759 retval = sctp_setsockopt_connectx(sk,
3760 (struct sockaddr __user *)optval,
3761 optlen);
3762 break;
3763
3764 case SCTP_DISABLE_FRAGMENTS:
3765 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3766 break;
3767
3768 case SCTP_EVENTS:
3769 retval = sctp_setsockopt_events(sk, optval, optlen);
3770 break;
3771
3772 case SCTP_AUTOCLOSE:
3773 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3774 break;
3775
3776 case SCTP_PEER_ADDR_PARAMS:
3777 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3778 break;
3779
3780 case SCTP_DELAYED_SACK:
3781 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3782 break;
3783 case SCTP_PARTIAL_DELIVERY_POINT:
3784 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3785 break;
3786
3787 case SCTP_INITMSG:
3788 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3789 break;
3790 case SCTP_DEFAULT_SEND_PARAM:
3791 retval = sctp_setsockopt_default_send_param(sk, optval,
3792 optlen);
3793 break;
3794 case SCTP_DEFAULT_SNDINFO:
3795 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen);
3796 break;
3797 case SCTP_PRIMARY_ADDR:
3798 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3799 break;
3800 case SCTP_SET_PEER_PRIMARY_ADDR:
3801 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3802 break;
3803 case SCTP_NODELAY:
3804 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3805 break;
3806 case SCTP_RTOINFO:
3807 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3808 break;
3809 case SCTP_ASSOCINFO:
3810 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3811 break;
3812 case SCTP_I_WANT_MAPPED_V4_ADDR:
3813 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3814 break;
3815 case SCTP_MAXSEG:
3816 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3817 break;
3818 case SCTP_ADAPTATION_LAYER:
3819 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3820 break;
3821 case SCTP_CONTEXT:
3822 retval = sctp_setsockopt_context(sk, optval, optlen);
3823 break;
3824 case SCTP_FRAGMENT_INTERLEAVE:
3825 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3826 break;
3827 case SCTP_MAX_BURST:
3828 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3829 break;
3830 case SCTP_AUTH_CHUNK:
3831 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3832 break;
3833 case SCTP_HMAC_IDENT:
3834 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3835 break;
3836 case SCTP_AUTH_KEY:
3837 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3838 break;
3839 case SCTP_AUTH_ACTIVE_KEY:
3840 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3841 break;
3842 case SCTP_AUTH_DELETE_KEY:
3843 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3844 break;
3845 case SCTP_AUTO_ASCONF:
3846 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3847 break;
3848 case SCTP_PEER_ADDR_THLDS:
3849 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3850 break;
3851 case SCTP_RECVRCVINFO:
3852 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen);
3853 break;
3854 case SCTP_RECVNXTINFO:
3855 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen);
3856 break;
3857 case SCTP_PR_SUPPORTED:
3858 retval = sctp_setsockopt_pr_supported(sk, optval, optlen);
3859 break;
3860 default:
3861 retval = -ENOPROTOOPT;
3862 break;
3863 }
3864
3865 release_sock(sk);
3866
3867 out_nounlock:
3868 return retval;
3869 }
3870
3871 /* API 3.1.6 connect() - UDP Style Syntax
3872 *
3873 * An application may use the connect() call in the UDP model to initiate an
3874 * association without sending data.
3875 *
3876 * The syntax is:
3877 *
3878 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3879 *
3880 * sd: the socket descriptor to have a new association added to.
3881 *
3882 * nam: the address structure (either struct sockaddr_in or struct
3883 * sockaddr_in6 defined in RFC2553 [7]).
3884 *
3885 * len: the size of the address.
3886 */
3887 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
3888 int addr_len)
3889 {
3890 int err = 0;
3891 struct sctp_af *af;
3892
3893 lock_sock(sk);
3894
3895 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
3896 addr, addr_len);
3897
3898 /* Validate addr_len before calling common connect/connectx routine. */
3899 af = sctp_get_af_specific(addr->sa_family);
3900 if (!af || addr_len < af->sockaddr_len) {
3901 err = -EINVAL;
3902 } else {
3903 /* Pass correct addr len to common routine (so it knows there
3904 * is only one address being passed.
3905 */
3906 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3907 }
3908
3909 release_sock(sk);
3910 return err;
3911 }
3912
3913 /* FIXME: Write comments. */
3914 static int sctp_disconnect(struct sock *sk, int flags)
3915 {
3916 return -EOPNOTSUPP; /* STUB */
3917 }
3918
3919 /* 4.1.4 accept() - TCP Style Syntax
3920 *
3921 * Applications use accept() call to remove an established SCTP
3922 * association from the accept queue of the endpoint. A new socket
3923 * descriptor will be returned from accept() to represent the newly
3924 * formed association.
3925 */
3926 static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3927 {
3928 struct sctp_sock *sp;
3929 struct sctp_endpoint *ep;
3930 struct sock *newsk = NULL;
3931 struct sctp_association *asoc;
3932 long timeo;
3933 int error = 0;
3934
3935 lock_sock(sk);
3936
3937 sp = sctp_sk(sk);
3938 ep = sp->ep;
3939
3940 if (!sctp_style(sk, TCP)) {
3941 error = -EOPNOTSUPP;
3942 goto out;
3943 }
3944
3945 if (!sctp_sstate(sk, LISTENING)) {
3946 error = -EINVAL;
3947 goto out;
3948 }
3949
3950 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3951
3952 error = sctp_wait_for_accept(sk, timeo);
3953 if (error)
3954 goto out;
3955
3956 /* We treat the list of associations on the endpoint as the accept
3957 * queue and pick the first association on the list.
3958 */
3959 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3960
3961 newsk = sp->pf->create_accept_sk(sk, asoc);
3962 if (!newsk) {
3963 error = -ENOMEM;
3964 goto out;
3965 }
3966
3967 /* Populate the fields of the newsk from the oldsk and migrate the
3968 * asoc to the newsk.
3969 */
3970 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3971
3972 out:
3973 release_sock(sk);
3974 *err = error;
3975 return newsk;
3976 }
3977
3978 /* The SCTP ioctl handler. */
3979 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3980 {
3981 int rc = -ENOTCONN;
3982
3983 lock_sock(sk);
3984
3985 /*
3986 * SEQPACKET-style sockets in LISTENING state are valid, for
3987 * SCTP, so only discard TCP-style sockets in LISTENING state.
3988 */
3989 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3990 goto out;
3991
3992 switch (cmd) {
3993 case SIOCINQ: {
3994 struct sk_buff *skb;
3995 unsigned int amount = 0;
3996
3997 skb = skb_peek(&sk->sk_receive_queue);
3998 if (skb != NULL) {
3999 /*
4000 * We will only return the amount of this packet since
4001 * that is all that will be read.
4002 */
4003 amount = skb->len;
4004 }
4005 rc = put_user(amount, (int __user *)arg);
4006 break;
4007 }
4008 default:
4009 rc = -ENOIOCTLCMD;
4010 break;
4011 }
4012 out:
4013 release_sock(sk);
4014 return rc;
4015 }
4016
4017 /* This is the function which gets called during socket creation to
4018 * initialized the SCTP-specific portion of the sock.
4019 * The sock structure should already be zero-filled memory.
4020 */
4021 static int sctp_init_sock(struct sock *sk)
4022 {
4023 struct net *net = sock_net(sk);
4024 struct sctp_sock *sp;
4025
4026 pr_debug("%s: sk:%p\n", __func__, sk);
4027
4028 sp = sctp_sk(sk);
4029
4030 /* Initialize the SCTP per socket area. */
4031 switch (sk->sk_type) {
4032 case SOCK_SEQPACKET:
4033 sp->type = SCTP_SOCKET_UDP;
4034 break;
4035 case SOCK_STREAM:
4036 sp->type = SCTP_SOCKET_TCP;
4037 break;
4038 default:
4039 return -ESOCKTNOSUPPORT;
4040 }
4041
4042 sk->sk_gso_type = SKB_GSO_SCTP;
4043
4044 /* Initialize default send parameters. These parameters can be
4045 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4046 */
4047 sp->default_stream = 0;
4048 sp->default_ppid = 0;
4049 sp->default_flags = 0;
4050 sp->default_context = 0;
4051 sp->default_timetolive = 0;
4052
4053 sp->default_rcv_context = 0;
4054 sp->max_burst = net->sctp.max_burst;
4055
4056 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
4057
4058 /* Initialize default setup parameters. These parameters
4059 * can be modified with the SCTP_INITMSG socket option or
4060 * overridden by the SCTP_INIT CMSG.
4061 */
4062 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
4063 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
4064 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
4065 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
4066
4067 /* Initialize default RTO related parameters. These parameters can
4068 * be modified for with the SCTP_RTOINFO socket option.
4069 */
4070 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
4071 sp->rtoinfo.srto_max = net->sctp.rto_max;
4072 sp->rtoinfo.srto_min = net->sctp.rto_min;
4073
4074 /* Initialize default association related parameters. These parameters
4075 * can be modified with the SCTP_ASSOCINFO socket option.
4076 */
4077 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
4078 sp->assocparams.sasoc_number_peer_destinations = 0;
4079 sp->assocparams.sasoc_peer_rwnd = 0;
4080 sp->assocparams.sasoc_local_rwnd = 0;
4081 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
4082
4083 /* Initialize default event subscriptions. By default, all the
4084 * options are off.
4085 */
4086 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
4087
4088 /* Default Peer Address Parameters. These defaults can
4089 * be modified via SCTP_PEER_ADDR_PARAMS
4090 */
4091 sp->hbinterval = net->sctp.hb_interval;
4092 sp->pathmaxrxt = net->sctp.max_retrans_path;
4093 sp->pathmtu = 0; /* allow default discovery */
4094 sp->sackdelay = net->sctp.sack_timeout;
4095 sp->sackfreq = 2;
4096 sp->param_flags = SPP_HB_ENABLE |
4097 SPP_PMTUD_ENABLE |
4098 SPP_SACKDELAY_ENABLE;
4099
4100 /* If enabled no SCTP message fragmentation will be performed.
4101 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4102 */
4103 sp->disable_fragments = 0;
4104
4105 /* Enable Nagle algorithm by default. */
4106 sp->nodelay = 0;
4107
4108 sp->recvrcvinfo = 0;
4109 sp->recvnxtinfo = 0;
4110
4111 /* Enable by default. */
4112 sp->v4mapped = 1;
4113
4114 /* Auto-close idle associations after the configured
4115 * number of seconds. A value of 0 disables this
4116 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4117 * for UDP-style sockets only.
4118 */
4119 sp->autoclose = 0;
4120
4121 /* User specified fragmentation limit. */
4122 sp->user_frag = 0;
4123
4124 sp->adaptation_ind = 0;
4125
4126 sp->pf = sctp_get_pf_specific(sk->sk_family);
4127
4128 /* Control variables for partial data delivery. */
4129 atomic_set(&sp->pd_mode, 0);
4130 skb_queue_head_init(&sp->pd_lobby);
4131 sp->frag_interleave = 0;
4132
4133 /* Create a per socket endpoint structure. Even if we
4134 * change the data structure relationships, this may still
4135 * be useful for storing pre-connect address information.
4136 */
4137 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4138 if (!sp->ep)
4139 return -ENOMEM;
4140
4141 sp->hmac = NULL;
4142
4143 sk->sk_destruct = sctp_destruct_sock;
4144
4145 SCTP_DBG_OBJCNT_INC(sock);
4146
4147 local_bh_disable();
4148 percpu_counter_inc(&sctp_sockets_allocated);
4149 sock_prot_inuse_add(net, sk->sk_prot, 1);
4150
4151 /* Nothing can fail after this block, otherwise
4152 * sctp_destroy_sock() will be called without addr_wq_lock held
4153 */
4154 if (net->sctp.default_auto_asconf) {
4155 spin_lock(&sock_net(sk)->sctp.addr_wq_lock);
4156 list_add_tail(&sp->auto_asconf_list,
4157 &net->sctp.auto_asconf_splist);
4158 sp->do_auto_asconf = 1;
4159 spin_unlock(&sock_net(sk)->sctp.addr_wq_lock);
4160 } else {
4161 sp->do_auto_asconf = 0;
4162 }
4163
4164 local_bh_enable();
4165
4166 return 0;
4167 }
4168
4169 /* Cleanup any SCTP per socket resources. Must be called with
4170 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4171 */
4172 static void sctp_destroy_sock(struct sock *sk)
4173 {
4174 struct sctp_sock *sp;
4175
4176 pr_debug("%s: sk:%p\n", __func__, sk);
4177
4178 /* Release our hold on the endpoint. */
4179 sp = sctp_sk(sk);
4180 /* This could happen during socket init, thus we bail out
4181 * early, since the rest of the below is not setup either.
4182 */
4183 if (sp->ep == NULL)
4184 return;
4185
4186 if (sp->do_auto_asconf) {
4187 sp->do_auto_asconf = 0;
4188 list_del(&sp->auto_asconf_list);
4189 }
4190 sctp_endpoint_free(sp->ep);
4191 local_bh_disable();
4192 percpu_counter_dec(&sctp_sockets_allocated);
4193 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4194 local_bh_enable();
4195 }
4196
4197 /* Triggered when there are no references on the socket anymore */
4198 static void sctp_destruct_sock(struct sock *sk)
4199 {
4200 struct sctp_sock *sp = sctp_sk(sk);
4201
4202 /* Free up the HMAC transform. */
4203 crypto_free_shash(sp->hmac);
4204
4205 inet_sock_destruct(sk);
4206 }
4207
4208 /* API 4.1.7 shutdown() - TCP Style Syntax
4209 * int shutdown(int socket, int how);
4210 *
4211 * sd - the socket descriptor of the association to be closed.
4212 * how - Specifies the type of shutdown. The values are
4213 * as follows:
4214 * SHUT_RD
4215 * Disables further receive operations. No SCTP
4216 * protocol action is taken.
4217 * SHUT_WR
4218 * Disables further send operations, and initiates
4219 * the SCTP shutdown sequence.
4220 * SHUT_RDWR
4221 * Disables further send and receive operations
4222 * and initiates the SCTP shutdown sequence.
4223 */
4224 static void sctp_shutdown(struct sock *sk, int how)
4225 {
4226 struct net *net = sock_net(sk);
4227 struct sctp_endpoint *ep;
4228 struct sctp_association *asoc;
4229
4230 if (!sctp_style(sk, TCP))
4231 return;
4232
4233 if (how & SEND_SHUTDOWN) {
4234 sk->sk_state = SCTP_SS_CLOSING;
4235 ep = sctp_sk(sk)->ep;
4236 if (!list_empty(&ep->asocs)) {
4237 asoc = list_entry(ep->asocs.next,
4238 struct sctp_association, asocs);
4239 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4240 }
4241 }
4242 }
4243
4244 int sctp_get_sctp_info(struct sock *sk, struct sctp_association *asoc,
4245 struct sctp_info *info)
4246 {
4247 struct sctp_transport *prim;
4248 struct list_head *pos;
4249 int mask;
4250
4251 memset(info, 0, sizeof(*info));
4252 if (!asoc) {
4253 struct sctp_sock *sp = sctp_sk(sk);
4254
4255 info->sctpi_s_autoclose = sp->autoclose;
4256 info->sctpi_s_adaptation_ind = sp->adaptation_ind;
4257 info->sctpi_s_pd_point = sp->pd_point;
4258 info->sctpi_s_nodelay = sp->nodelay;
4259 info->sctpi_s_disable_fragments = sp->disable_fragments;
4260 info->sctpi_s_v4mapped = sp->v4mapped;
4261 info->sctpi_s_frag_interleave = sp->frag_interleave;
4262 info->sctpi_s_type = sp->type;
4263
4264 return 0;
4265 }
4266
4267 info->sctpi_tag = asoc->c.my_vtag;
4268 info->sctpi_state = asoc->state;
4269 info->sctpi_rwnd = asoc->a_rwnd;
4270 info->sctpi_unackdata = asoc->unack_data;
4271 info->sctpi_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4272 info->sctpi_instrms = asoc->c.sinit_max_instreams;
4273 info->sctpi_outstrms = asoc->c.sinit_num_ostreams;
4274 list_for_each(pos, &asoc->base.inqueue.in_chunk_list)
4275 info->sctpi_inqueue++;
4276 list_for_each(pos, &asoc->outqueue.out_chunk_list)
4277 info->sctpi_outqueue++;
4278 info->sctpi_overall_error = asoc->overall_error_count;
4279 info->sctpi_max_burst = asoc->max_burst;
4280 info->sctpi_maxseg = asoc->frag_point;
4281 info->sctpi_peer_rwnd = asoc->peer.rwnd;
4282 info->sctpi_peer_tag = asoc->c.peer_vtag;
4283
4284 mask = asoc->peer.ecn_capable << 1;
4285 mask = (mask | asoc->peer.ipv4_address) << 1;
4286 mask = (mask | asoc->peer.ipv6_address) << 1;
4287 mask = (mask | asoc->peer.hostname_address) << 1;
4288 mask = (mask | asoc->peer.asconf_capable) << 1;
4289 mask = (mask | asoc->peer.prsctp_capable) << 1;
4290 mask = (mask | asoc->peer.auth_capable);
4291 info->sctpi_peer_capable = mask;
4292 mask = asoc->peer.sack_needed << 1;
4293 mask = (mask | asoc->peer.sack_generation) << 1;
4294 mask = (mask | asoc->peer.zero_window_announced);
4295 info->sctpi_peer_sack = mask;
4296
4297 info->sctpi_isacks = asoc->stats.isacks;
4298 info->sctpi_osacks = asoc->stats.osacks;
4299 info->sctpi_opackets = asoc->stats.opackets;
4300 info->sctpi_ipackets = asoc->stats.ipackets;
4301 info->sctpi_rtxchunks = asoc->stats.rtxchunks;
4302 info->sctpi_outofseqtsns = asoc->stats.outofseqtsns;
4303 info->sctpi_idupchunks = asoc->stats.idupchunks;
4304 info->sctpi_gapcnt = asoc->stats.gapcnt;
4305 info->sctpi_ouodchunks = asoc->stats.ouodchunks;
4306 info->sctpi_iuodchunks = asoc->stats.iuodchunks;
4307 info->sctpi_oodchunks = asoc->stats.oodchunks;
4308 info->sctpi_iodchunks = asoc->stats.iodchunks;
4309 info->sctpi_octrlchunks = asoc->stats.octrlchunks;
4310 info->sctpi_ictrlchunks = asoc->stats.ictrlchunks;
4311
4312 prim = asoc->peer.primary_path;
4313 memcpy(&info->sctpi_p_address, &prim->ipaddr,
4314 sizeof(struct sockaddr_storage));
4315 info->sctpi_p_state = prim->state;
4316 info->sctpi_p_cwnd = prim->cwnd;
4317 info->sctpi_p_srtt = prim->srtt;
4318 info->sctpi_p_rto = jiffies_to_msecs(prim->rto);
4319 info->sctpi_p_hbinterval = prim->hbinterval;
4320 info->sctpi_p_pathmaxrxt = prim->pathmaxrxt;
4321 info->sctpi_p_sackdelay = jiffies_to_msecs(prim->sackdelay);
4322 info->sctpi_p_ssthresh = prim->ssthresh;
4323 info->sctpi_p_partial_bytes_acked = prim->partial_bytes_acked;
4324 info->sctpi_p_flight_size = prim->flight_size;
4325 info->sctpi_p_error = prim->error_count;
4326
4327 return 0;
4328 }
4329 EXPORT_SYMBOL_GPL(sctp_get_sctp_info);
4330
4331 /* use callback to avoid exporting the core structure */
4332 int sctp_transport_walk_start(struct rhashtable_iter *iter)
4333 {
4334 int err;
4335
4336 err = rhashtable_walk_init(&sctp_transport_hashtable, iter,
4337 GFP_KERNEL);
4338 if (err)
4339 return err;
4340
4341 err = rhashtable_walk_start(iter);
4342 if (err && err != -EAGAIN) {
4343 rhashtable_walk_exit(iter);
4344 return err;
4345 }
4346
4347 return 0;
4348 }
4349
4350 void sctp_transport_walk_stop(struct rhashtable_iter *iter)
4351 {
4352 rhashtable_walk_stop(iter);
4353 rhashtable_walk_exit(iter);
4354 }
4355
4356 struct sctp_transport *sctp_transport_get_next(struct net *net,
4357 struct rhashtable_iter *iter)
4358 {
4359 struct sctp_transport *t;
4360
4361 t = rhashtable_walk_next(iter);
4362 for (; t; t = rhashtable_walk_next(iter)) {
4363 if (IS_ERR(t)) {
4364 if (PTR_ERR(t) == -EAGAIN)
4365 continue;
4366 break;
4367 }
4368
4369 if (net_eq(sock_net(t->asoc->base.sk), net) &&
4370 t->asoc->peer.primary_path == t)
4371 break;
4372 }
4373
4374 return t;
4375 }
4376
4377 struct sctp_transport *sctp_transport_get_idx(struct net *net,
4378 struct rhashtable_iter *iter,
4379 int pos)
4380 {
4381 void *obj = SEQ_START_TOKEN;
4382
4383 while (pos && (obj = sctp_transport_get_next(net, iter)) &&
4384 !IS_ERR(obj))
4385 pos--;
4386
4387 return obj;
4388 }
4389
4390 int sctp_for_each_endpoint(int (*cb)(struct sctp_endpoint *, void *),
4391 void *p) {
4392 int err = 0;
4393 int hash = 0;
4394 struct sctp_ep_common *epb;
4395 struct sctp_hashbucket *head;
4396
4397 for (head = sctp_ep_hashtable; hash < sctp_ep_hashsize;
4398 hash++, head++) {
4399 read_lock(&head->lock);
4400 sctp_for_each_hentry(epb, &head->chain) {
4401 err = cb(sctp_ep(epb), p);
4402 if (err)
4403 break;
4404 }
4405 read_unlock(&head->lock);
4406 }
4407
4408 return err;
4409 }
4410 EXPORT_SYMBOL_GPL(sctp_for_each_endpoint);
4411
4412 int sctp_transport_lookup_process(int (*cb)(struct sctp_transport *, void *),
4413 struct net *net,
4414 const union sctp_addr *laddr,
4415 const union sctp_addr *paddr, void *p)
4416 {
4417 struct sctp_transport *transport;
4418 int err = 0;
4419
4420 rcu_read_lock();
4421 transport = sctp_addrs_lookup_transport(net, laddr, paddr);
4422 if (!transport || !sctp_transport_hold(transport))
4423 goto out;
4424 err = cb(transport, p);
4425 sctp_transport_put(transport);
4426
4427 out:
4428 rcu_read_unlock();
4429 return err;
4430 }
4431 EXPORT_SYMBOL_GPL(sctp_transport_lookup_process);
4432
4433 int sctp_for_each_transport(int (*cb)(struct sctp_transport *, void *),
4434 struct net *net, int pos, void *p) {
4435 struct rhashtable_iter hti;
4436 void *obj;
4437 int err;
4438
4439 err = sctp_transport_walk_start(&hti);
4440 if (err)
4441 return err;
4442
4443 sctp_transport_get_idx(net, &hti, pos);
4444 obj = sctp_transport_get_next(net, &hti);
4445 for (; obj && !IS_ERR(obj); obj = sctp_transport_get_next(net, &hti)) {
4446 struct sctp_transport *transport = obj;
4447
4448 if (!sctp_transport_hold(transport))
4449 continue;
4450 err = cb(transport, p);
4451 sctp_transport_put(transport);
4452 if (err)
4453 break;
4454 }
4455 sctp_transport_walk_stop(&hti);
4456
4457 return err;
4458 }
4459 EXPORT_SYMBOL_GPL(sctp_for_each_transport);
4460
4461 /* 7.2.1 Association Status (SCTP_STATUS)
4462
4463 * Applications can retrieve current status information about an
4464 * association, including association state, peer receiver window size,
4465 * number of unacked data chunks, and number of data chunks pending
4466 * receipt. This information is read-only.
4467 */
4468 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4469 char __user *optval,
4470 int __user *optlen)
4471 {
4472 struct sctp_status status;
4473 struct sctp_association *asoc = NULL;
4474 struct sctp_transport *transport;
4475 sctp_assoc_t associd;
4476 int retval = 0;
4477
4478 if (len < sizeof(status)) {
4479 retval = -EINVAL;
4480 goto out;
4481 }
4482
4483 len = sizeof(status);
4484 if (copy_from_user(&status, optval, len)) {
4485 retval = -EFAULT;
4486 goto out;
4487 }
4488
4489 associd = status.sstat_assoc_id;
4490 asoc = sctp_id2assoc(sk, associd);
4491 if (!asoc) {
4492 retval = -EINVAL;
4493 goto out;
4494 }
4495
4496 transport = asoc->peer.primary_path;
4497
4498 status.sstat_assoc_id = sctp_assoc2id(asoc);
4499 status.sstat_state = sctp_assoc_to_state(asoc);
4500 status.sstat_rwnd = asoc->peer.rwnd;
4501 status.sstat_unackdata = asoc->unack_data;
4502
4503 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4504 status.sstat_instrms = asoc->c.sinit_max_instreams;
4505 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4506 status.sstat_fragmentation_point = asoc->frag_point;
4507 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4508 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4509 transport->af_specific->sockaddr_len);
4510 /* Map ipv4 address into v4-mapped-on-v6 address. */
4511 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
4512 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4513 status.sstat_primary.spinfo_state = transport->state;
4514 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4515 status.sstat_primary.spinfo_srtt = transport->srtt;
4516 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4517 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4518
4519 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4520 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4521
4522 if (put_user(len, optlen)) {
4523 retval = -EFAULT;
4524 goto out;
4525 }
4526
4527 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4528 __func__, len, status.sstat_state, status.sstat_rwnd,
4529 status.sstat_assoc_id);
4530
4531 if (copy_to_user(optval, &status, len)) {
4532 retval = -EFAULT;
4533 goto out;
4534 }
4535
4536 out:
4537 return retval;
4538 }
4539
4540
4541 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4542 *
4543 * Applications can retrieve information about a specific peer address
4544 * of an association, including its reachability state, congestion
4545 * window, and retransmission timer values. This information is
4546 * read-only.
4547 */
4548 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4549 char __user *optval,
4550 int __user *optlen)
4551 {
4552 struct sctp_paddrinfo pinfo;
4553 struct sctp_transport *transport;
4554 int retval = 0;
4555
4556 if (len < sizeof(pinfo)) {
4557 retval = -EINVAL;
4558 goto out;
4559 }
4560
4561 len = sizeof(pinfo);
4562 if (copy_from_user(&pinfo, optval, len)) {
4563 retval = -EFAULT;
4564 goto out;
4565 }
4566
4567 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4568 pinfo.spinfo_assoc_id);
4569 if (!transport)
4570 return -EINVAL;
4571
4572 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4573 pinfo.spinfo_state = transport->state;
4574 pinfo.spinfo_cwnd = transport->cwnd;
4575 pinfo.spinfo_srtt = transport->srtt;
4576 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4577 pinfo.spinfo_mtu = transport->pathmtu;
4578
4579 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4580 pinfo.spinfo_state = SCTP_ACTIVE;
4581
4582 if (put_user(len, optlen)) {
4583 retval = -EFAULT;
4584 goto out;
4585 }
4586
4587 if (copy_to_user(optval, &pinfo, len)) {
4588 retval = -EFAULT;
4589 goto out;
4590 }
4591
4592 out:
4593 return retval;
4594 }
4595
4596 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4597 *
4598 * This option is a on/off flag. If enabled no SCTP message
4599 * fragmentation will be performed. Instead if a message being sent
4600 * exceeds the current PMTU size, the message will NOT be sent and
4601 * instead a error will be indicated to the user.
4602 */
4603 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4604 char __user *optval, int __user *optlen)
4605 {
4606 int val;
4607
4608 if (len < sizeof(int))
4609 return -EINVAL;
4610
4611 len = sizeof(int);
4612 val = (sctp_sk(sk)->disable_fragments == 1);
4613 if (put_user(len, optlen))
4614 return -EFAULT;
4615 if (copy_to_user(optval, &val, len))
4616 return -EFAULT;
4617 return 0;
4618 }
4619
4620 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4621 *
4622 * This socket option is used to specify various notifications and
4623 * ancillary data the user wishes to receive.
4624 */
4625 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4626 int __user *optlen)
4627 {
4628 if (len <= 0)
4629 return -EINVAL;
4630 if (len > sizeof(struct sctp_event_subscribe))
4631 len = sizeof(struct sctp_event_subscribe);
4632 if (put_user(len, optlen))
4633 return -EFAULT;
4634 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4635 return -EFAULT;
4636 return 0;
4637 }
4638
4639 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4640 *
4641 * This socket option is applicable to the UDP-style socket only. When
4642 * set it will cause associations that are idle for more than the
4643 * specified number of seconds to automatically close. An association
4644 * being idle is defined an association that has NOT sent or received
4645 * user data. The special value of '0' indicates that no automatic
4646 * close of any associations should be performed. The option expects an
4647 * integer defining the number of seconds of idle time before an
4648 * association is closed.
4649 */
4650 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4651 {
4652 /* Applicable to UDP-style socket only */
4653 if (sctp_style(sk, TCP))
4654 return -EOPNOTSUPP;
4655 if (len < sizeof(int))
4656 return -EINVAL;
4657 len = sizeof(int);
4658 if (put_user(len, optlen))
4659 return -EFAULT;
4660 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4661 return -EFAULT;
4662 return 0;
4663 }
4664
4665 /* Helper routine to branch off an association to a new socket. */
4666 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4667 {
4668 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4669 struct sctp_sock *sp = sctp_sk(sk);
4670 struct socket *sock;
4671 int err = 0;
4672
4673 if (!asoc)
4674 return -EINVAL;
4675
4676 /* An association cannot be branched off from an already peeled-off
4677 * socket, nor is this supported for tcp style sockets.
4678 */
4679 if (!sctp_style(sk, UDP))
4680 return -EINVAL;
4681
4682 /* Create a new socket. */
4683 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4684 if (err < 0)
4685 return err;
4686
4687 sctp_copy_sock(sock->sk, sk, asoc);
4688
4689 /* Make peeled-off sockets more like 1-1 accepted sockets.
4690 * Set the daddr and initialize id to something more random
4691 */
4692 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
4693
4694 /* Populate the fields of the newsk from the oldsk and migrate the
4695 * asoc to the newsk.
4696 */
4697 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4698
4699 *sockp = sock;
4700
4701 return err;
4702 }
4703 EXPORT_SYMBOL(sctp_do_peeloff);
4704
4705 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4706 {
4707 sctp_peeloff_arg_t peeloff;
4708 struct socket *newsock;
4709 struct file *newfile;
4710 int retval = 0;
4711
4712 if (len < sizeof(sctp_peeloff_arg_t))
4713 return -EINVAL;
4714 len = sizeof(sctp_peeloff_arg_t);
4715 if (copy_from_user(&peeloff, optval, len))
4716 return -EFAULT;
4717
4718 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4719 if (retval < 0)
4720 goto out;
4721
4722 /* Map the socket to an unused fd that can be returned to the user. */
4723 retval = get_unused_fd_flags(0);
4724 if (retval < 0) {
4725 sock_release(newsock);
4726 goto out;
4727 }
4728
4729 newfile = sock_alloc_file(newsock, 0, NULL);
4730 if (IS_ERR(newfile)) {
4731 put_unused_fd(retval);
4732 sock_release(newsock);
4733 return PTR_ERR(newfile);
4734 }
4735
4736 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4737 retval);
4738
4739 /* Return the fd mapped to the new socket. */
4740 if (put_user(len, optlen)) {
4741 fput(newfile);
4742 put_unused_fd(retval);
4743 return -EFAULT;
4744 }
4745 peeloff.sd = retval;
4746 if (copy_to_user(optval, &peeloff, len)) {
4747 fput(newfile);
4748 put_unused_fd(retval);
4749 return -EFAULT;
4750 }
4751 fd_install(retval, newfile);
4752 out:
4753 return retval;
4754 }
4755
4756 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4757 *
4758 * Applications can enable or disable heartbeats for any peer address of
4759 * an association, modify an address's heartbeat interval, force a
4760 * heartbeat to be sent immediately, and adjust the address's maximum
4761 * number of retransmissions sent before an address is considered
4762 * unreachable. The following structure is used to access and modify an
4763 * address's parameters:
4764 *
4765 * struct sctp_paddrparams {
4766 * sctp_assoc_t spp_assoc_id;
4767 * struct sockaddr_storage spp_address;
4768 * uint32_t spp_hbinterval;
4769 * uint16_t spp_pathmaxrxt;
4770 * uint32_t spp_pathmtu;
4771 * uint32_t spp_sackdelay;
4772 * uint32_t spp_flags;
4773 * };
4774 *
4775 * spp_assoc_id - (one-to-many style socket) This is filled in the
4776 * application, and identifies the association for
4777 * this query.
4778 * spp_address - This specifies which address is of interest.
4779 * spp_hbinterval - This contains the value of the heartbeat interval,
4780 * in milliseconds. If a value of zero
4781 * is present in this field then no changes are to
4782 * be made to this parameter.
4783 * spp_pathmaxrxt - This contains the maximum number of
4784 * retransmissions before this address shall be
4785 * considered unreachable. If a value of zero
4786 * is present in this field then no changes are to
4787 * be made to this parameter.
4788 * spp_pathmtu - When Path MTU discovery is disabled the value
4789 * specified here will be the "fixed" path mtu.
4790 * Note that if the spp_address field is empty
4791 * then all associations on this address will
4792 * have this fixed path mtu set upon them.
4793 *
4794 * spp_sackdelay - When delayed sack is enabled, this value specifies
4795 * the number of milliseconds that sacks will be delayed
4796 * for. This value will apply to all addresses of an
4797 * association if the spp_address field is empty. Note
4798 * also, that if delayed sack is enabled and this
4799 * value is set to 0, no change is made to the last
4800 * recorded delayed sack timer value.
4801 *
4802 * spp_flags - These flags are used to control various features
4803 * on an association. The flag field may contain
4804 * zero or more of the following options.
4805 *
4806 * SPP_HB_ENABLE - Enable heartbeats on the
4807 * specified address. Note that if the address
4808 * field is empty all addresses for the association
4809 * have heartbeats enabled upon them.
4810 *
4811 * SPP_HB_DISABLE - Disable heartbeats on the
4812 * speicifed address. Note that if the address
4813 * field is empty all addresses for the association
4814 * will have their heartbeats disabled. Note also
4815 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4816 * mutually exclusive, only one of these two should
4817 * be specified. Enabling both fields will have
4818 * undetermined results.
4819 *
4820 * SPP_HB_DEMAND - Request a user initiated heartbeat
4821 * to be made immediately.
4822 *
4823 * SPP_PMTUD_ENABLE - This field will enable PMTU
4824 * discovery upon the specified address. Note that
4825 * if the address feild is empty then all addresses
4826 * on the association are effected.
4827 *
4828 * SPP_PMTUD_DISABLE - This field will disable PMTU
4829 * discovery upon the specified address. Note that
4830 * if the address feild is empty then all addresses
4831 * on the association are effected. Not also that
4832 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4833 * exclusive. Enabling both will have undetermined
4834 * results.
4835 *
4836 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4837 * on delayed sack. The time specified in spp_sackdelay
4838 * is used to specify the sack delay for this address. Note
4839 * that if spp_address is empty then all addresses will
4840 * enable delayed sack and take on the sack delay
4841 * value specified in spp_sackdelay.
4842 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4843 * off delayed sack. If the spp_address field is blank then
4844 * delayed sack is disabled for the entire association. Note
4845 * also that this field is mutually exclusive to
4846 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4847 * results.
4848 */
4849 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4850 char __user *optval, int __user *optlen)
4851 {
4852 struct sctp_paddrparams params;
4853 struct sctp_transport *trans = NULL;
4854 struct sctp_association *asoc = NULL;
4855 struct sctp_sock *sp = sctp_sk(sk);
4856
4857 if (len < sizeof(struct sctp_paddrparams))
4858 return -EINVAL;
4859 len = sizeof(struct sctp_paddrparams);
4860 if (copy_from_user(&params, optval, len))
4861 return -EFAULT;
4862
4863 /* If an address other than INADDR_ANY is specified, and
4864 * no transport is found, then the request is invalid.
4865 */
4866 if (!sctp_is_any(sk, (union sctp_addr *)&params.spp_address)) {
4867 trans = sctp_addr_id2transport(sk, &params.spp_address,
4868 params.spp_assoc_id);
4869 if (!trans) {
4870 pr_debug("%s: failed no transport\n", __func__);
4871 return -EINVAL;
4872 }
4873 }
4874
4875 /* Get association, if assoc_id != 0 and the socket is a one
4876 * to many style socket, and an association was not found, then
4877 * the id was invalid.
4878 */
4879 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4880 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4881 pr_debug("%s: failed no association\n", __func__);
4882 return -EINVAL;
4883 }
4884
4885 if (trans) {
4886 /* Fetch transport values. */
4887 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4888 params.spp_pathmtu = trans->pathmtu;
4889 params.spp_pathmaxrxt = trans->pathmaxrxt;
4890 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4891
4892 /*draft-11 doesn't say what to return in spp_flags*/
4893 params.spp_flags = trans->param_flags;
4894 } else if (asoc) {
4895 /* Fetch association values. */
4896 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4897 params.spp_pathmtu = asoc->pathmtu;
4898 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4899 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4900
4901 /*draft-11 doesn't say what to return in spp_flags*/
4902 params.spp_flags = asoc->param_flags;
4903 } else {
4904 /* Fetch socket values. */
4905 params.spp_hbinterval = sp->hbinterval;
4906 params.spp_pathmtu = sp->pathmtu;
4907 params.spp_sackdelay = sp->sackdelay;
4908 params.spp_pathmaxrxt = sp->pathmaxrxt;
4909
4910 /*draft-11 doesn't say what to return in spp_flags*/
4911 params.spp_flags = sp->param_flags;
4912 }
4913
4914 if (copy_to_user(optval, &params, len))
4915 return -EFAULT;
4916
4917 if (put_user(len, optlen))
4918 return -EFAULT;
4919
4920 return 0;
4921 }
4922
4923 /*
4924 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4925 *
4926 * This option will effect the way delayed acks are performed. This
4927 * option allows you to get or set the delayed ack time, in
4928 * milliseconds. It also allows changing the delayed ack frequency.
4929 * Changing the frequency to 1 disables the delayed sack algorithm. If
4930 * the assoc_id is 0, then this sets or gets the endpoints default
4931 * values. If the assoc_id field is non-zero, then the set or get
4932 * effects the specified association for the one to many model (the
4933 * assoc_id field is ignored by the one to one model). Note that if
4934 * sack_delay or sack_freq are 0 when setting this option, then the
4935 * current values will remain unchanged.
4936 *
4937 * struct sctp_sack_info {
4938 * sctp_assoc_t sack_assoc_id;
4939 * uint32_t sack_delay;
4940 * uint32_t sack_freq;
4941 * };
4942 *
4943 * sack_assoc_id - This parameter, indicates which association the user
4944 * is performing an action upon. Note that if this field's value is
4945 * zero then the endpoints default value is changed (effecting future
4946 * associations only).
4947 *
4948 * sack_delay - This parameter contains the number of milliseconds that
4949 * the user is requesting the delayed ACK timer be set to. Note that
4950 * this value is defined in the standard to be between 200 and 500
4951 * milliseconds.
4952 *
4953 * sack_freq - This parameter contains the number of packets that must
4954 * be received before a sack is sent without waiting for the delay
4955 * timer to expire. The default value for this is 2, setting this
4956 * value to 1 will disable the delayed sack algorithm.
4957 */
4958 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4959 char __user *optval,
4960 int __user *optlen)
4961 {
4962 struct sctp_sack_info params;
4963 struct sctp_association *asoc = NULL;
4964 struct sctp_sock *sp = sctp_sk(sk);
4965
4966 if (len >= sizeof(struct sctp_sack_info)) {
4967 len = sizeof(struct sctp_sack_info);
4968
4969 if (copy_from_user(&params, optval, len))
4970 return -EFAULT;
4971 } else if (len == sizeof(struct sctp_assoc_value)) {
4972 pr_warn_ratelimited(DEPRECATED
4973 "%s (pid %d) "
4974 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
4975 "Use struct sctp_sack_info instead\n",
4976 current->comm, task_pid_nr(current));
4977 if (copy_from_user(&params, optval, len))
4978 return -EFAULT;
4979 } else
4980 return -EINVAL;
4981
4982 /* Get association, if sack_assoc_id != 0 and the socket is a one
4983 * to many style socket, and an association was not found, then
4984 * the id was invalid.
4985 */
4986 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4987 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4988 return -EINVAL;
4989
4990 if (asoc) {
4991 /* Fetch association values. */
4992 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4993 params.sack_delay = jiffies_to_msecs(
4994 asoc->sackdelay);
4995 params.sack_freq = asoc->sackfreq;
4996
4997 } else {
4998 params.sack_delay = 0;
4999 params.sack_freq = 1;
5000 }
5001 } else {
5002 /* Fetch socket values. */
5003 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
5004 params.sack_delay = sp->sackdelay;
5005 params.sack_freq = sp->sackfreq;
5006 } else {
5007 params.sack_delay = 0;
5008 params.sack_freq = 1;
5009 }
5010 }
5011
5012 if (copy_to_user(optval, &params, len))
5013 return -EFAULT;
5014
5015 if (put_user(len, optlen))
5016 return -EFAULT;
5017
5018 return 0;
5019 }
5020
5021 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
5022 *
5023 * Applications can specify protocol parameters for the default association
5024 * initialization. The option name argument to setsockopt() and getsockopt()
5025 * is SCTP_INITMSG.
5026 *
5027 * Setting initialization parameters is effective only on an unconnected
5028 * socket (for UDP-style sockets only future associations are effected
5029 * by the change). With TCP-style sockets, this option is inherited by
5030 * sockets derived from a listener socket.
5031 */
5032 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
5033 {
5034 if (len < sizeof(struct sctp_initmsg))
5035 return -EINVAL;
5036 len = sizeof(struct sctp_initmsg);
5037 if (put_user(len, optlen))
5038 return -EFAULT;
5039 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
5040 return -EFAULT;
5041 return 0;
5042 }
5043
5044
5045 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
5046 char __user *optval, int __user *optlen)
5047 {
5048 struct sctp_association *asoc;
5049 int cnt = 0;
5050 struct sctp_getaddrs getaddrs;
5051 struct sctp_transport *from;
5052 void __user *to;
5053 union sctp_addr temp;
5054 struct sctp_sock *sp = sctp_sk(sk);
5055 int addrlen;
5056 size_t space_left;
5057 int bytes_copied;
5058
5059 if (len < sizeof(struct sctp_getaddrs))
5060 return -EINVAL;
5061
5062 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5063 return -EFAULT;
5064
5065 /* For UDP-style sockets, id specifies the association to query. */
5066 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5067 if (!asoc)
5068 return -EINVAL;
5069
5070 to = optval + offsetof(struct sctp_getaddrs, addrs);
5071 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5072
5073 list_for_each_entry(from, &asoc->peer.transport_addr_list,
5074 transports) {
5075 memcpy(&temp, &from->ipaddr, sizeof(temp));
5076 addrlen = sctp_get_pf_specific(sk->sk_family)
5077 ->addr_to_user(sp, &temp);
5078 if (space_left < addrlen)
5079 return -ENOMEM;
5080 if (copy_to_user(to, &temp, addrlen))
5081 return -EFAULT;
5082 to += addrlen;
5083 cnt++;
5084 space_left -= addrlen;
5085 }
5086
5087 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
5088 return -EFAULT;
5089 bytes_copied = ((char __user *)to) - optval;
5090 if (put_user(bytes_copied, optlen))
5091 return -EFAULT;
5092
5093 return 0;
5094 }
5095
5096 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
5097 size_t space_left, int *bytes_copied)
5098 {
5099 struct sctp_sockaddr_entry *addr;
5100 union sctp_addr temp;
5101 int cnt = 0;
5102 int addrlen;
5103 struct net *net = sock_net(sk);
5104
5105 rcu_read_lock();
5106 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
5107 if (!addr->valid)
5108 continue;
5109
5110 if ((PF_INET == sk->sk_family) &&
5111 (AF_INET6 == addr->a.sa.sa_family))
5112 continue;
5113 if ((PF_INET6 == sk->sk_family) &&
5114 inet_v6_ipv6only(sk) &&
5115 (AF_INET == addr->a.sa.sa_family))
5116 continue;
5117 memcpy(&temp, &addr->a, sizeof(temp));
5118 if (!temp.v4.sin_port)
5119 temp.v4.sin_port = htons(port);
5120
5121 addrlen = sctp_get_pf_specific(sk->sk_family)
5122 ->addr_to_user(sctp_sk(sk), &temp);
5123
5124 if (space_left < addrlen) {
5125 cnt = -ENOMEM;
5126 break;
5127 }
5128 memcpy(to, &temp, addrlen);
5129
5130 to += addrlen;
5131 cnt++;
5132 space_left -= addrlen;
5133 *bytes_copied += addrlen;
5134 }
5135 rcu_read_unlock();
5136
5137 return cnt;
5138 }
5139
5140
5141 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
5142 char __user *optval, int __user *optlen)
5143 {
5144 struct sctp_bind_addr *bp;
5145 struct sctp_association *asoc;
5146 int cnt = 0;
5147 struct sctp_getaddrs getaddrs;
5148 struct sctp_sockaddr_entry *addr;
5149 void __user *to;
5150 union sctp_addr temp;
5151 struct sctp_sock *sp = sctp_sk(sk);
5152 int addrlen;
5153 int err = 0;
5154 size_t space_left;
5155 int bytes_copied = 0;
5156 void *addrs;
5157 void *buf;
5158
5159 if (len < sizeof(struct sctp_getaddrs))
5160 return -EINVAL;
5161
5162 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5163 return -EFAULT;
5164
5165 /*
5166 * For UDP-style sockets, id specifies the association to query.
5167 * If the id field is set to the value '0' then the locally bound
5168 * addresses are returned without regard to any particular
5169 * association.
5170 */
5171 if (0 == getaddrs.assoc_id) {
5172 bp = &sctp_sk(sk)->ep->base.bind_addr;
5173 } else {
5174 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5175 if (!asoc)
5176 return -EINVAL;
5177 bp = &asoc->base.bind_addr;
5178 }
5179
5180 to = optval + offsetof(struct sctp_getaddrs, addrs);
5181 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5182
5183 addrs = kmalloc(space_left, GFP_USER | __GFP_NOWARN);
5184 if (!addrs)
5185 return -ENOMEM;
5186
5187 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
5188 * addresses from the global local address list.
5189 */
5190 if (sctp_list_single_entry(&bp->address_list)) {
5191 addr = list_entry(bp->address_list.next,
5192 struct sctp_sockaddr_entry, list);
5193 if (sctp_is_any(sk, &addr->a)) {
5194 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
5195 space_left, &bytes_copied);
5196 if (cnt < 0) {
5197 err = cnt;
5198 goto out;
5199 }
5200 goto copy_getaddrs;
5201 }
5202 }
5203
5204 buf = addrs;
5205 /* Protection on the bound address list is not needed since
5206 * in the socket option context we hold a socket lock and
5207 * thus the bound address list can't change.
5208 */
5209 list_for_each_entry(addr, &bp->address_list, list) {
5210 memcpy(&temp, &addr->a, sizeof(temp));
5211 addrlen = sctp_get_pf_specific(sk->sk_family)
5212 ->addr_to_user(sp, &temp);
5213 if (space_left < addrlen) {
5214 err = -ENOMEM; /*fixme: right error?*/
5215 goto out;
5216 }
5217 memcpy(buf, &temp, addrlen);
5218 buf += addrlen;
5219 bytes_copied += addrlen;
5220 cnt++;
5221 space_left -= addrlen;
5222 }
5223
5224 copy_getaddrs:
5225 if (copy_to_user(to, addrs, bytes_copied)) {
5226 err = -EFAULT;
5227 goto out;
5228 }
5229 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
5230 err = -EFAULT;
5231 goto out;
5232 }
5233 if (put_user(bytes_copied, optlen))
5234 err = -EFAULT;
5235 out:
5236 kfree(addrs);
5237 return err;
5238 }
5239
5240 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
5241 *
5242 * Requests that the local SCTP stack use the enclosed peer address as
5243 * the association primary. The enclosed address must be one of the
5244 * association peer's addresses.
5245 */
5246 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
5247 char __user *optval, int __user *optlen)
5248 {
5249 struct sctp_prim prim;
5250 struct sctp_association *asoc;
5251 struct sctp_sock *sp = sctp_sk(sk);
5252
5253 if (len < sizeof(struct sctp_prim))
5254 return -EINVAL;
5255
5256 len = sizeof(struct sctp_prim);
5257
5258 if (copy_from_user(&prim, optval, len))
5259 return -EFAULT;
5260
5261 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
5262 if (!asoc)
5263 return -EINVAL;
5264
5265 if (!asoc->peer.primary_path)
5266 return -ENOTCONN;
5267
5268 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
5269 asoc->peer.primary_path->af_specific->sockaddr_len);
5270
5271 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp,
5272 (union sctp_addr *)&prim.ssp_addr);
5273
5274 if (put_user(len, optlen))
5275 return -EFAULT;
5276 if (copy_to_user(optval, &prim, len))
5277 return -EFAULT;
5278
5279 return 0;
5280 }
5281
5282 /*
5283 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
5284 *
5285 * Requests that the local endpoint set the specified Adaptation Layer
5286 * Indication parameter for all future INIT and INIT-ACK exchanges.
5287 */
5288 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
5289 char __user *optval, int __user *optlen)
5290 {
5291 struct sctp_setadaptation adaptation;
5292
5293 if (len < sizeof(struct sctp_setadaptation))
5294 return -EINVAL;
5295
5296 len = sizeof(struct sctp_setadaptation);
5297
5298 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
5299
5300 if (put_user(len, optlen))
5301 return -EFAULT;
5302 if (copy_to_user(optval, &adaptation, len))
5303 return -EFAULT;
5304
5305 return 0;
5306 }
5307
5308 /*
5309 *
5310 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
5311 *
5312 * Applications that wish to use the sendto() system call may wish to
5313 * specify a default set of parameters that would normally be supplied
5314 * through the inclusion of ancillary data. This socket option allows
5315 * such an application to set the default sctp_sndrcvinfo structure.
5316
5317
5318 * The application that wishes to use this socket option simply passes
5319 * in to this call the sctp_sndrcvinfo structure defined in Section
5320 * 5.2.2) The input parameters accepted by this call include
5321 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
5322 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
5323 * to this call if the caller is using the UDP model.
5324 *
5325 * For getsockopt, it get the default sctp_sndrcvinfo structure.
5326 */
5327 static int sctp_getsockopt_default_send_param(struct sock *sk,
5328 int len, char __user *optval,
5329 int __user *optlen)
5330 {
5331 struct sctp_sock *sp = sctp_sk(sk);
5332 struct sctp_association *asoc;
5333 struct sctp_sndrcvinfo info;
5334
5335 if (len < sizeof(info))
5336 return -EINVAL;
5337
5338 len = sizeof(info);
5339
5340 if (copy_from_user(&info, optval, len))
5341 return -EFAULT;
5342
5343 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
5344 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
5345 return -EINVAL;
5346 if (asoc) {
5347 info.sinfo_stream = asoc->default_stream;
5348 info.sinfo_flags = asoc->default_flags;
5349 info.sinfo_ppid = asoc->default_ppid;
5350 info.sinfo_context = asoc->default_context;
5351 info.sinfo_timetolive = asoc->default_timetolive;
5352 } else {
5353 info.sinfo_stream = sp->default_stream;
5354 info.sinfo_flags = sp->default_flags;
5355 info.sinfo_ppid = sp->default_ppid;
5356 info.sinfo_context = sp->default_context;
5357 info.sinfo_timetolive = sp->default_timetolive;
5358 }
5359
5360 if (put_user(len, optlen))
5361 return -EFAULT;
5362 if (copy_to_user(optval, &info, len))
5363 return -EFAULT;
5364
5365 return 0;
5366 }
5367
5368 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
5369 * (SCTP_DEFAULT_SNDINFO)
5370 */
5371 static int sctp_getsockopt_default_sndinfo(struct sock *sk, int len,
5372 char __user *optval,
5373 int __user *optlen)
5374 {
5375 struct sctp_sock *sp = sctp_sk(sk);
5376 struct sctp_association *asoc;
5377 struct sctp_sndinfo info;
5378
5379 if (len < sizeof(info))
5380 return -EINVAL;
5381
5382 len = sizeof(info);
5383
5384 if (copy_from_user(&info, optval, len))
5385 return -EFAULT;
5386
5387 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
5388 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
5389 return -EINVAL;
5390 if (asoc) {
5391 info.snd_sid = asoc->default_stream;
5392 info.snd_flags = asoc->default_flags;
5393 info.snd_ppid = asoc->default_ppid;
5394 info.snd_context = asoc->default_context;
5395 } else {
5396 info.snd_sid = sp->default_stream;
5397 info.snd_flags = sp->default_flags;
5398 info.snd_ppid = sp->default_ppid;
5399 info.snd_context = sp->default_context;
5400 }
5401
5402 if (put_user(len, optlen))
5403 return -EFAULT;
5404 if (copy_to_user(optval, &info, len))
5405 return -EFAULT;
5406
5407 return 0;
5408 }
5409
5410 /*
5411 *
5412 * 7.1.5 SCTP_NODELAY
5413 *
5414 * Turn on/off any Nagle-like algorithm. This means that packets are
5415 * generally sent as soon as possible and no unnecessary delays are
5416 * introduced, at the cost of more packets in the network. Expects an
5417 * integer boolean flag.
5418 */
5419
5420 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
5421 char __user *optval, int __user *optlen)
5422 {
5423 int val;
5424
5425 if (len < sizeof(int))
5426 return -EINVAL;
5427
5428 len = sizeof(int);
5429 val = (sctp_sk(sk)->nodelay == 1);
5430 if (put_user(len, optlen))
5431 return -EFAULT;
5432 if (copy_to_user(optval, &val, len))
5433 return -EFAULT;
5434 return 0;
5435 }
5436
5437 /*
5438 *
5439 * 7.1.1 SCTP_RTOINFO
5440 *
5441 * The protocol parameters used to initialize and bound retransmission
5442 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5443 * and modify these parameters.
5444 * All parameters are time values, in milliseconds. A value of 0, when
5445 * modifying the parameters, indicates that the current value should not
5446 * be changed.
5447 *
5448 */
5449 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5450 char __user *optval,
5451 int __user *optlen) {
5452 struct sctp_rtoinfo rtoinfo;
5453 struct sctp_association *asoc;
5454
5455 if (len < sizeof (struct sctp_rtoinfo))
5456 return -EINVAL;
5457
5458 len = sizeof(struct sctp_rtoinfo);
5459
5460 if (copy_from_user(&rtoinfo, optval, len))
5461 return -EFAULT;
5462
5463 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5464
5465 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5466 return -EINVAL;
5467
5468 /* Values corresponding to the specific association. */
5469 if (asoc) {
5470 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5471 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5472 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5473 } else {
5474 /* Values corresponding to the endpoint. */
5475 struct sctp_sock *sp = sctp_sk(sk);
5476
5477 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5478 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5479 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5480 }
5481
5482 if (put_user(len, optlen))
5483 return -EFAULT;
5484
5485 if (copy_to_user(optval, &rtoinfo, len))
5486 return -EFAULT;
5487
5488 return 0;
5489 }
5490
5491 /*
5492 *
5493 * 7.1.2 SCTP_ASSOCINFO
5494 *
5495 * This option is used to tune the maximum retransmission attempts
5496 * of the association.
5497 * Returns an error if the new association retransmission value is
5498 * greater than the sum of the retransmission value of the peer.
5499 * See [SCTP] for more information.
5500 *
5501 */
5502 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5503 char __user *optval,
5504 int __user *optlen)
5505 {
5506
5507 struct sctp_assocparams assocparams;
5508 struct sctp_association *asoc;
5509 struct list_head *pos;
5510 int cnt = 0;
5511
5512 if (len < sizeof (struct sctp_assocparams))
5513 return -EINVAL;
5514
5515 len = sizeof(struct sctp_assocparams);
5516
5517 if (copy_from_user(&assocparams, optval, len))
5518 return -EFAULT;
5519
5520 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5521
5522 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5523 return -EINVAL;
5524
5525 /* Values correspoinding to the specific association */
5526 if (asoc) {
5527 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5528 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5529 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5530 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5531
5532 list_for_each(pos, &asoc->peer.transport_addr_list) {
5533 cnt++;
5534 }
5535
5536 assocparams.sasoc_number_peer_destinations = cnt;
5537 } else {
5538 /* Values corresponding to the endpoint */
5539 struct sctp_sock *sp = sctp_sk(sk);
5540
5541 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5542 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5543 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5544 assocparams.sasoc_cookie_life =
5545 sp->assocparams.sasoc_cookie_life;
5546 assocparams.sasoc_number_peer_destinations =
5547 sp->assocparams.
5548 sasoc_number_peer_destinations;
5549 }
5550
5551 if (put_user(len, optlen))
5552 return -EFAULT;
5553
5554 if (copy_to_user(optval, &assocparams, len))
5555 return -EFAULT;
5556
5557 return 0;
5558 }
5559
5560 /*
5561 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5562 *
5563 * This socket option is a boolean flag which turns on or off mapped V4
5564 * addresses. If this option is turned on and the socket is type
5565 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5566 * If this option is turned off, then no mapping will be done of V4
5567 * addresses and a user will receive both PF_INET6 and PF_INET type
5568 * addresses on the socket.
5569 */
5570 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5571 char __user *optval, int __user *optlen)
5572 {
5573 int val;
5574 struct sctp_sock *sp = sctp_sk(sk);
5575
5576 if (len < sizeof(int))
5577 return -EINVAL;
5578
5579 len = sizeof(int);
5580 val = sp->v4mapped;
5581 if (put_user(len, optlen))
5582 return -EFAULT;
5583 if (copy_to_user(optval, &val, len))
5584 return -EFAULT;
5585
5586 return 0;
5587 }
5588
5589 /*
5590 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5591 * (chapter and verse is quoted at sctp_setsockopt_context())
5592 */
5593 static int sctp_getsockopt_context(struct sock *sk, int len,
5594 char __user *optval, int __user *optlen)
5595 {
5596 struct sctp_assoc_value params;
5597 struct sctp_sock *sp;
5598 struct sctp_association *asoc;
5599
5600 if (len < sizeof(struct sctp_assoc_value))
5601 return -EINVAL;
5602
5603 len = sizeof(struct sctp_assoc_value);
5604
5605 if (copy_from_user(&params, optval, len))
5606 return -EFAULT;
5607
5608 sp = sctp_sk(sk);
5609
5610 if (params.assoc_id != 0) {
5611 asoc = sctp_id2assoc(sk, params.assoc_id);
5612 if (!asoc)
5613 return -EINVAL;
5614 params.assoc_value = asoc->default_rcv_context;
5615 } else {
5616 params.assoc_value = sp->default_rcv_context;
5617 }
5618
5619 if (put_user(len, optlen))
5620 return -EFAULT;
5621 if (copy_to_user(optval, &params, len))
5622 return -EFAULT;
5623
5624 return 0;
5625 }
5626
5627 /*
5628 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5629 * This option will get or set the maximum size to put in any outgoing
5630 * SCTP DATA chunk. If a message is larger than this size it will be
5631 * fragmented by SCTP into the specified size. Note that the underlying
5632 * SCTP implementation may fragment into smaller sized chunks when the
5633 * PMTU of the underlying association is smaller than the value set by
5634 * the user. The default value for this option is '0' which indicates
5635 * the user is NOT limiting fragmentation and only the PMTU will effect
5636 * SCTP's choice of DATA chunk size. Note also that values set larger
5637 * than the maximum size of an IP datagram will effectively let SCTP
5638 * control fragmentation (i.e. the same as setting this option to 0).
5639 *
5640 * The following structure is used to access and modify this parameter:
5641 *
5642 * struct sctp_assoc_value {
5643 * sctp_assoc_t assoc_id;
5644 * uint32_t assoc_value;
5645 * };
5646 *
5647 * assoc_id: This parameter is ignored for one-to-one style sockets.
5648 * For one-to-many style sockets this parameter indicates which
5649 * association the user is performing an action upon. Note that if
5650 * this field's value is zero then the endpoints default value is
5651 * changed (effecting future associations only).
5652 * assoc_value: This parameter specifies the maximum size in bytes.
5653 */
5654 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5655 char __user *optval, int __user *optlen)
5656 {
5657 struct sctp_assoc_value params;
5658 struct sctp_association *asoc;
5659
5660 if (len == sizeof(int)) {
5661 pr_warn_ratelimited(DEPRECATED
5662 "%s (pid %d) "
5663 "Use of int in maxseg socket option.\n"
5664 "Use struct sctp_assoc_value instead\n",
5665 current->comm, task_pid_nr(current));
5666 params.assoc_id = 0;
5667 } else if (len >= sizeof(struct sctp_assoc_value)) {
5668 len = sizeof(struct sctp_assoc_value);
5669 if (copy_from_user(&params, optval, sizeof(params)))
5670 return -EFAULT;
5671 } else
5672 return -EINVAL;
5673
5674 asoc = sctp_id2assoc(sk, params.assoc_id);
5675 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5676 return -EINVAL;
5677
5678 if (asoc)
5679 params.assoc_value = asoc->frag_point;
5680 else
5681 params.assoc_value = sctp_sk(sk)->user_frag;
5682
5683 if (put_user(len, optlen))
5684 return -EFAULT;
5685 if (len == sizeof(int)) {
5686 if (copy_to_user(optval, &params.assoc_value, len))
5687 return -EFAULT;
5688 } else {
5689 if (copy_to_user(optval, &params, len))
5690 return -EFAULT;
5691 }
5692
5693 return 0;
5694 }
5695
5696 /*
5697 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5698 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5699 */
5700 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5701 char __user *optval, int __user *optlen)
5702 {
5703 int val;
5704
5705 if (len < sizeof(int))
5706 return -EINVAL;
5707
5708 len = sizeof(int);
5709
5710 val = sctp_sk(sk)->frag_interleave;
5711 if (put_user(len, optlen))
5712 return -EFAULT;
5713 if (copy_to_user(optval, &val, len))
5714 return -EFAULT;
5715
5716 return 0;
5717 }
5718
5719 /*
5720 * 7.1.25. Set or Get the sctp partial delivery point
5721 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5722 */
5723 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5724 char __user *optval,
5725 int __user *optlen)
5726 {
5727 u32 val;
5728
5729 if (len < sizeof(u32))
5730 return -EINVAL;
5731
5732 len = sizeof(u32);
5733
5734 val = sctp_sk(sk)->pd_point;
5735 if (put_user(len, optlen))
5736 return -EFAULT;
5737 if (copy_to_user(optval, &val, len))
5738 return -EFAULT;
5739
5740 return 0;
5741 }
5742
5743 /*
5744 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5745 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5746 */
5747 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5748 char __user *optval,
5749 int __user *optlen)
5750 {
5751 struct sctp_assoc_value params;
5752 struct sctp_sock *sp;
5753 struct sctp_association *asoc;
5754
5755 if (len == sizeof(int)) {
5756 pr_warn_ratelimited(DEPRECATED
5757 "%s (pid %d) "
5758 "Use of int in max_burst socket option.\n"
5759 "Use struct sctp_assoc_value instead\n",
5760 current->comm, task_pid_nr(current));
5761 params.assoc_id = 0;
5762 } else if (len >= sizeof(struct sctp_assoc_value)) {
5763 len = sizeof(struct sctp_assoc_value);
5764 if (copy_from_user(&params, optval, len))
5765 return -EFAULT;
5766 } else
5767 return -EINVAL;
5768
5769 sp = sctp_sk(sk);
5770
5771 if (params.assoc_id != 0) {
5772 asoc = sctp_id2assoc(sk, params.assoc_id);
5773 if (!asoc)
5774 return -EINVAL;
5775 params.assoc_value = asoc->max_burst;
5776 } else
5777 params.assoc_value = sp->max_burst;
5778
5779 if (len == sizeof(int)) {
5780 if (copy_to_user(optval, &params.assoc_value, len))
5781 return -EFAULT;
5782 } else {
5783 if (copy_to_user(optval, &params, len))
5784 return -EFAULT;
5785 }
5786
5787 return 0;
5788
5789 }
5790
5791 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5792 char __user *optval, int __user *optlen)
5793 {
5794 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5795 struct sctp_hmacalgo __user *p = (void __user *)optval;
5796 struct sctp_hmac_algo_param *hmacs;
5797 __u16 data_len = 0;
5798 u32 num_idents;
5799 int i;
5800
5801 if (!ep->auth_enable)
5802 return -EACCES;
5803
5804 hmacs = ep->auth_hmacs_list;
5805 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5806
5807 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5808 return -EINVAL;
5809
5810 len = sizeof(struct sctp_hmacalgo) + data_len;
5811 num_idents = data_len / sizeof(u16);
5812
5813 if (put_user(len, optlen))
5814 return -EFAULT;
5815 if (put_user(num_idents, &p->shmac_num_idents))
5816 return -EFAULT;
5817 for (i = 0; i < num_idents; i++) {
5818 __u16 hmacid = ntohs(hmacs->hmac_ids[i]);
5819
5820 if (copy_to_user(&p->shmac_idents[i], &hmacid, sizeof(__u16)))
5821 return -EFAULT;
5822 }
5823 return 0;
5824 }
5825
5826 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5827 char __user *optval, int __user *optlen)
5828 {
5829 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5830 struct sctp_authkeyid val;
5831 struct sctp_association *asoc;
5832
5833 if (!ep->auth_enable)
5834 return -EACCES;
5835
5836 if (len < sizeof(struct sctp_authkeyid))
5837 return -EINVAL;
5838 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5839 return -EFAULT;
5840
5841 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5842 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5843 return -EINVAL;
5844
5845 if (asoc)
5846 val.scact_keynumber = asoc->active_key_id;
5847 else
5848 val.scact_keynumber = ep->active_key_id;
5849
5850 len = sizeof(struct sctp_authkeyid);
5851 if (put_user(len, optlen))
5852 return -EFAULT;
5853 if (copy_to_user(optval, &val, len))
5854 return -EFAULT;
5855
5856 return 0;
5857 }
5858
5859 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5860 char __user *optval, int __user *optlen)
5861 {
5862 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5863 struct sctp_authchunks __user *p = (void __user *)optval;
5864 struct sctp_authchunks val;
5865 struct sctp_association *asoc;
5866 struct sctp_chunks_param *ch;
5867 u32 num_chunks = 0;
5868 char __user *to;
5869
5870 if (!ep->auth_enable)
5871 return -EACCES;
5872
5873 if (len < sizeof(struct sctp_authchunks))
5874 return -EINVAL;
5875
5876 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5877 return -EFAULT;
5878
5879 to = p->gauth_chunks;
5880 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5881 if (!asoc)
5882 return -EINVAL;
5883
5884 ch = asoc->peer.peer_chunks;
5885 if (!ch)
5886 goto num;
5887
5888 /* See if the user provided enough room for all the data */
5889 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5890 if (len < num_chunks)
5891 return -EINVAL;
5892
5893 if (copy_to_user(to, ch->chunks, num_chunks))
5894 return -EFAULT;
5895 num:
5896 len = sizeof(struct sctp_authchunks) + num_chunks;
5897 if (put_user(len, optlen))
5898 return -EFAULT;
5899 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5900 return -EFAULT;
5901 return 0;
5902 }
5903
5904 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5905 char __user *optval, int __user *optlen)
5906 {
5907 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5908 struct sctp_authchunks __user *p = (void __user *)optval;
5909 struct sctp_authchunks val;
5910 struct sctp_association *asoc;
5911 struct sctp_chunks_param *ch;
5912 u32 num_chunks = 0;
5913 char __user *to;
5914
5915 if (!ep->auth_enable)
5916 return -EACCES;
5917
5918 if (len < sizeof(struct sctp_authchunks))
5919 return -EINVAL;
5920
5921 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5922 return -EFAULT;
5923
5924 to = p->gauth_chunks;
5925 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5926 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5927 return -EINVAL;
5928
5929 if (asoc)
5930 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
5931 else
5932 ch = ep->auth_chunk_list;
5933
5934 if (!ch)
5935 goto num;
5936
5937 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5938 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5939 return -EINVAL;
5940
5941 if (copy_to_user(to, ch->chunks, num_chunks))
5942 return -EFAULT;
5943 num:
5944 len = sizeof(struct sctp_authchunks) + num_chunks;
5945 if (put_user(len, optlen))
5946 return -EFAULT;
5947 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5948 return -EFAULT;
5949
5950 return 0;
5951 }
5952
5953 /*
5954 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5955 * This option gets the current number of associations that are attached
5956 * to a one-to-many style socket. The option value is an uint32_t.
5957 */
5958 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5959 char __user *optval, int __user *optlen)
5960 {
5961 struct sctp_sock *sp = sctp_sk(sk);
5962 struct sctp_association *asoc;
5963 u32 val = 0;
5964
5965 if (sctp_style(sk, TCP))
5966 return -EOPNOTSUPP;
5967
5968 if (len < sizeof(u32))
5969 return -EINVAL;
5970
5971 len = sizeof(u32);
5972
5973 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5974 val++;
5975 }
5976
5977 if (put_user(len, optlen))
5978 return -EFAULT;
5979 if (copy_to_user(optval, &val, len))
5980 return -EFAULT;
5981
5982 return 0;
5983 }
5984
5985 /*
5986 * 8.1.23 SCTP_AUTO_ASCONF
5987 * See the corresponding setsockopt entry as description
5988 */
5989 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5990 char __user *optval, int __user *optlen)
5991 {
5992 int val = 0;
5993
5994 if (len < sizeof(int))
5995 return -EINVAL;
5996
5997 len = sizeof(int);
5998 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5999 val = 1;
6000 if (put_user(len, optlen))
6001 return -EFAULT;
6002 if (copy_to_user(optval, &val, len))
6003 return -EFAULT;
6004 return 0;
6005 }
6006
6007 /*
6008 * 8.2.6. Get the Current Identifiers of Associations
6009 * (SCTP_GET_ASSOC_ID_LIST)
6010 *
6011 * This option gets the current list of SCTP association identifiers of
6012 * the SCTP associations handled by a one-to-many style socket.
6013 */
6014 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
6015 char __user *optval, int __user *optlen)
6016 {
6017 struct sctp_sock *sp = sctp_sk(sk);
6018 struct sctp_association *asoc;
6019 struct sctp_assoc_ids *ids;
6020 u32 num = 0;
6021
6022 if (sctp_style(sk, TCP))
6023 return -EOPNOTSUPP;
6024
6025 if (len < sizeof(struct sctp_assoc_ids))
6026 return -EINVAL;
6027
6028 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6029 num++;
6030 }
6031
6032 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
6033 return -EINVAL;
6034
6035 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
6036
6037 ids = kmalloc(len, GFP_USER | __GFP_NOWARN);
6038 if (unlikely(!ids))
6039 return -ENOMEM;
6040
6041 ids->gaids_number_of_ids = num;
6042 num = 0;
6043 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6044 ids->gaids_assoc_id[num++] = asoc->assoc_id;
6045 }
6046
6047 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
6048 kfree(ids);
6049 return -EFAULT;
6050 }
6051
6052 kfree(ids);
6053 return 0;
6054 }
6055
6056 /*
6057 * SCTP_PEER_ADDR_THLDS
6058 *
6059 * This option allows us to fetch the partially failed threshold for one or all
6060 * transports in an association. See Section 6.1 of:
6061 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
6062 */
6063 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
6064 char __user *optval,
6065 int len,
6066 int __user *optlen)
6067 {
6068 struct sctp_paddrthlds val;
6069 struct sctp_transport *trans;
6070 struct sctp_association *asoc;
6071
6072 if (len < sizeof(struct sctp_paddrthlds))
6073 return -EINVAL;
6074 len = sizeof(struct sctp_paddrthlds);
6075 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
6076 return -EFAULT;
6077
6078 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
6079 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
6080 if (!asoc)
6081 return -ENOENT;
6082
6083 val.spt_pathpfthld = asoc->pf_retrans;
6084 val.spt_pathmaxrxt = asoc->pathmaxrxt;
6085 } else {
6086 trans = sctp_addr_id2transport(sk, &val.spt_address,
6087 val.spt_assoc_id);
6088 if (!trans)
6089 return -ENOENT;
6090
6091 val.spt_pathmaxrxt = trans->pathmaxrxt;
6092 val.spt_pathpfthld = trans->pf_retrans;
6093 }
6094
6095 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
6096 return -EFAULT;
6097
6098 return 0;
6099 }
6100
6101 /*
6102 * SCTP_GET_ASSOC_STATS
6103 *
6104 * This option retrieves local per endpoint statistics. It is modeled
6105 * after OpenSolaris' implementation
6106 */
6107 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
6108 char __user *optval,
6109 int __user *optlen)
6110 {
6111 struct sctp_assoc_stats sas;
6112 struct sctp_association *asoc = NULL;
6113
6114 /* User must provide at least the assoc id */
6115 if (len < sizeof(sctp_assoc_t))
6116 return -EINVAL;
6117
6118 /* Allow the struct to grow and fill in as much as possible */
6119 len = min_t(size_t, len, sizeof(sas));
6120
6121 if (copy_from_user(&sas, optval, len))
6122 return -EFAULT;
6123
6124 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
6125 if (!asoc)
6126 return -EINVAL;
6127
6128 sas.sas_rtxchunks = asoc->stats.rtxchunks;
6129 sas.sas_gapcnt = asoc->stats.gapcnt;
6130 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
6131 sas.sas_osacks = asoc->stats.osacks;
6132 sas.sas_isacks = asoc->stats.isacks;
6133 sas.sas_octrlchunks = asoc->stats.octrlchunks;
6134 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
6135 sas.sas_oodchunks = asoc->stats.oodchunks;
6136 sas.sas_iodchunks = asoc->stats.iodchunks;
6137 sas.sas_ouodchunks = asoc->stats.ouodchunks;
6138 sas.sas_iuodchunks = asoc->stats.iuodchunks;
6139 sas.sas_idupchunks = asoc->stats.idupchunks;
6140 sas.sas_opackets = asoc->stats.opackets;
6141 sas.sas_ipackets = asoc->stats.ipackets;
6142
6143 /* New high max rto observed, will return 0 if not a single
6144 * RTO update took place. obs_rto_ipaddr will be bogus
6145 * in such a case
6146 */
6147 sas.sas_maxrto = asoc->stats.max_obs_rto;
6148 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
6149 sizeof(struct sockaddr_storage));
6150
6151 /* Mark beginning of a new observation period */
6152 asoc->stats.max_obs_rto = asoc->rto_min;
6153
6154 if (put_user(len, optlen))
6155 return -EFAULT;
6156
6157 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
6158
6159 if (copy_to_user(optval, &sas, len))
6160 return -EFAULT;
6161
6162 return 0;
6163 }
6164
6165 static int sctp_getsockopt_recvrcvinfo(struct sock *sk, int len,
6166 char __user *optval,
6167 int __user *optlen)
6168 {
6169 int val = 0;
6170
6171 if (len < sizeof(int))
6172 return -EINVAL;
6173
6174 len = sizeof(int);
6175 if (sctp_sk(sk)->recvrcvinfo)
6176 val = 1;
6177 if (put_user(len, optlen))
6178 return -EFAULT;
6179 if (copy_to_user(optval, &val, len))
6180 return -EFAULT;
6181
6182 return 0;
6183 }
6184
6185 static int sctp_getsockopt_recvnxtinfo(struct sock *sk, int len,
6186 char __user *optval,
6187 int __user *optlen)
6188 {
6189 int val = 0;
6190
6191 if (len < sizeof(int))
6192 return -EINVAL;
6193
6194 len = sizeof(int);
6195 if (sctp_sk(sk)->recvnxtinfo)
6196 val = 1;
6197 if (put_user(len, optlen))
6198 return -EFAULT;
6199 if (copy_to_user(optval, &val, len))
6200 return -EFAULT;
6201
6202 return 0;
6203 }
6204
6205 static int sctp_getsockopt_pr_supported(struct sock *sk, int len,
6206 char __user *optval,
6207 int __user *optlen)
6208 {
6209 struct sctp_assoc_value params;
6210 struct sctp_association *asoc;
6211 int retval = -EFAULT;
6212
6213 if (len < sizeof(params)) {
6214 retval = -EINVAL;
6215 goto out;
6216 }
6217
6218 len = sizeof(params);
6219 if (copy_from_user(&params, optval, len))
6220 goto out;
6221
6222 asoc = sctp_id2assoc(sk, params.assoc_id);
6223 if (asoc) {
6224 params.assoc_value = asoc->prsctp_enable;
6225 } else if (!params.assoc_id) {
6226 struct sctp_sock *sp = sctp_sk(sk);
6227
6228 params.assoc_value = sp->ep->prsctp_enable;
6229 } else {
6230 retval = -EINVAL;
6231 goto out;
6232 }
6233
6234 if (put_user(len, optlen))
6235 goto out;
6236
6237 if (copy_to_user(optval, &params, len))
6238 goto out;
6239
6240 retval = 0;
6241
6242 out:
6243 return retval;
6244 }
6245
6246 static int sctp_getsockopt(struct sock *sk, int level, int optname,
6247 char __user *optval, int __user *optlen)
6248 {
6249 int retval = 0;
6250 int len;
6251
6252 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
6253
6254 /* I can hardly begin to describe how wrong this is. This is
6255 * so broken as to be worse than useless. The API draft
6256 * REALLY is NOT helpful here... I am not convinced that the
6257 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
6258 * are at all well-founded.
6259 */
6260 if (level != SOL_SCTP) {
6261 struct sctp_af *af = sctp_sk(sk)->pf->af;
6262
6263 retval = af->getsockopt(sk, level, optname, optval, optlen);
6264 return retval;
6265 }
6266
6267 if (get_user(len, optlen))
6268 return -EFAULT;
6269
6270 lock_sock(sk);
6271
6272 switch (optname) {
6273 case SCTP_STATUS:
6274 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
6275 break;
6276 case SCTP_DISABLE_FRAGMENTS:
6277 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
6278 optlen);
6279 break;
6280 case SCTP_EVENTS:
6281 retval = sctp_getsockopt_events(sk, len, optval, optlen);
6282 break;
6283 case SCTP_AUTOCLOSE:
6284 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
6285 break;
6286 case SCTP_SOCKOPT_PEELOFF:
6287 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
6288 break;
6289 case SCTP_PEER_ADDR_PARAMS:
6290 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
6291 optlen);
6292 break;
6293 case SCTP_DELAYED_SACK:
6294 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
6295 optlen);
6296 break;
6297 case SCTP_INITMSG:
6298 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
6299 break;
6300 case SCTP_GET_PEER_ADDRS:
6301 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
6302 optlen);
6303 break;
6304 case SCTP_GET_LOCAL_ADDRS:
6305 retval = sctp_getsockopt_local_addrs(sk, len, optval,
6306 optlen);
6307 break;
6308 case SCTP_SOCKOPT_CONNECTX3:
6309 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
6310 break;
6311 case SCTP_DEFAULT_SEND_PARAM:
6312 retval = sctp_getsockopt_default_send_param(sk, len,
6313 optval, optlen);
6314 break;
6315 case SCTP_DEFAULT_SNDINFO:
6316 retval = sctp_getsockopt_default_sndinfo(sk, len,
6317 optval, optlen);
6318 break;
6319 case SCTP_PRIMARY_ADDR:
6320 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
6321 break;
6322 case SCTP_NODELAY:
6323 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
6324 break;
6325 case SCTP_RTOINFO:
6326 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
6327 break;
6328 case SCTP_ASSOCINFO:
6329 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
6330 break;
6331 case SCTP_I_WANT_MAPPED_V4_ADDR:
6332 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
6333 break;
6334 case SCTP_MAXSEG:
6335 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
6336 break;
6337 case SCTP_GET_PEER_ADDR_INFO:
6338 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
6339 optlen);
6340 break;
6341 case SCTP_ADAPTATION_LAYER:
6342 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
6343 optlen);
6344 break;
6345 case SCTP_CONTEXT:
6346 retval = sctp_getsockopt_context(sk, len, optval, optlen);
6347 break;
6348 case SCTP_FRAGMENT_INTERLEAVE:
6349 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
6350 optlen);
6351 break;
6352 case SCTP_PARTIAL_DELIVERY_POINT:
6353 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
6354 optlen);
6355 break;
6356 case SCTP_MAX_BURST:
6357 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
6358 break;
6359 case SCTP_AUTH_KEY:
6360 case SCTP_AUTH_CHUNK:
6361 case SCTP_AUTH_DELETE_KEY:
6362 retval = -EOPNOTSUPP;
6363 break;
6364 case SCTP_HMAC_IDENT:
6365 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
6366 break;
6367 case SCTP_AUTH_ACTIVE_KEY:
6368 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
6369 break;
6370 case SCTP_PEER_AUTH_CHUNKS:
6371 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
6372 optlen);
6373 break;
6374 case SCTP_LOCAL_AUTH_CHUNKS:
6375 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
6376 optlen);
6377 break;
6378 case SCTP_GET_ASSOC_NUMBER:
6379 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
6380 break;
6381 case SCTP_GET_ASSOC_ID_LIST:
6382 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
6383 break;
6384 case SCTP_AUTO_ASCONF:
6385 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
6386 break;
6387 case SCTP_PEER_ADDR_THLDS:
6388 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
6389 break;
6390 case SCTP_GET_ASSOC_STATS:
6391 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
6392 break;
6393 case SCTP_RECVRCVINFO:
6394 retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen);
6395 break;
6396 case SCTP_RECVNXTINFO:
6397 retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen);
6398 break;
6399 case SCTP_PR_SUPPORTED:
6400 retval = sctp_getsockopt_pr_supported(sk, len, optval, optlen);
6401 break;
6402 default:
6403 retval = -ENOPROTOOPT;
6404 break;
6405 }
6406
6407 release_sock(sk);
6408 return retval;
6409 }
6410
6411 static int sctp_hash(struct sock *sk)
6412 {
6413 /* STUB */
6414 return 0;
6415 }
6416
6417 static void sctp_unhash(struct sock *sk)
6418 {
6419 /* STUB */
6420 }
6421
6422 /* Check if port is acceptable. Possibly find first available port.
6423 *
6424 * The port hash table (contained in the 'global' SCTP protocol storage
6425 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
6426 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
6427 * list (the list number is the port number hashed out, so as you
6428 * would expect from a hash function, all the ports in a given list have
6429 * such a number that hashes out to the same list number; you were
6430 * expecting that, right?); so each list has a set of ports, with a
6431 * link to the socket (struct sock) that uses it, the port number and
6432 * a fastreuse flag (FIXME: NPI ipg).
6433 */
6434 static struct sctp_bind_bucket *sctp_bucket_create(
6435 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
6436
6437 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
6438 {
6439 struct sctp_bind_hashbucket *head; /* hash list */
6440 struct sctp_bind_bucket *pp;
6441 unsigned short snum;
6442 int ret;
6443
6444 snum = ntohs(addr->v4.sin_port);
6445
6446 pr_debug("%s: begins, snum:%d\n", __func__, snum);
6447
6448 local_bh_disable();
6449
6450 if (snum == 0) {
6451 /* Search for an available port. */
6452 int low, high, remaining, index;
6453 unsigned int rover;
6454 struct net *net = sock_net(sk);
6455
6456 inet_get_local_port_range(net, &low, &high);
6457 remaining = (high - low) + 1;
6458 rover = prandom_u32() % remaining + low;
6459
6460 do {
6461 rover++;
6462 if ((rover < low) || (rover > high))
6463 rover = low;
6464 if (inet_is_local_reserved_port(net, rover))
6465 continue;
6466 index = sctp_phashfn(sock_net(sk), rover);
6467 head = &sctp_port_hashtable[index];
6468 spin_lock(&head->lock);
6469 sctp_for_each_hentry(pp, &head->chain)
6470 if ((pp->port == rover) &&
6471 net_eq(sock_net(sk), pp->net))
6472 goto next;
6473 break;
6474 next:
6475 spin_unlock(&head->lock);
6476 } while (--remaining > 0);
6477
6478 /* Exhausted local port range during search? */
6479 ret = 1;
6480 if (remaining <= 0)
6481 goto fail;
6482
6483 /* OK, here is the one we will use. HEAD (the port
6484 * hash table list entry) is non-NULL and we hold it's
6485 * mutex.
6486 */
6487 snum = rover;
6488 } else {
6489 /* We are given an specific port number; we verify
6490 * that it is not being used. If it is used, we will
6491 * exahust the search in the hash list corresponding
6492 * to the port number (snum) - we detect that with the
6493 * port iterator, pp being NULL.
6494 */
6495 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
6496 spin_lock(&head->lock);
6497 sctp_for_each_hentry(pp, &head->chain) {
6498 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
6499 goto pp_found;
6500 }
6501 }
6502 pp = NULL;
6503 goto pp_not_found;
6504 pp_found:
6505 if (!hlist_empty(&pp->owner)) {
6506 /* We had a port hash table hit - there is an
6507 * available port (pp != NULL) and it is being
6508 * used by other socket (pp->owner not empty); that other
6509 * socket is going to be sk2.
6510 */
6511 int reuse = sk->sk_reuse;
6512 struct sock *sk2;
6513
6514 pr_debug("%s: found a possible match\n", __func__);
6515
6516 if (pp->fastreuse && sk->sk_reuse &&
6517 sk->sk_state != SCTP_SS_LISTENING)
6518 goto success;
6519
6520 /* Run through the list of sockets bound to the port
6521 * (pp->port) [via the pointers bind_next and
6522 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
6523 * we get the endpoint they describe and run through
6524 * the endpoint's list of IP (v4 or v6) addresses,
6525 * comparing each of the addresses with the address of
6526 * the socket sk. If we find a match, then that means
6527 * that this port/socket (sk) combination are already
6528 * in an endpoint.
6529 */
6530 sk_for_each_bound(sk2, &pp->owner) {
6531 struct sctp_endpoint *ep2;
6532 ep2 = sctp_sk(sk2)->ep;
6533
6534 if (sk == sk2 ||
6535 (reuse && sk2->sk_reuse &&
6536 sk2->sk_state != SCTP_SS_LISTENING))
6537 continue;
6538
6539 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
6540 sctp_sk(sk2), sctp_sk(sk))) {
6541 ret = (long)sk2;
6542 goto fail_unlock;
6543 }
6544 }
6545
6546 pr_debug("%s: found a match\n", __func__);
6547 }
6548 pp_not_found:
6549 /* If there was a hash table miss, create a new port. */
6550 ret = 1;
6551 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
6552 goto fail_unlock;
6553
6554 /* In either case (hit or miss), make sure fastreuse is 1 only
6555 * if sk->sk_reuse is too (that is, if the caller requested
6556 * SO_REUSEADDR on this socket -sk-).
6557 */
6558 if (hlist_empty(&pp->owner)) {
6559 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
6560 pp->fastreuse = 1;
6561 else
6562 pp->fastreuse = 0;
6563 } else if (pp->fastreuse &&
6564 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6565 pp->fastreuse = 0;
6566
6567 /* We are set, so fill up all the data in the hash table
6568 * entry, tie the socket list information with the rest of the
6569 * sockets FIXME: Blurry, NPI (ipg).
6570 */
6571 success:
6572 if (!sctp_sk(sk)->bind_hash) {
6573 inet_sk(sk)->inet_num = snum;
6574 sk_add_bind_node(sk, &pp->owner);
6575 sctp_sk(sk)->bind_hash = pp;
6576 }
6577 ret = 0;
6578
6579 fail_unlock:
6580 spin_unlock(&head->lock);
6581
6582 fail:
6583 local_bh_enable();
6584 return ret;
6585 }
6586
6587 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6588 * port is requested.
6589 */
6590 static int sctp_get_port(struct sock *sk, unsigned short snum)
6591 {
6592 union sctp_addr addr;
6593 struct sctp_af *af = sctp_sk(sk)->pf->af;
6594
6595 /* Set up a dummy address struct from the sk. */
6596 af->from_sk(&addr, sk);
6597 addr.v4.sin_port = htons(snum);
6598
6599 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6600 return !!sctp_get_port_local(sk, &addr);
6601 }
6602
6603 /*
6604 * Move a socket to LISTENING state.
6605 */
6606 static int sctp_listen_start(struct sock *sk, int backlog)
6607 {
6608 struct sctp_sock *sp = sctp_sk(sk);
6609 struct sctp_endpoint *ep = sp->ep;
6610 struct crypto_shash *tfm = NULL;
6611 char alg[32];
6612
6613 /* Allocate HMAC for generating cookie. */
6614 if (!sp->hmac && sp->sctp_hmac_alg) {
6615 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6616 tfm = crypto_alloc_shash(alg, 0, 0);
6617 if (IS_ERR(tfm)) {
6618 net_info_ratelimited("failed to load transform for %s: %ld\n",
6619 sp->sctp_hmac_alg, PTR_ERR(tfm));
6620 return -ENOSYS;
6621 }
6622 sctp_sk(sk)->hmac = tfm;
6623 }
6624
6625 /*
6626 * If a bind() or sctp_bindx() is not called prior to a listen()
6627 * call that allows new associations to be accepted, the system
6628 * picks an ephemeral port and will choose an address set equivalent
6629 * to binding with a wildcard address.
6630 *
6631 * This is not currently spelled out in the SCTP sockets
6632 * extensions draft, but follows the practice as seen in TCP
6633 * sockets.
6634 *
6635 */
6636 sk->sk_state = SCTP_SS_LISTENING;
6637 if (!ep->base.bind_addr.port) {
6638 if (sctp_autobind(sk))
6639 return -EAGAIN;
6640 } else {
6641 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6642 sk->sk_state = SCTP_SS_CLOSED;
6643 return -EADDRINUSE;
6644 }
6645 }
6646
6647 sk->sk_max_ack_backlog = backlog;
6648 sctp_hash_endpoint(ep);
6649 return 0;
6650 }
6651
6652 /*
6653 * 4.1.3 / 5.1.3 listen()
6654 *
6655 * By default, new associations are not accepted for UDP style sockets.
6656 * An application uses listen() to mark a socket as being able to
6657 * accept new associations.
6658 *
6659 * On TCP style sockets, applications use listen() to ready the SCTP
6660 * endpoint for accepting inbound associations.
6661 *
6662 * On both types of endpoints a backlog of '0' disables listening.
6663 *
6664 * Move a socket to LISTENING state.
6665 */
6666 int sctp_inet_listen(struct socket *sock, int backlog)
6667 {
6668 struct sock *sk = sock->sk;
6669 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6670 int err = -EINVAL;
6671
6672 if (unlikely(backlog < 0))
6673 return err;
6674
6675 lock_sock(sk);
6676
6677 /* Peeled-off sockets are not allowed to listen(). */
6678 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6679 goto out;
6680
6681 if (sock->state != SS_UNCONNECTED)
6682 goto out;
6683
6684 /* If backlog is zero, disable listening. */
6685 if (!backlog) {
6686 if (sctp_sstate(sk, CLOSED))
6687 goto out;
6688
6689 err = 0;
6690 sctp_unhash_endpoint(ep);
6691 sk->sk_state = SCTP_SS_CLOSED;
6692 if (sk->sk_reuse)
6693 sctp_sk(sk)->bind_hash->fastreuse = 1;
6694 goto out;
6695 }
6696
6697 /* If we are already listening, just update the backlog */
6698 if (sctp_sstate(sk, LISTENING))
6699 sk->sk_max_ack_backlog = backlog;
6700 else {
6701 err = sctp_listen_start(sk, backlog);
6702 if (err)
6703 goto out;
6704 }
6705
6706 err = 0;
6707 out:
6708 release_sock(sk);
6709 return err;
6710 }
6711
6712 /*
6713 * This function is done by modeling the current datagram_poll() and the
6714 * tcp_poll(). Note that, based on these implementations, we don't
6715 * lock the socket in this function, even though it seems that,
6716 * ideally, locking or some other mechanisms can be used to ensure
6717 * the integrity of the counters (sndbuf and wmem_alloc) used
6718 * in this place. We assume that we don't need locks either until proven
6719 * otherwise.
6720 *
6721 * Another thing to note is that we include the Async I/O support
6722 * here, again, by modeling the current TCP/UDP code. We don't have
6723 * a good way to test with it yet.
6724 */
6725 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6726 {
6727 struct sock *sk = sock->sk;
6728 struct sctp_sock *sp = sctp_sk(sk);
6729 unsigned int mask;
6730
6731 poll_wait(file, sk_sleep(sk), wait);
6732
6733 sock_rps_record_flow(sk);
6734
6735 /* A TCP-style listening socket becomes readable when the accept queue
6736 * is not empty.
6737 */
6738 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6739 return (!list_empty(&sp->ep->asocs)) ?
6740 (POLLIN | POLLRDNORM) : 0;
6741
6742 mask = 0;
6743
6744 /* Is there any exceptional events? */
6745 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6746 mask |= POLLERR |
6747 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
6748 if (sk->sk_shutdown & RCV_SHUTDOWN)
6749 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6750 if (sk->sk_shutdown == SHUTDOWN_MASK)
6751 mask |= POLLHUP;
6752
6753 /* Is it readable? Reconsider this code with TCP-style support. */
6754 if (!skb_queue_empty(&sk->sk_receive_queue))
6755 mask |= POLLIN | POLLRDNORM;
6756
6757 /* The association is either gone or not ready. */
6758 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6759 return mask;
6760
6761 /* Is it writable? */
6762 if (sctp_writeable(sk)) {
6763 mask |= POLLOUT | POLLWRNORM;
6764 } else {
6765 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
6766 /*
6767 * Since the socket is not locked, the buffer
6768 * might be made available after the writeable check and
6769 * before the bit is set. This could cause a lost I/O
6770 * signal. tcp_poll() has a race breaker for this race
6771 * condition. Based on their implementation, we put
6772 * in the following code to cover it as well.
6773 */
6774 if (sctp_writeable(sk))
6775 mask |= POLLOUT | POLLWRNORM;
6776 }
6777 return mask;
6778 }
6779
6780 /********************************************************************
6781 * 2nd Level Abstractions
6782 ********************************************************************/
6783
6784 static struct sctp_bind_bucket *sctp_bucket_create(
6785 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6786 {
6787 struct sctp_bind_bucket *pp;
6788
6789 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6790 if (pp) {
6791 SCTP_DBG_OBJCNT_INC(bind_bucket);
6792 pp->port = snum;
6793 pp->fastreuse = 0;
6794 INIT_HLIST_HEAD(&pp->owner);
6795 pp->net = net;
6796 hlist_add_head(&pp->node, &head->chain);
6797 }
6798 return pp;
6799 }
6800
6801 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6802 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6803 {
6804 if (pp && hlist_empty(&pp->owner)) {
6805 __hlist_del(&pp->node);
6806 kmem_cache_free(sctp_bucket_cachep, pp);
6807 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6808 }
6809 }
6810
6811 /* Release this socket's reference to a local port. */
6812 static inline void __sctp_put_port(struct sock *sk)
6813 {
6814 struct sctp_bind_hashbucket *head =
6815 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6816 inet_sk(sk)->inet_num)];
6817 struct sctp_bind_bucket *pp;
6818
6819 spin_lock(&head->lock);
6820 pp = sctp_sk(sk)->bind_hash;
6821 __sk_del_bind_node(sk);
6822 sctp_sk(sk)->bind_hash = NULL;
6823 inet_sk(sk)->inet_num = 0;
6824 sctp_bucket_destroy(pp);
6825 spin_unlock(&head->lock);
6826 }
6827
6828 void sctp_put_port(struct sock *sk)
6829 {
6830 local_bh_disable();
6831 __sctp_put_port(sk);
6832 local_bh_enable();
6833 }
6834
6835 /*
6836 * The system picks an ephemeral port and choose an address set equivalent
6837 * to binding with a wildcard address.
6838 * One of those addresses will be the primary address for the association.
6839 * This automatically enables the multihoming capability of SCTP.
6840 */
6841 static int sctp_autobind(struct sock *sk)
6842 {
6843 union sctp_addr autoaddr;
6844 struct sctp_af *af;
6845 __be16 port;
6846
6847 /* Initialize a local sockaddr structure to INADDR_ANY. */
6848 af = sctp_sk(sk)->pf->af;
6849
6850 port = htons(inet_sk(sk)->inet_num);
6851 af->inaddr_any(&autoaddr, port);
6852
6853 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6854 }
6855
6856 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6857 *
6858 * From RFC 2292
6859 * 4.2 The cmsghdr Structure *
6860 *
6861 * When ancillary data is sent or received, any number of ancillary data
6862 * objects can be specified by the msg_control and msg_controllen members of
6863 * the msghdr structure, because each object is preceded by
6864 * a cmsghdr structure defining the object's length (the cmsg_len member).
6865 * Historically Berkeley-derived implementations have passed only one object
6866 * at a time, but this API allows multiple objects to be
6867 * passed in a single call to sendmsg() or recvmsg(). The following example
6868 * shows two ancillary data objects in a control buffer.
6869 *
6870 * |<--------------------------- msg_controllen -------------------------->|
6871 * | |
6872 *
6873 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6874 *
6875 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6876 * | | |
6877 *
6878 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6879 *
6880 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6881 * | | | | |
6882 *
6883 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6884 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6885 *
6886 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6887 *
6888 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6889 * ^
6890 * |
6891 *
6892 * msg_control
6893 * points here
6894 */
6895 static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
6896 {
6897 struct cmsghdr *cmsg;
6898 struct msghdr *my_msg = (struct msghdr *)msg;
6899
6900 for_each_cmsghdr(cmsg, my_msg) {
6901 if (!CMSG_OK(my_msg, cmsg))
6902 return -EINVAL;
6903
6904 /* Should we parse this header or ignore? */
6905 if (cmsg->cmsg_level != IPPROTO_SCTP)
6906 continue;
6907
6908 /* Strictly check lengths following example in SCM code. */
6909 switch (cmsg->cmsg_type) {
6910 case SCTP_INIT:
6911 /* SCTP Socket API Extension
6912 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
6913 *
6914 * This cmsghdr structure provides information for
6915 * initializing new SCTP associations with sendmsg().
6916 * The SCTP_INITMSG socket option uses this same data
6917 * structure. This structure is not used for
6918 * recvmsg().
6919 *
6920 * cmsg_level cmsg_type cmsg_data[]
6921 * ------------ ------------ ----------------------
6922 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6923 */
6924 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg)))
6925 return -EINVAL;
6926
6927 cmsgs->init = CMSG_DATA(cmsg);
6928 break;
6929
6930 case SCTP_SNDRCV:
6931 /* SCTP Socket API Extension
6932 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
6933 *
6934 * This cmsghdr structure specifies SCTP options for
6935 * sendmsg() and describes SCTP header information
6936 * about a received message through recvmsg().
6937 *
6938 * cmsg_level cmsg_type cmsg_data[]
6939 * ------------ ------------ ----------------------
6940 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6941 */
6942 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6943 return -EINVAL;
6944
6945 cmsgs->srinfo = CMSG_DATA(cmsg);
6946
6947 if (cmsgs->srinfo->sinfo_flags &
6948 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6949 SCTP_SACK_IMMEDIATELY |
6950 SCTP_ABORT | SCTP_EOF))
6951 return -EINVAL;
6952 break;
6953
6954 case SCTP_SNDINFO:
6955 /* SCTP Socket API Extension
6956 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
6957 *
6958 * This cmsghdr structure specifies SCTP options for
6959 * sendmsg(). This structure and SCTP_RCVINFO replaces
6960 * SCTP_SNDRCV which has been deprecated.
6961 *
6962 * cmsg_level cmsg_type cmsg_data[]
6963 * ------------ ------------ ---------------------
6964 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
6965 */
6966 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo)))
6967 return -EINVAL;
6968
6969 cmsgs->sinfo = CMSG_DATA(cmsg);
6970
6971 if (cmsgs->sinfo->snd_flags &
6972 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6973 SCTP_SACK_IMMEDIATELY |
6974 SCTP_ABORT | SCTP_EOF))
6975 return -EINVAL;
6976 break;
6977 default:
6978 return -EINVAL;
6979 }
6980 }
6981
6982 return 0;
6983 }
6984
6985 /*
6986 * Wait for a packet..
6987 * Note: This function is the same function as in core/datagram.c
6988 * with a few modifications to make lksctp work.
6989 */
6990 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
6991 {
6992 int error;
6993 DEFINE_WAIT(wait);
6994
6995 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6996
6997 /* Socket errors? */
6998 error = sock_error(sk);
6999 if (error)
7000 goto out;
7001
7002 if (!skb_queue_empty(&sk->sk_receive_queue))
7003 goto ready;
7004
7005 /* Socket shut down? */
7006 if (sk->sk_shutdown & RCV_SHUTDOWN)
7007 goto out;
7008
7009 /* Sequenced packets can come disconnected. If so we report the
7010 * problem.
7011 */
7012 error = -ENOTCONN;
7013
7014 /* Is there a good reason to think that we may receive some data? */
7015 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
7016 goto out;
7017
7018 /* Handle signals. */
7019 if (signal_pending(current))
7020 goto interrupted;
7021
7022 /* Let another process have a go. Since we are going to sleep
7023 * anyway. Note: This may cause odd behaviors if the message
7024 * does not fit in the user's buffer, but this seems to be the
7025 * only way to honor MSG_DONTWAIT realistically.
7026 */
7027 release_sock(sk);
7028 *timeo_p = schedule_timeout(*timeo_p);
7029 lock_sock(sk);
7030
7031 ready:
7032 finish_wait(sk_sleep(sk), &wait);
7033 return 0;
7034
7035 interrupted:
7036 error = sock_intr_errno(*timeo_p);
7037
7038 out:
7039 finish_wait(sk_sleep(sk), &wait);
7040 *err = error;
7041 return error;
7042 }
7043
7044 /* Receive a datagram.
7045 * Note: This is pretty much the same routine as in core/datagram.c
7046 * with a few changes to make lksctp work.
7047 */
7048 struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
7049 int noblock, int *err)
7050 {
7051 int error;
7052 struct sk_buff *skb;
7053 long timeo;
7054
7055 timeo = sock_rcvtimeo(sk, noblock);
7056
7057 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
7058 MAX_SCHEDULE_TIMEOUT);
7059
7060 do {
7061 /* Again only user level code calls this function,
7062 * so nothing interrupt level
7063 * will suddenly eat the receive_queue.
7064 *
7065 * Look at current nfs client by the way...
7066 * However, this function was correct in any case. 8)
7067 */
7068 if (flags & MSG_PEEK) {
7069 skb = skb_peek(&sk->sk_receive_queue);
7070 if (skb)
7071 atomic_inc(&skb->users);
7072 } else {
7073 skb = __skb_dequeue(&sk->sk_receive_queue);
7074 }
7075
7076 if (skb)
7077 return skb;
7078
7079 /* Caller is allowed not to check sk->sk_err before calling. */
7080 error = sock_error(sk);
7081 if (error)
7082 goto no_packet;
7083
7084 if (sk->sk_shutdown & RCV_SHUTDOWN)
7085 break;
7086
7087 if (sk_can_busy_loop(sk) &&
7088 sk_busy_loop(sk, noblock))
7089 continue;
7090
7091 /* User doesn't want to wait. */
7092 error = -EAGAIN;
7093 if (!timeo)
7094 goto no_packet;
7095 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
7096
7097 return NULL;
7098
7099 no_packet:
7100 *err = error;
7101 return NULL;
7102 }
7103
7104 /* If sndbuf has changed, wake up per association sndbuf waiters. */
7105 static void __sctp_write_space(struct sctp_association *asoc)
7106 {
7107 struct sock *sk = asoc->base.sk;
7108
7109 if (sctp_wspace(asoc) <= 0)
7110 return;
7111
7112 if (waitqueue_active(&asoc->wait))
7113 wake_up_interruptible(&asoc->wait);
7114
7115 if (sctp_writeable(sk)) {
7116 struct socket_wq *wq;
7117
7118 rcu_read_lock();
7119 wq = rcu_dereference(sk->sk_wq);
7120 if (wq) {
7121 if (waitqueue_active(&wq->wait))
7122 wake_up_interruptible(&wq->wait);
7123
7124 /* Note that we try to include the Async I/O support
7125 * here by modeling from the current TCP/UDP code.
7126 * We have not tested with it yet.
7127 */
7128 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
7129 sock_wake_async(wq, SOCK_WAKE_SPACE, POLL_OUT);
7130 }
7131 rcu_read_unlock();
7132 }
7133 }
7134
7135 static void sctp_wake_up_waiters(struct sock *sk,
7136 struct sctp_association *asoc)
7137 {
7138 struct sctp_association *tmp = asoc;
7139
7140 /* We do accounting for the sndbuf space per association,
7141 * so we only need to wake our own association.
7142 */
7143 if (asoc->ep->sndbuf_policy)
7144 return __sctp_write_space(asoc);
7145
7146 /* If association goes down and is just flushing its
7147 * outq, then just normally notify others.
7148 */
7149 if (asoc->base.dead)
7150 return sctp_write_space(sk);
7151
7152 /* Accounting for the sndbuf space is per socket, so we
7153 * need to wake up others, try to be fair and in case of
7154 * other associations, let them have a go first instead
7155 * of just doing a sctp_write_space() call.
7156 *
7157 * Note that we reach sctp_wake_up_waiters() only when
7158 * associations free up queued chunks, thus we are under
7159 * lock and the list of associations on a socket is
7160 * guaranteed not to change.
7161 */
7162 for (tmp = list_next_entry(tmp, asocs); 1;
7163 tmp = list_next_entry(tmp, asocs)) {
7164 /* Manually skip the head element. */
7165 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
7166 continue;
7167 /* Wake up association. */
7168 __sctp_write_space(tmp);
7169 /* We've reached the end. */
7170 if (tmp == asoc)
7171 break;
7172 }
7173 }
7174
7175 /* Do accounting for the sndbuf space.
7176 * Decrement the used sndbuf space of the corresponding association by the
7177 * data size which was just transmitted(freed).
7178 */
7179 static void sctp_wfree(struct sk_buff *skb)
7180 {
7181 struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
7182 struct sctp_association *asoc = chunk->asoc;
7183 struct sock *sk = asoc->base.sk;
7184
7185 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
7186 sizeof(struct sk_buff) +
7187 sizeof(struct sctp_chunk);
7188
7189 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
7190
7191 /*
7192 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
7193 */
7194 sk->sk_wmem_queued -= skb->truesize;
7195 sk_mem_uncharge(sk, skb->truesize);
7196
7197 sock_wfree(skb);
7198 sctp_wake_up_waiters(sk, asoc);
7199
7200 sctp_association_put(asoc);
7201 }
7202
7203 /* Do accounting for the receive space on the socket.
7204 * Accounting for the association is done in ulpevent.c
7205 * We set this as a destructor for the cloned data skbs so that
7206 * accounting is done at the correct time.
7207 */
7208 void sctp_sock_rfree(struct sk_buff *skb)
7209 {
7210 struct sock *sk = skb->sk;
7211 struct sctp_ulpevent *event = sctp_skb2event(skb);
7212
7213 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
7214
7215 /*
7216 * Mimic the behavior of sock_rfree
7217 */
7218 sk_mem_uncharge(sk, event->rmem_len);
7219 }
7220
7221
7222 /* Helper function to wait for space in the sndbuf. */
7223 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
7224 size_t msg_len)
7225 {
7226 struct sock *sk = asoc->base.sk;
7227 int err = 0;
7228 long current_timeo = *timeo_p;
7229 DEFINE_WAIT(wait);
7230
7231 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
7232 *timeo_p, msg_len);
7233
7234 /* Increment the association's refcnt. */
7235 sctp_association_hold(asoc);
7236
7237 /* Wait on the association specific sndbuf space. */
7238 for (;;) {
7239 prepare_to_wait_exclusive(&asoc->wait, &wait,
7240 TASK_INTERRUPTIBLE);
7241 if (!*timeo_p)
7242 goto do_nonblock;
7243 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
7244 asoc->base.dead)
7245 goto do_error;
7246 if (signal_pending(current))
7247 goto do_interrupted;
7248 if (msg_len <= sctp_wspace(asoc))
7249 break;
7250
7251 /* Let another process have a go. Since we are going
7252 * to sleep anyway.
7253 */
7254 release_sock(sk);
7255 current_timeo = schedule_timeout(current_timeo);
7256 BUG_ON(sk != asoc->base.sk);
7257 lock_sock(sk);
7258
7259 *timeo_p = current_timeo;
7260 }
7261
7262 out:
7263 finish_wait(&asoc->wait, &wait);
7264
7265 /* Release the association's refcnt. */
7266 sctp_association_put(asoc);
7267
7268 return err;
7269
7270 do_error:
7271 err = -EPIPE;
7272 goto out;
7273
7274 do_interrupted:
7275 err = sock_intr_errno(*timeo_p);
7276 goto out;
7277
7278 do_nonblock:
7279 err = -EAGAIN;
7280 goto out;
7281 }
7282
7283 void sctp_data_ready(struct sock *sk)
7284 {
7285 struct socket_wq *wq;
7286
7287 rcu_read_lock();
7288 wq = rcu_dereference(sk->sk_wq);
7289 if (skwq_has_sleeper(wq))
7290 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
7291 POLLRDNORM | POLLRDBAND);
7292 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
7293 rcu_read_unlock();
7294 }
7295
7296 /* If socket sndbuf has changed, wake up all per association waiters. */
7297 void sctp_write_space(struct sock *sk)
7298 {
7299 struct sctp_association *asoc;
7300
7301 /* Wake up the tasks in each wait queue. */
7302 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
7303 __sctp_write_space(asoc);
7304 }
7305 }
7306
7307 /* Is there any sndbuf space available on the socket?
7308 *
7309 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
7310 * associations on the same socket. For a UDP-style socket with
7311 * multiple associations, it is possible for it to be "unwriteable"
7312 * prematurely. I assume that this is acceptable because
7313 * a premature "unwriteable" is better than an accidental "writeable" which
7314 * would cause an unwanted block under certain circumstances. For the 1-1
7315 * UDP-style sockets or TCP-style sockets, this code should work.
7316 * - Daisy
7317 */
7318 static int sctp_writeable(struct sock *sk)
7319 {
7320 int amt = 0;
7321
7322 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
7323 if (amt < 0)
7324 amt = 0;
7325 return amt;
7326 }
7327
7328 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
7329 * returns immediately with EINPROGRESS.
7330 */
7331 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
7332 {
7333 struct sock *sk = asoc->base.sk;
7334 int err = 0;
7335 long current_timeo = *timeo_p;
7336 DEFINE_WAIT(wait);
7337
7338 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
7339
7340 /* Increment the association's refcnt. */
7341 sctp_association_hold(asoc);
7342
7343 for (;;) {
7344 prepare_to_wait_exclusive(&asoc->wait, &wait,
7345 TASK_INTERRUPTIBLE);
7346 if (!*timeo_p)
7347 goto do_nonblock;
7348 if (sk->sk_shutdown & RCV_SHUTDOWN)
7349 break;
7350 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
7351 asoc->base.dead)
7352 goto do_error;
7353 if (signal_pending(current))
7354 goto do_interrupted;
7355
7356 if (sctp_state(asoc, ESTABLISHED))
7357 break;
7358
7359 /* Let another process have a go. Since we are going
7360 * to sleep anyway.
7361 */
7362 release_sock(sk);
7363 current_timeo = schedule_timeout(current_timeo);
7364 lock_sock(sk);
7365
7366 *timeo_p = current_timeo;
7367 }
7368
7369 out:
7370 finish_wait(&asoc->wait, &wait);
7371
7372 /* Release the association's refcnt. */
7373 sctp_association_put(asoc);
7374
7375 return err;
7376
7377 do_error:
7378 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
7379 err = -ETIMEDOUT;
7380 else
7381 err = -ECONNREFUSED;
7382 goto out;
7383
7384 do_interrupted:
7385 err = sock_intr_errno(*timeo_p);
7386 goto out;
7387
7388 do_nonblock:
7389 err = -EINPROGRESS;
7390 goto out;
7391 }
7392
7393 static int sctp_wait_for_accept(struct sock *sk, long timeo)
7394 {
7395 struct sctp_endpoint *ep;
7396 int err = 0;
7397 DEFINE_WAIT(wait);
7398
7399 ep = sctp_sk(sk)->ep;
7400
7401
7402 for (;;) {
7403 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
7404 TASK_INTERRUPTIBLE);
7405
7406 if (list_empty(&ep->asocs)) {
7407 release_sock(sk);
7408 timeo = schedule_timeout(timeo);
7409 lock_sock(sk);
7410 }
7411
7412 err = -EINVAL;
7413 if (!sctp_sstate(sk, LISTENING))
7414 break;
7415
7416 err = 0;
7417 if (!list_empty(&ep->asocs))
7418 break;
7419
7420 err = sock_intr_errno(timeo);
7421 if (signal_pending(current))
7422 break;
7423
7424 err = -EAGAIN;
7425 if (!timeo)
7426 break;
7427 }
7428
7429 finish_wait(sk_sleep(sk), &wait);
7430
7431 return err;
7432 }
7433
7434 static void sctp_wait_for_close(struct sock *sk, long timeout)
7435 {
7436 DEFINE_WAIT(wait);
7437
7438 do {
7439 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
7440 if (list_empty(&sctp_sk(sk)->ep->asocs))
7441 break;
7442 release_sock(sk);
7443 timeout = schedule_timeout(timeout);
7444 lock_sock(sk);
7445 } while (!signal_pending(current) && timeout);
7446
7447 finish_wait(sk_sleep(sk), &wait);
7448 }
7449
7450 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
7451 {
7452 struct sk_buff *frag;
7453
7454 if (!skb->data_len)
7455 goto done;
7456
7457 /* Don't forget the fragments. */
7458 skb_walk_frags(skb, frag)
7459 sctp_skb_set_owner_r_frag(frag, sk);
7460
7461 done:
7462 sctp_skb_set_owner_r(skb, sk);
7463 }
7464
7465 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
7466 struct sctp_association *asoc)
7467 {
7468 struct inet_sock *inet = inet_sk(sk);
7469 struct inet_sock *newinet;
7470
7471 newsk->sk_type = sk->sk_type;
7472 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
7473 newsk->sk_flags = sk->sk_flags;
7474 newsk->sk_tsflags = sk->sk_tsflags;
7475 newsk->sk_no_check_tx = sk->sk_no_check_tx;
7476 newsk->sk_no_check_rx = sk->sk_no_check_rx;
7477 newsk->sk_reuse = sk->sk_reuse;
7478
7479 newsk->sk_shutdown = sk->sk_shutdown;
7480 newsk->sk_destruct = sctp_destruct_sock;
7481 newsk->sk_family = sk->sk_family;
7482 newsk->sk_protocol = IPPROTO_SCTP;
7483 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
7484 newsk->sk_sndbuf = sk->sk_sndbuf;
7485 newsk->sk_rcvbuf = sk->sk_rcvbuf;
7486 newsk->sk_lingertime = sk->sk_lingertime;
7487 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
7488 newsk->sk_sndtimeo = sk->sk_sndtimeo;
7489 newsk->sk_rxhash = sk->sk_rxhash;
7490
7491 newinet = inet_sk(newsk);
7492
7493 /* Initialize sk's sport, dport, rcv_saddr and daddr for
7494 * getsockname() and getpeername()
7495 */
7496 newinet->inet_sport = inet->inet_sport;
7497 newinet->inet_saddr = inet->inet_saddr;
7498 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
7499 newinet->inet_dport = htons(asoc->peer.port);
7500 newinet->pmtudisc = inet->pmtudisc;
7501 newinet->inet_id = asoc->next_tsn ^ jiffies;
7502
7503 newinet->uc_ttl = inet->uc_ttl;
7504 newinet->mc_loop = 1;
7505 newinet->mc_ttl = 1;
7506 newinet->mc_index = 0;
7507 newinet->mc_list = NULL;
7508
7509 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
7510 net_enable_timestamp();
7511
7512 security_sk_clone(sk, newsk);
7513 }
7514
7515 static inline void sctp_copy_descendant(struct sock *sk_to,
7516 const struct sock *sk_from)
7517 {
7518 int ancestor_size = sizeof(struct inet_sock) +
7519 sizeof(struct sctp_sock) -
7520 offsetof(struct sctp_sock, auto_asconf_list);
7521
7522 if (sk_from->sk_family == PF_INET6)
7523 ancestor_size += sizeof(struct ipv6_pinfo);
7524
7525 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
7526 }
7527
7528 /* Populate the fields of the newsk from the oldsk and migrate the assoc
7529 * and its messages to the newsk.
7530 */
7531 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
7532 struct sctp_association *assoc,
7533 sctp_socket_type_t type)
7534 {
7535 struct sctp_sock *oldsp = sctp_sk(oldsk);
7536 struct sctp_sock *newsp = sctp_sk(newsk);
7537 struct sctp_bind_bucket *pp; /* hash list port iterator */
7538 struct sctp_endpoint *newep = newsp->ep;
7539 struct sk_buff *skb, *tmp;
7540 struct sctp_ulpevent *event;
7541 struct sctp_bind_hashbucket *head;
7542
7543 /* Migrate socket buffer sizes and all the socket level options to the
7544 * new socket.
7545 */
7546 newsk->sk_sndbuf = oldsk->sk_sndbuf;
7547 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
7548 /* Brute force copy old sctp opt. */
7549 sctp_copy_descendant(newsk, oldsk);
7550
7551 /* Restore the ep value that was overwritten with the above structure
7552 * copy.
7553 */
7554 newsp->ep = newep;
7555 newsp->hmac = NULL;
7556
7557 /* Hook this new socket in to the bind_hash list. */
7558 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
7559 inet_sk(oldsk)->inet_num)];
7560 spin_lock_bh(&head->lock);
7561 pp = sctp_sk(oldsk)->bind_hash;
7562 sk_add_bind_node(newsk, &pp->owner);
7563 sctp_sk(newsk)->bind_hash = pp;
7564 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
7565 spin_unlock_bh(&head->lock);
7566
7567 /* Copy the bind_addr list from the original endpoint to the new
7568 * endpoint so that we can handle restarts properly
7569 */
7570 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
7571 &oldsp->ep->base.bind_addr, GFP_KERNEL);
7572
7573 /* Move any messages in the old socket's receive queue that are for the
7574 * peeled off association to the new socket's receive queue.
7575 */
7576 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
7577 event = sctp_skb2event(skb);
7578 if (event->asoc == assoc) {
7579 __skb_unlink(skb, &oldsk->sk_receive_queue);
7580 __skb_queue_tail(&newsk->sk_receive_queue, skb);
7581 sctp_skb_set_owner_r_frag(skb, newsk);
7582 }
7583 }
7584
7585 /* Clean up any messages pending delivery due to partial
7586 * delivery. Three cases:
7587 * 1) No partial deliver; no work.
7588 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
7589 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
7590 */
7591 skb_queue_head_init(&newsp->pd_lobby);
7592 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
7593
7594 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
7595 struct sk_buff_head *queue;
7596
7597 /* Decide which queue to move pd_lobby skbs to. */
7598 if (assoc->ulpq.pd_mode) {
7599 queue = &newsp->pd_lobby;
7600 } else
7601 queue = &newsk->sk_receive_queue;
7602
7603 /* Walk through the pd_lobby, looking for skbs that
7604 * need moved to the new socket.
7605 */
7606 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
7607 event = sctp_skb2event(skb);
7608 if (event->asoc == assoc) {
7609 __skb_unlink(skb, &oldsp->pd_lobby);
7610 __skb_queue_tail(queue, skb);
7611 sctp_skb_set_owner_r_frag(skb, newsk);
7612 }
7613 }
7614
7615 /* Clear up any skbs waiting for the partial
7616 * delivery to finish.
7617 */
7618 if (assoc->ulpq.pd_mode)
7619 sctp_clear_pd(oldsk, NULL);
7620
7621 }
7622
7623 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
7624 sctp_skb_set_owner_r_frag(skb, newsk);
7625
7626 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
7627 sctp_skb_set_owner_r_frag(skb, newsk);
7628
7629 /* Set the type of socket to indicate that it is peeled off from the
7630 * original UDP-style socket or created with the accept() call on a
7631 * TCP-style socket..
7632 */
7633 newsp->type = type;
7634
7635 /* Mark the new socket "in-use" by the user so that any packets
7636 * that may arrive on the association after we've moved it are
7637 * queued to the backlog. This prevents a potential race between
7638 * backlog processing on the old socket and new-packet processing
7639 * on the new socket.
7640 *
7641 * The caller has just allocated newsk so we can guarantee that other
7642 * paths won't try to lock it and then oldsk.
7643 */
7644 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7645 sctp_assoc_migrate(assoc, newsk);
7646
7647 /* If the association on the newsk is already closed before accept()
7648 * is called, set RCV_SHUTDOWN flag.
7649 */
7650 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP)) {
7651 newsk->sk_state = SCTP_SS_CLOSED;
7652 newsk->sk_shutdown |= RCV_SHUTDOWN;
7653 } else {
7654 newsk->sk_state = SCTP_SS_ESTABLISHED;
7655 }
7656
7657 release_sock(newsk);
7658 }
7659
7660
7661 /* This proto struct describes the ULP interface for SCTP. */
7662 struct proto sctp_prot = {
7663 .name = "SCTP",
7664 .owner = THIS_MODULE,
7665 .close = sctp_close,
7666 .connect = sctp_connect,
7667 .disconnect = sctp_disconnect,
7668 .accept = sctp_accept,
7669 .ioctl = sctp_ioctl,
7670 .init = sctp_init_sock,
7671 .destroy = sctp_destroy_sock,
7672 .shutdown = sctp_shutdown,
7673 .setsockopt = sctp_setsockopt,
7674 .getsockopt = sctp_getsockopt,
7675 .sendmsg = sctp_sendmsg,
7676 .recvmsg = sctp_recvmsg,
7677 .bind = sctp_bind,
7678 .backlog_rcv = sctp_backlog_rcv,
7679 .hash = sctp_hash,
7680 .unhash = sctp_unhash,
7681 .get_port = sctp_get_port,
7682 .obj_size = sizeof(struct sctp_sock),
7683 .sysctl_mem = sysctl_sctp_mem,
7684 .sysctl_rmem = sysctl_sctp_rmem,
7685 .sysctl_wmem = sysctl_sctp_wmem,
7686 .memory_pressure = &sctp_memory_pressure,
7687 .enter_memory_pressure = sctp_enter_memory_pressure,
7688 .memory_allocated = &sctp_memory_allocated,
7689 .sockets_allocated = &sctp_sockets_allocated,
7690 };
7691
7692 #if IS_ENABLED(CONFIG_IPV6)
7693
7694 #include <net/transp_v6.h>
7695 static void sctp_v6_destroy_sock(struct sock *sk)
7696 {
7697 sctp_destroy_sock(sk);
7698 inet6_destroy_sock(sk);
7699 }
7700
7701 struct proto sctpv6_prot = {
7702 .name = "SCTPv6",
7703 .owner = THIS_MODULE,
7704 .close = sctp_close,
7705 .connect = sctp_connect,
7706 .disconnect = sctp_disconnect,
7707 .accept = sctp_accept,
7708 .ioctl = sctp_ioctl,
7709 .init = sctp_init_sock,
7710 .destroy = sctp_v6_destroy_sock,
7711 .shutdown = sctp_shutdown,
7712 .setsockopt = sctp_setsockopt,
7713 .getsockopt = sctp_getsockopt,
7714 .sendmsg = sctp_sendmsg,
7715 .recvmsg = sctp_recvmsg,
7716 .bind = sctp_bind,
7717 .backlog_rcv = sctp_backlog_rcv,
7718 .hash = sctp_hash,
7719 .unhash = sctp_unhash,
7720 .get_port = sctp_get_port,
7721 .obj_size = sizeof(struct sctp6_sock),
7722 .sysctl_mem = sysctl_sctp_mem,
7723 .sysctl_rmem = sysctl_sctp_rmem,
7724 .sysctl_wmem = sysctl_sctp_wmem,
7725 .memory_pressure = &sctp_memory_pressure,
7726 .enter_memory_pressure = sctp_enter_memory_pressure,
7727 .memory_allocated = &sctp_memory_allocated,
7728 .sockets_allocated = &sctp_sockets_allocated,
7729 };
7730 #endif /* IS_ENABLED(CONFIG_IPV6) */
This page took 0.239925 seconds and 5 git commands to generate.