]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blob - net/sctp/socket.c
Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
[mirror_ubuntu-jammy-kernel.git] / net / sctp / socket.c
1 /* SCTP kernel implementation
2 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001-2003 Intel Corp.
6 * Copyright (c) 2001-2002 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
8 *
9 * This file is part of the SCTP kernel implementation
10 *
11 * These functions interface with the sockets layer to implement the
12 * SCTP Extensions for the Sockets API.
13 *
14 * Note that the descriptions from the specification are USER level
15 * functions--this file is the functions which populate the struct proto
16 * for SCTP which is the BOTTOM of the sockets interface.
17 *
18 * This SCTP implementation is free software;
19 * you can redistribute it and/or modify it under the terms of
20 * the GNU General Public License as published by
21 * the Free Software Foundation; either version 2, or (at your option)
22 * any later version.
23 *
24 * This SCTP implementation is distributed in the hope that it
25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 * ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
29 *
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, write to
32 * the Free Software Foundation, 59 Temple Place - Suite 330,
33 * Boston, MA 02111-1307, USA.
34 *
35 * Please send any bug reports or fixes you make to the
36 * email address(es):
37 * lksctp developers <lksctp-developers@lists.sourceforge.net>
38 *
39 * Or submit a bug report through the following website:
40 * http://www.sf.net/projects/lksctp
41 *
42 * Written or modified by:
43 * La Monte H.P. Yarroll <piggy@acm.org>
44 * Narasimha Budihal <narsi@refcode.org>
45 * Karl Knutson <karl@athena.chicago.il.us>
46 * Jon Grimm <jgrimm@us.ibm.com>
47 * Xingang Guo <xingang.guo@intel.com>
48 * Daisy Chang <daisyc@us.ibm.com>
49 * Sridhar Samudrala <samudrala@us.ibm.com>
50 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
51 * Ardelle Fan <ardelle.fan@intel.com>
52 * Ryan Layer <rmlayer@us.ibm.com>
53 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
54 * Kevin Gao <kevin.gao@intel.com>
55 *
56 * Any bugs reported given to us we will try to fix... any fixes shared will
57 * be incorporated into the next SCTP release.
58 */
59
60 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
61
62 #include <linux/types.h>
63 #include <linux/kernel.h>
64 #include <linux/wait.h>
65 #include <linux/time.h>
66 #include <linux/ip.h>
67 #include <linux/capability.h>
68 #include <linux/fcntl.h>
69 #include <linux/poll.h>
70 #include <linux/init.h>
71 #include <linux/crypto.h>
72 #include <linux/slab.h>
73 #include <linux/file.h>
74
75 #include <net/ip.h>
76 #include <net/icmp.h>
77 #include <net/route.h>
78 #include <net/ipv6.h>
79 #include <net/inet_common.h>
80
81 #include <linux/socket.h> /* for sa_family_t */
82 #include <linux/export.h>
83 #include <net/sock.h>
84 #include <net/sctp/sctp.h>
85 #include <net/sctp/sm.h>
86
87 /* Forward declarations for internal helper functions. */
88 static int sctp_writeable(struct sock *sk);
89 static void sctp_wfree(struct sk_buff *skb);
90 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
91 size_t msg_len);
92 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
93 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
94 static int sctp_wait_for_accept(struct sock *sk, long timeo);
95 static void sctp_wait_for_close(struct sock *sk, long timeo);
96 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
97 union sctp_addr *addr, int len);
98 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
99 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
100 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
101 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
102 static int sctp_send_asconf(struct sctp_association *asoc,
103 struct sctp_chunk *chunk);
104 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
105 static int sctp_autobind(struct sock *sk);
106 static void sctp_sock_migrate(struct sock *, struct sock *,
107 struct sctp_association *, sctp_socket_type_t);
108
109 extern struct kmem_cache *sctp_bucket_cachep;
110 extern long sysctl_sctp_mem[3];
111 extern int sysctl_sctp_rmem[3];
112 extern int sysctl_sctp_wmem[3];
113
114 static int sctp_memory_pressure;
115 static atomic_long_t sctp_memory_allocated;
116 struct percpu_counter sctp_sockets_allocated;
117
118 static void sctp_enter_memory_pressure(struct sock *sk)
119 {
120 sctp_memory_pressure = 1;
121 }
122
123
124 /* Get the sndbuf space available at the time on the association. */
125 static inline int sctp_wspace(struct sctp_association *asoc)
126 {
127 int amt;
128
129 if (asoc->ep->sndbuf_policy)
130 amt = asoc->sndbuf_used;
131 else
132 amt = sk_wmem_alloc_get(asoc->base.sk);
133
134 if (amt >= asoc->base.sk->sk_sndbuf) {
135 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
136 amt = 0;
137 else {
138 amt = sk_stream_wspace(asoc->base.sk);
139 if (amt < 0)
140 amt = 0;
141 }
142 } else {
143 amt = asoc->base.sk->sk_sndbuf - amt;
144 }
145 return amt;
146 }
147
148 /* Increment the used sndbuf space count of the corresponding association by
149 * the size of the outgoing data chunk.
150 * Also, set the skb destructor for sndbuf accounting later.
151 *
152 * Since it is always 1-1 between chunk and skb, and also a new skb is always
153 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
154 * destructor in the data chunk skb for the purpose of the sndbuf space
155 * tracking.
156 */
157 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
158 {
159 struct sctp_association *asoc = chunk->asoc;
160 struct sock *sk = asoc->base.sk;
161
162 /* The sndbuf space is tracked per association. */
163 sctp_association_hold(asoc);
164
165 skb_set_owner_w(chunk->skb, sk);
166
167 chunk->skb->destructor = sctp_wfree;
168 /* Save the chunk pointer in skb for sctp_wfree to use later. */
169 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
170
171 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
172 sizeof(struct sk_buff) +
173 sizeof(struct sctp_chunk);
174
175 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
176 sk->sk_wmem_queued += chunk->skb->truesize;
177 sk_mem_charge(sk, chunk->skb->truesize);
178 }
179
180 /* Verify that this is a valid address. */
181 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
182 int len)
183 {
184 struct sctp_af *af;
185
186 /* Verify basic sockaddr. */
187 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
188 if (!af)
189 return -EINVAL;
190
191 /* Is this a valid SCTP address? */
192 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
193 return -EINVAL;
194
195 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
196 return -EINVAL;
197
198 return 0;
199 }
200
201 /* Look up the association by its id. If this is not a UDP-style
202 * socket, the ID field is always ignored.
203 */
204 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
205 {
206 struct sctp_association *asoc = NULL;
207
208 /* If this is not a UDP-style socket, assoc id should be ignored. */
209 if (!sctp_style(sk, UDP)) {
210 /* Return NULL if the socket state is not ESTABLISHED. It
211 * could be a TCP-style listening socket or a socket which
212 * hasn't yet called connect() to establish an association.
213 */
214 if (!sctp_sstate(sk, ESTABLISHED))
215 return NULL;
216
217 /* Get the first and the only association from the list. */
218 if (!list_empty(&sctp_sk(sk)->ep->asocs))
219 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
220 struct sctp_association, asocs);
221 return asoc;
222 }
223
224 /* Otherwise this is a UDP-style socket. */
225 if (!id || (id == (sctp_assoc_t)-1))
226 return NULL;
227
228 spin_lock_bh(&sctp_assocs_id_lock);
229 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
230 spin_unlock_bh(&sctp_assocs_id_lock);
231
232 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
233 return NULL;
234
235 return asoc;
236 }
237
238 /* Look up the transport from an address and an assoc id. If both address and
239 * id are specified, the associations matching the address and the id should be
240 * the same.
241 */
242 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
243 struct sockaddr_storage *addr,
244 sctp_assoc_t id)
245 {
246 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
247 struct sctp_transport *transport;
248 union sctp_addr *laddr = (union sctp_addr *)addr;
249
250 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
251 laddr,
252 &transport);
253
254 if (!addr_asoc)
255 return NULL;
256
257 id_asoc = sctp_id2assoc(sk, id);
258 if (id_asoc && (id_asoc != addr_asoc))
259 return NULL;
260
261 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
262 (union sctp_addr *)addr);
263
264 return transport;
265 }
266
267 /* API 3.1.2 bind() - UDP Style Syntax
268 * The syntax of bind() is,
269 *
270 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
271 *
272 * sd - the socket descriptor returned by socket().
273 * addr - the address structure (struct sockaddr_in or struct
274 * sockaddr_in6 [RFC 2553]),
275 * addr_len - the size of the address structure.
276 */
277 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
278 {
279 int retval = 0;
280
281 sctp_lock_sock(sk);
282
283 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
284 sk, addr, addr_len);
285
286 /* Disallow binding twice. */
287 if (!sctp_sk(sk)->ep->base.bind_addr.port)
288 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
289 addr_len);
290 else
291 retval = -EINVAL;
292
293 sctp_release_sock(sk);
294
295 return retval;
296 }
297
298 static long sctp_get_port_local(struct sock *, union sctp_addr *);
299
300 /* Verify this is a valid sockaddr. */
301 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
302 union sctp_addr *addr, int len)
303 {
304 struct sctp_af *af;
305
306 /* Check minimum size. */
307 if (len < sizeof (struct sockaddr))
308 return NULL;
309
310 /* V4 mapped address are really of AF_INET family */
311 if (addr->sa.sa_family == AF_INET6 &&
312 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
313 if (!opt->pf->af_supported(AF_INET, opt))
314 return NULL;
315 } else {
316 /* Does this PF support this AF? */
317 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
318 return NULL;
319 }
320
321 /* If we get this far, af is valid. */
322 af = sctp_get_af_specific(addr->sa.sa_family);
323
324 if (len < af->sockaddr_len)
325 return NULL;
326
327 return af;
328 }
329
330 /* Bind a local address either to an endpoint or to an association. */
331 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
332 {
333 struct net *net = sock_net(sk);
334 struct sctp_sock *sp = sctp_sk(sk);
335 struct sctp_endpoint *ep = sp->ep;
336 struct sctp_bind_addr *bp = &ep->base.bind_addr;
337 struct sctp_af *af;
338 unsigned short snum;
339 int ret = 0;
340
341 /* Common sockaddr verification. */
342 af = sctp_sockaddr_af(sp, addr, len);
343 if (!af) {
344 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
345 sk, addr, len);
346 return -EINVAL;
347 }
348
349 snum = ntohs(addr->v4.sin_port);
350
351 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
352 ", port: %d, new port: %d, len: %d)\n",
353 sk,
354 addr,
355 bp->port, snum,
356 len);
357
358 /* PF specific bind() address verification. */
359 if (!sp->pf->bind_verify(sp, addr))
360 return -EADDRNOTAVAIL;
361
362 /* We must either be unbound, or bind to the same port.
363 * It's OK to allow 0 ports if we are already bound.
364 * We'll just inhert an already bound port in this case
365 */
366 if (bp->port) {
367 if (!snum)
368 snum = bp->port;
369 else if (snum != bp->port) {
370 SCTP_DEBUG_PRINTK("sctp_do_bind:"
371 " New port %d does not match existing port "
372 "%d.\n", snum, bp->port);
373 return -EINVAL;
374 }
375 }
376
377 if (snum && snum < PROT_SOCK &&
378 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
379 return -EACCES;
380
381 /* See if the address matches any of the addresses we may have
382 * already bound before checking against other endpoints.
383 */
384 if (sctp_bind_addr_match(bp, addr, sp))
385 return -EINVAL;
386
387 /* Make sure we are allowed to bind here.
388 * The function sctp_get_port_local() does duplicate address
389 * detection.
390 */
391 addr->v4.sin_port = htons(snum);
392 if ((ret = sctp_get_port_local(sk, addr))) {
393 return -EADDRINUSE;
394 }
395
396 /* Refresh ephemeral port. */
397 if (!bp->port)
398 bp->port = inet_sk(sk)->inet_num;
399
400 /* Add the address to the bind address list.
401 * Use GFP_ATOMIC since BHs will be disabled.
402 */
403 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
404
405 /* Copy back into socket for getsockname() use. */
406 if (!ret) {
407 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
408 af->to_sk_saddr(addr, sk);
409 }
410
411 return ret;
412 }
413
414 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
415 *
416 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
417 * at any one time. If a sender, after sending an ASCONF chunk, decides
418 * it needs to transfer another ASCONF Chunk, it MUST wait until the
419 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
420 * subsequent ASCONF. Note this restriction binds each side, so at any
421 * time two ASCONF may be in-transit on any given association (one sent
422 * from each endpoint).
423 */
424 static int sctp_send_asconf(struct sctp_association *asoc,
425 struct sctp_chunk *chunk)
426 {
427 struct net *net = sock_net(asoc->base.sk);
428 int retval = 0;
429
430 /* If there is an outstanding ASCONF chunk, queue it for later
431 * transmission.
432 */
433 if (asoc->addip_last_asconf) {
434 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
435 goto out;
436 }
437
438 /* Hold the chunk until an ASCONF_ACK is received. */
439 sctp_chunk_hold(chunk);
440 retval = sctp_primitive_ASCONF(net, asoc, chunk);
441 if (retval)
442 sctp_chunk_free(chunk);
443 else
444 asoc->addip_last_asconf = chunk;
445
446 out:
447 return retval;
448 }
449
450 /* Add a list of addresses as bind addresses to local endpoint or
451 * association.
452 *
453 * Basically run through each address specified in the addrs/addrcnt
454 * array/length pair, determine if it is IPv6 or IPv4 and call
455 * sctp_do_bind() on it.
456 *
457 * If any of them fails, then the operation will be reversed and the
458 * ones that were added will be removed.
459 *
460 * Only sctp_setsockopt_bindx() is supposed to call this function.
461 */
462 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
463 {
464 int cnt;
465 int retval = 0;
466 void *addr_buf;
467 struct sockaddr *sa_addr;
468 struct sctp_af *af;
469
470 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
471 sk, addrs, addrcnt);
472
473 addr_buf = addrs;
474 for (cnt = 0; cnt < addrcnt; cnt++) {
475 /* The list may contain either IPv4 or IPv6 address;
476 * determine the address length for walking thru the list.
477 */
478 sa_addr = addr_buf;
479 af = sctp_get_af_specific(sa_addr->sa_family);
480 if (!af) {
481 retval = -EINVAL;
482 goto err_bindx_add;
483 }
484
485 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
486 af->sockaddr_len);
487
488 addr_buf += af->sockaddr_len;
489
490 err_bindx_add:
491 if (retval < 0) {
492 /* Failed. Cleanup the ones that have been added */
493 if (cnt > 0)
494 sctp_bindx_rem(sk, addrs, cnt);
495 return retval;
496 }
497 }
498
499 return retval;
500 }
501
502 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
503 * associations that are part of the endpoint indicating that a list of local
504 * addresses are added to the endpoint.
505 *
506 * If any of the addresses is already in the bind address list of the
507 * association, we do not send the chunk for that association. But it will not
508 * affect other associations.
509 *
510 * Only sctp_setsockopt_bindx() is supposed to call this function.
511 */
512 static int sctp_send_asconf_add_ip(struct sock *sk,
513 struct sockaddr *addrs,
514 int addrcnt)
515 {
516 struct net *net = sock_net(sk);
517 struct sctp_sock *sp;
518 struct sctp_endpoint *ep;
519 struct sctp_association *asoc;
520 struct sctp_bind_addr *bp;
521 struct sctp_chunk *chunk;
522 struct sctp_sockaddr_entry *laddr;
523 union sctp_addr *addr;
524 union sctp_addr saveaddr;
525 void *addr_buf;
526 struct sctp_af *af;
527 struct list_head *p;
528 int i;
529 int retval = 0;
530
531 if (!net->sctp.addip_enable)
532 return retval;
533
534 sp = sctp_sk(sk);
535 ep = sp->ep;
536
537 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
538 __func__, sk, addrs, addrcnt);
539
540 list_for_each_entry(asoc, &ep->asocs, asocs) {
541
542 if (!asoc->peer.asconf_capable)
543 continue;
544
545 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
546 continue;
547
548 if (!sctp_state(asoc, ESTABLISHED))
549 continue;
550
551 /* Check if any address in the packed array of addresses is
552 * in the bind address list of the association. If so,
553 * do not send the asconf chunk to its peer, but continue with
554 * other associations.
555 */
556 addr_buf = addrs;
557 for (i = 0; i < addrcnt; i++) {
558 addr = addr_buf;
559 af = sctp_get_af_specific(addr->v4.sin_family);
560 if (!af) {
561 retval = -EINVAL;
562 goto out;
563 }
564
565 if (sctp_assoc_lookup_laddr(asoc, addr))
566 break;
567
568 addr_buf += af->sockaddr_len;
569 }
570 if (i < addrcnt)
571 continue;
572
573 /* Use the first valid address in bind addr list of
574 * association as Address Parameter of ASCONF CHUNK.
575 */
576 bp = &asoc->base.bind_addr;
577 p = bp->address_list.next;
578 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
579 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
580 addrcnt, SCTP_PARAM_ADD_IP);
581 if (!chunk) {
582 retval = -ENOMEM;
583 goto out;
584 }
585
586 /* Add the new addresses to the bind address list with
587 * use_as_src set to 0.
588 */
589 addr_buf = addrs;
590 for (i = 0; i < addrcnt; i++) {
591 addr = addr_buf;
592 af = sctp_get_af_specific(addr->v4.sin_family);
593 memcpy(&saveaddr, addr, af->sockaddr_len);
594 retval = sctp_add_bind_addr(bp, &saveaddr,
595 SCTP_ADDR_NEW, GFP_ATOMIC);
596 addr_buf += af->sockaddr_len;
597 }
598 if (asoc->src_out_of_asoc_ok) {
599 struct sctp_transport *trans;
600
601 list_for_each_entry(trans,
602 &asoc->peer.transport_addr_list, transports) {
603 /* Clear the source and route cache */
604 dst_release(trans->dst);
605 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
606 2*asoc->pathmtu, 4380));
607 trans->ssthresh = asoc->peer.i.a_rwnd;
608 trans->rto = asoc->rto_initial;
609 sctp_max_rto(asoc, trans);
610 trans->rtt = trans->srtt = trans->rttvar = 0;
611 sctp_transport_route(trans, NULL,
612 sctp_sk(asoc->base.sk));
613 }
614 }
615 retval = sctp_send_asconf(asoc, chunk);
616 }
617
618 out:
619 return retval;
620 }
621
622 /* Remove a list of addresses from bind addresses list. Do not remove the
623 * last address.
624 *
625 * Basically run through each address specified in the addrs/addrcnt
626 * array/length pair, determine if it is IPv6 or IPv4 and call
627 * sctp_del_bind() on it.
628 *
629 * If any of them fails, then the operation will be reversed and the
630 * ones that were removed will be added back.
631 *
632 * At least one address has to be left; if only one address is
633 * available, the operation will return -EBUSY.
634 *
635 * Only sctp_setsockopt_bindx() is supposed to call this function.
636 */
637 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
638 {
639 struct sctp_sock *sp = sctp_sk(sk);
640 struct sctp_endpoint *ep = sp->ep;
641 int cnt;
642 struct sctp_bind_addr *bp = &ep->base.bind_addr;
643 int retval = 0;
644 void *addr_buf;
645 union sctp_addr *sa_addr;
646 struct sctp_af *af;
647
648 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
649 sk, addrs, addrcnt);
650
651 addr_buf = addrs;
652 for (cnt = 0; cnt < addrcnt; cnt++) {
653 /* If the bind address list is empty or if there is only one
654 * bind address, there is nothing more to be removed (we need
655 * at least one address here).
656 */
657 if (list_empty(&bp->address_list) ||
658 (sctp_list_single_entry(&bp->address_list))) {
659 retval = -EBUSY;
660 goto err_bindx_rem;
661 }
662
663 sa_addr = addr_buf;
664 af = sctp_get_af_specific(sa_addr->sa.sa_family);
665 if (!af) {
666 retval = -EINVAL;
667 goto err_bindx_rem;
668 }
669
670 if (!af->addr_valid(sa_addr, sp, NULL)) {
671 retval = -EADDRNOTAVAIL;
672 goto err_bindx_rem;
673 }
674
675 if (sa_addr->v4.sin_port &&
676 sa_addr->v4.sin_port != htons(bp->port)) {
677 retval = -EINVAL;
678 goto err_bindx_rem;
679 }
680
681 if (!sa_addr->v4.sin_port)
682 sa_addr->v4.sin_port = htons(bp->port);
683
684 /* FIXME - There is probably a need to check if sk->sk_saddr and
685 * sk->sk_rcv_addr are currently set to one of the addresses to
686 * be removed. This is something which needs to be looked into
687 * when we are fixing the outstanding issues with multi-homing
688 * socket routing and failover schemes. Refer to comments in
689 * sctp_do_bind(). -daisy
690 */
691 retval = sctp_del_bind_addr(bp, sa_addr);
692
693 addr_buf += af->sockaddr_len;
694 err_bindx_rem:
695 if (retval < 0) {
696 /* Failed. Add the ones that has been removed back */
697 if (cnt > 0)
698 sctp_bindx_add(sk, addrs, cnt);
699 return retval;
700 }
701 }
702
703 return retval;
704 }
705
706 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
707 * the associations that are part of the endpoint indicating that a list of
708 * local addresses are removed from the endpoint.
709 *
710 * If any of the addresses is already in the bind address list of the
711 * association, we do not send the chunk for that association. But it will not
712 * affect other associations.
713 *
714 * Only sctp_setsockopt_bindx() is supposed to call this function.
715 */
716 static int sctp_send_asconf_del_ip(struct sock *sk,
717 struct sockaddr *addrs,
718 int addrcnt)
719 {
720 struct net *net = sock_net(sk);
721 struct sctp_sock *sp;
722 struct sctp_endpoint *ep;
723 struct sctp_association *asoc;
724 struct sctp_transport *transport;
725 struct sctp_bind_addr *bp;
726 struct sctp_chunk *chunk;
727 union sctp_addr *laddr;
728 void *addr_buf;
729 struct sctp_af *af;
730 struct sctp_sockaddr_entry *saddr;
731 int i;
732 int retval = 0;
733 int stored = 0;
734
735 chunk = NULL;
736 if (!net->sctp.addip_enable)
737 return retval;
738
739 sp = sctp_sk(sk);
740 ep = sp->ep;
741
742 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
743 __func__, sk, addrs, addrcnt);
744
745 list_for_each_entry(asoc, &ep->asocs, asocs) {
746
747 if (!asoc->peer.asconf_capable)
748 continue;
749
750 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
751 continue;
752
753 if (!sctp_state(asoc, ESTABLISHED))
754 continue;
755
756 /* Check if any address in the packed array of addresses is
757 * not present in the bind address list of the association.
758 * If so, do not send the asconf chunk to its peer, but
759 * continue with other associations.
760 */
761 addr_buf = addrs;
762 for (i = 0; i < addrcnt; i++) {
763 laddr = addr_buf;
764 af = sctp_get_af_specific(laddr->v4.sin_family);
765 if (!af) {
766 retval = -EINVAL;
767 goto out;
768 }
769
770 if (!sctp_assoc_lookup_laddr(asoc, laddr))
771 break;
772
773 addr_buf += af->sockaddr_len;
774 }
775 if (i < addrcnt)
776 continue;
777
778 /* Find one address in the association's bind address list
779 * that is not in the packed array of addresses. This is to
780 * make sure that we do not delete all the addresses in the
781 * association.
782 */
783 bp = &asoc->base.bind_addr;
784 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
785 addrcnt, sp);
786 if ((laddr == NULL) && (addrcnt == 1)) {
787 if (asoc->asconf_addr_del_pending)
788 continue;
789 asoc->asconf_addr_del_pending =
790 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
791 if (asoc->asconf_addr_del_pending == NULL) {
792 retval = -ENOMEM;
793 goto out;
794 }
795 asoc->asconf_addr_del_pending->sa.sa_family =
796 addrs->sa_family;
797 asoc->asconf_addr_del_pending->v4.sin_port =
798 htons(bp->port);
799 if (addrs->sa_family == AF_INET) {
800 struct sockaddr_in *sin;
801
802 sin = (struct sockaddr_in *)addrs;
803 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
804 } else if (addrs->sa_family == AF_INET6) {
805 struct sockaddr_in6 *sin6;
806
807 sin6 = (struct sockaddr_in6 *)addrs;
808 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
809 }
810 SCTP_DEBUG_PRINTK_IPADDR("send_asconf_del_ip: keep the last address asoc: %p ",
811 " at %p\n", asoc, asoc->asconf_addr_del_pending,
812 asoc->asconf_addr_del_pending);
813 asoc->src_out_of_asoc_ok = 1;
814 stored = 1;
815 goto skip_mkasconf;
816 }
817
818 /* We do not need RCU protection throughout this loop
819 * because this is done under a socket lock from the
820 * setsockopt call.
821 */
822 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
823 SCTP_PARAM_DEL_IP);
824 if (!chunk) {
825 retval = -ENOMEM;
826 goto out;
827 }
828
829 skip_mkasconf:
830 /* Reset use_as_src flag for the addresses in the bind address
831 * list that are to be deleted.
832 */
833 addr_buf = addrs;
834 for (i = 0; i < addrcnt; i++) {
835 laddr = addr_buf;
836 af = sctp_get_af_specific(laddr->v4.sin_family);
837 list_for_each_entry(saddr, &bp->address_list, list) {
838 if (sctp_cmp_addr_exact(&saddr->a, laddr))
839 saddr->state = SCTP_ADDR_DEL;
840 }
841 addr_buf += af->sockaddr_len;
842 }
843
844 /* Update the route and saddr entries for all the transports
845 * as some of the addresses in the bind address list are
846 * about to be deleted and cannot be used as source addresses.
847 */
848 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
849 transports) {
850 dst_release(transport->dst);
851 sctp_transport_route(transport, NULL,
852 sctp_sk(asoc->base.sk));
853 }
854
855 if (stored)
856 /* We don't need to transmit ASCONF */
857 continue;
858 retval = sctp_send_asconf(asoc, chunk);
859 }
860 out:
861 return retval;
862 }
863
864 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
865 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
866 {
867 struct sock *sk = sctp_opt2sk(sp);
868 union sctp_addr *addr;
869 struct sctp_af *af;
870
871 /* It is safe to write port space in caller. */
872 addr = &addrw->a;
873 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
874 af = sctp_get_af_specific(addr->sa.sa_family);
875 if (!af)
876 return -EINVAL;
877 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
878 return -EINVAL;
879
880 if (addrw->state == SCTP_ADDR_NEW)
881 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
882 else
883 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
884 }
885
886 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
887 *
888 * API 8.1
889 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
890 * int flags);
891 *
892 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
893 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
894 * or IPv6 addresses.
895 *
896 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
897 * Section 3.1.2 for this usage.
898 *
899 * addrs is a pointer to an array of one or more socket addresses. Each
900 * address is contained in its appropriate structure (i.e. struct
901 * sockaddr_in or struct sockaddr_in6) the family of the address type
902 * must be used to distinguish the address length (note that this
903 * representation is termed a "packed array" of addresses). The caller
904 * specifies the number of addresses in the array with addrcnt.
905 *
906 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
907 * -1, and sets errno to the appropriate error code.
908 *
909 * For SCTP, the port given in each socket address must be the same, or
910 * sctp_bindx() will fail, setting errno to EINVAL.
911 *
912 * The flags parameter is formed from the bitwise OR of zero or more of
913 * the following currently defined flags:
914 *
915 * SCTP_BINDX_ADD_ADDR
916 *
917 * SCTP_BINDX_REM_ADDR
918 *
919 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
920 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
921 * addresses from the association. The two flags are mutually exclusive;
922 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
923 * not remove all addresses from an association; sctp_bindx() will
924 * reject such an attempt with EINVAL.
925 *
926 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
927 * additional addresses with an endpoint after calling bind(). Or use
928 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
929 * socket is associated with so that no new association accepted will be
930 * associated with those addresses. If the endpoint supports dynamic
931 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
932 * endpoint to send the appropriate message to the peer to change the
933 * peers address lists.
934 *
935 * Adding and removing addresses from a connected association is
936 * optional functionality. Implementations that do not support this
937 * functionality should return EOPNOTSUPP.
938 *
939 * Basically do nothing but copying the addresses from user to kernel
940 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
941 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
942 * from userspace.
943 *
944 * We don't use copy_from_user() for optimization: we first do the
945 * sanity checks (buffer size -fast- and access check-healthy
946 * pointer); if all of those succeed, then we can alloc the memory
947 * (expensive operation) needed to copy the data to kernel. Then we do
948 * the copying without checking the user space area
949 * (__copy_from_user()).
950 *
951 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
952 * it.
953 *
954 * sk The sk of the socket
955 * addrs The pointer to the addresses in user land
956 * addrssize Size of the addrs buffer
957 * op Operation to perform (add or remove, see the flags of
958 * sctp_bindx)
959 *
960 * Returns 0 if ok, <0 errno code on error.
961 */
962 static int sctp_setsockopt_bindx(struct sock* sk,
963 struct sockaddr __user *addrs,
964 int addrs_size, int op)
965 {
966 struct sockaddr *kaddrs;
967 int err;
968 int addrcnt = 0;
969 int walk_size = 0;
970 struct sockaddr *sa_addr;
971 void *addr_buf;
972 struct sctp_af *af;
973
974 SCTP_DEBUG_PRINTK("sctp_setsockopt_bindx: sk %p addrs %p"
975 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
976
977 if (unlikely(addrs_size <= 0))
978 return -EINVAL;
979
980 /* Check the user passed a healthy pointer. */
981 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
982 return -EFAULT;
983
984 /* Alloc space for the address array in kernel memory. */
985 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
986 if (unlikely(!kaddrs))
987 return -ENOMEM;
988
989 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
990 kfree(kaddrs);
991 return -EFAULT;
992 }
993
994 /* Walk through the addrs buffer and count the number of addresses. */
995 addr_buf = kaddrs;
996 while (walk_size < addrs_size) {
997 if (walk_size + sizeof(sa_family_t) > addrs_size) {
998 kfree(kaddrs);
999 return -EINVAL;
1000 }
1001
1002 sa_addr = addr_buf;
1003 af = sctp_get_af_specific(sa_addr->sa_family);
1004
1005 /* If the address family is not supported or if this address
1006 * causes the address buffer to overflow return EINVAL.
1007 */
1008 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1009 kfree(kaddrs);
1010 return -EINVAL;
1011 }
1012 addrcnt++;
1013 addr_buf += af->sockaddr_len;
1014 walk_size += af->sockaddr_len;
1015 }
1016
1017 /* Do the work. */
1018 switch (op) {
1019 case SCTP_BINDX_ADD_ADDR:
1020 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1021 if (err)
1022 goto out;
1023 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1024 break;
1025
1026 case SCTP_BINDX_REM_ADDR:
1027 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1028 if (err)
1029 goto out;
1030 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1031 break;
1032
1033 default:
1034 err = -EINVAL;
1035 break;
1036 }
1037
1038 out:
1039 kfree(kaddrs);
1040
1041 return err;
1042 }
1043
1044 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1045 *
1046 * Common routine for handling connect() and sctp_connectx().
1047 * Connect will come in with just a single address.
1048 */
1049 static int __sctp_connect(struct sock* sk,
1050 struct sockaddr *kaddrs,
1051 int addrs_size,
1052 sctp_assoc_t *assoc_id)
1053 {
1054 struct net *net = sock_net(sk);
1055 struct sctp_sock *sp;
1056 struct sctp_endpoint *ep;
1057 struct sctp_association *asoc = NULL;
1058 struct sctp_association *asoc2;
1059 struct sctp_transport *transport;
1060 union sctp_addr to;
1061 struct sctp_af *af;
1062 sctp_scope_t scope;
1063 long timeo;
1064 int err = 0;
1065 int addrcnt = 0;
1066 int walk_size = 0;
1067 union sctp_addr *sa_addr = NULL;
1068 void *addr_buf;
1069 unsigned short port;
1070 unsigned int f_flags = 0;
1071
1072 sp = sctp_sk(sk);
1073 ep = sp->ep;
1074
1075 /* connect() cannot be done on a socket that is already in ESTABLISHED
1076 * state - UDP-style peeled off socket or a TCP-style socket that
1077 * is already connected.
1078 * It cannot be done even on a TCP-style listening socket.
1079 */
1080 if (sctp_sstate(sk, ESTABLISHED) ||
1081 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1082 err = -EISCONN;
1083 goto out_free;
1084 }
1085
1086 /* Walk through the addrs buffer and count the number of addresses. */
1087 addr_buf = kaddrs;
1088 while (walk_size < addrs_size) {
1089 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1090 err = -EINVAL;
1091 goto out_free;
1092 }
1093
1094 sa_addr = addr_buf;
1095 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1096
1097 /* If the address family is not supported or if this address
1098 * causes the address buffer to overflow return EINVAL.
1099 */
1100 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1101 err = -EINVAL;
1102 goto out_free;
1103 }
1104
1105 port = ntohs(sa_addr->v4.sin_port);
1106
1107 /* Save current address so we can work with it */
1108 memcpy(&to, sa_addr, af->sockaddr_len);
1109
1110 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1111 if (err)
1112 goto out_free;
1113
1114 /* Make sure the destination port is correctly set
1115 * in all addresses.
1116 */
1117 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1118 err = -EINVAL;
1119 goto out_free;
1120 }
1121
1122 /* Check if there already is a matching association on the
1123 * endpoint (other than the one created here).
1124 */
1125 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1126 if (asoc2 && asoc2 != asoc) {
1127 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1128 err = -EISCONN;
1129 else
1130 err = -EALREADY;
1131 goto out_free;
1132 }
1133
1134 /* If we could not find a matching association on the endpoint,
1135 * make sure that there is no peeled-off association matching
1136 * the peer address even on another socket.
1137 */
1138 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1139 err = -EADDRNOTAVAIL;
1140 goto out_free;
1141 }
1142
1143 if (!asoc) {
1144 /* If a bind() or sctp_bindx() is not called prior to
1145 * an sctp_connectx() call, the system picks an
1146 * ephemeral port and will choose an address set
1147 * equivalent to binding with a wildcard address.
1148 */
1149 if (!ep->base.bind_addr.port) {
1150 if (sctp_autobind(sk)) {
1151 err = -EAGAIN;
1152 goto out_free;
1153 }
1154 } else {
1155 /*
1156 * If an unprivileged user inherits a 1-many
1157 * style socket with open associations on a
1158 * privileged port, it MAY be permitted to
1159 * accept new associations, but it SHOULD NOT
1160 * be permitted to open new associations.
1161 */
1162 if (ep->base.bind_addr.port < PROT_SOCK &&
1163 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1164 err = -EACCES;
1165 goto out_free;
1166 }
1167 }
1168
1169 scope = sctp_scope(&to);
1170 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1171 if (!asoc) {
1172 err = -ENOMEM;
1173 goto out_free;
1174 }
1175
1176 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1177 GFP_KERNEL);
1178 if (err < 0) {
1179 goto out_free;
1180 }
1181
1182 }
1183
1184 /* Prime the peer's transport structures. */
1185 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1186 SCTP_UNKNOWN);
1187 if (!transport) {
1188 err = -ENOMEM;
1189 goto out_free;
1190 }
1191
1192 addrcnt++;
1193 addr_buf += af->sockaddr_len;
1194 walk_size += af->sockaddr_len;
1195 }
1196
1197 /* In case the user of sctp_connectx() wants an association
1198 * id back, assign one now.
1199 */
1200 if (assoc_id) {
1201 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1202 if (err < 0)
1203 goto out_free;
1204 }
1205
1206 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1207 if (err < 0) {
1208 goto out_free;
1209 }
1210
1211 /* Initialize sk's dport and daddr for getpeername() */
1212 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1213 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1214 af->to_sk_daddr(sa_addr, sk);
1215 sk->sk_err = 0;
1216
1217 /* in-kernel sockets don't generally have a file allocated to them
1218 * if all they do is call sock_create_kern().
1219 */
1220 if (sk->sk_socket->file)
1221 f_flags = sk->sk_socket->file->f_flags;
1222
1223 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1224
1225 err = sctp_wait_for_connect(asoc, &timeo);
1226 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1227 *assoc_id = asoc->assoc_id;
1228
1229 /* Don't free association on exit. */
1230 asoc = NULL;
1231
1232 out_free:
1233
1234 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1235 " kaddrs: %p err: %d\n",
1236 asoc, kaddrs, err);
1237 if (asoc) {
1238 /* sctp_primitive_ASSOCIATE may have added this association
1239 * To the hash table, try to unhash it, just in case, its a noop
1240 * if it wasn't hashed so we're safe
1241 */
1242 sctp_unhash_established(asoc);
1243 sctp_association_free(asoc);
1244 }
1245 return err;
1246 }
1247
1248 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1249 *
1250 * API 8.9
1251 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1252 * sctp_assoc_t *asoc);
1253 *
1254 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1255 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1256 * or IPv6 addresses.
1257 *
1258 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1259 * Section 3.1.2 for this usage.
1260 *
1261 * addrs is a pointer to an array of one or more socket addresses. Each
1262 * address is contained in its appropriate structure (i.e. struct
1263 * sockaddr_in or struct sockaddr_in6) the family of the address type
1264 * must be used to distengish the address length (note that this
1265 * representation is termed a "packed array" of addresses). The caller
1266 * specifies the number of addresses in the array with addrcnt.
1267 *
1268 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1269 * the association id of the new association. On failure, sctp_connectx()
1270 * returns -1, and sets errno to the appropriate error code. The assoc_id
1271 * is not touched by the kernel.
1272 *
1273 * For SCTP, the port given in each socket address must be the same, or
1274 * sctp_connectx() will fail, setting errno to EINVAL.
1275 *
1276 * An application can use sctp_connectx to initiate an association with
1277 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1278 * allows a caller to specify multiple addresses at which a peer can be
1279 * reached. The way the SCTP stack uses the list of addresses to set up
1280 * the association is implementation dependent. This function only
1281 * specifies that the stack will try to make use of all the addresses in
1282 * the list when needed.
1283 *
1284 * Note that the list of addresses passed in is only used for setting up
1285 * the association. It does not necessarily equal the set of addresses
1286 * the peer uses for the resulting association. If the caller wants to
1287 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1288 * retrieve them after the association has been set up.
1289 *
1290 * Basically do nothing but copying the addresses from user to kernel
1291 * land and invoking either sctp_connectx(). This is used for tunneling
1292 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1293 *
1294 * We don't use copy_from_user() for optimization: we first do the
1295 * sanity checks (buffer size -fast- and access check-healthy
1296 * pointer); if all of those succeed, then we can alloc the memory
1297 * (expensive operation) needed to copy the data to kernel. Then we do
1298 * the copying without checking the user space area
1299 * (__copy_from_user()).
1300 *
1301 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1302 * it.
1303 *
1304 * sk The sk of the socket
1305 * addrs The pointer to the addresses in user land
1306 * addrssize Size of the addrs buffer
1307 *
1308 * Returns >=0 if ok, <0 errno code on error.
1309 */
1310 static int __sctp_setsockopt_connectx(struct sock* sk,
1311 struct sockaddr __user *addrs,
1312 int addrs_size,
1313 sctp_assoc_t *assoc_id)
1314 {
1315 int err = 0;
1316 struct sockaddr *kaddrs;
1317
1318 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1319 __func__, sk, addrs, addrs_size);
1320
1321 if (unlikely(addrs_size <= 0))
1322 return -EINVAL;
1323
1324 /* Check the user passed a healthy pointer. */
1325 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1326 return -EFAULT;
1327
1328 /* Alloc space for the address array in kernel memory. */
1329 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1330 if (unlikely(!kaddrs))
1331 return -ENOMEM;
1332
1333 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1334 err = -EFAULT;
1335 } else {
1336 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1337 }
1338
1339 kfree(kaddrs);
1340
1341 return err;
1342 }
1343
1344 /*
1345 * This is an older interface. It's kept for backward compatibility
1346 * to the option that doesn't provide association id.
1347 */
1348 static int sctp_setsockopt_connectx_old(struct sock* sk,
1349 struct sockaddr __user *addrs,
1350 int addrs_size)
1351 {
1352 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1353 }
1354
1355 /*
1356 * New interface for the API. The since the API is done with a socket
1357 * option, to make it simple we feed back the association id is as a return
1358 * indication to the call. Error is always negative and association id is
1359 * always positive.
1360 */
1361 static int sctp_setsockopt_connectx(struct sock* sk,
1362 struct sockaddr __user *addrs,
1363 int addrs_size)
1364 {
1365 sctp_assoc_t assoc_id = 0;
1366 int err = 0;
1367
1368 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1369
1370 if (err)
1371 return err;
1372 else
1373 return assoc_id;
1374 }
1375
1376 /*
1377 * New (hopefully final) interface for the API.
1378 * We use the sctp_getaddrs_old structure so that use-space library
1379 * can avoid any unnecessary allocations. The only defferent part
1380 * is that we store the actual length of the address buffer into the
1381 * addrs_num structure member. That way we can re-use the existing
1382 * code.
1383 */
1384 static int sctp_getsockopt_connectx3(struct sock* sk, int len,
1385 char __user *optval,
1386 int __user *optlen)
1387 {
1388 struct sctp_getaddrs_old param;
1389 sctp_assoc_t assoc_id = 0;
1390 int err = 0;
1391
1392 if (len < sizeof(param))
1393 return -EINVAL;
1394
1395 if (copy_from_user(&param, optval, sizeof(param)))
1396 return -EFAULT;
1397
1398 err = __sctp_setsockopt_connectx(sk,
1399 (struct sockaddr __user *)param.addrs,
1400 param.addr_num, &assoc_id);
1401
1402 if (err == 0 || err == -EINPROGRESS) {
1403 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1404 return -EFAULT;
1405 if (put_user(sizeof(assoc_id), optlen))
1406 return -EFAULT;
1407 }
1408
1409 return err;
1410 }
1411
1412 /* API 3.1.4 close() - UDP Style Syntax
1413 * Applications use close() to perform graceful shutdown (as described in
1414 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1415 * by a UDP-style socket.
1416 *
1417 * The syntax is
1418 *
1419 * ret = close(int sd);
1420 *
1421 * sd - the socket descriptor of the associations to be closed.
1422 *
1423 * To gracefully shutdown a specific association represented by the
1424 * UDP-style socket, an application should use the sendmsg() call,
1425 * passing no user data, but including the appropriate flag in the
1426 * ancillary data (see Section xxxx).
1427 *
1428 * If sd in the close() call is a branched-off socket representing only
1429 * one association, the shutdown is performed on that association only.
1430 *
1431 * 4.1.6 close() - TCP Style Syntax
1432 *
1433 * Applications use close() to gracefully close down an association.
1434 *
1435 * The syntax is:
1436 *
1437 * int close(int sd);
1438 *
1439 * sd - the socket descriptor of the association to be closed.
1440 *
1441 * After an application calls close() on a socket descriptor, no further
1442 * socket operations will succeed on that descriptor.
1443 *
1444 * API 7.1.4 SO_LINGER
1445 *
1446 * An application using the TCP-style socket can use this option to
1447 * perform the SCTP ABORT primitive. The linger option structure is:
1448 *
1449 * struct linger {
1450 * int l_onoff; // option on/off
1451 * int l_linger; // linger time
1452 * };
1453 *
1454 * To enable the option, set l_onoff to 1. If the l_linger value is set
1455 * to 0, calling close() is the same as the ABORT primitive. If the
1456 * value is set to a negative value, the setsockopt() call will return
1457 * an error. If the value is set to a positive value linger_time, the
1458 * close() can be blocked for at most linger_time ms. If the graceful
1459 * shutdown phase does not finish during this period, close() will
1460 * return but the graceful shutdown phase continues in the system.
1461 */
1462 static void sctp_close(struct sock *sk, long timeout)
1463 {
1464 struct net *net = sock_net(sk);
1465 struct sctp_endpoint *ep;
1466 struct sctp_association *asoc;
1467 struct list_head *pos, *temp;
1468 unsigned int data_was_unread;
1469
1470 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1471
1472 sctp_lock_sock(sk);
1473 sk->sk_shutdown = SHUTDOWN_MASK;
1474 sk->sk_state = SCTP_SS_CLOSING;
1475
1476 ep = sctp_sk(sk)->ep;
1477
1478 /* Clean up any skbs sitting on the receive queue. */
1479 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1480 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1481
1482 /* Walk all associations on an endpoint. */
1483 list_for_each_safe(pos, temp, &ep->asocs) {
1484 asoc = list_entry(pos, struct sctp_association, asocs);
1485
1486 if (sctp_style(sk, TCP)) {
1487 /* A closed association can still be in the list if
1488 * it belongs to a TCP-style listening socket that is
1489 * not yet accepted. If so, free it. If not, send an
1490 * ABORT or SHUTDOWN based on the linger options.
1491 */
1492 if (sctp_state(asoc, CLOSED)) {
1493 sctp_unhash_established(asoc);
1494 sctp_association_free(asoc);
1495 continue;
1496 }
1497 }
1498
1499 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1500 !skb_queue_empty(&asoc->ulpq.reasm) ||
1501 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1502 struct sctp_chunk *chunk;
1503
1504 chunk = sctp_make_abort_user(asoc, NULL, 0);
1505 if (chunk)
1506 sctp_primitive_ABORT(net, asoc, chunk);
1507 } else
1508 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1509 }
1510
1511 /* On a TCP-style socket, block for at most linger_time if set. */
1512 if (sctp_style(sk, TCP) && timeout)
1513 sctp_wait_for_close(sk, timeout);
1514
1515 /* This will run the backlog queue. */
1516 sctp_release_sock(sk);
1517
1518 /* Supposedly, no process has access to the socket, but
1519 * the net layers still may.
1520 */
1521 sctp_local_bh_disable();
1522 sctp_bh_lock_sock(sk);
1523
1524 /* Hold the sock, since sk_common_release() will put sock_put()
1525 * and we have just a little more cleanup.
1526 */
1527 sock_hold(sk);
1528 sk_common_release(sk);
1529
1530 sctp_bh_unlock_sock(sk);
1531 sctp_local_bh_enable();
1532
1533 sock_put(sk);
1534
1535 SCTP_DBG_OBJCNT_DEC(sock);
1536 }
1537
1538 /* Handle EPIPE error. */
1539 static int sctp_error(struct sock *sk, int flags, int err)
1540 {
1541 if (err == -EPIPE)
1542 err = sock_error(sk) ? : -EPIPE;
1543 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1544 send_sig(SIGPIPE, current, 0);
1545 return err;
1546 }
1547
1548 /* API 3.1.3 sendmsg() - UDP Style Syntax
1549 *
1550 * An application uses sendmsg() and recvmsg() calls to transmit data to
1551 * and receive data from its peer.
1552 *
1553 * ssize_t sendmsg(int socket, const struct msghdr *message,
1554 * int flags);
1555 *
1556 * socket - the socket descriptor of the endpoint.
1557 * message - pointer to the msghdr structure which contains a single
1558 * user message and possibly some ancillary data.
1559 *
1560 * See Section 5 for complete description of the data
1561 * structures.
1562 *
1563 * flags - flags sent or received with the user message, see Section
1564 * 5 for complete description of the flags.
1565 *
1566 * Note: This function could use a rewrite especially when explicit
1567 * connect support comes in.
1568 */
1569 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1570
1571 static int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1572
1573 static int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1574 struct msghdr *msg, size_t msg_len)
1575 {
1576 struct net *net = sock_net(sk);
1577 struct sctp_sock *sp;
1578 struct sctp_endpoint *ep;
1579 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1580 struct sctp_transport *transport, *chunk_tp;
1581 struct sctp_chunk *chunk;
1582 union sctp_addr to;
1583 struct sockaddr *msg_name = NULL;
1584 struct sctp_sndrcvinfo default_sinfo;
1585 struct sctp_sndrcvinfo *sinfo;
1586 struct sctp_initmsg *sinit;
1587 sctp_assoc_t associd = 0;
1588 sctp_cmsgs_t cmsgs = { NULL };
1589 int err;
1590 sctp_scope_t scope;
1591 long timeo;
1592 __u16 sinfo_flags = 0;
1593 struct sctp_datamsg *datamsg;
1594 int msg_flags = msg->msg_flags;
1595
1596 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1597 sk, msg, msg_len);
1598
1599 err = 0;
1600 sp = sctp_sk(sk);
1601 ep = sp->ep;
1602
1603 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1604
1605 /* We cannot send a message over a TCP-style listening socket. */
1606 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1607 err = -EPIPE;
1608 goto out_nounlock;
1609 }
1610
1611 /* Parse out the SCTP CMSGs. */
1612 err = sctp_msghdr_parse(msg, &cmsgs);
1613
1614 if (err) {
1615 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1616 goto out_nounlock;
1617 }
1618
1619 /* Fetch the destination address for this packet. This
1620 * address only selects the association--it is not necessarily
1621 * the address we will send to.
1622 * For a peeled-off socket, msg_name is ignored.
1623 */
1624 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1625 int msg_namelen = msg->msg_namelen;
1626
1627 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1628 msg_namelen);
1629 if (err)
1630 return err;
1631
1632 if (msg_namelen > sizeof(to))
1633 msg_namelen = sizeof(to);
1634 memcpy(&to, msg->msg_name, msg_namelen);
1635 msg_name = msg->msg_name;
1636 }
1637
1638 sinfo = cmsgs.info;
1639 sinit = cmsgs.init;
1640
1641 /* Did the user specify SNDRCVINFO? */
1642 if (sinfo) {
1643 sinfo_flags = sinfo->sinfo_flags;
1644 associd = sinfo->sinfo_assoc_id;
1645 }
1646
1647 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1648 msg_len, sinfo_flags);
1649
1650 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1651 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1652 err = -EINVAL;
1653 goto out_nounlock;
1654 }
1655
1656 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1657 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1658 * If SCTP_ABORT is set, the message length could be non zero with
1659 * the msg_iov set to the user abort reason.
1660 */
1661 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1662 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1663 err = -EINVAL;
1664 goto out_nounlock;
1665 }
1666
1667 /* If SCTP_ADDR_OVER is set, there must be an address
1668 * specified in msg_name.
1669 */
1670 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1671 err = -EINVAL;
1672 goto out_nounlock;
1673 }
1674
1675 transport = NULL;
1676
1677 SCTP_DEBUG_PRINTK("About to look up association.\n");
1678
1679 sctp_lock_sock(sk);
1680
1681 /* If a msg_name has been specified, assume this is to be used. */
1682 if (msg_name) {
1683 /* Look for a matching association on the endpoint. */
1684 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1685 if (!asoc) {
1686 /* If we could not find a matching association on the
1687 * endpoint, make sure that it is not a TCP-style
1688 * socket that already has an association or there is
1689 * no peeled-off association on another socket.
1690 */
1691 if ((sctp_style(sk, TCP) &&
1692 sctp_sstate(sk, ESTABLISHED)) ||
1693 sctp_endpoint_is_peeled_off(ep, &to)) {
1694 err = -EADDRNOTAVAIL;
1695 goto out_unlock;
1696 }
1697 }
1698 } else {
1699 asoc = sctp_id2assoc(sk, associd);
1700 if (!asoc) {
1701 err = -EPIPE;
1702 goto out_unlock;
1703 }
1704 }
1705
1706 if (asoc) {
1707 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1708
1709 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1710 * socket that has an association in CLOSED state. This can
1711 * happen when an accepted socket has an association that is
1712 * already CLOSED.
1713 */
1714 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1715 err = -EPIPE;
1716 goto out_unlock;
1717 }
1718
1719 if (sinfo_flags & SCTP_EOF) {
1720 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1721 asoc);
1722 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1723 err = 0;
1724 goto out_unlock;
1725 }
1726 if (sinfo_flags & SCTP_ABORT) {
1727
1728 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1729 if (!chunk) {
1730 err = -ENOMEM;
1731 goto out_unlock;
1732 }
1733
1734 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1735 sctp_primitive_ABORT(net, asoc, chunk);
1736 err = 0;
1737 goto out_unlock;
1738 }
1739 }
1740
1741 /* Do we need to create the association? */
1742 if (!asoc) {
1743 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1744
1745 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1746 err = -EINVAL;
1747 goto out_unlock;
1748 }
1749
1750 /* Check for invalid stream against the stream counts,
1751 * either the default or the user specified stream counts.
1752 */
1753 if (sinfo) {
1754 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1755 /* Check against the defaults. */
1756 if (sinfo->sinfo_stream >=
1757 sp->initmsg.sinit_num_ostreams) {
1758 err = -EINVAL;
1759 goto out_unlock;
1760 }
1761 } else {
1762 /* Check against the requested. */
1763 if (sinfo->sinfo_stream >=
1764 sinit->sinit_num_ostreams) {
1765 err = -EINVAL;
1766 goto out_unlock;
1767 }
1768 }
1769 }
1770
1771 /*
1772 * API 3.1.2 bind() - UDP Style Syntax
1773 * If a bind() or sctp_bindx() is not called prior to a
1774 * sendmsg() call that initiates a new association, the
1775 * system picks an ephemeral port and will choose an address
1776 * set equivalent to binding with a wildcard address.
1777 */
1778 if (!ep->base.bind_addr.port) {
1779 if (sctp_autobind(sk)) {
1780 err = -EAGAIN;
1781 goto out_unlock;
1782 }
1783 } else {
1784 /*
1785 * If an unprivileged user inherits a one-to-many
1786 * style socket with open associations on a privileged
1787 * port, it MAY be permitted to accept new associations,
1788 * but it SHOULD NOT be permitted to open new
1789 * associations.
1790 */
1791 if (ep->base.bind_addr.port < PROT_SOCK &&
1792 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1793 err = -EACCES;
1794 goto out_unlock;
1795 }
1796 }
1797
1798 scope = sctp_scope(&to);
1799 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1800 if (!new_asoc) {
1801 err = -ENOMEM;
1802 goto out_unlock;
1803 }
1804 asoc = new_asoc;
1805 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1806 if (err < 0) {
1807 err = -ENOMEM;
1808 goto out_free;
1809 }
1810
1811 /* If the SCTP_INIT ancillary data is specified, set all
1812 * the association init values accordingly.
1813 */
1814 if (sinit) {
1815 if (sinit->sinit_num_ostreams) {
1816 asoc->c.sinit_num_ostreams =
1817 sinit->sinit_num_ostreams;
1818 }
1819 if (sinit->sinit_max_instreams) {
1820 asoc->c.sinit_max_instreams =
1821 sinit->sinit_max_instreams;
1822 }
1823 if (sinit->sinit_max_attempts) {
1824 asoc->max_init_attempts
1825 = sinit->sinit_max_attempts;
1826 }
1827 if (sinit->sinit_max_init_timeo) {
1828 asoc->max_init_timeo =
1829 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1830 }
1831 }
1832
1833 /* Prime the peer's transport structures. */
1834 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1835 if (!transport) {
1836 err = -ENOMEM;
1837 goto out_free;
1838 }
1839 }
1840
1841 /* ASSERT: we have a valid association at this point. */
1842 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1843
1844 if (!sinfo) {
1845 /* If the user didn't specify SNDRCVINFO, make up one with
1846 * some defaults.
1847 */
1848 memset(&default_sinfo, 0, sizeof(default_sinfo));
1849 default_sinfo.sinfo_stream = asoc->default_stream;
1850 default_sinfo.sinfo_flags = asoc->default_flags;
1851 default_sinfo.sinfo_ppid = asoc->default_ppid;
1852 default_sinfo.sinfo_context = asoc->default_context;
1853 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1854 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1855 sinfo = &default_sinfo;
1856 }
1857
1858 /* API 7.1.7, the sndbuf size per association bounds the
1859 * maximum size of data that can be sent in a single send call.
1860 */
1861 if (msg_len > sk->sk_sndbuf) {
1862 err = -EMSGSIZE;
1863 goto out_free;
1864 }
1865
1866 if (asoc->pmtu_pending)
1867 sctp_assoc_pending_pmtu(sk, asoc);
1868
1869 /* If fragmentation is disabled and the message length exceeds the
1870 * association fragmentation point, return EMSGSIZE. The I-D
1871 * does not specify what this error is, but this looks like
1872 * a great fit.
1873 */
1874 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1875 err = -EMSGSIZE;
1876 goto out_free;
1877 }
1878
1879 /* Check for invalid stream. */
1880 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1881 err = -EINVAL;
1882 goto out_free;
1883 }
1884
1885 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1886 if (!sctp_wspace(asoc)) {
1887 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1888 if (err)
1889 goto out_free;
1890 }
1891
1892 /* If an address is passed with the sendto/sendmsg call, it is used
1893 * to override the primary destination address in the TCP model, or
1894 * when SCTP_ADDR_OVER flag is set in the UDP model.
1895 */
1896 if ((sctp_style(sk, TCP) && msg_name) ||
1897 (sinfo_flags & SCTP_ADDR_OVER)) {
1898 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1899 if (!chunk_tp) {
1900 err = -EINVAL;
1901 goto out_free;
1902 }
1903 } else
1904 chunk_tp = NULL;
1905
1906 /* Auto-connect, if we aren't connected already. */
1907 if (sctp_state(asoc, CLOSED)) {
1908 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1909 if (err < 0)
1910 goto out_free;
1911 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1912 }
1913
1914 /* Break the message into multiple chunks of maximum size. */
1915 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1916 if (IS_ERR(datamsg)) {
1917 err = PTR_ERR(datamsg);
1918 goto out_free;
1919 }
1920
1921 /* Now send the (possibly) fragmented message. */
1922 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1923 sctp_chunk_hold(chunk);
1924
1925 /* Do accounting for the write space. */
1926 sctp_set_owner_w(chunk);
1927
1928 chunk->transport = chunk_tp;
1929 }
1930
1931 /* Send it to the lower layers. Note: all chunks
1932 * must either fail or succeed. The lower layer
1933 * works that way today. Keep it that way or this
1934 * breaks.
1935 */
1936 err = sctp_primitive_SEND(net, asoc, datamsg);
1937 /* Did the lower layer accept the chunk? */
1938 if (err)
1939 sctp_datamsg_free(datamsg);
1940 else
1941 sctp_datamsg_put(datamsg);
1942
1943 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1944
1945 if (err)
1946 goto out_free;
1947 else
1948 err = msg_len;
1949
1950 /* If we are already past ASSOCIATE, the lower
1951 * layers are responsible for association cleanup.
1952 */
1953 goto out_unlock;
1954
1955 out_free:
1956 if (new_asoc) {
1957 sctp_unhash_established(asoc);
1958 sctp_association_free(asoc);
1959 }
1960 out_unlock:
1961 sctp_release_sock(sk);
1962
1963 out_nounlock:
1964 return sctp_error(sk, msg_flags, err);
1965
1966 #if 0
1967 do_sock_err:
1968 if (msg_len)
1969 err = msg_len;
1970 else
1971 err = sock_error(sk);
1972 goto out;
1973
1974 do_interrupted:
1975 if (msg_len)
1976 err = msg_len;
1977 goto out;
1978 #endif /* 0 */
1979 }
1980
1981 /* This is an extended version of skb_pull() that removes the data from the
1982 * start of a skb even when data is spread across the list of skb's in the
1983 * frag_list. len specifies the total amount of data that needs to be removed.
1984 * when 'len' bytes could be removed from the skb, it returns 0.
1985 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1986 * could not be removed.
1987 */
1988 static int sctp_skb_pull(struct sk_buff *skb, int len)
1989 {
1990 struct sk_buff *list;
1991 int skb_len = skb_headlen(skb);
1992 int rlen;
1993
1994 if (len <= skb_len) {
1995 __skb_pull(skb, len);
1996 return 0;
1997 }
1998 len -= skb_len;
1999 __skb_pull(skb, skb_len);
2000
2001 skb_walk_frags(skb, list) {
2002 rlen = sctp_skb_pull(list, len);
2003 skb->len -= (len-rlen);
2004 skb->data_len -= (len-rlen);
2005
2006 if (!rlen)
2007 return 0;
2008
2009 len = rlen;
2010 }
2011
2012 return len;
2013 }
2014
2015 /* API 3.1.3 recvmsg() - UDP Style Syntax
2016 *
2017 * ssize_t recvmsg(int socket, struct msghdr *message,
2018 * int flags);
2019 *
2020 * socket - the socket descriptor of the endpoint.
2021 * message - pointer to the msghdr structure which contains a single
2022 * user message and possibly some ancillary data.
2023 *
2024 * See Section 5 for complete description of the data
2025 * structures.
2026 *
2027 * flags - flags sent or received with the user message, see Section
2028 * 5 for complete description of the flags.
2029 */
2030 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2031
2032 static int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2033 struct msghdr *msg, size_t len, int noblock,
2034 int flags, int *addr_len)
2035 {
2036 struct sctp_ulpevent *event = NULL;
2037 struct sctp_sock *sp = sctp_sk(sk);
2038 struct sk_buff *skb;
2039 int copied;
2040 int err = 0;
2041 int skb_len;
2042
2043 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
2044 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
2045 "len", len, "knoblauch", noblock,
2046 "flags", flags, "addr_len", addr_len);
2047
2048 sctp_lock_sock(sk);
2049
2050 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2051 err = -ENOTCONN;
2052 goto out;
2053 }
2054
2055 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2056 if (!skb)
2057 goto out;
2058
2059 /* Get the total length of the skb including any skb's in the
2060 * frag_list.
2061 */
2062 skb_len = skb->len;
2063
2064 copied = skb_len;
2065 if (copied > len)
2066 copied = len;
2067
2068 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2069
2070 event = sctp_skb2event(skb);
2071
2072 if (err)
2073 goto out_free;
2074
2075 sock_recv_ts_and_drops(msg, sk, skb);
2076 if (sctp_ulpevent_is_notification(event)) {
2077 msg->msg_flags |= MSG_NOTIFICATION;
2078 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2079 } else {
2080 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2081 }
2082
2083 /* Check if we allow SCTP_SNDRCVINFO. */
2084 if (sp->subscribe.sctp_data_io_event)
2085 sctp_ulpevent_read_sndrcvinfo(event, msg);
2086 #if 0
2087 /* FIXME: we should be calling IP/IPv6 layers. */
2088 if (sk->sk_protinfo.af_inet.cmsg_flags)
2089 ip_cmsg_recv(msg, skb);
2090 #endif
2091
2092 err = copied;
2093
2094 /* If skb's length exceeds the user's buffer, update the skb and
2095 * push it back to the receive_queue so that the next call to
2096 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2097 */
2098 if (skb_len > copied) {
2099 msg->msg_flags &= ~MSG_EOR;
2100 if (flags & MSG_PEEK)
2101 goto out_free;
2102 sctp_skb_pull(skb, copied);
2103 skb_queue_head(&sk->sk_receive_queue, skb);
2104
2105 /* When only partial message is copied to the user, increase
2106 * rwnd by that amount. If all the data in the skb is read,
2107 * rwnd is updated when the event is freed.
2108 */
2109 if (!sctp_ulpevent_is_notification(event))
2110 sctp_assoc_rwnd_increase(event->asoc, copied);
2111 goto out;
2112 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2113 (event->msg_flags & MSG_EOR))
2114 msg->msg_flags |= MSG_EOR;
2115 else
2116 msg->msg_flags &= ~MSG_EOR;
2117
2118 out_free:
2119 if (flags & MSG_PEEK) {
2120 /* Release the skb reference acquired after peeking the skb in
2121 * sctp_skb_recv_datagram().
2122 */
2123 kfree_skb(skb);
2124 } else {
2125 /* Free the event which includes releasing the reference to
2126 * the owner of the skb, freeing the skb and updating the
2127 * rwnd.
2128 */
2129 sctp_ulpevent_free(event);
2130 }
2131 out:
2132 sctp_release_sock(sk);
2133 return err;
2134 }
2135
2136 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2137 *
2138 * This option is a on/off flag. If enabled no SCTP message
2139 * fragmentation will be performed. Instead if a message being sent
2140 * exceeds the current PMTU size, the message will NOT be sent and
2141 * instead a error will be indicated to the user.
2142 */
2143 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2144 char __user *optval,
2145 unsigned int optlen)
2146 {
2147 int val;
2148
2149 if (optlen < sizeof(int))
2150 return -EINVAL;
2151
2152 if (get_user(val, (int __user *)optval))
2153 return -EFAULT;
2154
2155 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2156
2157 return 0;
2158 }
2159
2160 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2161 unsigned int optlen)
2162 {
2163 struct sctp_association *asoc;
2164 struct sctp_ulpevent *event;
2165
2166 if (optlen > sizeof(struct sctp_event_subscribe))
2167 return -EINVAL;
2168 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2169 return -EFAULT;
2170
2171 /*
2172 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2173 * if there is no data to be sent or retransmit, the stack will
2174 * immediately send up this notification.
2175 */
2176 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2177 &sctp_sk(sk)->subscribe)) {
2178 asoc = sctp_id2assoc(sk, 0);
2179
2180 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2181 event = sctp_ulpevent_make_sender_dry_event(asoc,
2182 GFP_ATOMIC);
2183 if (!event)
2184 return -ENOMEM;
2185
2186 sctp_ulpq_tail_event(&asoc->ulpq, event);
2187 }
2188 }
2189
2190 return 0;
2191 }
2192
2193 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2194 *
2195 * This socket option is applicable to the UDP-style socket only. When
2196 * set it will cause associations that are idle for more than the
2197 * specified number of seconds to automatically close. An association
2198 * being idle is defined an association that has NOT sent or received
2199 * user data. The special value of '0' indicates that no automatic
2200 * close of any associations should be performed. The option expects an
2201 * integer defining the number of seconds of idle time before an
2202 * association is closed.
2203 */
2204 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2205 unsigned int optlen)
2206 {
2207 struct sctp_sock *sp = sctp_sk(sk);
2208
2209 /* Applicable to UDP-style socket only */
2210 if (sctp_style(sk, TCP))
2211 return -EOPNOTSUPP;
2212 if (optlen != sizeof(int))
2213 return -EINVAL;
2214 if (copy_from_user(&sp->autoclose, optval, optlen))
2215 return -EFAULT;
2216
2217 return 0;
2218 }
2219
2220 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2221 *
2222 * Applications can enable or disable heartbeats for any peer address of
2223 * an association, modify an address's heartbeat interval, force a
2224 * heartbeat to be sent immediately, and adjust the address's maximum
2225 * number of retransmissions sent before an address is considered
2226 * unreachable. The following structure is used to access and modify an
2227 * address's parameters:
2228 *
2229 * struct sctp_paddrparams {
2230 * sctp_assoc_t spp_assoc_id;
2231 * struct sockaddr_storage spp_address;
2232 * uint32_t spp_hbinterval;
2233 * uint16_t spp_pathmaxrxt;
2234 * uint32_t spp_pathmtu;
2235 * uint32_t spp_sackdelay;
2236 * uint32_t spp_flags;
2237 * };
2238 *
2239 * spp_assoc_id - (one-to-many style socket) This is filled in the
2240 * application, and identifies the association for
2241 * this query.
2242 * spp_address - This specifies which address is of interest.
2243 * spp_hbinterval - This contains the value of the heartbeat interval,
2244 * in milliseconds. If a value of zero
2245 * is present in this field then no changes are to
2246 * be made to this parameter.
2247 * spp_pathmaxrxt - This contains the maximum number of
2248 * retransmissions before this address shall be
2249 * considered unreachable. If a value of zero
2250 * is present in this field then no changes are to
2251 * be made to this parameter.
2252 * spp_pathmtu - When Path MTU discovery is disabled the value
2253 * specified here will be the "fixed" path mtu.
2254 * Note that if the spp_address field is empty
2255 * then all associations on this address will
2256 * have this fixed path mtu set upon them.
2257 *
2258 * spp_sackdelay - When delayed sack is enabled, this value specifies
2259 * the number of milliseconds that sacks will be delayed
2260 * for. This value will apply to all addresses of an
2261 * association if the spp_address field is empty. Note
2262 * also, that if delayed sack is enabled and this
2263 * value is set to 0, no change is made to the last
2264 * recorded delayed sack timer value.
2265 *
2266 * spp_flags - These flags are used to control various features
2267 * on an association. The flag field may contain
2268 * zero or more of the following options.
2269 *
2270 * SPP_HB_ENABLE - Enable heartbeats on the
2271 * specified address. Note that if the address
2272 * field is empty all addresses for the association
2273 * have heartbeats enabled upon them.
2274 *
2275 * SPP_HB_DISABLE - Disable heartbeats on the
2276 * speicifed address. Note that if the address
2277 * field is empty all addresses for the association
2278 * will have their heartbeats disabled. Note also
2279 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2280 * mutually exclusive, only one of these two should
2281 * be specified. Enabling both fields will have
2282 * undetermined results.
2283 *
2284 * SPP_HB_DEMAND - Request a user initiated heartbeat
2285 * to be made immediately.
2286 *
2287 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2288 * heartbeat delayis to be set to the value of 0
2289 * milliseconds.
2290 *
2291 * SPP_PMTUD_ENABLE - This field will enable PMTU
2292 * discovery upon the specified address. Note that
2293 * if the address feild is empty then all addresses
2294 * on the association are effected.
2295 *
2296 * SPP_PMTUD_DISABLE - This field will disable PMTU
2297 * discovery upon the specified address. Note that
2298 * if the address feild is empty then all addresses
2299 * on the association are effected. Not also that
2300 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2301 * exclusive. Enabling both will have undetermined
2302 * results.
2303 *
2304 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2305 * on delayed sack. The time specified in spp_sackdelay
2306 * is used to specify the sack delay for this address. Note
2307 * that if spp_address is empty then all addresses will
2308 * enable delayed sack and take on the sack delay
2309 * value specified in spp_sackdelay.
2310 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2311 * off delayed sack. If the spp_address field is blank then
2312 * delayed sack is disabled for the entire association. Note
2313 * also that this field is mutually exclusive to
2314 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2315 * results.
2316 */
2317 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2318 struct sctp_transport *trans,
2319 struct sctp_association *asoc,
2320 struct sctp_sock *sp,
2321 int hb_change,
2322 int pmtud_change,
2323 int sackdelay_change)
2324 {
2325 int error;
2326
2327 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2328 struct net *net = sock_net(trans->asoc->base.sk);
2329
2330 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2331 if (error)
2332 return error;
2333 }
2334
2335 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2336 * this field is ignored. Note also that a value of zero indicates
2337 * the current setting should be left unchanged.
2338 */
2339 if (params->spp_flags & SPP_HB_ENABLE) {
2340
2341 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2342 * set. This lets us use 0 value when this flag
2343 * is set.
2344 */
2345 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2346 params->spp_hbinterval = 0;
2347
2348 if (params->spp_hbinterval ||
2349 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2350 if (trans) {
2351 trans->hbinterval =
2352 msecs_to_jiffies(params->spp_hbinterval);
2353 } else if (asoc) {
2354 asoc->hbinterval =
2355 msecs_to_jiffies(params->spp_hbinterval);
2356 } else {
2357 sp->hbinterval = params->spp_hbinterval;
2358 }
2359 }
2360 }
2361
2362 if (hb_change) {
2363 if (trans) {
2364 trans->param_flags =
2365 (trans->param_flags & ~SPP_HB) | hb_change;
2366 } else if (asoc) {
2367 asoc->param_flags =
2368 (asoc->param_flags & ~SPP_HB) | hb_change;
2369 } else {
2370 sp->param_flags =
2371 (sp->param_flags & ~SPP_HB) | hb_change;
2372 }
2373 }
2374
2375 /* When Path MTU discovery is disabled the value specified here will
2376 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2377 * include the flag SPP_PMTUD_DISABLE for this field to have any
2378 * effect).
2379 */
2380 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2381 if (trans) {
2382 trans->pathmtu = params->spp_pathmtu;
2383 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2384 } else if (asoc) {
2385 asoc->pathmtu = params->spp_pathmtu;
2386 sctp_frag_point(asoc, params->spp_pathmtu);
2387 } else {
2388 sp->pathmtu = params->spp_pathmtu;
2389 }
2390 }
2391
2392 if (pmtud_change) {
2393 if (trans) {
2394 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2395 (params->spp_flags & SPP_PMTUD_ENABLE);
2396 trans->param_flags =
2397 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2398 if (update) {
2399 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2400 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2401 }
2402 } else if (asoc) {
2403 asoc->param_flags =
2404 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2405 } else {
2406 sp->param_flags =
2407 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2408 }
2409 }
2410
2411 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2412 * value of this field is ignored. Note also that a value of zero
2413 * indicates the current setting should be left unchanged.
2414 */
2415 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2416 if (trans) {
2417 trans->sackdelay =
2418 msecs_to_jiffies(params->spp_sackdelay);
2419 } else if (asoc) {
2420 asoc->sackdelay =
2421 msecs_to_jiffies(params->spp_sackdelay);
2422 } else {
2423 sp->sackdelay = params->spp_sackdelay;
2424 }
2425 }
2426
2427 if (sackdelay_change) {
2428 if (trans) {
2429 trans->param_flags =
2430 (trans->param_flags & ~SPP_SACKDELAY) |
2431 sackdelay_change;
2432 } else if (asoc) {
2433 asoc->param_flags =
2434 (asoc->param_flags & ~SPP_SACKDELAY) |
2435 sackdelay_change;
2436 } else {
2437 sp->param_flags =
2438 (sp->param_flags & ~SPP_SACKDELAY) |
2439 sackdelay_change;
2440 }
2441 }
2442
2443 /* Note that a value of zero indicates the current setting should be
2444 left unchanged.
2445 */
2446 if (params->spp_pathmaxrxt) {
2447 if (trans) {
2448 trans->pathmaxrxt = params->spp_pathmaxrxt;
2449 } else if (asoc) {
2450 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2451 } else {
2452 sp->pathmaxrxt = params->spp_pathmaxrxt;
2453 }
2454 }
2455
2456 return 0;
2457 }
2458
2459 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2460 char __user *optval,
2461 unsigned int optlen)
2462 {
2463 struct sctp_paddrparams params;
2464 struct sctp_transport *trans = NULL;
2465 struct sctp_association *asoc = NULL;
2466 struct sctp_sock *sp = sctp_sk(sk);
2467 int error;
2468 int hb_change, pmtud_change, sackdelay_change;
2469
2470 if (optlen != sizeof(struct sctp_paddrparams))
2471 return - EINVAL;
2472
2473 if (copy_from_user(&params, optval, optlen))
2474 return -EFAULT;
2475
2476 /* Validate flags and value parameters. */
2477 hb_change = params.spp_flags & SPP_HB;
2478 pmtud_change = params.spp_flags & SPP_PMTUD;
2479 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2480
2481 if (hb_change == SPP_HB ||
2482 pmtud_change == SPP_PMTUD ||
2483 sackdelay_change == SPP_SACKDELAY ||
2484 params.spp_sackdelay > 500 ||
2485 (params.spp_pathmtu &&
2486 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2487 return -EINVAL;
2488
2489 /* If an address other than INADDR_ANY is specified, and
2490 * no transport is found, then the request is invalid.
2491 */
2492 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
2493 trans = sctp_addr_id2transport(sk, &params.spp_address,
2494 params.spp_assoc_id);
2495 if (!trans)
2496 return -EINVAL;
2497 }
2498
2499 /* Get association, if assoc_id != 0 and the socket is a one
2500 * to many style socket, and an association was not found, then
2501 * the id was invalid.
2502 */
2503 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2504 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2505 return -EINVAL;
2506
2507 /* Heartbeat demand can only be sent on a transport or
2508 * association, but not a socket.
2509 */
2510 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2511 return -EINVAL;
2512
2513 /* Process parameters. */
2514 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2515 hb_change, pmtud_change,
2516 sackdelay_change);
2517
2518 if (error)
2519 return error;
2520
2521 /* If changes are for association, also apply parameters to each
2522 * transport.
2523 */
2524 if (!trans && asoc) {
2525 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2526 transports) {
2527 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2528 hb_change, pmtud_change,
2529 sackdelay_change);
2530 }
2531 }
2532
2533 return 0;
2534 }
2535
2536 /*
2537 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2538 *
2539 * This option will effect the way delayed acks are performed. This
2540 * option allows you to get or set the delayed ack time, in
2541 * milliseconds. It also allows changing the delayed ack frequency.
2542 * Changing the frequency to 1 disables the delayed sack algorithm. If
2543 * the assoc_id is 0, then this sets or gets the endpoints default
2544 * values. If the assoc_id field is non-zero, then the set or get
2545 * effects the specified association for the one to many model (the
2546 * assoc_id field is ignored by the one to one model). Note that if
2547 * sack_delay or sack_freq are 0 when setting this option, then the
2548 * current values will remain unchanged.
2549 *
2550 * struct sctp_sack_info {
2551 * sctp_assoc_t sack_assoc_id;
2552 * uint32_t sack_delay;
2553 * uint32_t sack_freq;
2554 * };
2555 *
2556 * sack_assoc_id - This parameter, indicates which association the user
2557 * is performing an action upon. Note that if this field's value is
2558 * zero then the endpoints default value is changed (effecting future
2559 * associations only).
2560 *
2561 * sack_delay - This parameter contains the number of milliseconds that
2562 * the user is requesting the delayed ACK timer be set to. Note that
2563 * this value is defined in the standard to be between 200 and 500
2564 * milliseconds.
2565 *
2566 * sack_freq - This parameter contains the number of packets that must
2567 * be received before a sack is sent without waiting for the delay
2568 * timer to expire. The default value for this is 2, setting this
2569 * value to 1 will disable the delayed sack algorithm.
2570 */
2571
2572 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2573 char __user *optval, unsigned int optlen)
2574 {
2575 struct sctp_sack_info params;
2576 struct sctp_transport *trans = NULL;
2577 struct sctp_association *asoc = NULL;
2578 struct sctp_sock *sp = sctp_sk(sk);
2579
2580 if (optlen == sizeof(struct sctp_sack_info)) {
2581 if (copy_from_user(&params, optval, optlen))
2582 return -EFAULT;
2583
2584 if (params.sack_delay == 0 && params.sack_freq == 0)
2585 return 0;
2586 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2587 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2588 pr_warn("Use struct sctp_sack_info instead\n");
2589 if (copy_from_user(&params, optval, optlen))
2590 return -EFAULT;
2591
2592 if (params.sack_delay == 0)
2593 params.sack_freq = 1;
2594 else
2595 params.sack_freq = 0;
2596 } else
2597 return - EINVAL;
2598
2599 /* Validate value parameter. */
2600 if (params.sack_delay > 500)
2601 return -EINVAL;
2602
2603 /* Get association, if sack_assoc_id != 0 and the socket is a one
2604 * to many style socket, and an association was not found, then
2605 * the id was invalid.
2606 */
2607 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2608 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2609 return -EINVAL;
2610
2611 if (params.sack_delay) {
2612 if (asoc) {
2613 asoc->sackdelay =
2614 msecs_to_jiffies(params.sack_delay);
2615 asoc->param_flags =
2616 (asoc->param_flags & ~SPP_SACKDELAY) |
2617 SPP_SACKDELAY_ENABLE;
2618 } else {
2619 sp->sackdelay = params.sack_delay;
2620 sp->param_flags =
2621 (sp->param_flags & ~SPP_SACKDELAY) |
2622 SPP_SACKDELAY_ENABLE;
2623 }
2624 }
2625
2626 if (params.sack_freq == 1) {
2627 if (asoc) {
2628 asoc->param_flags =
2629 (asoc->param_flags & ~SPP_SACKDELAY) |
2630 SPP_SACKDELAY_DISABLE;
2631 } else {
2632 sp->param_flags =
2633 (sp->param_flags & ~SPP_SACKDELAY) |
2634 SPP_SACKDELAY_DISABLE;
2635 }
2636 } else if (params.sack_freq > 1) {
2637 if (asoc) {
2638 asoc->sackfreq = params.sack_freq;
2639 asoc->param_flags =
2640 (asoc->param_flags & ~SPP_SACKDELAY) |
2641 SPP_SACKDELAY_ENABLE;
2642 } else {
2643 sp->sackfreq = params.sack_freq;
2644 sp->param_flags =
2645 (sp->param_flags & ~SPP_SACKDELAY) |
2646 SPP_SACKDELAY_ENABLE;
2647 }
2648 }
2649
2650 /* If change is for association, also apply to each transport. */
2651 if (asoc) {
2652 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2653 transports) {
2654 if (params.sack_delay) {
2655 trans->sackdelay =
2656 msecs_to_jiffies(params.sack_delay);
2657 trans->param_flags =
2658 (trans->param_flags & ~SPP_SACKDELAY) |
2659 SPP_SACKDELAY_ENABLE;
2660 }
2661 if (params.sack_freq == 1) {
2662 trans->param_flags =
2663 (trans->param_flags & ~SPP_SACKDELAY) |
2664 SPP_SACKDELAY_DISABLE;
2665 } else if (params.sack_freq > 1) {
2666 trans->sackfreq = params.sack_freq;
2667 trans->param_flags =
2668 (trans->param_flags & ~SPP_SACKDELAY) |
2669 SPP_SACKDELAY_ENABLE;
2670 }
2671 }
2672 }
2673
2674 return 0;
2675 }
2676
2677 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2678 *
2679 * Applications can specify protocol parameters for the default association
2680 * initialization. The option name argument to setsockopt() and getsockopt()
2681 * is SCTP_INITMSG.
2682 *
2683 * Setting initialization parameters is effective only on an unconnected
2684 * socket (for UDP-style sockets only future associations are effected
2685 * by the change). With TCP-style sockets, this option is inherited by
2686 * sockets derived from a listener socket.
2687 */
2688 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2689 {
2690 struct sctp_initmsg sinit;
2691 struct sctp_sock *sp = sctp_sk(sk);
2692
2693 if (optlen != sizeof(struct sctp_initmsg))
2694 return -EINVAL;
2695 if (copy_from_user(&sinit, optval, optlen))
2696 return -EFAULT;
2697
2698 if (sinit.sinit_num_ostreams)
2699 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2700 if (sinit.sinit_max_instreams)
2701 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2702 if (sinit.sinit_max_attempts)
2703 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2704 if (sinit.sinit_max_init_timeo)
2705 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2706
2707 return 0;
2708 }
2709
2710 /*
2711 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2712 *
2713 * Applications that wish to use the sendto() system call may wish to
2714 * specify a default set of parameters that would normally be supplied
2715 * through the inclusion of ancillary data. This socket option allows
2716 * such an application to set the default sctp_sndrcvinfo structure.
2717 * The application that wishes to use this socket option simply passes
2718 * in to this call the sctp_sndrcvinfo structure defined in Section
2719 * 5.2.2) The input parameters accepted by this call include
2720 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2721 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2722 * to this call if the caller is using the UDP model.
2723 */
2724 static int sctp_setsockopt_default_send_param(struct sock *sk,
2725 char __user *optval,
2726 unsigned int optlen)
2727 {
2728 struct sctp_sndrcvinfo info;
2729 struct sctp_association *asoc;
2730 struct sctp_sock *sp = sctp_sk(sk);
2731
2732 if (optlen != sizeof(struct sctp_sndrcvinfo))
2733 return -EINVAL;
2734 if (copy_from_user(&info, optval, optlen))
2735 return -EFAULT;
2736
2737 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2738 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2739 return -EINVAL;
2740
2741 if (asoc) {
2742 asoc->default_stream = info.sinfo_stream;
2743 asoc->default_flags = info.sinfo_flags;
2744 asoc->default_ppid = info.sinfo_ppid;
2745 asoc->default_context = info.sinfo_context;
2746 asoc->default_timetolive = info.sinfo_timetolive;
2747 } else {
2748 sp->default_stream = info.sinfo_stream;
2749 sp->default_flags = info.sinfo_flags;
2750 sp->default_ppid = info.sinfo_ppid;
2751 sp->default_context = info.sinfo_context;
2752 sp->default_timetolive = info.sinfo_timetolive;
2753 }
2754
2755 return 0;
2756 }
2757
2758 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2759 *
2760 * Requests that the local SCTP stack use the enclosed peer address as
2761 * the association primary. The enclosed address must be one of the
2762 * association peer's addresses.
2763 */
2764 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2765 unsigned int optlen)
2766 {
2767 struct sctp_prim prim;
2768 struct sctp_transport *trans;
2769
2770 if (optlen != sizeof(struct sctp_prim))
2771 return -EINVAL;
2772
2773 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2774 return -EFAULT;
2775
2776 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2777 if (!trans)
2778 return -EINVAL;
2779
2780 sctp_assoc_set_primary(trans->asoc, trans);
2781
2782 return 0;
2783 }
2784
2785 /*
2786 * 7.1.5 SCTP_NODELAY
2787 *
2788 * Turn on/off any Nagle-like algorithm. This means that packets are
2789 * generally sent as soon as possible and no unnecessary delays are
2790 * introduced, at the cost of more packets in the network. Expects an
2791 * integer boolean flag.
2792 */
2793 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2794 unsigned int optlen)
2795 {
2796 int val;
2797
2798 if (optlen < sizeof(int))
2799 return -EINVAL;
2800 if (get_user(val, (int __user *)optval))
2801 return -EFAULT;
2802
2803 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2804 return 0;
2805 }
2806
2807 /*
2808 *
2809 * 7.1.1 SCTP_RTOINFO
2810 *
2811 * The protocol parameters used to initialize and bound retransmission
2812 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2813 * and modify these parameters.
2814 * All parameters are time values, in milliseconds. A value of 0, when
2815 * modifying the parameters, indicates that the current value should not
2816 * be changed.
2817 *
2818 */
2819 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2820 {
2821 struct sctp_rtoinfo rtoinfo;
2822 struct sctp_association *asoc;
2823
2824 if (optlen != sizeof (struct sctp_rtoinfo))
2825 return -EINVAL;
2826
2827 if (copy_from_user(&rtoinfo, optval, optlen))
2828 return -EFAULT;
2829
2830 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2831
2832 /* Set the values to the specific association */
2833 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2834 return -EINVAL;
2835
2836 if (asoc) {
2837 if (rtoinfo.srto_initial != 0)
2838 asoc->rto_initial =
2839 msecs_to_jiffies(rtoinfo.srto_initial);
2840 if (rtoinfo.srto_max != 0)
2841 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2842 if (rtoinfo.srto_min != 0)
2843 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2844 } else {
2845 /* If there is no association or the association-id = 0
2846 * set the values to the endpoint.
2847 */
2848 struct sctp_sock *sp = sctp_sk(sk);
2849
2850 if (rtoinfo.srto_initial != 0)
2851 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2852 if (rtoinfo.srto_max != 0)
2853 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2854 if (rtoinfo.srto_min != 0)
2855 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2856 }
2857
2858 return 0;
2859 }
2860
2861 /*
2862 *
2863 * 7.1.2 SCTP_ASSOCINFO
2864 *
2865 * This option is used to tune the maximum retransmission attempts
2866 * of the association.
2867 * Returns an error if the new association retransmission value is
2868 * greater than the sum of the retransmission value of the peer.
2869 * See [SCTP] for more information.
2870 *
2871 */
2872 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2873 {
2874
2875 struct sctp_assocparams assocparams;
2876 struct sctp_association *asoc;
2877
2878 if (optlen != sizeof(struct sctp_assocparams))
2879 return -EINVAL;
2880 if (copy_from_user(&assocparams, optval, optlen))
2881 return -EFAULT;
2882
2883 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2884
2885 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2886 return -EINVAL;
2887
2888 /* Set the values to the specific association */
2889 if (asoc) {
2890 if (assocparams.sasoc_asocmaxrxt != 0) {
2891 __u32 path_sum = 0;
2892 int paths = 0;
2893 struct sctp_transport *peer_addr;
2894
2895 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2896 transports) {
2897 path_sum += peer_addr->pathmaxrxt;
2898 paths++;
2899 }
2900
2901 /* Only validate asocmaxrxt if we have more than
2902 * one path/transport. We do this because path
2903 * retransmissions are only counted when we have more
2904 * then one path.
2905 */
2906 if (paths > 1 &&
2907 assocparams.sasoc_asocmaxrxt > path_sum)
2908 return -EINVAL;
2909
2910 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2911 }
2912
2913 if (assocparams.sasoc_cookie_life != 0) {
2914 asoc->cookie_life.tv_sec =
2915 assocparams.sasoc_cookie_life / 1000;
2916 asoc->cookie_life.tv_usec =
2917 (assocparams.sasoc_cookie_life % 1000)
2918 * 1000;
2919 }
2920 } else {
2921 /* Set the values to the endpoint */
2922 struct sctp_sock *sp = sctp_sk(sk);
2923
2924 if (assocparams.sasoc_asocmaxrxt != 0)
2925 sp->assocparams.sasoc_asocmaxrxt =
2926 assocparams.sasoc_asocmaxrxt;
2927 if (assocparams.sasoc_cookie_life != 0)
2928 sp->assocparams.sasoc_cookie_life =
2929 assocparams.sasoc_cookie_life;
2930 }
2931 return 0;
2932 }
2933
2934 /*
2935 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2936 *
2937 * This socket option is a boolean flag which turns on or off mapped V4
2938 * addresses. If this option is turned on and the socket is type
2939 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2940 * If this option is turned off, then no mapping will be done of V4
2941 * addresses and a user will receive both PF_INET6 and PF_INET type
2942 * addresses on the socket.
2943 */
2944 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2945 {
2946 int val;
2947 struct sctp_sock *sp = sctp_sk(sk);
2948
2949 if (optlen < sizeof(int))
2950 return -EINVAL;
2951 if (get_user(val, (int __user *)optval))
2952 return -EFAULT;
2953 if (val)
2954 sp->v4mapped = 1;
2955 else
2956 sp->v4mapped = 0;
2957
2958 return 0;
2959 }
2960
2961 /*
2962 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2963 * This option will get or set the maximum size to put in any outgoing
2964 * SCTP DATA chunk. If a message is larger than this size it will be
2965 * fragmented by SCTP into the specified size. Note that the underlying
2966 * SCTP implementation may fragment into smaller sized chunks when the
2967 * PMTU of the underlying association is smaller than the value set by
2968 * the user. The default value for this option is '0' which indicates
2969 * the user is NOT limiting fragmentation and only the PMTU will effect
2970 * SCTP's choice of DATA chunk size. Note also that values set larger
2971 * than the maximum size of an IP datagram will effectively let SCTP
2972 * control fragmentation (i.e. the same as setting this option to 0).
2973 *
2974 * The following structure is used to access and modify this parameter:
2975 *
2976 * struct sctp_assoc_value {
2977 * sctp_assoc_t assoc_id;
2978 * uint32_t assoc_value;
2979 * };
2980 *
2981 * assoc_id: This parameter is ignored for one-to-one style sockets.
2982 * For one-to-many style sockets this parameter indicates which
2983 * association the user is performing an action upon. Note that if
2984 * this field's value is zero then the endpoints default value is
2985 * changed (effecting future associations only).
2986 * assoc_value: This parameter specifies the maximum size in bytes.
2987 */
2988 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2989 {
2990 struct sctp_assoc_value params;
2991 struct sctp_association *asoc;
2992 struct sctp_sock *sp = sctp_sk(sk);
2993 int val;
2994
2995 if (optlen == sizeof(int)) {
2996 pr_warn("Use of int in maxseg socket option deprecated\n");
2997 pr_warn("Use struct sctp_assoc_value instead\n");
2998 if (copy_from_user(&val, optval, optlen))
2999 return -EFAULT;
3000 params.assoc_id = 0;
3001 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3002 if (copy_from_user(&params, optval, optlen))
3003 return -EFAULT;
3004 val = params.assoc_value;
3005 } else
3006 return -EINVAL;
3007
3008 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3009 return -EINVAL;
3010
3011 asoc = sctp_id2assoc(sk, params.assoc_id);
3012 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3013 return -EINVAL;
3014
3015 if (asoc) {
3016 if (val == 0) {
3017 val = asoc->pathmtu;
3018 val -= sp->pf->af->net_header_len;
3019 val -= sizeof(struct sctphdr) +
3020 sizeof(struct sctp_data_chunk);
3021 }
3022 asoc->user_frag = val;
3023 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3024 } else {
3025 sp->user_frag = val;
3026 }
3027
3028 return 0;
3029 }
3030
3031
3032 /*
3033 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3034 *
3035 * Requests that the peer mark the enclosed address as the association
3036 * primary. The enclosed address must be one of the association's
3037 * locally bound addresses. The following structure is used to make a
3038 * set primary request:
3039 */
3040 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3041 unsigned int optlen)
3042 {
3043 struct net *net = sock_net(sk);
3044 struct sctp_sock *sp;
3045 struct sctp_association *asoc = NULL;
3046 struct sctp_setpeerprim prim;
3047 struct sctp_chunk *chunk;
3048 struct sctp_af *af;
3049 int err;
3050
3051 sp = sctp_sk(sk);
3052
3053 if (!net->sctp.addip_enable)
3054 return -EPERM;
3055
3056 if (optlen != sizeof(struct sctp_setpeerprim))
3057 return -EINVAL;
3058
3059 if (copy_from_user(&prim, optval, optlen))
3060 return -EFAULT;
3061
3062 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3063 if (!asoc)
3064 return -EINVAL;
3065
3066 if (!asoc->peer.asconf_capable)
3067 return -EPERM;
3068
3069 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3070 return -EPERM;
3071
3072 if (!sctp_state(asoc, ESTABLISHED))
3073 return -ENOTCONN;
3074
3075 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3076 if (!af)
3077 return -EINVAL;
3078
3079 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3080 return -EADDRNOTAVAIL;
3081
3082 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3083 return -EADDRNOTAVAIL;
3084
3085 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3086 chunk = sctp_make_asconf_set_prim(asoc,
3087 (union sctp_addr *)&prim.sspp_addr);
3088 if (!chunk)
3089 return -ENOMEM;
3090
3091 err = sctp_send_asconf(asoc, chunk);
3092
3093 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
3094
3095 return err;
3096 }
3097
3098 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3099 unsigned int optlen)
3100 {
3101 struct sctp_setadaptation adaptation;
3102
3103 if (optlen != sizeof(struct sctp_setadaptation))
3104 return -EINVAL;
3105 if (copy_from_user(&adaptation, optval, optlen))
3106 return -EFAULT;
3107
3108 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3109
3110 return 0;
3111 }
3112
3113 /*
3114 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3115 *
3116 * The context field in the sctp_sndrcvinfo structure is normally only
3117 * used when a failed message is retrieved holding the value that was
3118 * sent down on the actual send call. This option allows the setting of
3119 * a default context on an association basis that will be received on
3120 * reading messages from the peer. This is especially helpful in the
3121 * one-2-many model for an application to keep some reference to an
3122 * internal state machine that is processing messages on the
3123 * association. Note that the setting of this value only effects
3124 * received messages from the peer and does not effect the value that is
3125 * saved with outbound messages.
3126 */
3127 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3128 unsigned int optlen)
3129 {
3130 struct sctp_assoc_value params;
3131 struct sctp_sock *sp;
3132 struct sctp_association *asoc;
3133
3134 if (optlen != sizeof(struct sctp_assoc_value))
3135 return -EINVAL;
3136 if (copy_from_user(&params, optval, optlen))
3137 return -EFAULT;
3138
3139 sp = sctp_sk(sk);
3140
3141 if (params.assoc_id != 0) {
3142 asoc = sctp_id2assoc(sk, params.assoc_id);
3143 if (!asoc)
3144 return -EINVAL;
3145 asoc->default_rcv_context = params.assoc_value;
3146 } else {
3147 sp->default_rcv_context = params.assoc_value;
3148 }
3149
3150 return 0;
3151 }
3152
3153 /*
3154 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3155 *
3156 * This options will at a minimum specify if the implementation is doing
3157 * fragmented interleave. Fragmented interleave, for a one to many
3158 * socket, is when subsequent calls to receive a message may return
3159 * parts of messages from different associations. Some implementations
3160 * may allow you to turn this value on or off. If so, when turned off,
3161 * no fragment interleave will occur (which will cause a head of line
3162 * blocking amongst multiple associations sharing the same one to many
3163 * socket). When this option is turned on, then each receive call may
3164 * come from a different association (thus the user must receive data
3165 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3166 * association each receive belongs to.
3167 *
3168 * This option takes a boolean value. A non-zero value indicates that
3169 * fragmented interleave is on. A value of zero indicates that
3170 * fragmented interleave is off.
3171 *
3172 * Note that it is important that an implementation that allows this
3173 * option to be turned on, have it off by default. Otherwise an unaware
3174 * application using the one to many model may become confused and act
3175 * incorrectly.
3176 */
3177 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3178 char __user *optval,
3179 unsigned int optlen)
3180 {
3181 int val;
3182
3183 if (optlen != sizeof(int))
3184 return -EINVAL;
3185 if (get_user(val, (int __user *)optval))
3186 return -EFAULT;
3187
3188 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3189
3190 return 0;
3191 }
3192
3193 /*
3194 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3195 * (SCTP_PARTIAL_DELIVERY_POINT)
3196 *
3197 * This option will set or get the SCTP partial delivery point. This
3198 * point is the size of a message where the partial delivery API will be
3199 * invoked to help free up rwnd space for the peer. Setting this to a
3200 * lower value will cause partial deliveries to happen more often. The
3201 * calls argument is an integer that sets or gets the partial delivery
3202 * point. Note also that the call will fail if the user attempts to set
3203 * this value larger than the socket receive buffer size.
3204 *
3205 * Note that any single message having a length smaller than or equal to
3206 * the SCTP partial delivery point will be delivered in one single read
3207 * call as long as the user provided buffer is large enough to hold the
3208 * message.
3209 */
3210 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3211 char __user *optval,
3212 unsigned int optlen)
3213 {
3214 u32 val;
3215
3216 if (optlen != sizeof(u32))
3217 return -EINVAL;
3218 if (get_user(val, (int __user *)optval))
3219 return -EFAULT;
3220
3221 /* Note: We double the receive buffer from what the user sets
3222 * it to be, also initial rwnd is based on rcvbuf/2.
3223 */
3224 if (val > (sk->sk_rcvbuf >> 1))
3225 return -EINVAL;
3226
3227 sctp_sk(sk)->pd_point = val;
3228
3229 return 0; /* is this the right error code? */
3230 }
3231
3232 /*
3233 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3234 *
3235 * This option will allow a user to change the maximum burst of packets
3236 * that can be emitted by this association. Note that the default value
3237 * is 4, and some implementations may restrict this setting so that it
3238 * can only be lowered.
3239 *
3240 * NOTE: This text doesn't seem right. Do this on a socket basis with
3241 * future associations inheriting the socket value.
3242 */
3243 static int sctp_setsockopt_maxburst(struct sock *sk,
3244 char __user *optval,
3245 unsigned int optlen)
3246 {
3247 struct sctp_assoc_value params;
3248 struct sctp_sock *sp;
3249 struct sctp_association *asoc;
3250 int val;
3251 int assoc_id = 0;
3252
3253 if (optlen == sizeof(int)) {
3254 pr_warn("Use of int in max_burst socket option deprecated\n");
3255 pr_warn("Use struct sctp_assoc_value instead\n");
3256 if (copy_from_user(&val, optval, optlen))
3257 return -EFAULT;
3258 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3259 if (copy_from_user(&params, optval, optlen))
3260 return -EFAULT;
3261 val = params.assoc_value;
3262 assoc_id = params.assoc_id;
3263 } else
3264 return -EINVAL;
3265
3266 sp = sctp_sk(sk);
3267
3268 if (assoc_id != 0) {
3269 asoc = sctp_id2assoc(sk, assoc_id);
3270 if (!asoc)
3271 return -EINVAL;
3272 asoc->max_burst = val;
3273 } else
3274 sp->max_burst = val;
3275
3276 return 0;
3277 }
3278
3279 /*
3280 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3281 *
3282 * This set option adds a chunk type that the user is requesting to be
3283 * received only in an authenticated way. Changes to the list of chunks
3284 * will only effect future associations on the socket.
3285 */
3286 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3287 char __user *optval,
3288 unsigned int optlen)
3289 {
3290 struct net *net = sock_net(sk);
3291 struct sctp_authchunk val;
3292
3293 if (!net->sctp.auth_enable)
3294 return -EACCES;
3295
3296 if (optlen != sizeof(struct sctp_authchunk))
3297 return -EINVAL;
3298 if (copy_from_user(&val, optval, optlen))
3299 return -EFAULT;
3300
3301 switch (val.sauth_chunk) {
3302 case SCTP_CID_INIT:
3303 case SCTP_CID_INIT_ACK:
3304 case SCTP_CID_SHUTDOWN_COMPLETE:
3305 case SCTP_CID_AUTH:
3306 return -EINVAL;
3307 }
3308
3309 /* add this chunk id to the endpoint */
3310 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3311 }
3312
3313 /*
3314 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3315 *
3316 * This option gets or sets the list of HMAC algorithms that the local
3317 * endpoint requires the peer to use.
3318 */
3319 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3320 char __user *optval,
3321 unsigned int optlen)
3322 {
3323 struct net *net = sock_net(sk);
3324 struct sctp_hmacalgo *hmacs;
3325 u32 idents;
3326 int err;
3327
3328 if (!net->sctp.auth_enable)
3329 return -EACCES;
3330
3331 if (optlen < sizeof(struct sctp_hmacalgo))
3332 return -EINVAL;
3333
3334 hmacs= memdup_user(optval, optlen);
3335 if (IS_ERR(hmacs))
3336 return PTR_ERR(hmacs);
3337
3338 idents = hmacs->shmac_num_idents;
3339 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3340 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3341 err = -EINVAL;
3342 goto out;
3343 }
3344
3345 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3346 out:
3347 kfree(hmacs);
3348 return err;
3349 }
3350
3351 /*
3352 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3353 *
3354 * This option will set a shared secret key which is used to build an
3355 * association shared key.
3356 */
3357 static int sctp_setsockopt_auth_key(struct sock *sk,
3358 char __user *optval,
3359 unsigned int optlen)
3360 {
3361 struct net *net = sock_net(sk);
3362 struct sctp_authkey *authkey;
3363 struct sctp_association *asoc;
3364 int ret;
3365
3366 if (!net->sctp.auth_enable)
3367 return -EACCES;
3368
3369 if (optlen <= sizeof(struct sctp_authkey))
3370 return -EINVAL;
3371
3372 authkey= memdup_user(optval, optlen);
3373 if (IS_ERR(authkey))
3374 return PTR_ERR(authkey);
3375
3376 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3377 ret = -EINVAL;
3378 goto out;
3379 }
3380
3381 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3382 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3383 ret = -EINVAL;
3384 goto out;
3385 }
3386
3387 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3388 out:
3389 kzfree(authkey);
3390 return ret;
3391 }
3392
3393 /*
3394 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3395 *
3396 * This option will get or set the active shared key to be used to build
3397 * the association shared key.
3398 */
3399 static int sctp_setsockopt_active_key(struct sock *sk,
3400 char __user *optval,
3401 unsigned int optlen)
3402 {
3403 struct net *net = sock_net(sk);
3404 struct sctp_authkeyid val;
3405 struct sctp_association *asoc;
3406
3407 if (!net->sctp.auth_enable)
3408 return -EACCES;
3409
3410 if (optlen != sizeof(struct sctp_authkeyid))
3411 return -EINVAL;
3412 if (copy_from_user(&val, optval, optlen))
3413 return -EFAULT;
3414
3415 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3416 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3417 return -EINVAL;
3418
3419 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3420 val.scact_keynumber);
3421 }
3422
3423 /*
3424 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3425 *
3426 * This set option will delete a shared secret key from use.
3427 */
3428 static int sctp_setsockopt_del_key(struct sock *sk,
3429 char __user *optval,
3430 unsigned int optlen)
3431 {
3432 struct net *net = sock_net(sk);
3433 struct sctp_authkeyid val;
3434 struct sctp_association *asoc;
3435
3436 if (!net->sctp.auth_enable)
3437 return -EACCES;
3438
3439 if (optlen != sizeof(struct sctp_authkeyid))
3440 return -EINVAL;
3441 if (copy_from_user(&val, optval, optlen))
3442 return -EFAULT;
3443
3444 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3445 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3446 return -EINVAL;
3447
3448 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3449 val.scact_keynumber);
3450
3451 }
3452
3453 /*
3454 * 8.1.23 SCTP_AUTO_ASCONF
3455 *
3456 * This option will enable or disable the use of the automatic generation of
3457 * ASCONF chunks to add and delete addresses to an existing association. Note
3458 * that this option has two caveats namely: a) it only affects sockets that
3459 * are bound to all addresses available to the SCTP stack, and b) the system
3460 * administrator may have an overriding control that turns the ASCONF feature
3461 * off no matter what setting the socket option may have.
3462 * This option expects an integer boolean flag, where a non-zero value turns on
3463 * the option, and a zero value turns off the option.
3464 * Note. In this implementation, socket operation overrides default parameter
3465 * being set by sysctl as well as FreeBSD implementation
3466 */
3467 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3468 unsigned int optlen)
3469 {
3470 int val;
3471 struct sctp_sock *sp = sctp_sk(sk);
3472
3473 if (optlen < sizeof(int))
3474 return -EINVAL;
3475 if (get_user(val, (int __user *)optval))
3476 return -EFAULT;
3477 if (!sctp_is_ep_boundall(sk) && val)
3478 return -EINVAL;
3479 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3480 return 0;
3481
3482 if (val == 0 && sp->do_auto_asconf) {
3483 list_del(&sp->auto_asconf_list);
3484 sp->do_auto_asconf = 0;
3485 } else if (val && !sp->do_auto_asconf) {
3486 list_add_tail(&sp->auto_asconf_list,
3487 &sock_net(sk)->sctp.auto_asconf_splist);
3488 sp->do_auto_asconf = 1;
3489 }
3490 return 0;
3491 }
3492
3493
3494 /*
3495 * SCTP_PEER_ADDR_THLDS
3496 *
3497 * This option allows us to alter the partially failed threshold for one or all
3498 * transports in an association. See Section 6.1 of:
3499 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3500 */
3501 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3502 char __user *optval,
3503 unsigned int optlen)
3504 {
3505 struct sctp_paddrthlds val;
3506 struct sctp_transport *trans;
3507 struct sctp_association *asoc;
3508
3509 if (optlen < sizeof(struct sctp_paddrthlds))
3510 return -EINVAL;
3511 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3512 sizeof(struct sctp_paddrthlds)))
3513 return -EFAULT;
3514
3515
3516 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3517 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3518 if (!asoc)
3519 return -ENOENT;
3520 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3521 transports) {
3522 if (val.spt_pathmaxrxt)
3523 trans->pathmaxrxt = val.spt_pathmaxrxt;
3524 trans->pf_retrans = val.spt_pathpfthld;
3525 }
3526
3527 if (val.spt_pathmaxrxt)
3528 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3529 asoc->pf_retrans = val.spt_pathpfthld;
3530 } else {
3531 trans = sctp_addr_id2transport(sk, &val.spt_address,
3532 val.spt_assoc_id);
3533 if (!trans)
3534 return -ENOENT;
3535
3536 if (val.spt_pathmaxrxt)
3537 trans->pathmaxrxt = val.spt_pathmaxrxt;
3538 trans->pf_retrans = val.spt_pathpfthld;
3539 }
3540
3541 return 0;
3542 }
3543
3544 /* API 6.2 setsockopt(), getsockopt()
3545 *
3546 * Applications use setsockopt() and getsockopt() to set or retrieve
3547 * socket options. Socket options are used to change the default
3548 * behavior of sockets calls. They are described in Section 7.
3549 *
3550 * The syntax is:
3551 *
3552 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3553 * int __user *optlen);
3554 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3555 * int optlen);
3556 *
3557 * sd - the socket descript.
3558 * level - set to IPPROTO_SCTP for all SCTP options.
3559 * optname - the option name.
3560 * optval - the buffer to store the value of the option.
3561 * optlen - the size of the buffer.
3562 */
3563 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3564 char __user *optval, unsigned int optlen)
3565 {
3566 int retval = 0;
3567
3568 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3569 sk, optname);
3570
3571 /* I can hardly begin to describe how wrong this is. This is
3572 * so broken as to be worse than useless. The API draft
3573 * REALLY is NOT helpful here... I am not convinced that the
3574 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3575 * are at all well-founded.
3576 */
3577 if (level != SOL_SCTP) {
3578 struct sctp_af *af = sctp_sk(sk)->pf->af;
3579 retval = af->setsockopt(sk, level, optname, optval, optlen);
3580 goto out_nounlock;
3581 }
3582
3583 sctp_lock_sock(sk);
3584
3585 switch (optname) {
3586 case SCTP_SOCKOPT_BINDX_ADD:
3587 /* 'optlen' is the size of the addresses buffer. */
3588 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3589 optlen, SCTP_BINDX_ADD_ADDR);
3590 break;
3591
3592 case SCTP_SOCKOPT_BINDX_REM:
3593 /* 'optlen' is the size of the addresses buffer. */
3594 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3595 optlen, SCTP_BINDX_REM_ADDR);
3596 break;
3597
3598 case SCTP_SOCKOPT_CONNECTX_OLD:
3599 /* 'optlen' is the size of the addresses buffer. */
3600 retval = sctp_setsockopt_connectx_old(sk,
3601 (struct sockaddr __user *)optval,
3602 optlen);
3603 break;
3604
3605 case SCTP_SOCKOPT_CONNECTX:
3606 /* 'optlen' is the size of the addresses buffer. */
3607 retval = sctp_setsockopt_connectx(sk,
3608 (struct sockaddr __user *)optval,
3609 optlen);
3610 break;
3611
3612 case SCTP_DISABLE_FRAGMENTS:
3613 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3614 break;
3615
3616 case SCTP_EVENTS:
3617 retval = sctp_setsockopt_events(sk, optval, optlen);
3618 break;
3619
3620 case SCTP_AUTOCLOSE:
3621 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3622 break;
3623
3624 case SCTP_PEER_ADDR_PARAMS:
3625 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3626 break;
3627
3628 case SCTP_DELAYED_SACK:
3629 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3630 break;
3631 case SCTP_PARTIAL_DELIVERY_POINT:
3632 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3633 break;
3634
3635 case SCTP_INITMSG:
3636 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3637 break;
3638 case SCTP_DEFAULT_SEND_PARAM:
3639 retval = sctp_setsockopt_default_send_param(sk, optval,
3640 optlen);
3641 break;
3642 case SCTP_PRIMARY_ADDR:
3643 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3644 break;
3645 case SCTP_SET_PEER_PRIMARY_ADDR:
3646 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3647 break;
3648 case SCTP_NODELAY:
3649 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3650 break;
3651 case SCTP_RTOINFO:
3652 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3653 break;
3654 case SCTP_ASSOCINFO:
3655 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3656 break;
3657 case SCTP_I_WANT_MAPPED_V4_ADDR:
3658 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3659 break;
3660 case SCTP_MAXSEG:
3661 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3662 break;
3663 case SCTP_ADAPTATION_LAYER:
3664 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3665 break;
3666 case SCTP_CONTEXT:
3667 retval = sctp_setsockopt_context(sk, optval, optlen);
3668 break;
3669 case SCTP_FRAGMENT_INTERLEAVE:
3670 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3671 break;
3672 case SCTP_MAX_BURST:
3673 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3674 break;
3675 case SCTP_AUTH_CHUNK:
3676 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3677 break;
3678 case SCTP_HMAC_IDENT:
3679 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3680 break;
3681 case SCTP_AUTH_KEY:
3682 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3683 break;
3684 case SCTP_AUTH_ACTIVE_KEY:
3685 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3686 break;
3687 case SCTP_AUTH_DELETE_KEY:
3688 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3689 break;
3690 case SCTP_AUTO_ASCONF:
3691 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3692 break;
3693 case SCTP_PEER_ADDR_THLDS:
3694 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3695 break;
3696 default:
3697 retval = -ENOPROTOOPT;
3698 break;
3699 }
3700
3701 sctp_release_sock(sk);
3702
3703 out_nounlock:
3704 return retval;
3705 }
3706
3707 /* API 3.1.6 connect() - UDP Style Syntax
3708 *
3709 * An application may use the connect() call in the UDP model to initiate an
3710 * association without sending data.
3711 *
3712 * The syntax is:
3713 *
3714 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3715 *
3716 * sd: the socket descriptor to have a new association added to.
3717 *
3718 * nam: the address structure (either struct sockaddr_in or struct
3719 * sockaddr_in6 defined in RFC2553 [7]).
3720 *
3721 * len: the size of the address.
3722 */
3723 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
3724 int addr_len)
3725 {
3726 int err = 0;
3727 struct sctp_af *af;
3728
3729 sctp_lock_sock(sk);
3730
3731 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3732 __func__, sk, addr, addr_len);
3733
3734 /* Validate addr_len before calling common connect/connectx routine. */
3735 af = sctp_get_af_specific(addr->sa_family);
3736 if (!af || addr_len < af->sockaddr_len) {
3737 err = -EINVAL;
3738 } else {
3739 /* Pass correct addr len to common routine (so it knows there
3740 * is only one address being passed.
3741 */
3742 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3743 }
3744
3745 sctp_release_sock(sk);
3746 return err;
3747 }
3748
3749 /* FIXME: Write comments. */
3750 static int sctp_disconnect(struct sock *sk, int flags)
3751 {
3752 return -EOPNOTSUPP; /* STUB */
3753 }
3754
3755 /* 4.1.4 accept() - TCP Style Syntax
3756 *
3757 * Applications use accept() call to remove an established SCTP
3758 * association from the accept queue of the endpoint. A new socket
3759 * descriptor will be returned from accept() to represent the newly
3760 * formed association.
3761 */
3762 static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3763 {
3764 struct sctp_sock *sp;
3765 struct sctp_endpoint *ep;
3766 struct sock *newsk = NULL;
3767 struct sctp_association *asoc;
3768 long timeo;
3769 int error = 0;
3770
3771 sctp_lock_sock(sk);
3772
3773 sp = sctp_sk(sk);
3774 ep = sp->ep;
3775
3776 if (!sctp_style(sk, TCP)) {
3777 error = -EOPNOTSUPP;
3778 goto out;
3779 }
3780
3781 if (!sctp_sstate(sk, LISTENING)) {
3782 error = -EINVAL;
3783 goto out;
3784 }
3785
3786 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3787
3788 error = sctp_wait_for_accept(sk, timeo);
3789 if (error)
3790 goto out;
3791
3792 /* We treat the list of associations on the endpoint as the accept
3793 * queue and pick the first association on the list.
3794 */
3795 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3796
3797 newsk = sp->pf->create_accept_sk(sk, asoc);
3798 if (!newsk) {
3799 error = -ENOMEM;
3800 goto out;
3801 }
3802
3803 /* Populate the fields of the newsk from the oldsk and migrate the
3804 * asoc to the newsk.
3805 */
3806 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3807
3808 out:
3809 sctp_release_sock(sk);
3810 *err = error;
3811 return newsk;
3812 }
3813
3814 /* The SCTP ioctl handler. */
3815 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3816 {
3817 int rc = -ENOTCONN;
3818
3819 sctp_lock_sock(sk);
3820
3821 /*
3822 * SEQPACKET-style sockets in LISTENING state are valid, for
3823 * SCTP, so only discard TCP-style sockets in LISTENING state.
3824 */
3825 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3826 goto out;
3827
3828 switch (cmd) {
3829 case SIOCINQ: {
3830 struct sk_buff *skb;
3831 unsigned int amount = 0;
3832
3833 skb = skb_peek(&sk->sk_receive_queue);
3834 if (skb != NULL) {
3835 /*
3836 * We will only return the amount of this packet since
3837 * that is all that will be read.
3838 */
3839 amount = skb->len;
3840 }
3841 rc = put_user(amount, (int __user *)arg);
3842 break;
3843 }
3844 default:
3845 rc = -ENOIOCTLCMD;
3846 break;
3847 }
3848 out:
3849 sctp_release_sock(sk);
3850 return rc;
3851 }
3852
3853 /* This is the function which gets called during socket creation to
3854 * initialized the SCTP-specific portion of the sock.
3855 * The sock structure should already be zero-filled memory.
3856 */
3857 static int sctp_init_sock(struct sock *sk)
3858 {
3859 struct net *net = sock_net(sk);
3860 struct sctp_sock *sp;
3861
3862 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3863
3864 sp = sctp_sk(sk);
3865
3866 /* Initialize the SCTP per socket area. */
3867 switch (sk->sk_type) {
3868 case SOCK_SEQPACKET:
3869 sp->type = SCTP_SOCKET_UDP;
3870 break;
3871 case SOCK_STREAM:
3872 sp->type = SCTP_SOCKET_TCP;
3873 break;
3874 default:
3875 return -ESOCKTNOSUPPORT;
3876 }
3877
3878 /* Initialize default send parameters. These parameters can be
3879 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3880 */
3881 sp->default_stream = 0;
3882 sp->default_ppid = 0;
3883 sp->default_flags = 0;
3884 sp->default_context = 0;
3885 sp->default_timetolive = 0;
3886
3887 sp->default_rcv_context = 0;
3888 sp->max_burst = net->sctp.max_burst;
3889
3890 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
3891
3892 /* Initialize default setup parameters. These parameters
3893 * can be modified with the SCTP_INITMSG socket option or
3894 * overridden by the SCTP_INIT CMSG.
3895 */
3896 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3897 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3898 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
3899 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
3900
3901 /* Initialize default RTO related parameters. These parameters can
3902 * be modified for with the SCTP_RTOINFO socket option.
3903 */
3904 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
3905 sp->rtoinfo.srto_max = net->sctp.rto_max;
3906 sp->rtoinfo.srto_min = net->sctp.rto_min;
3907
3908 /* Initialize default association related parameters. These parameters
3909 * can be modified with the SCTP_ASSOCINFO socket option.
3910 */
3911 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
3912 sp->assocparams.sasoc_number_peer_destinations = 0;
3913 sp->assocparams.sasoc_peer_rwnd = 0;
3914 sp->assocparams.sasoc_local_rwnd = 0;
3915 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
3916
3917 /* Initialize default event subscriptions. By default, all the
3918 * options are off.
3919 */
3920 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3921
3922 /* Default Peer Address Parameters. These defaults can
3923 * be modified via SCTP_PEER_ADDR_PARAMS
3924 */
3925 sp->hbinterval = net->sctp.hb_interval;
3926 sp->pathmaxrxt = net->sctp.max_retrans_path;
3927 sp->pathmtu = 0; // allow default discovery
3928 sp->sackdelay = net->sctp.sack_timeout;
3929 sp->sackfreq = 2;
3930 sp->param_flags = SPP_HB_ENABLE |
3931 SPP_PMTUD_ENABLE |
3932 SPP_SACKDELAY_ENABLE;
3933
3934 /* If enabled no SCTP message fragmentation will be performed.
3935 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3936 */
3937 sp->disable_fragments = 0;
3938
3939 /* Enable Nagle algorithm by default. */
3940 sp->nodelay = 0;
3941
3942 /* Enable by default. */
3943 sp->v4mapped = 1;
3944
3945 /* Auto-close idle associations after the configured
3946 * number of seconds. A value of 0 disables this
3947 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3948 * for UDP-style sockets only.
3949 */
3950 sp->autoclose = 0;
3951
3952 /* User specified fragmentation limit. */
3953 sp->user_frag = 0;
3954
3955 sp->adaptation_ind = 0;
3956
3957 sp->pf = sctp_get_pf_specific(sk->sk_family);
3958
3959 /* Control variables for partial data delivery. */
3960 atomic_set(&sp->pd_mode, 0);
3961 skb_queue_head_init(&sp->pd_lobby);
3962 sp->frag_interleave = 0;
3963
3964 /* Create a per socket endpoint structure. Even if we
3965 * change the data structure relationships, this may still
3966 * be useful for storing pre-connect address information.
3967 */
3968 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
3969 if (!sp->ep)
3970 return -ENOMEM;
3971
3972 sp->hmac = NULL;
3973
3974 SCTP_DBG_OBJCNT_INC(sock);
3975
3976 local_bh_disable();
3977 percpu_counter_inc(&sctp_sockets_allocated);
3978 sock_prot_inuse_add(net, sk->sk_prot, 1);
3979 if (net->sctp.default_auto_asconf) {
3980 list_add_tail(&sp->auto_asconf_list,
3981 &net->sctp.auto_asconf_splist);
3982 sp->do_auto_asconf = 1;
3983 } else
3984 sp->do_auto_asconf = 0;
3985 local_bh_enable();
3986
3987 return 0;
3988 }
3989
3990 /* Cleanup any SCTP per socket resources. */
3991 static void sctp_destroy_sock(struct sock *sk)
3992 {
3993 struct sctp_sock *sp;
3994
3995 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3996
3997 /* Release our hold on the endpoint. */
3998 sp = sctp_sk(sk);
3999 /* This could happen during socket init, thus we bail out
4000 * early, since the rest of the below is not setup either.
4001 */
4002 if (sp->ep == NULL)
4003 return;
4004
4005 if (sp->do_auto_asconf) {
4006 sp->do_auto_asconf = 0;
4007 list_del(&sp->auto_asconf_list);
4008 }
4009 sctp_endpoint_free(sp->ep);
4010 local_bh_disable();
4011 percpu_counter_dec(&sctp_sockets_allocated);
4012 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4013 local_bh_enable();
4014 }
4015
4016 /* API 4.1.7 shutdown() - TCP Style Syntax
4017 * int shutdown(int socket, int how);
4018 *
4019 * sd - the socket descriptor of the association to be closed.
4020 * how - Specifies the type of shutdown. The values are
4021 * as follows:
4022 * SHUT_RD
4023 * Disables further receive operations. No SCTP
4024 * protocol action is taken.
4025 * SHUT_WR
4026 * Disables further send operations, and initiates
4027 * the SCTP shutdown sequence.
4028 * SHUT_RDWR
4029 * Disables further send and receive operations
4030 * and initiates the SCTP shutdown sequence.
4031 */
4032 static void sctp_shutdown(struct sock *sk, int how)
4033 {
4034 struct net *net = sock_net(sk);
4035 struct sctp_endpoint *ep;
4036 struct sctp_association *asoc;
4037
4038 if (!sctp_style(sk, TCP))
4039 return;
4040
4041 if (how & SEND_SHUTDOWN) {
4042 ep = sctp_sk(sk)->ep;
4043 if (!list_empty(&ep->asocs)) {
4044 asoc = list_entry(ep->asocs.next,
4045 struct sctp_association, asocs);
4046 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4047 }
4048 }
4049 }
4050
4051 /* 7.2.1 Association Status (SCTP_STATUS)
4052
4053 * Applications can retrieve current status information about an
4054 * association, including association state, peer receiver window size,
4055 * number of unacked data chunks, and number of data chunks pending
4056 * receipt. This information is read-only.
4057 */
4058 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4059 char __user *optval,
4060 int __user *optlen)
4061 {
4062 struct sctp_status status;
4063 struct sctp_association *asoc = NULL;
4064 struct sctp_transport *transport;
4065 sctp_assoc_t associd;
4066 int retval = 0;
4067
4068 if (len < sizeof(status)) {
4069 retval = -EINVAL;
4070 goto out;
4071 }
4072
4073 len = sizeof(status);
4074 if (copy_from_user(&status, optval, len)) {
4075 retval = -EFAULT;
4076 goto out;
4077 }
4078
4079 associd = status.sstat_assoc_id;
4080 asoc = sctp_id2assoc(sk, associd);
4081 if (!asoc) {
4082 retval = -EINVAL;
4083 goto out;
4084 }
4085
4086 transport = asoc->peer.primary_path;
4087
4088 status.sstat_assoc_id = sctp_assoc2id(asoc);
4089 status.sstat_state = asoc->state;
4090 status.sstat_rwnd = asoc->peer.rwnd;
4091 status.sstat_unackdata = asoc->unack_data;
4092
4093 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4094 status.sstat_instrms = asoc->c.sinit_max_instreams;
4095 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4096 status.sstat_fragmentation_point = asoc->frag_point;
4097 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4098 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4099 transport->af_specific->sockaddr_len);
4100 /* Map ipv4 address into v4-mapped-on-v6 address. */
4101 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4102 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4103 status.sstat_primary.spinfo_state = transport->state;
4104 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4105 status.sstat_primary.spinfo_srtt = transport->srtt;
4106 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4107 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4108
4109 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4110 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4111
4112 if (put_user(len, optlen)) {
4113 retval = -EFAULT;
4114 goto out;
4115 }
4116
4117 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
4118 len, status.sstat_state, status.sstat_rwnd,
4119 status.sstat_assoc_id);
4120
4121 if (copy_to_user(optval, &status, len)) {
4122 retval = -EFAULT;
4123 goto out;
4124 }
4125
4126 out:
4127 return retval;
4128 }
4129
4130
4131 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4132 *
4133 * Applications can retrieve information about a specific peer address
4134 * of an association, including its reachability state, congestion
4135 * window, and retransmission timer values. This information is
4136 * read-only.
4137 */
4138 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4139 char __user *optval,
4140 int __user *optlen)
4141 {
4142 struct sctp_paddrinfo pinfo;
4143 struct sctp_transport *transport;
4144 int retval = 0;
4145
4146 if (len < sizeof(pinfo)) {
4147 retval = -EINVAL;
4148 goto out;
4149 }
4150
4151 len = sizeof(pinfo);
4152 if (copy_from_user(&pinfo, optval, len)) {
4153 retval = -EFAULT;
4154 goto out;
4155 }
4156
4157 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4158 pinfo.spinfo_assoc_id);
4159 if (!transport)
4160 return -EINVAL;
4161
4162 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4163 pinfo.spinfo_state = transport->state;
4164 pinfo.spinfo_cwnd = transport->cwnd;
4165 pinfo.spinfo_srtt = transport->srtt;
4166 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4167 pinfo.spinfo_mtu = transport->pathmtu;
4168
4169 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4170 pinfo.spinfo_state = SCTP_ACTIVE;
4171
4172 if (put_user(len, optlen)) {
4173 retval = -EFAULT;
4174 goto out;
4175 }
4176
4177 if (copy_to_user(optval, &pinfo, len)) {
4178 retval = -EFAULT;
4179 goto out;
4180 }
4181
4182 out:
4183 return retval;
4184 }
4185
4186 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4187 *
4188 * This option is a on/off flag. If enabled no SCTP message
4189 * fragmentation will be performed. Instead if a message being sent
4190 * exceeds the current PMTU size, the message will NOT be sent and
4191 * instead a error will be indicated to the user.
4192 */
4193 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4194 char __user *optval, int __user *optlen)
4195 {
4196 int val;
4197
4198 if (len < sizeof(int))
4199 return -EINVAL;
4200
4201 len = sizeof(int);
4202 val = (sctp_sk(sk)->disable_fragments == 1);
4203 if (put_user(len, optlen))
4204 return -EFAULT;
4205 if (copy_to_user(optval, &val, len))
4206 return -EFAULT;
4207 return 0;
4208 }
4209
4210 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4211 *
4212 * This socket option is used to specify various notifications and
4213 * ancillary data the user wishes to receive.
4214 */
4215 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4216 int __user *optlen)
4217 {
4218 if (len <= 0)
4219 return -EINVAL;
4220 if (len > sizeof(struct sctp_event_subscribe))
4221 len = sizeof(struct sctp_event_subscribe);
4222 if (put_user(len, optlen))
4223 return -EFAULT;
4224 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4225 return -EFAULT;
4226 return 0;
4227 }
4228
4229 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4230 *
4231 * This socket option is applicable to the UDP-style socket only. When
4232 * set it will cause associations that are idle for more than the
4233 * specified number of seconds to automatically close. An association
4234 * being idle is defined an association that has NOT sent or received
4235 * user data. The special value of '0' indicates that no automatic
4236 * close of any associations should be performed. The option expects an
4237 * integer defining the number of seconds of idle time before an
4238 * association is closed.
4239 */
4240 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4241 {
4242 /* Applicable to UDP-style socket only */
4243 if (sctp_style(sk, TCP))
4244 return -EOPNOTSUPP;
4245 if (len < sizeof(int))
4246 return -EINVAL;
4247 len = sizeof(int);
4248 if (put_user(len, optlen))
4249 return -EFAULT;
4250 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4251 return -EFAULT;
4252 return 0;
4253 }
4254
4255 /* Helper routine to branch off an association to a new socket. */
4256 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4257 {
4258 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4259 struct socket *sock;
4260 struct sctp_af *af;
4261 int err = 0;
4262
4263 if (!asoc)
4264 return -EINVAL;
4265
4266 /* An association cannot be branched off from an already peeled-off
4267 * socket, nor is this supported for tcp style sockets.
4268 */
4269 if (!sctp_style(sk, UDP))
4270 return -EINVAL;
4271
4272 /* Create a new socket. */
4273 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4274 if (err < 0)
4275 return err;
4276
4277 sctp_copy_sock(sock->sk, sk, asoc);
4278
4279 /* Make peeled-off sockets more like 1-1 accepted sockets.
4280 * Set the daddr and initialize id to something more random
4281 */
4282 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4283 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4284
4285 /* Populate the fields of the newsk from the oldsk and migrate the
4286 * asoc to the newsk.
4287 */
4288 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4289
4290 *sockp = sock;
4291
4292 return err;
4293 }
4294 EXPORT_SYMBOL(sctp_do_peeloff);
4295
4296 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4297 {
4298 sctp_peeloff_arg_t peeloff;
4299 struct socket *newsock;
4300 struct file *newfile;
4301 int retval = 0;
4302
4303 if (len < sizeof(sctp_peeloff_arg_t))
4304 return -EINVAL;
4305 len = sizeof(sctp_peeloff_arg_t);
4306 if (copy_from_user(&peeloff, optval, len))
4307 return -EFAULT;
4308
4309 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4310 if (retval < 0)
4311 goto out;
4312
4313 /* Map the socket to an unused fd that can be returned to the user. */
4314 retval = get_unused_fd();
4315 if (retval < 0) {
4316 sock_release(newsock);
4317 goto out;
4318 }
4319
4320 newfile = sock_alloc_file(newsock, 0, NULL);
4321 if (unlikely(IS_ERR(newfile))) {
4322 put_unused_fd(retval);
4323 sock_release(newsock);
4324 return PTR_ERR(newfile);
4325 }
4326
4327 SCTP_DEBUG_PRINTK("%s: sk: %p newsk: %p sd: %d\n",
4328 __func__, sk, newsock->sk, retval);
4329
4330 /* Return the fd mapped to the new socket. */
4331 if (put_user(len, optlen)) {
4332 fput(newfile);
4333 put_unused_fd(retval);
4334 return -EFAULT;
4335 }
4336 peeloff.sd = retval;
4337 if (copy_to_user(optval, &peeloff, len)) {
4338 fput(newfile);
4339 put_unused_fd(retval);
4340 return -EFAULT;
4341 }
4342 fd_install(retval, newfile);
4343 out:
4344 return retval;
4345 }
4346
4347 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4348 *
4349 * Applications can enable or disable heartbeats for any peer address of
4350 * an association, modify an address's heartbeat interval, force a
4351 * heartbeat to be sent immediately, and adjust the address's maximum
4352 * number of retransmissions sent before an address is considered
4353 * unreachable. The following structure is used to access and modify an
4354 * address's parameters:
4355 *
4356 * struct sctp_paddrparams {
4357 * sctp_assoc_t spp_assoc_id;
4358 * struct sockaddr_storage spp_address;
4359 * uint32_t spp_hbinterval;
4360 * uint16_t spp_pathmaxrxt;
4361 * uint32_t spp_pathmtu;
4362 * uint32_t spp_sackdelay;
4363 * uint32_t spp_flags;
4364 * };
4365 *
4366 * spp_assoc_id - (one-to-many style socket) This is filled in the
4367 * application, and identifies the association for
4368 * this query.
4369 * spp_address - This specifies which address is of interest.
4370 * spp_hbinterval - This contains the value of the heartbeat interval,
4371 * in milliseconds. If a value of zero
4372 * is present in this field then no changes are to
4373 * be made to this parameter.
4374 * spp_pathmaxrxt - This contains the maximum number of
4375 * retransmissions before this address shall be
4376 * considered unreachable. If a value of zero
4377 * is present in this field then no changes are to
4378 * be made to this parameter.
4379 * spp_pathmtu - When Path MTU discovery is disabled the value
4380 * specified here will be the "fixed" path mtu.
4381 * Note that if the spp_address field is empty
4382 * then all associations on this address will
4383 * have this fixed path mtu set upon them.
4384 *
4385 * spp_sackdelay - When delayed sack is enabled, this value specifies
4386 * the number of milliseconds that sacks will be delayed
4387 * for. This value will apply to all addresses of an
4388 * association if the spp_address field is empty. Note
4389 * also, that if delayed sack is enabled and this
4390 * value is set to 0, no change is made to the last
4391 * recorded delayed sack timer value.
4392 *
4393 * spp_flags - These flags are used to control various features
4394 * on an association. The flag field may contain
4395 * zero or more of the following options.
4396 *
4397 * SPP_HB_ENABLE - Enable heartbeats on the
4398 * specified address. Note that if the address
4399 * field is empty all addresses for the association
4400 * have heartbeats enabled upon them.
4401 *
4402 * SPP_HB_DISABLE - Disable heartbeats on the
4403 * speicifed address. Note that if the address
4404 * field is empty all addresses for the association
4405 * will have their heartbeats disabled. Note also
4406 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4407 * mutually exclusive, only one of these two should
4408 * be specified. Enabling both fields will have
4409 * undetermined results.
4410 *
4411 * SPP_HB_DEMAND - Request a user initiated heartbeat
4412 * to be made immediately.
4413 *
4414 * SPP_PMTUD_ENABLE - This field will enable PMTU
4415 * discovery upon the specified address. Note that
4416 * if the address feild is empty then all addresses
4417 * on the association are effected.
4418 *
4419 * SPP_PMTUD_DISABLE - This field will disable PMTU
4420 * discovery upon the specified address. Note that
4421 * if the address feild is empty then all addresses
4422 * on the association are effected. Not also that
4423 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4424 * exclusive. Enabling both will have undetermined
4425 * results.
4426 *
4427 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4428 * on delayed sack. The time specified in spp_sackdelay
4429 * is used to specify the sack delay for this address. Note
4430 * that if spp_address is empty then all addresses will
4431 * enable delayed sack and take on the sack delay
4432 * value specified in spp_sackdelay.
4433 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4434 * off delayed sack. If the spp_address field is blank then
4435 * delayed sack is disabled for the entire association. Note
4436 * also that this field is mutually exclusive to
4437 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4438 * results.
4439 */
4440 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4441 char __user *optval, int __user *optlen)
4442 {
4443 struct sctp_paddrparams params;
4444 struct sctp_transport *trans = NULL;
4445 struct sctp_association *asoc = NULL;
4446 struct sctp_sock *sp = sctp_sk(sk);
4447
4448 if (len < sizeof(struct sctp_paddrparams))
4449 return -EINVAL;
4450 len = sizeof(struct sctp_paddrparams);
4451 if (copy_from_user(&params, optval, len))
4452 return -EFAULT;
4453
4454 /* If an address other than INADDR_ANY is specified, and
4455 * no transport is found, then the request is invalid.
4456 */
4457 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
4458 trans = sctp_addr_id2transport(sk, &params.spp_address,
4459 params.spp_assoc_id);
4460 if (!trans) {
4461 SCTP_DEBUG_PRINTK("Failed no transport\n");
4462 return -EINVAL;
4463 }
4464 }
4465
4466 /* Get association, if assoc_id != 0 and the socket is a one
4467 * to many style socket, and an association was not found, then
4468 * the id was invalid.
4469 */
4470 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4471 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4472 SCTP_DEBUG_PRINTK("Failed no association\n");
4473 return -EINVAL;
4474 }
4475
4476 if (trans) {
4477 /* Fetch transport values. */
4478 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4479 params.spp_pathmtu = trans->pathmtu;
4480 params.spp_pathmaxrxt = trans->pathmaxrxt;
4481 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4482
4483 /*draft-11 doesn't say what to return in spp_flags*/
4484 params.spp_flags = trans->param_flags;
4485 } else if (asoc) {
4486 /* Fetch association values. */
4487 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4488 params.spp_pathmtu = asoc->pathmtu;
4489 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4490 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4491
4492 /*draft-11 doesn't say what to return in spp_flags*/
4493 params.spp_flags = asoc->param_flags;
4494 } else {
4495 /* Fetch socket values. */
4496 params.spp_hbinterval = sp->hbinterval;
4497 params.spp_pathmtu = sp->pathmtu;
4498 params.spp_sackdelay = sp->sackdelay;
4499 params.spp_pathmaxrxt = sp->pathmaxrxt;
4500
4501 /*draft-11 doesn't say what to return in spp_flags*/
4502 params.spp_flags = sp->param_flags;
4503 }
4504
4505 if (copy_to_user(optval, &params, len))
4506 return -EFAULT;
4507
4508 if (put_user(len, optlen))
4509 return -EFAULT;
4510
4511 return 0;
4512 }
4513
4514 /*
4515 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4516 *
4517 * This option will effect the way delayed acks are performed. This
4518 * option allows you to get or set the delayed ack time, in
4519 * milliseconds. It also allows changing the delayed ack frequency.
4520 * Changing the frequency to 1 disables the delayed sack algorithm. If
4521 * the assoc_id is 0, then this sets or gets the endpoints default
4522 * values. If the assoc_id field is non-zero, then the set or get
4523 * effects the specified association for the one to many model (the
4524 * assoc_id field is ignored by the one to one model). Note that if
4525 * sack_delay or sack_freq are 0 when setting this option, then the
4526 * current values will remain unchanged.
4527 *
4528 * struct sctp_sack_info {
4529 * sctp_assoc_t sack_assoc_id;
4530 * uint32_t sack_delay;
4531 * uint32_t sack_freq;
4532 * };
4533 *
4534 * sack_assoc_id - This parameter, indicates which association the user
4535 * is performing an action upon. Note that if this field's value is
4536 * zero then the endpoints default value is changed (effecting future
4537 * associations only).
4538 *
4539 * sack_delay - This parameter contains the number of milliseconds that
4540 * the user is requesting the delayed ACK timer be set to. Note that
4541 * this value is defined in the standard to be between 200 and 500
4542 * milliseconds.
4543 *
4544 * sack_freq - This parameter contains the number of packets that must
4545 * be received before a sack is sent without waiting for the delay
4546 * timer to expire. The default value for this is 2, setting this
4547 * value to 1 will disable the delayed sack algorithm.
4548 */
4549 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4550 char __user *optval,
4551 int __user *optlen)
4552 {
4553 struct sctp_sack_info params;
4554 struct sctp_association *asoc = NULL;
4555 struct sctp_sock *sp = sctp_sk(sk);
4556
4557 if (len >= sizeof(struct sctp_sack_info)) {
4558 len = sizeof(struct sctp_sack_info);
4559
4560 if (copy_from_user(&params, optval, len))
4561 return -EFAULT;
4562 } else if (len == sizeof(struct sctp_assoc_value)) {
4563 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4564 pr_warn("Use struct sctp_sack_info instead\n");
4565 if (copy_from_user(&params, optval, len))
4566 return -EFAULT;
4567 } else
4568 return - EINVAL;
4569
4570 /* Get association, if sack_assoc_id != 0 and the socket is a one
4571 * to many style socket, and an association was not found, then
4572 * the id was invalid.
4573 */
4574 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4575 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4576 return -EINVAL;
4577
4578 if (asoc) {
4579 /* Fetch association values. */
4580 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4581 params.sack_delay = jiffies_to_msecs(
4582 asoc->sackdelay);
4583 params.sack_freq = asoc->sackfreq;
4584
4585 } else {
4586 params.sack_delay = 0;
4587 params.sack_freq = 1;
4588 }
4589 } else {
4590 /* Fetch socket values. */
4591 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4592 params.sack_delay = sp->sackdelay;
4593 params.sack_freq = sp->sackfreq;
4594 } else {
4595 params.sack_delay = 0;
4596 params.sack_freq = 1;
4597 }
4598 }
4599
4600 if (copy_to_user(optval, &params, len))
4601 return -EFAULT;
4602
4603 if (put_user(len, optlen))
4604 return -EFAULT;
4605
4606 return 0;
4607 }
4608
4609 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4610 *
4611 * Applications can specify protocol parameters for the default association
4612 * initialization. The option name argument to setsockopt() and getsockopt()
4613 * is SCTP_INITMSG.
4614 *
4615 * Setting initialization parameters is effective only on an unconnected
4616 * socket (for UDP-style sockets only future associations are effected
4617 * by the change). With TCP-style sockets, this option is inherited by
4618 * sockets derived from a listener socket.
4619 */
4620 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4621 {
4622 if (len < sizeof(struct sctp_initmsg))
4623 return -EINVAL;
4624 len = sizeof(struct sctp_initmsg);
4625 if (put_user(len, optlen))
4626 return -EFAULT;
4627 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4628 return -EFAULT;
4629 return 0;
4630 }
4631
4632
4633 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4634 char __user *optval, int __user *optlen)
4635 {
4636 struct sctp_association *asoc;
4637 int cnt = 0;
4638 struct sctp_getaddrs getaddrs;
4639 struct sctp_transport *from;
4640 void __user *to;
4641 union sctp_addr temp;
4642 struct sctp_sock *sp = sctp_sk(sk);
4643 int addrlen;
4644 size_t space_left;
4645 int bytes_copied;
4646
4647 if (len < sizeof(struct sctp_getaddrs))
4648 return -EINVAL;
4649
4650 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4651 return -EFAULT;
4652
4653 /* For UDP-style sockets, id specifies the association to query. */
4654 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4655 if (!asoc)
4656 return -EINVAL;
4657
4658 to = optval + offsetof(struct sctp_getaddrs,addrs);
4659 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4660
4661 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4662 transports) {
4663 memcpy(&temp, &from->ipaddr, sizeof(temp));
4664 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4665 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4666 if (space_left < addrlen)
4667 return -ENOMEM;
4668 if (copy_to_user(to, &temp, addrlen))
4669 return -EFAULT;
4670 to += addrlen;
4671 cnt++;
4672 space_left -= addrlen;
4673 }
4674
4675 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4676 return -EFAULT;
4677 bytes_copied = ((char __user *)to) - optval;
4678 if (put_user(bytes_copied, optlen))
4679 return -EFAULT;
4680
4681 return 0;
4682 }
4683
4684 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4685 size_t space_left, int *bytes_copied)
4686 {
4687 struct sctp_sockaddr_entry *addr;
4688 union sctp_addr temp;
4689 int cnt = 0;
4690 int addrlen;
4691 struct net *net = sock_net(sk);
4692
4693 rcu_read_lock();
4694 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
4695 if (!addr->valid)
4696 continue;
4697
4698 if ((PF_INET == sk->sk_family) &&
4699 (AF_INET6 == addr->a.sa.sa_family))
4700 continue;
4701 if ((PF_INET6 == sk->sk_family) &&
4702 inet_v6_ipv6only(sk) &&
4703 (AF_INET == addr->a.sa.sa_family))
4704 continue;
4705 memcpy(&temp, &addr->a, sizeof(temp));
4706 if (!temp.v4.sin_port)
4707 temp.v4.sin_port = htons(port);
4708
4709 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4710 &temp);
4711 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4712 if (space_left < addrlen) {
4713 cnt = -ENOMEM;
4714 break;
4715 }
4716 memcpy(to, &temp, addrlen);
4717
4718 to += addrlen;
4719 cnt ++;
4720 space_left -= addrlen;
4721 *bytes_copied += addrlen;
4722 }
4723 rcu_read_unlock();
4724
4725 return cnt;
4726 }
4727
4728
4729 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4730 char __user *optval, int __user *optlen)
4731 {
4732 struct sctp_bind_addr *bp;
4733 struct sctp_association *asoc;
4734 int cnt = 0;
4735 struct sctp_getaddrs getaddrs;
4736 struct sctp_sockaddr_entry *addr;
4737 void __user *to;
4738 union sctp_addr temp;
4739 struct sctp_sock *sp = sctp_sk(sk);
4740 int addrlen;
4741 int err = 0;
4742 size_t space_left;
4743 int bytes_copied = 0;
4744 void *addrs;
4745 void *buf;
4746
4747 if (len < sizeof(struct sctp_getaddrs))
4748 return -EINVAL;
4749
4750 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4751 return -EFAULT;
4752
4753 /*
4754 * For UDP-style sockets, id specifies the association to query.
4755 * If the id field is set to the value '0' then the locally bound
4756 * addresses are returned without regard to any particular
4757 * association.
4758 */
4759 if (0 == getaddrs.assoc_id) {
4760 bp = &sctp_sk(sk)->ep->base.bind_addr;
4761 } else {
4762 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4763 if (!asoc)
4764 return -EINVAL;
4765 bp = &asoc->base.bind_addr;
4766 }
4767
4768 to = optval + offsetof(struct sctp_getaddrs,addrs);
4769 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4770
4771 addrs = kmalloc(space_left, GFP_KERNEL);
4772 if (!addrs)
4773 return -ENOMEM;
4774
4775 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4776 * addresses from the global local address list.
4777 */
4778 if (sctp_list_single_entry(&bp->address_list)) {
4779 addr = list_entry(bp->address_list.next,
4780 struct sctp_sockaddr_entry, list);
4781 if (sctp_is_any(sk, &addr->a)) {
4782 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4783 space_left, &bytes_copied);
4784 if (cnt < 0) {
4785 err = cnt;
4786 goto out;
4787 }
4788 goto copy_getaddrs;
4789 }
4790 }
4791
4792 buf = addrs;
4793 /* Protection on the bound address list is not needed since
4794 * in the socket option context we hold a socket lock and
4795 * thus the bound address list can't change.
4796 */
4797 list_for_each_entry(addr, &bp->address_list, list) {
4798 memcpy(&temp, &addr->a, sizeof(temp));
4799 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4800 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4801 if (space_left < addrlen) {
4802 err = -ENOMEM; /*fixme: right error?*/
4803 goto out;
4804 }
4805 memcpy(buf, &temp, addrlen);
4806 buf += addrlen;
4807 bytes_copied += addrlen;
4808 cnt ++;
4809 space_left -= addrlen;
4810 }
4811
4812 copy_getaddrs:
4813 if (copy_to_user(to, addrs, bytes_copied)) {
4814 err = -EFAULT;
4815 goto out;
4816 }
4817 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4818 err = -EFAULT;
4819 goto out;
4820 }
4821 if (put_user(bytes_copied, optlen))
4822 err = -EFAULT;
4823 out:
4824 kfree(addrs);
4825 return err;
4826 }
4827
4828 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4829 *
4830 * Requests that the local SCTP stack use the enclosed peer address as
4831 * the association primary. The enclosed address must be one of the
4832 * association peer's addresses.
4833 */
4834 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4835 char __user *optval, int __user *optlen)
4836 {
4837 struct sctp_prim prim;
4838 struct sctp_association *asoc;
4839 struct sctp_sock *sp = sctp_sk(sk);
4840
4841 if (len < sizeof(struct sctp_prim))
4842 return -EINVAL;
4843
4844 len = sizeof(struct sctp_prim);
4845
4846 if (copy_from_user(&prim, optval, len))
4847 return -EFAULT;
4848
4849 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4850 if (!asoc)
4851 return -EINVAL;
4852
4853 if (!asoc->peer.primary_path)
4854 return -ENOTCONN;
4855
4856 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4857 asoc->peer.primary_path->af_specific->sockaddr_len);
4858
4859 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4860 (union sctp_addr *)&prim.ssp_addr);
4861
4862 if (put_user(len, optlen))
4863 return -EFAULT;
4864 if (copy_to_user(optval, &prim, len))
4865 return -EFAULT;
4866
4867 return 0;
4868 }
4869
4870 /*
4871 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4872 *
4873 * Requests that the local endpoint set the specified Adaptation Layer
4874 * Indication parameter for all future INIT and INIT-ACK exchanges.
4875 */
4876 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4877 char __user *optval, int __user *optlen)
4878 {
4879 struct sctp_setadaptation adaptation;
4880
4881 if (len < sizeof(struct sctp_setadaptation))
4882 return -EINVAL;
4883
4884 len = sizeof(struct sctp_setadaptation);
4885
4886 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4887
4888 if (put_user(len, optlen))
4889 return -EFAULT;
4890 if (copy_to_user(optval, &adaptation, len))
4891 return -EFAULT;
4892
4893 return 0;
4894 }
4895
4896 /*
4897 *
4898 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4899 *
4900 * Applications that wish to use the sendto() system call may wish to
4901 * specify a default set of parameters that would normally be supplied
4902 * through the inclusion of ancillary data. This socket option allows
4903 * such an application to set the default sctp_sndrcvinfo structure.
4904
4905
4906 * The application that wishes to use this socket option simply passes
4907 * in to this call the sctp_sndrcvinfo structure defined in Section
4908 * 5.2.2) The input parameters accepted by this call include
4909 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4910 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4911 * to this call if the caller is using the UDP model.
4912 *
4913 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4914 */
4915 static int sctp_getsockopt_default_send_param(struct sock *sk,
4916 int len, char __user *optval,
4917 int __user *optlen)
4918 {
4919 struct sctp_sndrcvinfo info;
4920 struct sctp_association *asoc;
4921 struct sctp_sock *sp = sctp_sk(sk);
4922
4923 if (len < sizeof(struct sctp_sndrcvinfo))
4924 return -EINVAL;
4925
4926 len = sizeof(struct sctp_sndrcvinfo);
4927
4928 if (copy_from_user(&info, optval, len))
4929 return -EFAULT;
4930
4931 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4932 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4933 return -EINVAL;
4934
4935 if (asoc) {
4936 info.sinfo_stream = asoc->default_stream;
4937 info.sinfo_flags = asoc->default_flags;
4938 info.sinfo_ppid = asoc->default_ppid;
4939 info.sinfo_context = asoc->default_context;
4940 info.sinfo_timetolive = asoc->default_timetolive;
4941 } else {
4942 info.sinfo_stream = sp->default_stream;
4943 info.sinfo_flags = sp->default_flags;
4944 info.sinfo_ppid = sp->default_ppid;
4945 info.sinfo_context = sp->default_context;
4946 info.sinfo_timetolive = sp->default_timetolive;
4947 }
4948
4949 if (put_user(len, optlen))
4950 return -EFAULT;
4951 if (copy_to_user(optval, &info, len))
4952 return -EFAULT;
4953
4954 return 0;
4955 }
4956
4957 /*
4958 *
4959 * 7.1.5 SCTP_NODELAY
4960 *
4961 * Turn on/off any Nagle-like algorithm. This means that packets are
4962 * generally sent as soon as possible and no unnecessary delays are
4963 * introduced, at the cost of more packets in the network. Expects an
4964 * integer boolean flag.
4965 */
4966
4967 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4968 char __user *optval, int __user *optlen)
4969 {
4970 int val;
4971
4972 if (len < sizeof(int))
4973 return -EINVAL;
4974
4975 len = sizeof(int);
4976 val = (sctp_sk(sk)->nodelay == 1);
4977 if (put_user(len, optlen))
4978 return -EFAULT;
4979 if (copy_to_user(optval, &val, len))
4980 return -EFAULT;
4981 return 0;
4982 }
4983
4984 /*
4985 *
4986 * 7.1.1 SCTP_RTOINFO
4987 *
4988 * The protocol parameters used to initialize and bound retransmission
4989 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4990 * and modify these parameters.
4991 * All parameters are time values, in milliseconds. A value of 0, when
4992 * modifying the parameters, indicates that the current value should not
4993 * be changed.
4994 *
4995 */
4996 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4997 char __user *optval,
4998 int __user *optlen) {
4999 struct sctp_rtoinfo rtoinfo;
5000 struct sctp_association *asoc;
5001
5002 if (len < sizeof (struct sctp_rtoinfo))
5003 return -EINVAL;
5004
5005 len = sizeof(struct sctp_rtoinfo);
5006
5007 if (copy_from_user(&rtoinfo, optval, len))
5008 return -EFAULT;
5009
5010 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5011
5012 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5013 return -EINVAL;
5014
5015 /* Values corresponding to the specific association. */
5016 if (asoc) {
5017 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5018 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5019 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5020 } else {
5021 /* Values corresponding to the endpoint. */
5022 struct sctp_sock *sp = sctp_sk(sk);
5023
5024 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5025 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5026 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5027 }
5028
5029 if (put_user(len, optlen))
5030 return -EFAULT;
5031
5032 if (copy_to_user(optval, &rtoinfo, len))
5033 return -EFAULT;
5034
5035 return 0;
5036 }
5037
5038 /*
5039 *
5040 * 7.1.2 SCTP_ASSOCINFO
5041 *
5042 * This option is used to tune the maximum retransmission attempts
5043 * of the association.
5044 * Returns an error if the new association retransmission value is
5045 * greater than the sum of the retransmission value of the peer.
5046 * See [SCTP] for more information.
5047 *
5048 */
5049 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5050 char __user *optval,
5051 int __user *optlen)
5052 {
5053
5054 struct sctp_assocparams assocparams;
5055 struct sctp_association *asoc;
5056 struct list_head *pos;
5057 int cnt = 0;
5058
5059 if (len < sizeof (struct sctp_assocparams))
5060 return -EINVAL;
5061
5062 len = sizeof(struct sctp_assocparams);
5063
5064 if (copy_from_user(&assocparams, optval, len))
5065 return -EFAULT;
5066
5067 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5068
5069 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5070 return -EINVAL;
5071
5072 /* Values correspoinding to the specific association */
5073 if (asoc) {
5074 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5075 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5076 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5077 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
5078 * 1000) +
5079 (asoc->cookie_life.tv_usec
5080 / 1000);
5081
5082 list_for_each(pos, &asoc->peer.transport_addr_list) {
5083 cnt ++;
5084 }
5085
5086 assocparams.sasoc_number_peer_destinations = cnt;
5087 } else {
5088 /* Values corresponding to the endpoint */
5089 struct sctp_sock *sp = sctp_sk(sk);
5090
5091 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5092 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5093 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5094 assocparams.sasoc_cookie_life =
5095 sp->assocparams.sasoc_cookie_life;
5096 assocparams.sasoc_number_peer_destinations =
5097 sp->assocparams.
5098 sasoc_number_peer_destinations;
5099 }
5100
5101 if (put_user(len, optlen))
5102 return -EFAULT;
5103
5104 if (copy_to_user(optval, &assocparams, len))
5105 return -EFAULT;
5106
5107 return 0;
5108 }
5109
5110 /*
5111 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5112 *
5113 * This socket option is a boolean flag which turns on or off mapped V4
5114 * addresses. If this option is turned on and the socket is type
5115 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5116 * If this option is turned off, then no mapping will be done of V4
5117 * addresses and a user will receive both PF_INET6 and PF_INET type
5118 * addresses on the socket.
5119 */
5120 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5121 char __user *optval, int __user *optlen)
5122 {
5123 int val;
5124 struct sctp_sock *sp = sctp_sk(sk);
5125
5126 if (len < sizeof(int))
5127 return -EINVAL;
5128
5129 len = sizeof(int);
5130 val = sp->v4mapped;
5131 if (put_user(len, optlen))
5132 return -EFAULT;
5133 if (copy_to_user(optval, &val, len))
5134 return -EFAULT;
5135
5136 return 0;
5137 }
5138
5139 /*
5140 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5141 * (chapter and verse is quoted at sctp_setsockopt_context())
5142 */
5143 static int sctp_getsockopt_context(struct sock *sk, int len,
5144 char __user *optval, int __user *optlen)
5145 {
5146 struct sctp_assoc_value params;
5147 struct sctp_sock *sp;
5148 struct sctp_association *asoc;
5149
5150 if (len < sizeof(struct sctp_assoc_value))
5151 return -EINVAL;
5152
5153 len = sizeof(struct sctp_assoc_value);
5154
5155 if (copy_from_user(&params, optval, len))
5156 return -EFAULT;
5157
5158 sp = sctp_sk(sk);
5159
5160 if (params.assoc_id != 0) {
5161 asoc = sctp_id2assoc(sk, params.assoc_id);
5162 if (!asoc)
5163 return -EINVAL;
5164 params.assoc_value = asoc->default_rcv_context;
5165 } else {
5166 params.assoc_value = sp->default_rcv_context;
5167 }
5168
5169 if (put_user(len, optlen))
5170 return -EFAULT;
5171 if (copy_to_user(optval, &params, len))
5172 return -EFAULT;
5173
5174 return 0;
5175 }
5176
5177 /*
5178 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5179 * This option will get or set the maximum size to put in any outgoing
5180 * SCTP DATA chunk. If a message is larger than this size it will be
5181 * fragmented by SCTP into the specified size. Note that the underlying
5182 * SCTP implementation may fragment into smaller sized chunks when the
5183 * PMTU of the underlying association is smaller than the value set by
5184 * the user. The default value for this option is '0' which indicates
5185 * the user is NOT limiting fragmentation and only the PMTU will effect
5186 * SCTP's choice of DATA chunk size. Note also that values set larger
5187 * than the maximum size of an IP datagram will effectively let SCTP
5188 * control fragmentation (i.e. the same as setting this option to 0).
5189 *
5190 * The following structure is used to access and modify this parameter:
5191 *
5192 * struct sctp_assoc_value {
5193 * sctp_assoc_t assoc_id;
5194 * uint32_t assoc_value;
5195 * };
5196 *
5197 * assoc_id: This parameter is ignored for one-to-one style sockets.
5198 * For one-to-many style sockets this parameter indicates which
5199 * association the user is performing an action upon. Note that if
5200 * this field's value is zero then the endpoints default value is
5201 * changed (effecting future associations only).
5202 * assoc_value: This parameter specifies the maximum size in bytes.
5203 */
5204 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5205 char __user *optval, int __user *optlen)
5206 {
5207 struct sctp_assoc_value params;
5208 struct sctp_association *asoc;
5209
5210 if (len == sizeof(int)) {
5211 pr_warn("Use of int in maxseg socket option deprecated\n");
5212 pr_warn("Use struct sctp_assoc_value instead\n");
5213 params.assoc_id = 0;
5214 } else if (len >= sizeof(struct sctp_assoc_value)) {
5215 len = sizeof(struct sctp_assoc_value);
5216 if (copy_from_user(&params, optval, sizeof(params)))
5217 return -EFAULT;
5218 } else
5219 return -EINVAL;
5220
5221 asoc = sctp_id2assoc(sk, params.assoc_id);
5222 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5223 return -EINVAL;
5224
5225 if (asoc)
5226 params.assoc_value = asoc->frag_point;
5227 else
5228 params.assoc_value = sctp_sk(sk)->user_frag;
5229
5230 if (put_user(len, optlen))
5231 return -EFAULT;
5232 if (len == sizeof(int)) {
5233 if (copy_to_user(optval, &params.assoc_value, len))
5234 return -EFAULT;
5235 } else {
5236 if (copy_to_user(optval, &params, len))
5237 return -EFAULT;
5238 }
5239
5240 return 0;
5241 }
5242
5243 /*
5244 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5245 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5246 */
5247 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5248 char __user *optval, int __user *optlen)
5249 {
5250 int val;
5251
5252 if (len < sizeof(int))
5253 return -EINVAL;
5254
5255 len = sizeof(int);
5256
5257 val = sctp_sk(sk)->frag_interleave;
5258 if (put_user(len, optlen))
5259 return -EFAULT;
5260 if (copy_to_user(optval, &val, len))
5261 return -EFAULT;
5262
5263 return 0;
5264 }
5265
5266 /*
5267 * 7.1.25. Set or Get the sctp partial delivery point
5268 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5269 */
5270 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5271 char __user *optval,
5272 int __user *optlen)
5273 {
5274 u32 val;
5275
5276 if (len < sizeof(u32))
5277 return -EINVAL;
5278
5279 len = sizeof(u32);
5280
5281 val = sctp_sk(sk)->pd_point;
5282 if (put_user(len, optlen))
5283 return -EFAULT;
5284 if (copy_to_user(optval, &val, len))
5285 return -EFAULT;
5286
5287 return 0;
5288 }
5289
5290 /*
5291 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5292 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5293 */
5294 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5295 char __user *optval,
5296 int __user *optlen)
5297 {
5298 struct sctp_assoc_value params;
5299 struct sctp_sock *sp;
5300 struct sctp_association *asoc;
5301
5302 if (len == sizeof(int)) {
5303 pr_warn("Use of int in max_burst socket option deprecated\n");
5304 pr_warn("Use struct sctp_assoc_value instead\n");
5305 params.assoc_id = 0;
5306 } else if (len >= sizeof(struct sctp_assoc_value)) {
5307 len = sizeof(struct sctp_assoc_value);
5308 if (copy_from_user(&params, optval, len))
5309 return -EFAULT;
5310 } else
5311 return -EINVAL;
5312
5313 sp = sctp_sk(sk);
5314
5315 if (params.assoc_id != 0) {
5316 asoc = sctp_id2assoc(sk, params.assoc_id);
5317 if (!asoc)
5318 return -EINVAL;
5319 params.assoc_value = asoc->max_burst;
5320 } else
5321 params.assoc_value = sp->max_burst;
5322
5323 if (len == sizeof(int)) {
5324 if (copy_to_user(optval, &params.assoc_value, len))
5325 return -EFAULT;
5326 } else {
5327 if (copy_to_user(optval, &params, len))
5328 return -EFAULT;
5329 }
5330
5331 return 0;
5332
5333 }
5334
5335 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5336 char __user *optval, int __user *optlen)
5337 {
5338 struct net *net = sock_net(sk);
5339 struct sctp_hmacalgo __user *p = (void __user *)optval;
5340 struct sctp_hmac_algo_param *hmacs;
5341 __u16 data_len = 0;
5342 u32 num_idents;
5343
5344 if (!net->sctp.auth_enable)
5345 return -EACCES;
5346
5347 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5348 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5349
5350 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5351 return -EINVAL;
5352
5353 len = sizeof(struct sctp_hmacalgo) + data_len;
5354 num_idents = data_len / sizeof(u16);
5355
5356 if (put_user(len, optlen))
5357 return -EFAULT;
5358 if (put_user(num_idents, &p->shmac_num_idents))
5359 return -EFAULT;
5360 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5361 return -EFAULT;
5362 return 0;
5363 }
5364
5365 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5366 char __user *optval, int __user *optlen)
5367 {
5368 struct net *net = sock_net(sk);
5369 struct sctp_authkeyid val;
5370 struct sctp_association *asoc;
5371
5372 if (!net->sctp.auth_enable)
5373 return -EACCES;
5374
5375 if (len < sizeof(struct sctp_authkeyid))
5376 return -EINVAL;
5377 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5378 return -EFAULT;
5379
5380 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5381 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5382 return -EINVAL;
5383
5384 if (asoc)
5385 val.scact_keynumber = asoc->active_key_id;
5386 else
5387 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5388
5389 len = sizeof(struct sctp_authkeyid);
5390 if (put_user(len, optlen))
5391 return -EFAULT;
5392 if (copy_to_user(optval, &val, len))
5393 return -EFAULT;
5394
5395 return 0;
5396 }
5397
5398 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5399 char __user *optval, int __user *optlen)
5400 {
5401 struct net *net = sock_net(sk);
5402 struct sctp_authchunks __user *p = (void __user *)optval;
5403 struct sctp_authchunks val;
5404 struct sctp_association *asoc;
5405 struct sctp_chunks_param *ch;
5406 u32 num_chunks = 0;
5407 char __user *to;
5408
5409 if (!net->sctp.auth_enable)
5410 return -EACCES;
5411
5412 if (len < sizeof(struct sctp_authchunks))
5413 return -EINVAL;
5414
5415 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5416 return -EFAULT;
5417
5418 to = p->gauth_chunks;
5419 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5420 if (!asoc)
5421 return -EINVAL;
5422
5423 ch = asoc->peer.peer_chunks;
5424 if (!ch)
5425 goto num;
5426
5427 /* See if the user provided enough room for all the data */
5428 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5429 if (len < num_chunks)
5430 return -EINVAL;
5431
5432 if (copy_to_user(to, ch->chunks, num_chunks))
5433 return -EFAULT;
5434 num:
5435 len = sizeof(struct sctp_authchunks) + num_chunks;
5436 if (put_user(len, optlen)) return -EFAULT;
5437 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5438 return -EFAULT;
5439 return 0;
5440 }
5441
5442 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5443 char __user *optval, int __user *optlen)
5444 {
5445 struct net *net = sock_net(sk);
5446 struct sctp_authchunks __user *p = (void __user *)optval;
5447 struct sctp_authchunks val;
5448 struct sctp_association *asoc;
5449 struct sctp_chunks_param *ch;
5450 u32 num_chunks = 0;
5451 char __user *to;
5452
5453 if (!net->sctp.auth_enable)
5454 return -EACCES;
5455
5456 if (len < sizeof(struct sctp_authchunks))
5457 return -EINVAL;
5458
5459 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5460 return -EFAULT;
5461
5462 to = p->gauth_chunks;
5463 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5464 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5465 return -EINVAL;
5466
5467 if (asoc)
5468 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5469 else
5470 ch = sctp_sk(sk)->ep->auth_chunk_list;
5471
5472 if (!ch)
5473 goto num;
5474
5475 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5476 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5477 return -EINVAL;
5478
5479 if (copy_to_user(to, ch->chunks, num_chunks))
5480 return -EFAULT;
5481 num:
5482 len = sizeof(struct sctp_authchunks) + num_chunks;
5483 if (put_user(len, optlen))
5484 return -EFAULT;
5485 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5486 return -EFAULT;
5487
5488 return 0;
5489 }
5490
5491 /*
5492 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5493 * This option gets the current number of associations that are attached
5494 * to a one-to-many style socket. The option value is an uint32_t.
5495 */
5496 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5497 char __user *optval, int __user *optlen)
5498 {
5499 struct sctp_sock *sp = sctp_sk(sk);
5500 struct sctp_association *asoc;
5501 u32 val = 0;
5502
5503 if (sctp_style(sk, TCP))
5504 return -EOPNOTSUPP;
5505
5506 if (len < sizeof(u32))
5507 return -EINVAL;
5508
5509 len = sizeof(u32);
5510
5511 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5512 val++;
5513 }
5514
5515 if (put_user(len, optlen))
5516 return -EFAULT;
5517 if (copy_to_user(optval, &val, len))
5518 return -EFAULT;
5519
5520 return 0;
5521 }
5522
5523 /*
5524 * 8.1.23 SCTP_AUTO_ASCONF
5525 * See the corresponding setsockopt entry as description
5526 */
5527 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5528 char __user *optval, int __user *optlen)
5529 {
5530 int val = 0;
5531
5532 if (len < sizeof(int))
5533 return -EINVAL;
5534
5535 len = sizeof(int);
5536 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5537 val = 1;
5538 if (put_user(len, optlen))
5539 return -EFAULT;
5540 if (copy_to_user(optval, &val, len))
5541 return -EFAULT;
5542 return 0;
5543 }
5544
5545 /*
5546 * 8.2.6. Get the Current Identifiers of Associations
5547 * (SCTP_GET_ASSOC_ID_LIST)
5548 *
5549 * This option gets the current list of SCTP association identifiers of
5550 * the SCTP associations handled by a one-to-many style socket.
5551 */
5552 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5553 char __user *optval, int __user *optlen)
5554 {
5555 struct sctp_sock *sp = sctp_sk(sk);
5556 struct sctp_association *asoc;
5557 struct sctp_assoc_ids *ids;
5558 u32 num = 0;
5559
5560 if (sctp_style(sk, TCP))
5561 return -EOPNOTSUPP;
5562
5563 if (len < sizeof(struct sctp_assoc_ids))
5564 return -EINVAL;
5565
5566 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5567 num++;
5568 }
5569
5570 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5571 return -EINVAL;
5572
5573 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5574
5575 ids = kmalloc(len, GFP_KERNEL);
5576 if (unlikely(!ids))
5577 return -ENOMEM;
5578
5579 ids->gaids_number_of_ids = num;
5580 num = 0;
5581 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5582 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5583 }
5584
5585 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5586 kfree(ids);
5587 return -EFAULT;
5588 }
5589
5590 kfree(ids);
5591 return 0;
5592 }
5593
5594 /*
5595 * SCTP_PEER_ADDR_THLDS
5596 *
5597 * This option allows us to fetch the partially failed threshold for one or all
5598 * transports in an association. See Section 6.1 of:
5599 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5600 */
5601 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
5602 char __user *optval,
5603 int len,
5604 int __user *optlen)
5605 {
5606 struct sctp_paddrthlds val;
5607 struct sctp_transport *trans;
5608 struct sctp_association *asoc;
5609
5610 if (len < sizeof(struct sctp_paddrthlds))
5611 return -EINVAL;
5612 len = sizeof(struct sctp_paddrthlds);
5613 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
5614 return -EFAULT;
5615
5616 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
5617 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
5618 if (!asoc)
5619 return -ENOENT;
5620
5621 val.spt_pathpfthld = asoc->pf_retrans;
5622 val.spt_pathmaxrxt = asoc->pathmaxrxt;
5623 } else {
5624 trans = sctp_addr_id2transport(sk, &val.spt_address,
5625 val.spt_assoc_id);
5626 if (!trans)
5627 return -ENOENT;
5628
5629 val.spt_pathmaxrxt = trans->pathmaxrxt;
5630 val.spt_pathpfthld = trans->pf_retrans;
5631 }
5632
5633 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
5634 return -EFAULT;
5635
5636 return 0;
5637 }
5638
5639 /*
5640 * SCTP_GET_ASSOC_STATS
5641 *
5642 * This option retrieves local per endpoint statistics. It is modeled
5643 * after OpenSolaris' implementation
5644 */
5645 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
5646 char __user *optval,
5647 int __user *optlen)
5648 {
5649 struct sctp_assoc_stats sas;
5650 struct sctp_association *asoc = NULL;
5651
5652 /* User must provide at least the assoc id */
5653 if (len < sizeof(sctp_assoc_t))
5654 return -EINVAL;
5655
5656 /* Allow the struct to grow and fill in as much as possible */
5657 len = min_t(size_t, len, sizeof(sas));
5658
5659 if (copy_from_user(&sas, optval, len))
5660 return -EFAULT;
5661
5662 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
5663 if (!asoc)
5664 return -EINVAL;
5665
5666 sas.sas_rtxchunks = asoc->stats.rtxchunks;
5667 sas.sas_gapcnt = asoc->stats.gapcnt;
5668 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
5669 sas.sas_osacks = asoc->stats.osacks;
5670 sas.sas_isacks = asoc->stats.isacks;
5671 sas.sas_octrlchunks = asoc->stats.octrlchunks;
5672 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
5673 sas.sas_oodchunks = asoc->stats.oodchunks;
5674 sas.sas_iodchunks = asoc->stats.iodchunks;
5675 sas.sas_ouodchunks = asoc->stats.ouodchunks;
5676 sas.sas_iuodchunks = asoc->stats.iuodchunks;
5677 sas.sas_idupchunks = asoc->stats.idupchunks;
5678 sas.sas_opackets = asoc->stats.opackets;
5679 sas.sas_ipackets = asoc->stats.ipackets;
5680
5681 /* New high max rto observed, will return 0 if not a single
5682 * RTO update took place. obs_rto_ipaddr will be bogus
5683 * in such a case
5684 */
5685 sas.sas_maxrto = asoc->stats.max_obs_rto;
5686 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
5687 sizeof(struct sockaddr_storage));
5688
5689 /* Mark beginning of a new observation period */
5690 asoc->stats.max_obs_rto = asoc->rto_min;
5691
5692 if (put_user(len, optlen))
5693 return -EFAULT;
5694
5695 SCTP_DEBUG_PRINTK("sctp_getsockopt_assoc_stat(%d): %d\n",
5696 len, sas.sas_assoc_id);
5697
5698 if (copy_to_user(optval, &sas, len))
5699 return -EFAULT;
5700
5701 return 0;
5702 }
5703
5704 static int sctp_getsockopt(struct sock *sk, int level, int optname,
5705 char __user *optval, int __user *optlen)
5706 {
5707 int retval = 0;
5708 int len;
5709
5710 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5711 sk, optname);
5712
5713 /* I can hardly begin to describe how wrong this is. This is
5714 * so broken as to be worse than useless. The API draft
5715 * REALLY is NOT helpful here... I am not convinced that the
5716 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5717 * are at all well-founded.
5718 */
5719 if (level != SOL_SCTP) {
5720 struct sctp_af *af = sctp_sk(sk)->pf->af;
5721
5722 retval = af->getsockopt(sk, level, optname, optval, optlen);
5723 return retval;
5724 }
5725
5726 if (get_user(len, optlen))
5727 return -EFAULT;
5728
5729 sctp_lock_sock(sk);
5730
5731 switch (optname) {
5732 case SCTP_STATUS:
5733 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5734 break;
5735 case SCTP_DISABLE_FRAGMENTS:
5736 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5737 optlen);
5738 break;
5739 case SCTP_EVENTS:
5740 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5741 break;
5742 case SCTP_AUTOCLOSE:
5743 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5744 break;
5745 case SCTP_SOCKOPT_PEELOFF:
5746 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5747 break;
5748 case SCTP_PEER_ADDR_PARAMS:
5749 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5750 optlen);
5751 break;
5752 case SCTP_DELAYED_SACK:
5753 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5754 optlen);
5755 break;
5756 case SCTP_INITMSG:
5757 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5758 break;
5759 case SCTP_GET_PEER_ADDRS:
5760 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5761 optlen);
5762 break;
5763 case SCTP_GET_LOCAL_ADDRS:
5764 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5765 optlen);
5766 break;
5767 case SCTP_SOCKOPT_CONNECTX3:
5768 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5769 break;
5770 case SCTP_DEFAULT_SEND_PARAM:
5771 retval = sctp_getsockopt_default_send_param(sk, len,
5772 optval, optlen);
5773 break;
5774 case SCTP_PRIMARY_ADDR:
5775 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5776 break;
5777 case SCTP_NODELAY:
5778 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5779 break;
5780 case SCTP_RTOINFO:
5781 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5782 break;
5783 case SCTP_ASSOCINFO:
5784 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5785 break;
5786 case SCTP_I_WANT_MAPPED_V4_ADDR:
5787 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5788 break;
5789 case SCTP_MAXSEG:
5790 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5791 break;
5792 case SCTP_GET_PEER_ADDR_INFO:
5793 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5794 optlen);
5795 break;
5796 case SCTP_ADAPTATION_LAYER:
5797 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5798 optlen);
5799 break;
5800 case SCTP_CONTEXT:
5801 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5802 break;
5803 case SCTP_FRAGMENT_INTERLEAVE:
5804 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5805 optlen);
5806 break;
5807 case SCTP_PARTIAL_DELIVERY_POINT:
5808 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5809 optlen);
5810 break;
5811 case SCTP_MAX_BURST:
5812 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5813 break;
5814 case SCTP_AUTH_KEY:
5815 case SCTP_AUTH_CHUNK:
5816 case SCTP_AUTH_DELETE_KEY:
5817 retval = -EOPNOTSUPP;
5818 break;
5819 case SCTP_HMAC_IDENT:
5820 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5821 break;
5822 case SCTP_AUTH_ACTIVE_KEY:
5823 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5824 break;
5825 case SCTP_PEER_AUTH_CHUNKS:
5826 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5827 optlen);
5828 break;
5829 case SCTP_LOCAL_AUTH_CHUNKS:
5830 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5831 optlen);
5832 break;
5833 case SCTP_GET_ASSOC_NUMBER:
5834 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5835 break;
5836 case SCTP_GET_ASSOC_ID_LIST:
5837 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5838 break;
5839 case SCTP_AUTO_ASCONF:
5840 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5841 break;
5842 case SCTP_PEER_ADDR_THLDS:
5843 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
5844 break;
5845 case SCTP_GET_ASSOC_STATS:
5846 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
5847 break;
5848 default:
5849 retval = -ENOPROTOOPT;
5850 break;
5851 }
5852
5853 sctp_release_sock(sk);
5854 return retval;
5855 }
5856
5857 static void sctp_hash(struct sock *sk)
5858 {
5859 /* STUB */
5860 }
5861
5862 static void sctp_unhash(struct sock *sk)
5863 {
5864 /* STUB */
5865 }
5866
5867 /* Check if port is acceptable. Possibly find first available port.
5868 *
5869 * The port hash table (contained in the 'global' SCTP protocol storage
5870 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5871 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5872 * list (the list number is the port number hashed out, so as you
5873 * would expect from a hash function, all the ports in a given list have
5874 * such a number that hashes out to the same list number; you were
5875 * expecting that, right?); so each list has a set of ports, with a
5876 * link to the socket (struct sock) that uses it, the port number and
5877 * a fastreuse flag (FIXME: NPI ipg).
5878 */
5879 static struct sctp_bind_bucket *sctp_bucket_create(
5880 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
5881
5882 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5883 {
5884 struct sctp_bind_hashbucket *head; /* hash list */
5885 struct sctp_bind_bucket *pp;
5886 unsigned short snum;
5887 int ret;
5888
5889 snum = ntohs(addr->v4.sin_port);
5890
5891 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5892 sctp_local_bh_disable();
5893
5894 if (snum == 0) {
5895 /* Search for an available port. */
5896 int low, high, remaining, index;
5897 unsigned int rover;
5898
5899 inet_get_local_port_range(&low, &high);
5900 remaining = (high - low) + 1;
5901 rover = net_random() % remaining + low;
5902
5903 do {
5904 rover++;
5905 if ((rover < low) || (rover > high))
5906 rover = low;
5907 if (inet_is_reserved_local_port(rover))
5908 continue;
5909 index = sctp_phashfn(sock_net(sk), rover);
5910 head = &sctp_port_hashtable[index];
5911 sctp_spin_lock(&head->lock);
5912 sctp_for_each_hentry(pp, &head->chain)
5913 if ((pp->port == rover) &&
5914 net_eq(sock_net(sk), pp->net))
5915 goto next;
5916 break;
5917 next:
5918 sctp_spin_unlock(&head->lock);
5919 } while (--remaining > 0);
5920
5921 /* Exhausted local port range during search? */
5922 ret = 1;
5923 if (remaining <= 0)
5924 goto fail;
5925
5926 /* OK, here is the one we will use. HEAD (the port
5927 * hash table list entry) is non-NULL and we hold it's
5928 * mutex.
5929 */
5930 snum = rover;
5931 } else {
5932 /* We are given an specific port number; we verify
5933 * that it is not being used. If it is used, we will
5934 * exahust the search in the hash list corresponding
5935 * to the port number (snum) - we detect that with the
5936 * port iterator, pp being NULL.
5937 */
5938 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
5939 sctp_spin_lock(&head->lock);
5940 sctp_for_each_hentry(pp, &head->chain) {
5941 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
5942 goto pp_found;
5943 }
5944 }
5945 pp = NULL;
5946 goto pp_not_found;
5947 pp_found:
5948 if (!hlist_empty(&pp->owner)) {
5949 /* We had a port hash table hit - there is an
5950 * available port (pp != NULL) and it is being
5951 * used by other socket (pp->owner not empty); that other
5952 * socket is going to be sk2.
5953 */
5954 int reuse = sk->sk_reuse;
5955 struct sock *sk2;
5956
5957 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5958 if (pp->fastreuse && sk->sk_reuse &&
5959 sk->sk_state != SCTP_SS_LISTENING)
5960 goto success;
5961
5962 /* Run through the list of sockets bound to the port
5963 * (pp->port) [via the pointers bind_next and
5964 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5965 * we get the endpoint they describe and run through
5966 * the endpoint's list of IP (v4 or v6) addresses,
5967 * comparing each of the addresses with the address of
5968 * the socket sk. If we find a match, then that means
5969 * that this port/socket (sk) combination are already
5970 * in an endpoint.
5971 */
5972 sk_for_each_bound(sk2, &pp->owner) {
5973 struct sctp_endpoint *ep2;
5974 ep2 = sctp_sk(sk2)->ep;
5975
5976 if (sk == sk2 ||
5977 (reuse && sk2->sk_reuse &&
5978 sk2->sk_state != SCTP_SS_LISTENING))
5979 continue;
5980
5981 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5982 sctp_sk(sk2), sctp_sk(sk))) {
5983 ret = (long)sk2;
5984 goto fail_unlock;
5985 }
5986 }
5987 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5988 }
5989 pp_not_found:
5990 /* If there was a hash table miss, create a new port. */
5991 ret = 1;
5992 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
5993 goto fail_unlock;
5994
5995 /* In either case (hit or miss), make sure fastreuse is 1 only
5996 * if sk->sk_reuse is too (that is, if the caller requested
5997 * SO_REUSEADDR on this socket -sk-).
5998 */
5999 if (hlist_empty(&pp->owner)) {
6000 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
6001 pp->fastreuse = 1;
6002 else
6003 pp->fastreuse = 0;
6004 } else if (pp->fastreuse &&
6005 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6006 pp->fastreuse = 0;
6007
6008 /* We are set, so fill up all the data in the hash table
6009 * entry, tie the socket list information with the rest of the
6010 * sockets FIXME: Blurry, NPI (ipg).
6011 */
6012 success:
6013 if (!sctp_sk(sk)->bind_hash) {
6014 inet_sk(sk)->inet_num = snum;
6015 sk_add_bind_node(sk, &pp->owner);
6016 sctp_sk(sk)->bind_hash = pp;
6017 }
6018 ret = 0;
6019
6020 fail_unlock:
6021 sctp_spin_unlock(&head->lock);
6022
6023 fail:
6024 sctp_local_bh_enable();
6025 return ret;
6026 }
6027
6028 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6029 * port is requested.
6030 */
6031 static int sctp_get_port(struct sock *sk, unsigned short snum)
6032 {
6033 long ret;
6034 union sctp_addr addr;
6035 struct sctp_af *af = sctp_sk(sk)->pf->af;
6036
6037 /* Set up a dummy address struct from the sk. */
6038 af->from_sk(&addr, sk);
6039 addr.v4.sin_port = htons(snum);
6040
6041 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6042 ret = sctp_get_port_local(sk, &addr);
6043
6044 return ret ? 1 : 0;
6045 }
6046
6047 /*
6048 * Move a socket to LISTENING state.
6049 */
6050 static int sctp_listen_start(struct sock *sk, int backlog)
6051 {
6052 struct sctp_sock *sp = sctp_sk(sk);
6053 struct sctp_endpoint *ep = sp->ep;
6054 struct crypto_hash *tfm = NULL;
6055 char alg[32];
6056
6057 /* Allocate HMAC for generating cookie. */
6058 if (!sp->hmac && sp->sctp_hmac_alg) {
6059 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6060 tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC);
6061 if (IS_ERR(tfm)) {
6062 net_info_ratelimited("failed to load transform for %s: %ld\n",
6063 sp->sctp_hmac_alg, PTR_ERR(tfm));
6064 return -ENOSYS;
6065 }
6066 sctp_sk(sk)->hmac = tfm;
6067 }
6068
6069 /*
6070 * If a bind() or sctp_bindx() is not called prior to a listen()
6071 * call that allows new associations to be accepted, the system
6072 * picks an ephemeral port and will choose an address set equivalent
6073 * to binding with a wildcard address.
6074 *
6075 * This is not currently spelled out in the SCTP sockets
6076 * extensions draft, but follows the practice as seen in TCP
6077 * sockets.
6078 *
6079 */
6080 sk->sk_state = SCTP_SS_LISTENING;
6081 if (!ep->base.bind_addr.port) {
6082 if (sctp_autobind(sk))
6083 return -EAGAIN;
6084 } else {
6085 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6086 sk->sk_state = SCTP_SS_CLOSED;
6087 return -EADDRINUSE;
6088 }
6089 }
6090
6091 sk->sk_max_ack_backlog = backlog;
6092 sctp_hash_endpoint(ep);
6093 return 0;
6094 }
6095
6096 /*
6097 * 4.1.3 / 5.1.3 listen()
6098 *
6099 * By default, new associations are not accepted for UDP style sockets.
6100 * An application uses listen() to mark a socket as being able to
6101 * accept new associations.
6102 *
6103 * On TCP style sockets, applications use listen() to ready the SCTP
6104 * endpoint for accepting inbound associations.
6105 *
6106 * On both types of endpoints a backlog of '0' disables listening.
6107 *
6108 * Move a socket to LISTENING state.
6109 */
6110 int sctp_inet_listen(struct socket *sock, int backlog)
6111 {
6112 struct sock *sk = sock->sk;
6113 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6114 int err = -EINVAL;
6115
6116 if (unlikely(backlog < 0))
6117 return err;
6118
6119 sctp_lock_sock(sk);
6120
6121 /* Peeled-off sockets are not allowed to listen(). */
6122 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6123 goto out;
6124
6125 if (sock->state != SS_UNCONNECTED)
6126 goto out;
6127
6128 /* If backlog is zero, disable listening. */
6129 if (!backlog) {
6130 if (sctp_sstate(sk, CLOSED))
6131 goto out;
6132
6133 err = 0;
6134 sctp_unhash_endpoint(ep);
6135 sk->sk_state = SCTP_SS_CLOSED;
6136 if (sk->sk_reuse)
6137 sctp_sk(sk)->bind_hash->fastreuse = 1;
6138 goto out;
6139 }
6140
6141 /* If we are already listening, just update the backlog */
6142 if (sctp_sstate(sk, LISTENING))
6143 sk->sk_max_ack_backlog = backlog;
6144 else {
6145 err = sctp_listen_start(sk, backlog);
6146 if (err)
6147 goto out;
6148 }
6149
6150 err = 0;
6151 out:
6152 sctp_release_sock(sk);
6153 return err;
6154 }
6155
6156 /*
6157 * This function is done by modeling the current datagram_poll() and the
6158 * tcp_poll(). Note that, based on these implementations, we don't
6159 * lock the socket in this function, even though it seems that,
6160 * ideally, locking or some other mechanisms can be used to ensure
6161 * the integrity of the counters (sndbuf and wmem_alloc) used
6162 * in this place. We assume that we don't need locks either until proven
6163 * otherwise.
6164 *
6165 * Another thing to note is that we include the Async I/O support
6166 * here, again, by modeling the current TCP/UDP code. We don't have
6167 * a good way to test with it yet.
6168 */
6169 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6170 {
6171 struct sock *sk = sock->sk;
6172 struct sctp_sock *sp = sctp_sk(sk);
6173 unsigned int mask;
6174
6175 poll_wait(file, sk_sleep(sk), wait);
6176
6177 /* A TCP-style listening socket becomes readable when the accept queue
6178 * is not empty.
6179 */
6180 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6181 return (!list_empty(&sp->ep->asocs)) ?
6182 (POLLIN | POLLRDNORM) : 0;
6183
6184 mask = 0;
6185
6186 /* Is there any exceptional events? */
6187 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6188 mask |= POLLERR |
6189 sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0;
6190 if (sk->sk_shutdown & RCV_SHUTDOWN)
6191 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6192 if (sk->sk_shutdown == SHUTDOWN_MASK)
6193 mask |= POLLHUP;
6194
6195 /* Is it readable? Reconsider this code with TCP-style support. */
6196 if (!skb_queue_empty(&sk->sk_receive_queue))
6197 mask |= POLLIN | POLLRDNORM;
6198
6199 /* The association is either gone or not ready. */
6200 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6201 return mask;
6202
6203 /* Is it writable? */
6204 if (sctp_writeable(sk)) {
6205 mask |= POLLOUT | POLLWRNORM;
6206 } else {
6207 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6208 /*
6209 * Since the socket is not locked, the buffer
6210 * might be made available after the writeable check and
6211 * before the bit is set. This could cause a lost I/O
6212 * signal. tcp_poll() has a race breaker for this race
6213 * condition. Based on their implementation, we put
6214 * in the following code to cover it as well.
6215 */
6216 if (sctp_writeable(sk))
6217 mask |= POLLOUT | POLLWRNORM;
6218 }
6219 return mask;
6220 }
6221
6222 /********************************************************************
6223 * 2nd Level Abstractions
6224 ********************************************************************/
6225
6226 static struct sctp_bind_bucket *sctp_bucket_create(
6227 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6228 {
6229 struct sctp_bind_bucket *pp;
6230
6231 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6232 if (pp) {
6233 SCTP_DBG_OBJCNT_INC(bind_bucket);
6234 pp->port = snum;
6235 pp->fastreuse = 0;
6236 INIT_HLIST_HEAD(&pp->owner);
6237 pp->net = net;
6238 hlist_add_head(&pp->node, &head->chain);
6239 }
6240 return pp;
6241 }
6242
6243 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6244 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6245 {
6246 if (pp && hlist_empty(&pp->owner)) {
6247 __hlist_del(&pp->node);
6248 kmem_cache_free(sctp_bucket_cachep, pp);
6249 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6250 }
6251 }
6252
6253 /* Release this socket's reference to a local port. */
6254 static inline void __sctp_put_port(struct sock *sk)
6255 {
6256 struct sctp_bind_hashbucket *head =
6257 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6258 inet_sk(sk)->inet_num)];
6259 struct sctp_bind_bucket *pp;
6260
6261 sctp_spin_lock(&head->lock);
6262 pp = sctp_sk(sk)->bind_hash;
6263 __sk_del_bind_node(sk);
6264 sctp_sk(sk)->bind_hash = NULL;
6265 inet_sk(sk)->inet_num = 0;
6266 sctp_bucket_destroy(pp);
6267 sctp_spin_unlock(&head->lock);
6268 }
6269
6270 void sctp_put_port(struct sock *sk)
6271 {
6272 sctp_local_bh_disable();
6273 __sctp_put_port(sk);
6274 sctp_local_bh_enable();
6275 }
6276
6277 /*
6278 * The system picks an ephemeral port and choose an address set equivalent
6279 * to binding with a wildcard address.
6280 * One of those addresses will be the primary address for the association.
6281 * This automatically enables the multihoming capability of SCTP.
6282 */
6283 static int sctp_autobind(struct sock *sk)
6284 {
6285 union sctp_addr autoaddr;
6286 struct sctp_af *af;
6287 __be16 port;
6288
6289 /* Initialize a local sockaddr structure to INADDR_ANY. */
6290 af = sctp_sk(sk)->pf->af;
6291
6292 port = htons(inet_sk(sk)->inet_num);
6293 af->inaddr_any(&autoaddr, port);
6294
6295 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6296 }
6297
6298 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6299 *
6300 * From RFC 2292
6301 * 4.2 The cmsghdr Structure *
6302 *
6303 * When ancillary data is sent or received, any number of ancillary data
6304 * objects can be specified by the msg_control and msg_controllen members of
6305 * the msghdr structure, because each object is preceded by
6306 * a cmsghdr structure defining the object's length (the cmsg_len member).
6307 * Historically Berkeley-derived implementations have passed only one object
6308 * at a time, but this API allows multiple objects to be
6309 * passed in a single call to sendmsg() or recvmsg(). The following example
6310 * shows two ancillary data objects in a control buffer.
6311 *
6312 * |<--------------------------- msg_controllen -------------------------->|
6313 * | |
6314 *
6315 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6316 *
6317 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6318 * | | |
6319 *
6320 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6321 *
6322 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6323 * | | | | |
6324 *
6325 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6326 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6327 *
6328 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6329 *
6330 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6331 * ^
6332 * |
6333 *
6334 * msg_control
6335 * points here
6336 */
6337 static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
6338 {
6339 struct cmsghdr *cmsg;
6340 struct msghdr *my_msg = (struct msghdr *)msg;
6341
6342 for (cmsg = CMSG_FIRSTHDR(msg);
6343 cmsg != NULL;
6344 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6345 if (!CMSG_OK(my_msg, cmsg))
6346 return -EINVAL;
6347
6348 /* Should we parse this header or ignore? */
6349 if (cmsg->cmsg_level != IPPROTO_SCTP)
6350 continue;
6351
6352 /* Strictly check lengths following example in SCM code. */
6353 switch (cmsg->cmsg_type) {
6354 case SCTP_INIT:
6355 /* SCTP Socket API Extension
6356 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6357 *
6358 * This cmsghdr structure provides information for
6359 * initializing new SCTP associations with sendmsg().
6360 * The SCTP_INITMSG socket option uses this same data
6361 * structure. This structure is not used for
6362 * recvmsg().
6363 *
6364 * cmsg_level cmsg_type cmsg_data[]
6365 * ------------ ------------ ----------------------
6366 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6367 */
6368 if (cmsg->cmsg_len !=
6369 CMSG_LEN(sizeof(struct sctp_initmsg)))
6370 return -EINVAL;
6371 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6372 break;
6373
6374 case SCTP_SNDRCV:
6375 /* SCTP Socket API Extension
6376 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6377 *
6378 * This cmsghdr structure specifies SCTP options for
6379 * sendmsg() and describes SCTP header information
6380 * about a received message through recvmsg().
6381 *
6382 * cmsg_level cmsg_type cmsg_data[]
6383 * ------------ ------------ ----------------------
6384 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6385 */
6386 if (cmsg->cmsg_len !=
6387 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6388 return -EINVAL;
6389
6390 cmsgs->info =
6391 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6392
6393 /* Minimally, validate the sinfo_flags. */
6394 if (cmsgs->info->sinfo_flags &
6395 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6396 SCTP_ABORT | SCTP_EOF))
6397 return -EINVAL;
6398 break;
6399
6400 default:
6401 return -EINVAL;
6402 }
6403 }
6404 return 0;
6405 }
6406
6407 /*
6408 * Wait for a packet..
6409 * Note: This function is the same function as in core/datagram.c
6410 * with a few modifications to make lksctp work.
6411 */
6412 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6413 {
6414 int error;
6415 DEFINE_WAIT(wait);
6416
6417 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6418
6419 /* Socket errors? */
6420 error = sock_error(sk);
6421 if (error)
6422 goto out;
6423
6424 if (!skb_queue_empty(&sk->sk_receive_queue))
6425 goto ready;
6426
6427 /* Socket shut down? */
6428 if (sk->sk_shutdown & RCV_SHUTDOWN)
6429 goto out;
6430
6431 /* Sequenced packets can come disconnected. If so we report the
6432 * problem.
6433 */
6434 error = -ENOTCONN;
6435
6436 /* Is there a good reason to think that we may receive some data? */
6437 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6438 goto out;
6439
6440 /* Handle signals. */
6441 if (signal_pending(current))
6442 goto interrupted;
6443
6444 /* Let another process have a go. Since we are going to sleep
6445 * anyway. Note: This may cause odd behaviors if the message
6446 * does not fit in the user's buffer, but this seems to be the
6447 * only way to honor MSG_DONTWAIT realistically.
6448 */
6449 sctp_release_sock(sk);
6450 *timeo_p = schedule_timeout(*timeo_p);
6451 sctp_lock_sock(sk);
6452
6453 ready:
6454 finish_wait(sk_sleep(sk), &wait);
6455 return 0;
6456
6457 interrupted:
6458 error = sock_intr_errno(*timeo_p);
6459
6460 out:
6461 finish_wait(sk_sleep(sk), &wait);
6462 *err = error;
6463 return error;
6464 }
6465
6466 /* Receive a datagram.
6467 * Note: This is pretty much the same routine as in core/datagram.c
6468 * with a few changes to make lksctp work.
6469 */
6470 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6471 int noblock, int *err)
6472 {
6473 int error;
6474 struct sk_buff *skb;
6475 long timeo;
6476
6477 timeo = sock_rcvtimeo(sk, noblock);
6478
6479 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6480 timeo, MAX_SCHEDULE_TIMEOUT);
6481
6482 do {
6483 /* Again only user level code calls this function,
6484 * so nothing interrupt level
6485 * will suddenly eat the receive_queue.
6486 *
6487 * Look at current nfs client by the way...
6488 * However, this function was correct in any case. 8)
6489 */
6490 if (flags & MSG_PEEK) {
6491 spin_lock_bh(&sk->sk_receive_queue.lock);
6492 skb = skb_peek(&sk->sk_receive_queue);
6493 if (skb)
6494 atomic_inc(&skb->users);
6495 spin_unlock_bh(&sk->sk_receive_queue.lock);
6496 } else {
6497 skb = skb_dequeue(&sk->sk_receive_queue);
6498 }
6499
6500 if (skb)
6501 return skb;
6502
6503 /* Caller is allowed not to check sk->sk_err before calling. */
6504 error = sock_error(sk);
6505 if (error)
6506 goto no_packet;
6507
6508 if (sk->sk_shutdown & RCV_SHUTDOWN)
6509 break;
6510
6511 /* User doesn't want to wait. */
6512 error = -EAGAIN;
6513 if (!timeo)
6514 goto no_packet;
6515 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6516
6517 return NULL;
6518
6519 no_packet:
6520 *err = error;
6521 return NULL;
6522 }
6523
6524 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6525 static void __sctp_write_space(struct sctp_association *asoc)
6526 {
6527 struct sock *sk = asoc->base.sk;
6528 struct socket *sock = sk->sk_socket;
6529
6530 if ((sctp_wspace(asoc) > 0) && sock) {
6531 if (waitqueue_active(&asoc->wait))
6532 wake_up_interruptible(&asoc->wait);
6533
6534 if (sctp_writeable(sk)) {
6535 wait_queue_head_t *wq = sk_sleep(sk);
6536
6537 if (wq && waitqueue_active(wq))
6538 wake_up_interruptible(wq);
6539
6540 /* Note that we try to include the Async I/O support
6541 * here by modeling from the current TCP/UDP code.
6542 * We have not tested with it yet.
6543 */
6544 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6545 sock_wake_async(sock,
6546 SOCK_WAKE_SPACE, POLL_OUT);
6547 }
6548 }
6549 }
6550
6551 /* Do accounting for the sndbuf space.
6552 * Decrement the used sndbuf space of the corresponding association by the
6553 * data size which was just transmitted(freed).
6554 */
6555 static void sctp_wfree(struct sk_buff *skb)
6556 {
6557 struct sctp_association *asoc;
6558 struct sctp_chunk *chunk;
6559 struct sock *sk;
6560
6561 /* Get the saved chunk pointer. */
6562 chunk = *((struct sctp_chunk **)(skb->cb));
6563 asoc = chunk->asoc;
6564 sk = asoc->base.sk;
6565 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6566 sizeof(struct sk_buff) +
6567 sizeof(struct sctp_chunk);
6568
6569 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6570
6571 /*
6572 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6573 */
6574 sk->sk_wmem_queued -= skb->truesize;
6575 sk_mem_uncharge(sk, skb->truesize);
6576
6577 sock_wfree(skb);
6578 __sctp_write_space(asoc);
6579
6580 sctp_association_put(asoc);
6581 }
6582
6583 /* Do accounting for the receive space on the socket.
6584 * Accounting for the association is done in ulpevent.c
6585 * We set this as a destructor for the cloned data skbs so that
6586 * accounting is done at the correct time.
6587 */
6588 void sctp_sock_rfree(struct sk_buff *skb)
6589 {
6590 struct sock *sk = skb->sk;
6591 struct sctp_ulpevent *event = sctp_skb2event(skb);
6592
6593 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6594
6595 /*
6596 * Mimic the behavior of sock_rfree
6597 */
6598 sk_mem_uncharge(sk, event->rmem_len);
6599 }
6600
6601
6602 /* Helper function to wait for space in the sndbuf. */
6603 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6604 size_t msg_len)
6605 {
6606 struct sock *sk = asoc->base.sk;
6607 int err = 0;
6608 long current_timeo = *timeo_p;
6609 DEFINE_WAIT(wait);
6610
6611 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6612 asoc, (long)(*timeo_p), msg_len);
6613
6614 /* Increment the association's refcnt. */
6615 sctp_association_hold(asoc);
6616
6617 /* Wait on the association specific sndbuf space. */
6618 for (;;) {
6619 prepare_to_wait_exclusive(&asoc->wait, &wait,
6620 TASK_INTERRUPTIBLE);
6621 if (!*timeo_p)
6622 goto do_nonblock;
6623 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6624 asoc->base.dead)
6625 goto do_error;
6626 if (signal_pending(current))
6627 goto do_interrupted;
6628 if (msg_len <= sctp_wspace(asoc))
6629 break;
6630
6631 /* Let another process have a go. Since we are going
6632 * to sleep anyway.
6633 */
6634 sctp_release_sock(sk);
6635 current_timeo = schedule_timeout(current_timeo);
6636 BUG_ON(sk != asoc->base.sk);
6637 sctp_lock_sock(sk);
6638
6639 *timeo_p = current_timeo;
6640 }
6641
6642 out:
6643 finish_wait(&asoc->wait, &wait);
6644
6645 /* Release the association's refcnt. */
6646 sctp_association_put(asoc);
6647
6648 return err;
6649
6650 do_error:
6651 err = -EPIPE;
6652 goto out;
6653
6654 do_interrupted:
6655 err = sock_intr_errno(*timeo_p);
6656 goto out;
6657
6658 do_nonblock:
6659 err = -EAGAIN;
6660 goto out;
6661 }
6662
6663 void sctp_data_ready(struct sock *sk, int len)
6664 {
6665 struct socket_wq *wq;
6666
6667 rcu_read_lock();
6668 wq = rcu_dereference(sk->sk_wq);
6669 if (wq_has_sleeper(wq))
6670 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6671 POLLRDNORM | POLLRDBAND);
6672 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6673 rcu_read_unlock();
6674 }
6675
6676 /* If socket sndbuf has changed, wake up all per association waiters. */
6677 void sctp_write_space(struct sock *sk)
6678 {
6679 struct sctp_association *asoc;
6680
6681 /* Wake up the tasks in each wait queue. */
6682 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6683 __sctp_write_space(asoc);
6684 }
6685 }
6686
6687 /* Is there any sndbuf space available on the socket?
6688 *
6689 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6690 * associations on the same socket. For a UDP-style socket with
6691 * multiple associations, it is possible for it to be "unwriteable"
6692 * prematurely. I assume that this is acceptable because
6693 * a premature "unwriteable" is better than an accidental "writeable" which
6694 * would cause an unwanted block under certain circumstances. For the 1-1
6695 * UDP-style sockets or TCP-style sockets, this code should work.
6696 * - Daisy
6697 */
6698 static int sctp_writeable(struct sock *sk)
6699 {
6700 int amt = 0;
6701
6702 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6703 if (amt < 0)
6704 amt = 0;
6705 return amt;
6706 }
6707
6708 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6709 * returns immediately with EINPROGRESS.
6710 */
6711 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6712 {
6713 struct sock *sk = asoc->base.sk;
6714 int err = 0;
6715 long current_timeo = *timeo_p;
6716 DEFINE_WAIT(wait);
6717
6718 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6719 (long)(*timeo_p));
6720
6721 /* Increment the association's refcnt. */
6722 sctp_association_hold(asoc);
6723
6724 for (;;) {
6725 prepare_to_wait_exclusive(&asoc->wait, &wait,
6726 TASK_INTERRUPTIBLE);
6727 if (!*timeo_p)
6728 goto do_nonblock;
6729 if (sk->sk_shutdown & RCV_SHUTDOWN)
6730 break;
6731 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6732 asoc->base.dead)
6733 goto do_error;
6734 if (signal_pending(current))
6735 goto do_interrupted;
6736
6737 if (sctp_state(asoc, ESTABLISHED))
6738 break;
6739
6740 /* Let another process have a go. Since we are going
6741 * to sleep anyway.
6742 */
6743 sctp_release_sock(sk);
6744 current_timeo = schedule_timeout(current_timeo);
6745 sctp_lock_sock(sk);
6746
6747 *timeo_p = current_timeo;
6748 }
6749
6750 out:
6751 finish_wait(&asoc->wait, &wait);
6752
6753 /* Release the association's refcnt. */
6754 sctp_association_put(asoc);
6755
6756 return err;
6757
6758 do_error:
6759 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6760 err = -ETIMEDOUT;
6761 else
6762 err = -ECONNREFUSED;
6763 goto out;
6764
6765 do_interrupted:
6766 err = sock_intr_errno(*timeo_p);
6767 goto out;
6768
6769 do_nonblock:
6770 err = -EINPROGRESS;
6771 goto out;
6772 }
6773
6774 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6775 {
6776 struct sctp_endpoint *ep;
6777 int err = 0;
6778 DEFINE_WAIT(wait);
6779
6780 ep = sctp_sk(sk)->ep;
6781
6782
6783 for (;;) {
6784 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6785 TASK_INTERRUPTIBLE);
6786
6787 if (list_empty(&ep->asocs)) {
6788 sctp_release_sock(sk);
6789 timeo = schedule_timeout(timeo);
6790 sctp_lock_sock(sk);
6791 }
6792
6793 err = -EINVAL;
6794 if (!sctp_sstate(sk, LISTENING))
6795 break;
6796
6797 err = 0;
6798 if (!list_empty(&ep->asocs))
6799 break;
6800
6801 err = sock_intr_errno(timeo);
6802 if (signal_pending(current))
6803 break;
6804
6805 err = -EAGAIN;
6806 if (!timeo)
6807 break;
6808 }
6809
6810 finish_wait(sk_sleep(sk), &wait);
6811
6812 return err;
6813 }
6814
6815 static void sctp_wait_for_close(struct sock *sk, long timeout)
6816 {
6817 DEFINE_WAIT(wait);
6818
6819 do {
6820 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6821 if (list_empty(&sctp_sk(sk)->ep->asocs))
6822 break;
6823 sctp_release_sock(sk);
6824 timeout = schedule_timeout(timeout);
6825 sctp_lock_sock(sk);
6826 } while (!signal_pending(current) && timeout);
6827
6828 finish_wait(sk_sleep(sk), &wait);
6829 }
6830
6831 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6832 {
6833 struct sk_buff *frag;
6834
6835 if (!skb->data_len)
6836 goto done;
6837
6838 /* Don't forget the fragments. */
6839 skb_walk_frags(skb, frag)
6840 sctp_skb_set_owner_r_frag(frag, sk);
6841
6842 done:
6843 sctp_skb_set_owner_r(skb, sk);
6844 }
6845
6846 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6847 struct sctp_association *asoc)
6848 {
6849 struct inet_sock *inet = inet_sk(sk);
6850 struct inet_sock *newinet;
6851
6852 newsk->sk_type = sk->sk_type;
6853 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6854 newsk->sk_flags = sk->sk_flags;
6855 newsk->sk_no_check = sk->sk_no_check;
6856 newsk->sk_reuse = sk->sk_reuse;
6857
6858 newsk->sk_shutdown = sk->sk_shutdown;
6859 newsk->sk_destruct = inet_sock_destruct;
6860 newsk->sk_family = sk->sk_family;
6861 newsk->sk_protocol = IPPROTO_SCTP;
6862 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6863 newsk->sk_sndbuf = sk->sk_sndbuf;
6864 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6865 newsk->sk_lingertime = sk->sk_lingertime;
6866 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6867 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6868
6869 newinet = inet_sk(newsk);
6870
6871 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6872 * getsockname() and getpeername()
6873 */
6874 newinet->inet_sport = inet->inet_sport;
6875 newinet->inet_saddr = inet->inet_saddr;
6876 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6877 newinet->inet_dport = htons(asoc->peer.port);
6878 newinet->pmtudisc = inet->pmtudisc;
6879 newinet->inet_id = asoc->next_tsn ^ jiffies;
6880
6881 newinet->uc_ttl = inet->uc_ttl;
6882 newinet->mc_loop = 1;
6883 newinet->mc_ttl = 1;
6884 newinet->mc_index = 0;
6885 newinet->mc_list = NULL;
6886 }
6887
6888 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6889 * and its messages to the newsk.
6890 */
6891 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6892 struct sctp_association *assoc,
6893 sctp_socket_type_t type)
6894 {
6895 struct sctp_sock *oldsp = sctp_sk(oldsk);
6896 struct sctp_sock *newsp = sctp_sk(newsk);
6897 struct sctp_bind_bucket *pp; /* hash list port iterator */
6898 struct sctp_endpoint *newep = newsp->ep;
6899 struct sk_buff *skb, *tmp;
6900 struct sctp_ulpevent *event;
6901 struct sctp_bind_hashbucket *head;
6902 struct list_head tmplist;
6903
6904 /* Migrate socket buffer sizes and all the socket level options to the
6905 * new socket.
6906 */
6907 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6908 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6909 /* Brute force copy old sctp opt. */
6910 if (oldsp->do_auto_asconf) {
6911 memcpy(&tmplist, &newsp->auto_asconf_list, sizeof(tmplist));
6912 inet_sk_copy_descendant(newsk, oldsk);
6913 memcpy(&newsp->auto_asconf_list, &tmplist, sizeof(tmplist));
6914 } else
6915 inet_sk_copy_descendant(newsk, oldsk);
6916
6917 /* Restore the ep value that was overwritten with the above structure
6918 * copy.
6919 */
6920 newsp->ep = newep;
6921 newsp->hmac = NULL;
6922
6923 /* Hook this new socket in to the bind_hash list. */
6924 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
6925 inet_sk(oldsk)->inet_num)];
6926 sctp_local_bh_disable();
6927 sctp_spin_lock(&head->lock);
6928 pp = sctp_sk(oldsk)->bind_hash;
6929 sk_add_bind_node(newsk, &pp->owner);
6930 sctp_sk(newsk)->bind_hash = pp;
6931 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6932 sctp_spin_unlock(&head->lock);
6933 sctp_local_bh_enable();
6934
6935 /* Copy the bind_addr list from the original endpoint to the new
6936 * endpoint so that we can handle restarts properly
6937 */
6938 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6939 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6940
6941 /* Move any messages in the old socket's receive queue that are for the
6942 * peeled off association to the new socket's receive queue.
6943 */
6944 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6945 event = sctp_skb2event(skb);
6946 if (event->asoc == assoc) {
6947 __skb_unlink(skb, &oldsk->sk_receive_queue);
6948 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6949 sctp_skb_set_owner_r_frag(skb, newsk);
6950 }
6951 }
6952
6953 /* Clean up any messages pending delivery due to partial
6954 * delivery. Three cases:
6955 * 1) No partial deliver; no work.
6956 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6957 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6958 */
6959 skb_queue_head_init(&newsp->pd_lobby);
6960 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6961
6962 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6963 struct sk_buff_head *queue;
6964
6965 /* Decide which queue to move pd_lobby skbs to. */
6966 if (assoc->ulpq.pd_mode) {
6967 queue = &newsp->pd_lobby;
6968 } else
6969 queue = &newsk->sk_receive_queue;
6970
6971 /* Walk through the pd_lobby, looking for skbs that
6972 * need moved to the new socket.
6973 */
6974 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6975 event = sctp_skb2event(skb);
6976 if (event->asoc == assoc) {
6977 __skb_unlink(skb, &oldsp->pd_lobby);
6978 __skb_queue_tail(queue, skb);
6979 sctp_skb_set_owner_r_frag(skb, newsk);
6980 }
6981 }
6982
6983 /* Clear up any skbs waiting for the partial
6984 * delivery to finish.
6985 */
6986 if (assoc->ulpq.pd_mode)
6987 sctp_clear_pd(oldsk, NULL);
6988
6989 }
6990
6991 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6992 sctp_skb_set_owner_r_frag(skb, newsk);
6993
6994 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6995 sctp_skb_set_owner_r_frag(skb, newsk);
6996
6997 /* Set the type of socket to indicate that it is peeled off from the
6998 * original UDP-style socket or created with the accept() call on a
6999 * TCP-style socket..
7000 */
7001 newsp->type = type;
7002
7003 /* Mark the new socket "in-use" by the user so that any packets
7004 * that may arrive on the association after we've moved it are
7005 * queued to the backlog. This prevents a potential race between
7006 * backlog processing on the old socket and new-packet processing
7007 * on the new socket.
7008 *
7009 * The caller has just allocated newsk so we can guarantee that other
7010 * paths won't try to lock it and then oldsk.
7011 */
7012 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7013 sctp_assoc_migrate(assoc, newsk);
7014
7015 /* If the association on the newsk is already closed before accept()
7016 * is called, set RCV_SHUTDOWN flag.
7017 */
7018 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
7019 newsk->sk_shutdown |= RCV_SHUTDOWN;
7020
7021 newsk->sk_state = SCTP_SS_ESTABLISHED;
7022 sctp_release_sock(newsk);
7023 }
7024
7025
7026 /* This proto struct describes the ULP interface for SCTP. */
7027 struct proto sctp_prot = {
7028 .name = "SCTP",
7029 .owner = THIS_MODULE,
7030 .close = sctp_close,
7031 .connect = sctp_connect,
7032 .disconnect = sctp_disconnect,
7033 .accept = sctp_accept,
7034 .ioctl = sctp_ioctl,
7035 .init = sctp_init_sock,
7036 .destroy = sctp_destroy_sock,
7037 .shutdown = sctp_shutdown,
7038 .setsockopt = sctp_setsockopt,
7039 .getsockopt = sctp_getsockopt,
7040 .sendmsg = sctp_sendmsg,
7041 .recvmsg = sctp_recvmsg,
7042 .bind = sctp_bind,
7043 .backlog_rcv = sctp_backlog_rcv,
7044 .hash = sctp_hash,
7045 .unhash = sctp_unhash,
7046 .get_port = sctp_get_port,
7047 .obj_size = sizeof(struct sctp_sock),
7048 .sysctl_mem = sysctl_sctp_mem,
7049 .sysctl_rmem = sysctl_sctp_rmem,
7050 .sysctl_wmem = sysctl_sctp_wmem,
7051 .memory_pressure = &sctp_memory_pressure,
7052 .enter_memory_pressure = sctp_enter_memory_pressure,
7053 .memory_allocated = &sctp_memory_allocated,
7054 .sockets_allocated = &sctp_sockets_allocated,
7055 };
7056
7057 #if IS_ENABLED(CONFIG_IPV6)
7058
7059 struct proto sctpv6_prot = {
7060 .name = "SCTPv6",
7061 .owner = THIS_MODULE,
7062 .close = sctp_close,
7063 .connect = sctp_connect,
7064 .disconnect = sctp_disconnect,
7065 .accept = sctp_accept,
7066 .ioctl = sctp_ioctl,
7067 .init = sctp_init_sock,
7068 .destroy = sctp_destroy_sock,
7069 .shutdown = sctp_shutdown,
7070 .setsockopt = sctp_setsockopt,
7071 .getsockopt = sctp_getsockopt,
7072 .sendmsg = sctp_sendmsg,
7073 .recvmsg = sctp_recvmsg,
7074 .bind = sctp_bind,
7075 .backlog_rcv = sctp_backlog_rcv,
7076 .hash = sctp_hash,
7077 .unhash = sctp_unhash,
7078 .get_port = sctp_get_port,
7079 .obj_size = sizeof(struct sctp6_sock),
7080 .sysctl_mem = sysctl_sctp_mem,
7081 .sysctl_rmem = sysctl_sctp_rmem,
7082 .sysctl_wmem = sysctl_sctp_wmem,
7083 .memory_pressure = &sctp_memory_pressure,
7084 .enter_memory_pressure = sctp_enter_memory_pressure,
7085 .memory_allocated = &sctp_memory_allocated,
7086 .sockets_allocated = &sctp_sockets_allocated,
7087 };
7088 #endif /* IS_ENABLED(CONFIG_IPV6) */