drivers: power: report battery voltage in AOSP compatible format
[GitHub/mt8127/android_kernel_alcatel_ttab.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 #include <linux/compat.h>
75
76 #include <net/ip.h>
77 #include <net/icmp.h>
78 #include <net/route.h>
79 #include <net/ipv6.h>
80 #include <net/inet_common.h>
81
82 #include <linux/socket.h> /* for sa_family_t */
83 #include <linux/export.h>
84 #include <net/sock.h>
85 #include <net/sctp/sctp.h>
86 #include <net/sctp/sm.h>
87
88 /* WARNING: Please do not remove the SCTP_STATIC attribute to
89 * any of the functions below as they are used to export functions
90 * used by a project regression testsuite.
91 */
92
93 /* Forward declarations for internal helper functions. */
94 static int sctp_writeable(struct sock *sk);
95 static void sctp_wfree(struct sk_buff *skb);
96 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
97 size_t msg_len);
98 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
99 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
100 static int sctp_wait_for_accept(struct sock *sk, long timeo);
101 static void sctp_wait_for_close(struct sock *sk, long timeo);
102 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
103 union sctp_addr *addr, int len);
104 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
105 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
106 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
107 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
108 static int sctp_send_asconf(struct sctp_association *asoc,
109 struct sctp_chunk *chunk);
110 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
111 static int sctp_autobind(struct sock *sk);
112 static void sctp_sock_migrate(struct sock *, struct sock *,
113 struct sctp_association *, sctp_socket_type_t);
114
115 extern struct kmem_cache *sctp_bucket_cachep;
116 extern long sysctl_sctp_mem[3];
117 extern int sysctl_sctp_rmem[3];
118 extern int sysctl_sctp_wmem[3];
119
120 static int sctp_memory_pressure;
121 static atomic_long_t sctp_memory_allocated;
122 struct percpu_counter sctp_sockets_allocated;
123
124 static void sctp_enter_memory_pressure(struct sock *sk)
125 {
126 sctp_memory_pressure = 1;
127 }
128
129
130 /* Get the sndbuf space available at the time on the association. */
131 static inline int sctp_wspace(struct sctp_association *asoc)
132 {
133 int amt;
134
135 if (asoc->ep->sndbuf_policy)
136 amt = asoc->sndbuf_used;
137 else
138 amt = sk_wmem_alloc_get(asoc->base.sk);
139
140 if (amt >= asoc->base.sk->sk_sndbuf) {
141 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
142 amt = 0;
143 else {
144 amt = sk_stream_wspace(asoc->base.sk);
145 if (amt < 0)
146 amt = 0;
147 }
148 } else {
149 amt = asoc->base.sk->sk_sndbuf - amt;
150 }
151 return amt;
152 }
153
154 /* Increment the used sndbuf space count of the corresponding association by
155 * the size of the outgoing data chunk.
156 * Also, set the skb destructor for sndbuf accounting later.
157 *
158 * Since it is always 1-1 between chunk and skb, and also a new skb is always
159 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
160 * destructor in the data chunk skb for the purpose of the sndbuf space
161 * tracking.
162 */
163 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
164 {
165 struct sctp_association *asoc = chunk->asoc;
166 struct sock *sk = asoc->base.sk;
167
168 /* The sndbuf space is tracked per association. */
169 sctp_association_hold(asoc);
170
171 skb_set_owner_w(chunk->skb, sk);
172
173 chunk->skb->destructor = sctp_wfree;
174 /* Save the chunk pointer in skb for sctp_wfree to use later. */
175 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
176
177 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
178 sizeof(struct sk_buff) +
179 sizeof(struct sctp_chunk);
180
181 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
182 sk->sk_wmem_queued += chunk->skb->truesize;
183 sk_mem_charge(sk, chunk->skb->truesize);
184 }
185
186 /* Verify that this is a valid address. */
187 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
188 int len)
189 {
190 struct sctp_af *af;
191
192 /* Verify basic sockaddr. */
193 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
194 if (!af)
195 return -EINVAL;
196
197 /* Is this a valid SCTP address? */
198 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
199 return -EINVAL;
200
201 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
202 return -EINVAL;
203
204 return 0;
205 }
206
207 /* Look up the association by its id. If this is not a UDP-style
208 * socket, the ID field is always ignored.
209 */
210 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
211 {
212 struct sctp_association *asoc = NULL;
213
214 /* If this is not a UDP-style socket, assoc id should be ignored. */
215 if (!sctp_style(sk, UDP)) {
216 /* Return NULL if the socket state is not ESTABLISHED. It
217 * could be a TCP-style listening socket or a socket which
218 * hasn't yet called connect() to establish an association.
219 */
220 if (!sctp_sstate(sk, ESTABLISHED))
221 return NULL;
222
223 /* Get the first and the only association from the list. */
224 if (!list_empty(&sctp_sk(sk)->ep->asocs))
225 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
226 struct sctp_association, asocs);
227 return asoc;
228 }
229
230 /* Otherwise this is a UDP-style socket. */
231 if (!id || (id == (sctp_assoc_t)-1))
232 return NULL;
233
234 spin_lock_bh(&sctp_assocs_id_lock);
235 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
236 spin_unlock_bh(&sctp_assocs_id_lock);
237
238 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
239 return NULL;
240
241 return asoc;
242 }
243
244 /* Look up the transport from an address and an assoc id. If both address and
245 * id are specified, the associations matching the address and the id should be
246 * the same.
247 */
248 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
249 struct sockaddr_storage *addr,
250 sctp_assoc_t id)
251 {
252 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
253 struct sctp_transport *transport;
254 union sctp_addr *laddr = (union sctp_addr *)addr;
255
256 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
257 laddr,
258 &transport);
259
260 if (!addr_asoc)
261 return NULL;
262
263 id_asoc = sctp_id2assoc(sk, id);
264 if (id_asoc && (id_asoc != addr_asoc))
265 return NULL;
266
267 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
268 (union sctp_addr *)addr);
269
270 return transport;
271 }
272
273 /* API 3.1.2 bind() - UDP Style Syntax
274 * The syntax of bind() is,
275 *
276 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
277 *
278 * sd - the socket descriptor returned by socket().
279 * addr - the address structure (struct sockaddr_in or struct
280 * sockaddr_in6 [RFC 2553]),
281 * addr_len - the size of the address structure.
282 */
283 SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
284 {
285 int retval = 0;
286
287 sctp_lock_sock(sk);
288
289 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
290 sk, addr, addr_len);
291
292 /* Disallow binding twice. */
293 if (!sctp_sk(sk)->ep->base.bind_addr.port)
294 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
295 addr_len);
296 else
297 retval = -EINVAL;
298
299 sctp_release_sock(sk);
300
301 return retval;
302 }
303
304 static long sctp_get_port_local(struct sock *, union sctp_addr *);
305
306 /* Verify this is a valid sockaddr. */
307 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
308 union sctp_addr *addr, int len)
309 {
310 struct sctp_af *af;
311
312 /* Check minimum size. */
313 if (len < sizeof (struct sockaddr))
314 return NULL;
315
316 /* V4 mapped address are really of AF_INET family */
317 if (addr->sa.sa_family == AF_INET6 &&
318 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
319 if (!opt->pf->af_supported(AF_INET, opt))
320 return NULL;
321 } else {
322 /* Does this PF support this AF? */
323 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
324 return NULL;
325 }
326
327 /* If we get this far, af is valid. */
328 af = sctp_get_af_specific(addr->sa.sa_family);
329
330 if (len < af->sockaddr_len)
331 return NULL;
332
333 return af;
334 }
335
336 /* Bind a local address either to an endpoint or to an association. */
337 SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
338 {
339 struct net *net = sock_net(sk);
340 struct sctp_sock *sp = sctp_sk(sk);
341 struct sctp_endpoint *ep = sp->ep;
342 struct sctp_bind_addr *bp = &ep->base.bind_addr;
343 struct sctp_af *af;
344 unsigned short snum;
345 int ret = 0;
346
347 /* Common sockaddr verification. */
348 af = sctp_sockaddr_af(sp, addr, len);
349 if (!af) {
350 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
351 sk, addr, len);
352 return -EINVAL;
353 }
354
355 snum = ntohs(addr->v4.sin_port);
356
357 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
358 ", port: %d, new port: %d, len: %d)\n",
359 sk,
360 addr,
361 bp->port, snum,
362 len);
363
364 /* PF specific bind() address verification. */
365 if (!sp->pf->bind_verify(sp, addr))
366 return -EADDRNOTAVAIL;
367
368 /* We must either be unbound, or bind to the same port.
369 * It's OK to allow 0 ports if we are already bound.
370 * We'll just inhert an already bound port in this case
371 */
372 if (bp->port) {
373 if (!snum)
374 snum = bp->port;
375 else if (snum != bp->port) {
376 SCTP_DEBUG_PRINTK("sctp_do_bind:"
377 " New port %d does not match existing port "
378 "%d.\n", snum, bp->port);
379 return -EINVAL;
380 }
381 }
382
383 if (snum && snum < PROT_SOCK &&
384 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
385 return -EACCES;
386
387 /* See if the address matches any of the addresses we may have
388 * already bound before checking against other endpoints.
389 */
390 if (sctp_bind_addr_match(bp, addr, sp))
391 return -EINVAL;
392
393 /* Make sure we are allowed to bind here.
394 * The function sctp_get_port_local() does duplicate address
395 * detection.
396 */
397 addr->v4.sin_port = htons(snum);
398 if ((ret = sctp_get_port_local(sk, addr))) {
399 return -EADDRINUSE;
400 }
401
402 /* Refresh ephemeral port. */
403 if (!bp->port)
404 bp->port = inet_sk(sk)->inet_num;
405
406 /* Add the address to the bind address list.
407 * Use GFP_ATOMIC since BHs will be disabled.
408 */
409 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
410
411 /* Copy back into socket for getsockname() use. */
412 if (!ret) {
413 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
414 af->to_sk_saddr(addr, sk);
415 }
416
417 return ret;
418 }
419
420 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
421 *
422 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
423 * at any one time. If a sender, after sending an ASCONF chunk, decides
424 * it needs to transfer another ASCONF Chunk, it MUST wait until the
425 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
426 * subsequent ASCONF. Note this restriction binds each side, so at any
427 * time two ASCONF may be in-transit on any given association (one sent
428 * from each endpoint).
429 */
430 static int sctp_send_asconf(struct sctp_association *asoc,
431 struct sctp_chunk *chunk)
432 {
433 struct net *net = sock_net(asoc->base.sk);
434 int retval = 0;
435
436 /* If there is an outstanding ASCONF chunk, queue it for later
437 * transmission.
438 */
439 if (asoc->addip_last_asconf) {
440 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
441 goto out;
442 }
443
444 /* Hold the chunk until an ASCONF_ACK is received. */
445 sctp_chunk_hold(chunk);
446 retval = sctp_primitive_ASCONF(net, asoc, chunk);
447 if (retval)
448 sctp_chunk_free(chunk);
449 else
450 asoc->addip_last_asconf = chunk;
451
452 out:
453 return retval;
454 }
455
456 /* Add a list of addresses as bind addresses to local endpoint or
457 * association.
458 *
459 * Basically run through each address specified in the addrs/addrcnt
460 * array/length pair, determine if it is IPv6 or IPv4 and call
461 * sctp_do_bind() on it.
462 *
463 * If any of them fails, then the operation will be reversed and the
464 * ones that were added will be removed.
465 *
466 * Only sctp_setsockopt_bindx() is supposed to call this function.
467 */
468 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
469 {
470 int cnt;
471 int retval = 0;
472 void *addr_buf;
473 struct sockaddr *sa_addr;
474 struct sctp_af *af;
475
476 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
477 sk, addrs, addrcnt);
478
479 addr_buf = addrs;
480 for (cnt = 0; cnt < addrcnt; cnt++) {
481 /* The list may contain either IPv4 or IPv6 address;
482 * determine the address length for walking thru the list.
483 */
484 sa_addr = addr_buf;
485 af = sctp_get_af_specific(sa_addr->sa_family);
486 if (!af) {
487 retval = -EINVAL;
488 goto err_bindx_add;
489 }
490
491 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
492 af->sockaddr_len);
493
494 addr_buf += af->sockaddr_len;
495
496 err_bindx_add:
497 if (retval < 0) {
498 /* Failed. Cleanup the ones that have been added */
499 if (cnt > 0)
500 sctp_bindx_rem(sk, addrs, cnt);
501 return retval;
502 }
503 }
504
505 return retval;
506 }
507
508 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
509 * associations that are part of the endpoint indicating that a list of local
510 * addresses are added to the endpoint.
511 *
512 * If any of the addresses is already in the bind address list of the
513 * association, we do not send the chunk for that association. But it will not
514 * affect other associations.
515 *
516 * Only sctp_setsockopt_bindx() is supposed to call this function.
517 */
518 static int sctp_send_asconf_add_ip(struct sock *sk,
519 struct sockaddr *addrs,
520 int addrcnt)
521 {
522 struct net *net = sock_net(sk);
523 struct sctp_sock *sp;
524 struct sctp_endpoint *ep;
525 struct sctp_association *asoc;
526 struct sctp_bind_addr *bp;
527 struct sctp_chunk *chunk;
528 struct sctp_sockaddr_entry *laddr;
529 union sctp_addr *addr;
530 union sctp_addr saveaddr;
531 void *addr_buf;
532 struct sctp_af *af;
533 struct list_head *p;
534 int i;
535 int retval = 0;
536
537 if (!net->sctp.addip_enable)
538 return retval;
539
540 sp = sctp_sk(sk);
541 ep = sp->ep;
542
543 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
544 __func__, sk, addrs, addrcnt);
545
546 list_for_each_entry(asoc, &ep->asocs, asocs) {
547
548 if (!asoc->peer.asconf_capable)
549 continue;
550
551 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
552 continue;
553
554 if (!sctp_state(asoc, ESTABLISHED))
555 continue;
556
557 /* Check if any address in the packed array of addresses is
558 * in the bind address list of the association. If so,
559 * do not send the asconf chunk to its peer, but continue with
560 * other associations.
561 */
562 addr_buf = addrs;
563 for (i = 0; i < addrcnt; i++) {
564 addr = addr_buf;
565 af = sctp_get_af_specific(addr->v4.sin_family);
566 if (!af) {
567 retval = -EINVAL;
568 goto out;
569 }
570
571 if (sctp_assoc_lookup_laddr(asoc, addr))
572 break;
573
574 addr_buf += af->sockaddr_len;
575 }
576 if (i < addrcnt)
577 continue;
578
579 /* Use the first valid address in bind addr list of
580 * association as Address Parameter of ASCONF CHUNK.
581 */
582 bp = &asoc->base.bind_addr;
583 p = bp->address_list.next;
584 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
585 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
586 addrcnt, SCTP_PARAM_ADD_IP);
587 if (!chunk) {
588 retval = -ENOMEM;
589 goto out;
590 }
591
592 /* Add the new addresses to the bind address list with
593 * use_as_src set to 0.
594 */
595 addr_buf = addrs;
596 for (i = 0; i < addrcnt; i++) {
597 addr = addr_buf;
598 af = sctp_get_af_specific(addr->v4.sin_family);
599 memcpy(&saveaddr, addr, af->sockaddr_len);
600 retval = sctp_add_bind_addr(bp, &saveaddr,
601 SCTP_ADDR_NEW, GFP_ATOMIC);
602 addr_buf += af->sockaddr_len;
603 }
604 if (asoc->src_out_of_asoc_ok) {
605 struct sctp_transport *trans;
606
607 list_for_each_entry(trans,
608 &asoc->peer.transport_addr_list, transports) {
609 /* Clear the source and route cache */
610 dst_release(trans->dst);
611 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
612 2*asoc->pathmtu, 4380));
613 trans->ssthresh = asoc->peer.i.a_rwnd;
614 trans->rto = asoc->rto_initial;
615 sctp_max_rto(asoc, trans);
616 trans->rtt = trans->srtt = trans->rttvar = 0;
617 sctp_transport_route(trans, NULL,
618 sctp_sk(asoc->base.sk));
619 }
620 }
621 retval = sctp_send_asconf(asoc, chunk);
622 }
623
624 out:
625 return retval;
626 }
627
628 /* Remove a list of addresses from bind addresses list. Do not remove the
629 * last address.
630 *
631 * Basically run through each address specified in the addrs/addrcnt
632 * array/length pair, determine if it is IPv6 or IPv4 and call
633 * sctp_del_bind() on it.
634 *
635 * If any of them fails, then the operation will be reversed and the
636 * ones that were removed will be added back.
637 *
638 * At least one address has to be left; if only one address is
639 * available, the operation will return -EBUSY.
640 *
641 * Only sctp_setsockopt_bindx() is supposed to call this function.
642 */
643 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
644 {
645 struct sctp_sock *sp = sctp_sk(sk);
646 struct sctp_endpoint *ep = sp->ep;
647 int cnt;
648 struct sctp_bind_addr *bp = &ep->base.bind_addr;
649 int retval = 0;
650 void *addr_buf;
651 union sctp_addr *sa_addr;
652 struct sctp_af *af;
653
654 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
655 sk, addrs, addrcnt);
656
657 addr_buf = addrs;
658 for (cnt = 0; cnt < addrcnt; cnt++) {
659 /* If the bind address list is empty or if there is only one
660 * bind address, there is nothing more to be removed (we need
661 * at least one address here).
662 */
663 if (list_empty(&bp->address_list) ||
664 (sctp_list_single_entry(&bp->address_list))) {
665 retval = -EBUSY;
666 goto err_bindx_rem;
667 }
668
669 sa_addr = addr_buf;
670 af = sctp_get_af_specific(sa_addr->sa.sa_family);
671 if (!af) {
672 retval = -EINVAL;
673 goto err_bindx_rem;
674 }
675
676 if (!af->addr_valid(sa_addr, sp, NULL)) {
677 retval = -EADDRNOTAVAIL;
678 goto err_bindx_rem;
679 }
680
681 if (sa_addr->v4.sin_port &&
682 sa_addr->v4.sin_port != htons(bp->port)) {
683 retval = -EINVAL;
684 goto err_bindx_rem;
685 }
686
687 if (!sa_addr->v4.sin_port)
688 sa_addr->v4.sin_port = htons(bp->port);
689
690 /* FIXME - There is probably a need to check if sk->sk_saddr and
691 * sk->sk_rcv_addr are currently set to one of the addresses to
692 * be removed. This is something which needs to be looked into
693 * when we are fixing the outstanding issues with multi-homing
694 * socket routing and failover schemes. Refer to comments in
695 * sctp_do_bind(). -daisy
696 */
697 retval = sctp_del_bind_addr(bp, sa_addr);
698
699 addr_buf += af->sockaddr_len;
700 err_bindx_rem:
701 if (retval < 0) {
702 /* Failed. Add the ones that has been removed back */
703 if (cnt > 0)
704 sctp_bindx_add(sk, addrs, cnt);
705 return retval;
706 }
707 }
708
709 return retval;
710 }
711
712 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
713 * the associations that are part of the endpoint indicating that a list of
714 * local addresses are removed from the endpoint.
715 *
716 * If any of the addresses is already in the bind address list of the
717 * association, we do not send the chunk for that association. But it will not
718 * affect other associations.
719 *
720 * Only sctp_setsockopt_bindx() is supposed to call this function.
721 */
722 static int sctp_send_asconf_del_ip(struct sock *sk,
723 struct sockaddr *addrs,
724 int addrcnt)
725 {
726 struct net *net = sock_net(sk);
727 struct sctp_sock *sp;
728 struct sctp_endpoint *ep;
729 struct sctp_association *asoc;
730 struct sctp_transport *transport;
731 struct sctp_bind_addr *bp;
732 struct sctp_chunk *chunk;
733 union sctp_addr *laddr;
734 void *addr_buf;
735 struct sctp_af *af;
736 struct sctp_sockaddr_entry *saddr;
737 int i;
738 int retval = 0;
739 int stored = 0;
740
741 chunk = NULL;
742 if (!net->sctp.addip_enable)
743 return retval;
744
745 sp = sctp_sk(sk);
746 ep = sp->ep;
747
748 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
749 __func__, sk, addrs, addrcnt);
750
751 list_for_each_entry(asoc, &ep->asocs, asocs) {
752
753 if (!asoc->peer.asconf_capable)
754 continue;
755
756 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
757 continue;
758
759 if (!sctp_state(asoc, ESTABLISHED))
760 continue;
761
762 /* Check if any address in the packed array of addresses is
763 * not present in the bind address list of the association.
764 * If so, do not send the asconf chunk to its peer, but
765 * continue with other associations.
766 */
767 addr_buf = addrs;
768 for (i = 0; i < addrcnt; i++) {
769 laddr = addr_buf;
770 af = sctp_get_af_specific(laddr->v4.sin_family);
771 if (!af) {
772 retval = -EINVAL;
773 goto out;
774 }
775
776 if (!sctp_assoc_lookup_laddr(asoc, laddr))
777 break;
778
779 addr_buf += af->sockaddr_len;
780 }
781 if (i < addrcnt)
782 continue;
783
784 /* Find one address in the association's bind address list
785 * that is not in the packed array of addresses. This is to
786 * make sure that we do not delete all the addresses in the
787 * association.
788 */
789 bp = &asoc->base.bind_addr;
790 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
791 addrcnt, sp);
792 if ((laddr == NULL) && (addrcnt == 1)) {
793 if (asoc->asconf_addr_del_pending)
794 continue;
795 asoc->asconf_addr_del_pending =
796 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
797 if (asoc->asconf_addr_del_pending == NULL) {
798 retval = -ENOMEM;
799 goto out;
800 }
801 asoc->asconf_addr_del_pending->sa.sa_family =
802 addrs->sa_family;
803 asoc->asconf_addr_del_pending->v4.sin_port =
804 htons(bp->port);
805 if (addrs->sa_family == AF_INET) {
806 struct sockaddr_in *sin;
807
808 sin = (struct sockaddr_in *)addrs;
809 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
810 } else if (addrs->sa_family == AF_INET6) {
811 struct sockaddr_in6 *sin6;
812
813 sin6 = (struct sockaddr_in6 *)addrs;
814 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
815 }
816 SCTP_DEBUG_PRINTK_IPADDR("send_asconf_del_ip: keep the last address asoc: %p ",
817 " at %p\n", asoc, asoc->asconf_addr_del_pending,
818 asoc->asconf_addr_del_pending);
819 asoc->src_out_of_asoc_ok = 1;
820 stored = 1;
821 goto skip_mkasconf;
822 }
823
824 if (laddr == NULL)
825 return -EINVAL;
826
827 /* We do not need RCU protection throughout this loop
828 * because this is done under a socket lock from the
829 * setsockopt call.
830 */
831 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
832 SCTP_PARAM_DEL_IP);
833 if (!chunk) {
834 retval = -ENOMEM;
835 goto out;
836 }
837
838 skip_mkasconf:
839 /* Reset use_as_src flag for the addresses in the bind address
840 * list that are to be deleted.
841 */
842 addr_buf = addrs;
843 for (i = 0; i < addrcnt; i++) {
844 laddr = addr_buf;
845 af = sctp_get_af_specific(laddr->v4.sin_family);
846 list_for_each_entry(saddr, &bp->address_list, list) {
847 if (sctp_cmp_addr_exact(&saddr->a, laddr))
848 saddr->state = SCTP_ADDR_DEL;
849 }
850 addr_buf += af->sockaddr_len;
851 }
852
853 /* Update the route and saddr entries for all the transports
854 * as some of the addresses in the bind address list are
855 * about to be deleted and cannot be used as source addresses.
856 */
857 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
858 transports) {
859 dst_release(transport->dst);
860 sctp_transport_route(transport, NULL,
861 sctp_sk(asoc->base.sk));
862 }
863
864 if (stored)
865 /* We don't need to transmit ASCONF */
866 continue;
867 retval = sctp_send_asconf(asoc, chunk);
868 }
869 out:
870 return retval;
871 }
872
873 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
874 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
875 {
876 struct sock *sk = sctp_opt2sk(sp);
877 union sctp_addr *addr;
878 struct sctp_af *af;
879
880 /* It is safe to write port space in caller. */
881 addr = &addrw->a;
882 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
883 af = sctp_get_af_specific(addr->sa.sa_family);
884 if (!af)
885 return -EINVAL;
886 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
887 return -EINVAL;
888
889 if (addrw->state == SCTP_ADDR_NEW)
890 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
891 else
892 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
893 }
894
895 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
896 *
897 * API 8.1
898 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
899 * int flags);
900 *
901 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
902 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
903 * or IPv6 addresses.
904 *
905 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
906 * Section 3.1.2 for this usage.
907 *
908 * addrs is a pointer to an array of one or more socket addresses. Each
909 * address is contained in its appropriate structure (i.e. struct
910 * sockaddr_in or struct sockaddr_in6) the family of the address type
911 * must be used to distinguish the address length (note that this
912 * representation is termed a "packed array" of addresses). The caller
913 * specifies the number of addresses in the array with addrcnt.
914 *
915 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
916 * -1, and sets errno to the appropriate error code.
917 *
918 * For SCTP, the port given in each socket address must be the same, or
919 * sctp_bindx() will fail, setting errno to EINVAL.
920 *
921 * The flags parameter is formed from the bitwise OR of zero or more of
922 * the following currently defined flags:
923 *
924 * SCTP_BINDX_ADD_ADDR
925 *
926 * SCTP_BINDX_REM_ADDR
927 *
928 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
929 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
930 * addresses from the association. The two flags are mutually exclusive;
931 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
932 * not remove all addresses from an association; sctp_bindx() will
933 * reject such an attempt with EINVAL.
934 *
935 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
936 * additional addresses with an endpoint after calling bind(). Or use
937 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
938 * socket is associated with so that no new association accepted will be
939 * associated with those addresses. If the endpoint supports dynamic
940 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
941 * endpoint to send the appropriate message to the peer to change the
942 * peers address lists.
943 *
944 * Adding and removing addresses from a connected association is
945 * optional functionality. Implementations that do not support this
946 * functionality should return EOPNOTSUPP.
947 *
948 * Basically do nothing but copying the addresses from user to kernel
949 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
950 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
951 * from userspace.
952 *
953 * We don't use copy_from_user() for optimization: we first do the
954 * sanity checks (buffer size -fast- and access check-healthy
955 * pointer); if all of those succeed, then we can alloc the memory
956 * (expensive operation) needed to copy the data to kernel. Then we do
957 * the copying without checking the user space area
958 * (__copy_from_user()).
959 *
960 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
961 * it.
962 *
963 * sk The sk of the socket
964 * addrs The pointer to the addresses in user land
965 * addrssize Size of the addrs buffer
966 * op Operation to perform (add or remove, see the flags of
967 * sctp_bindx)
968 *
969 * Returns 0 if ok, <0 errno code on error.
970 */
971 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
972 struct sockaddr __user *addrs,
973 int addrs_size, int op)
974 {
975 struct sockaddr *kaddrs;
976 int err;
977 int addrcnt = 0;
978 int walk_size = 0;
979 struct sockaddr *sa_addr;
980 void *addr_buf;
981 struct sctp_af *af;
982
983 SCTP_DEBUG_PRINTK("sctp_setsockopt_bindx: sk %p addrs %p"
984 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
985
986 if (unlikely(addrs_size <= 0))
987 return -EINVAL;
988
989 /* Check the user passed a healthy pointer. */
990 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
991 return -EFAULT;
992
993 /* Alloc space for the address array in kernel memory. */
994 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
995 if (unlikely(!kaddrs))
996 return -ENOMEM;
997
998 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
999 kfree(kaddrs);
1000 return -EFAULT;
1001 }
1002
1003 /* Walk through the addrs buffer and count the number of addresses. */
1004 addr_buf = kaddrs;
1005 while (walk_size < addrs_size) {
1006 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1007 kfree(kaddrs);
1008 return -EINVAL;
1009 }
1010
1011 sa_addr = addr_buf;
1012 af = sctp_get_af_specific(sa_addr->sa_family);
1013
1014 /* If the address family is not supported or if this address
1015 * causes the address buffer to overflow return EINVAL.
1016 */
1017 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1018 kfree(kaddrs);
1019 return -EINVAL;
1020 }
1021 addrcnt++;
1022 addr_buf += af->sockaddr_len;
1023 walk_size += af->sockaddr_len;
1024 }
1025
1026 /* Do the work. */
1027 switch (op) {
1028 case SCTP_BINDX_ADD_ADDR:
1029 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1030 if (err)
1031 goto out;
1032 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1033 break;
1034
1035 case SCTP_BINDX_REM_ADDR:
1036 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1037 if (err)
1038 goto out;
1039 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1040 break;
1041
1042 default:
1043 err = -EINVAL;
1044 break;
1045 }
1046
1047 out:
1048 kfree(kaddrs);
1049
1050 return err;
1051 }
1052
1053 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1054 *
1055 * Common routine for handling connect() and sctp_connectx().
1056 * Connect will come in with just a single address.
1057 */
1058 static int __sctp_connect(struct sock* sk,
1059 struct sockaddr *kaddrs,
1060 int addrs_size,
1061 sctp_assoc_t *assoc_id)
1062 {
1063 struct net *net = sock_net(sk);
1064 struct sctp_sock *sp;
1065 struct sctp_endpoint *ep;
1066 struct sctp_association *asoc = NULL;
1067 struct sctp_association *asoc2;
1068 struct sctp_transport *transport;
1069 union sctp_addr to;
1070 struct sctp_af *af;
1071 sctp_scope_t scope;
1072 long timeo;
1073 int err = 0;
1074 int addrcnt = 0;
1075 int walk_size = 0;
1076 union sctp_addr *sa_addr = NULL;
1077 void *addr_buf;
1078 unsigned short port;
1079 unsigned int f_flags = 0;
1080
1081 sp = sctp_sk(sk);
1082 ep = sp->ep;
1083
1084 /* connect() cannot be done on a socket that is already in ESTABLISHED
1085 * state - UDP-style peeled off socket or a TCP-style socket that
1086 * is already connected.
1087 * It cannot be done even on a TCP-style listening socket.
1088 */
1089 if (sctp_sstate(sk, ESTABLISHED) ||
1090 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1091 err = -EISCONN;
1092 goto out_free;
1093 }
1094
1095 /* Walk through the addrs buffer and count the number of addresses. */
1096 addr_buf = kaddrs;
1097 while (walk_size < addrs_size) {
1098 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1099 err = -EINVAL;
1100 goto out_free;
1101 }
1102
1103 sa_addr = addr_buf;
1104 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1105
1106 /* If the address family is not supported or if this address
1107 * causes the address buffer to overflow return EINVAL.
1108 */
1109 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1110 err = -EINVAL;
1111 goto out_free;
1112 }
1113
1114 port = ntohs(sa_addr->v4.sin_port);
1115
1116 /* Save current address so we can work with it */
1117 memcpy(&to, sa_addr, af->sockaddr_len);
1118
1119 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1120 if (err)
1121 goto out_free;
1122
1123 /* Make sure the destination port is correctly set
1124 * in all addresses.
1125 */
1126 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1127 err = -EINVAL;
1128 goto out_free;
1129 }
1130
1131 /* Check if there already is a matching association on the
1132 * endpoint (other than the one created here).
1133 */
1134 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1135 if (asoc2 && asoc2 != asoc) {
1136 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1137 err = -EISCONN;
1138 else
1139 err = -EALREADY;
1140 goto out_free;
1141 }
1142
1143 /* If we could not find a matching association on the endpoint,
1144 * make sure that there is no peeled-off association matching
1145 * the peer address even on another socket.
1146 */
1147 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1148 err = -EADDRNOTAVAIL;
1149 goto out_free;
1150 }
1151
1152 if (!asoc) {
1153 /* If a bind() or sctp_bindx() is not called prior to
1154 * an sctp_connectx() call, the system picks an
1155 * ephemeral port and will choose an address set
1156 * equivalent to binding with a wildcard address.
1157 */
1158 if (!ep->base.bind_addr.port) {
1159 if (sctp_autobind(sk)) {
1160 err = -EAGAIN;
1161 goto out_free;
1162 }
1163 } else {
1164 /*
1165 * If an unprivileged user inherits a 1-many
1166 * style socket with open associations on a
1167 * privileged port, it MAY be permitted to
1168 * accept new associations, but it SHOULD NOT
1169 * be permitted to open new associations.
1170 */
1171 if (ep->base.bind_addr.port < PROT_SOCK &&
1172 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1173 err = -EACCES;
1174 goto out_free;
1175 }
1176 }
1177
1178 scope = sctp_scope(&to);
1179 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1180 if (!asoc) {
1181 err = -ENOMEM;
1182 goto out_free;
1183 }
1184
1185 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1186 GFP_KERNEL);
1187 if (err < 0) {
1188 goto out_free;
1189 }
1190
1191 }
1192
1193 /* Prime the peer's transport structures. */
1194 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1195 SCTP_UNKNOWN);
1196 if (!transport) {
1197 err = -ENOMEM;
1198 goto out_free;
1199 }
1200
1201 addrcnt++;
1202 addr_buf += af->sockaddr_len;
1203 walk_size += af->sockaddr_len;
1204 }
1205
1206 /* In case the user of sctp_connectx() wants an association
1207 * id back, assign one now.
1208 */
1209 if (assoc_id) {
1210 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1211 if (err < 0)
1212 goto out_free;
1213 }
1214
1215 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1216 if (err < 0) {
1217 goto out_free;
1218 }
1219
1220 /* Initialize sk's dport and daddr for getpeername() */
1221 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1222 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1223 af->to_sk_daddr(sa_addr, sk);
1224 sk->sk_err = 0;
1225
1226 /* in-kernel sockets don't generally have a file allocated to them
1227 * if all they do is call sock_create_kern().
1228 */
1229 if (sk->sk_socket->file)
1230 f_flags = sk->sk_socket->file->f_flags;
1231
1232 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1233
1234 if (assoc_id)
1235 *assoc_id = asoc->assoc_id;
1236 err = sctp_wait_for_connect(asoc, &timeo);
1237 /* Note: the asoc may be freed after the return of
1238 * sctp_wait_for_connect.
1239 */
1240
1241 /* Don't free association on exit. */
1242 asoc = NULL;
1243
1244 out_free:
1245
1246 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1247 " kaddrs: %p err: %d\n",
1248 asoc, kaddrs, err);
1249 if (asoc) {
1250 /* sctp_primitive_ASSOCIATE may have added this association
1251 * To the hash table, try to unhash it, just in case, its a noop
1252 * if it wasn't hashed so we're safe
1253 */
1254 sctp_unhash_established(asoc);
1255 sctp_association_free(asoc);
1256 }
1257 return err;
1258 }
1259
1260 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1261 *
1262 * API 8.9
1263 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1264 * sctp_assoc_t *asoc);
1265 *
1266 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1267 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1268 * or IPv6 addresses.
1269 *
1270 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1271 * Section 3.1.2 for this usage.
1272 *
1273 * addrs is a pointer to an array of one or more socket addresses. Each
1274 * address is contained in its appropriate structure (i.e. struct
1275 * sockaddr_in or struct sockaddr_in6) the family of the address type
1276 * must be used to distengish the address length (note that this
1277 * representation is termed a "packed array" of addresses). The caller
1278 * specifies the number of addresses in the array with addrcnt.
1279 *
1280 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1281 * the association id of the new association. On failure, sctp_connectx()
1282 * returns -1, and sets errno to the appropriate error code. The assoc_id
1283 * is not touched by the kernel.
1284 *
1285 * For SCTP, the port given in each socket address must be the same, or
1286 * sctp_connectx() will fail, setting errno to EINVAL.
1287 *
1288 * An application can use sctp_connectx to initiate an association with
1289 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1290 * allows a caller to specify multiple addresses at which a peer can be
1291 * reached. The way the SCTP stack uses the list of addresses to set up
1292 * the association is implementation dependent. This function only
1293 * specifies that the stack will try to make use of all the addresses in
1294 * the list when needed.
1295 *
1296 * Note that the list of addresses passed in is only used for setting up
1297 * the association. It does not necessarily equal the set of addresses
1298 * the peer uses for the resulting association. If the caller wants to
1299 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1300 * retrieve them after the association has been set up.
1301 *
1302 * Basically do nothing but copying the addresses from user to kernel
1303 * land and invoking either sctp_connectx(). This is used for tunneling
1304 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1305 *
1306 * We don't use copy_from_user() for optimization: we first do the
1307 * sanity checks (buffer size -fast- and access check-healthy
1308 * pointer); if all of those succeed, then we can alloc the memory
1309 * (expensive operation) needed to copy the data to kernel. Then we do
1310 * the copying without checking the user space area
1311 * (__copy_from_user()).
1312 *
1313 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1314 * it.
1315 *
1316 * sk The sk of the socket
1317 * addrs The pointer to the addresses in user land
1318 * addrssize Size of the addrs buffer
1319 *
1320 * Returns >=0 if ok, <0 errno code on error.
1321 */
1322 SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
1323 struct sockaddr __user *addrs,
1324 int addrs_size,
1325 sctp_assoc_t *assoc_id)
1326 {
1327 int err = 0;
1328 struct sockaddr *kaddrs;
1329
1330 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1331 __func__, sk, addrs, addrs_size);
1332
1333 if (unlikely(addrs_size <= 0))
1334 return -EINVAL;
1335
1336 /* Check the user passed a healthy pointer. */
1337 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1338 return -EFAULT;
1339
1340 /* Alloc space for the address array in kernel memory. */
1341 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1342 if (unlikely(!kaddrs))
1343 return -ENOMEM;
1344
1345 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1346 err = -EFAULT;
1347 } else {
1348 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1349 }
1350
1351 kfree(kaddrs);
1352
1353 return err;
1354 }
1355
1356 /*
1357 * This is an older interface. It's kept for backward compatibility
1358 * to the option that doesn't provide association id.
1359 */
1360 SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1361 struct sockaddr __user *addrs,
1362 int addrs_size)
1363 {
1364 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1365 }
1366
1367 /*
1368 * New interface for the API. The since the API is done with a socket
1369 * option, to make it simple we feed back the association id is as a return
1370 * indication to the call. Error is always negative and association id is
1371 * always positive.
1372 */
1373 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1374 struct sockaddr __user *addrs,
1375 int addrs_size)
1376 {
1377 sctp_assoc_t assoc_id = 0;
1378 int err = 0;
1379
1380 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1381
1382 if (err)
1383 return err;
1384 else
1385 return assoc_id;
1386 }
1387
1388 /*
1389 * New (hopefully final) interface for the API.
1390 * We use the sctp_getaddrs_old structure so that use-space library
1391 * can avoid any unnecessary allocations. The only different part
1392 * is that we store the actual length of the address buffer into the
1393 * addrs_num structure member. That way we can re-use the existing
1394 * code.
1395 */
1396 #ifdef CONFIG_COMPAT
1397 struct compat_sctp_getaddrs_old {
1398 sctp_assoc_t assoc_id;
1399 s32 addr_num;
1400 compat_uptr_t addrs; /* struct sockaddr * */
1401 };
1402 #endif
1403
1404 SCTP_STATIC int sctp_getsockopt_connectx3(struct sock* sk, int len,
1405 char __user *optval,
1406 int __user *optlen)
1407 {
1408 struct sctp_getaddrs_old param;
1409 sctp_assoc_t assoc_id = 0;
1410 int err = 0;
1411
1412 #ifdef CONFIG_COMPAT
1413 if (is_compat_task()) {
1414 struct compat_sctp_getaddrs_old param32;
1415
1416 if (len < sizeof(param32))
1417 return -EINVAL;
1418 if (copy_from_user(&param32, optval, sizeof(param32)))
1419 return -EFAULT;
1420
1421 param.assoc_id = param32.assoc_id;
1422 param.addr_num = param32.addr_num;
1423 param.addrs = compat_ptr(param32.addrs);
1424 } else
1425 #endif
1426 {
1427 if (len < sizeof(param))
1428 return -EINVAL;
1429 if (copy_from_user(&param, optval, sizeof(param)))
1430 return -EFAULT;
1431 }
1432
1433 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1434 param.addrs, param.addr_num,
1435 &assoc_id);
1436 if (err == 0 || err == -EINPROGRESS) {
1437 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1438 return -EFAULT;
1439 if (put_user(sizeof(assoc_id), optlen))
1440 return -EFAULT;
1441 }
1442
1443 return err;
1444 }
1445
1446 /* API 3.1.4 close() - UDP Style Syntax
1447 * Applications use close() to perform graceful shutdown (as described in
1448 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1449 * by a UDP-style socket.
1450 *
1451 * The syntax is
1452 *
1453 * ret = close(int sd);
1454 *
1455 * sd - the socket descriptor of the associations to be closed.
1456 *
1457 * To gracefully shutdown a specific association represented by the
1458 * UDP-style socket, an application should use the sendmsg() call,
1459 * passing no user data, but including the appropriate flag in the
1460 * ancillary data (see Section xxxx).
1461 *
1462 * If sd in the close() call is a branched-off socket representing only
1463 * one association, the shutdown is performed on that association only.
1464 *
1465 * 4.1.6 close() - TCP Style Syntax
1466 *
1467 * Applications use close() to gracefully close down an association.
1468 *
1469 * The syntax is:
1470 *
1471 * int close(int sd);
1472 *
1473 * sd - the socket descriptor of the association to be closed.
1474 *
1475 * After an application calls close() on a socket descriptor, no further
1476 * socket operations will succeed on that descriptor.
1477 *
1478 * API 7.1.4 SO_LINGER
1479 *
1480 * An application using the TCP-style socket can use this option to
1481 * perform the SCTP ABORT primitive. The linger option structure is:
1482 *
1483 * struct linger {
1484 * int l_onoff; // option on/off
1485 * int l_linger; // linger time
1486 * };
1487 *
1488 * To enable the option, set l_onoff to 1. If the l_linger value is set
1489 * to 0, calling close() is the same as the ABORT primitive. If the
1490 * value is set to a negative value, the setsockopt() call will return
1491 * an error. If the value is set to a positive value linger_time, the
1492 * close() can be blocked for at most linger_time ms. If the graceful
1493 * shutdown phase does not finish during this period, close() will
1494 * return but the graceful shutdown phase continues in the system.
1495 */
1496 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1497 {
1498 struct net *net = sock_net(sk);
1499 struct sctp_endpoint *ep;
1500 struct sctp_association *asoc;
1501 struct list_head *pos, *temp;
1502 unsigned int data_was_unread;
1503
1504 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1505
1506 sctp_lock_sock(sk);
1507 sk->sk_shutdown = SHUTDOWN_MASK;
1508 sk->sk_state = SCTP_SS_CLOSING;
1509
1510 ep = sctp_sk(sk)->ep;
1511
1512 /* Clean up any skbs sitting on the receive queue. */
1513 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1514 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1515
1516 /* Walk all associations on an endpoint. */
1517 list_for_each_safe(pos, temp, &ep->asocs) {
1518 asoc = list_entry(pos, struct sctp_association, asocs);
1519
1520 if (sctp_style(sk, TCP)) {
1521 /* A closed association can still be in the list if
1522 * it belongs to a TCP-style listening socket that is
1523 * not yet accepted. If so, free it. If not, send an
1524 * ABORT or SHUTDOWN based on the linger options.
1525 */
1526 if (sctp_state(asoc, CLOSED)) {
1527 sctp_unhash_established(asoc);
1528 sctp_association_free(asoc);
1529 continue;
1530 }
1531 }
1532
1533 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1534 !skb_queue_empty(&asoc->ulpq.reasm) ||
1535 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1536 struct sctp_chunk *chunk;
1537
1538 chunk = sctp_make_abort_user(asoc, NULL, 0);
1539 sctp_primitive_ABORT(net, asoc, chunk);
1540 } else
1541 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1542 }
1543
1544 /* On a TCP-style socket, block for at most linger_time if set. */
1545 if (sctp_style(sk, TCP) && timeout)
1546 sctp_wait_for_close(sk, timeout);
1547
1548 /* This will run the backlog queue. */
1549 sctp_release_sock(sk);
1550
1551 /* Supposedly, no process has access to the socket, but
1552 * the net layers still may.
1553 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1554 * held and that should be grabbed before socket lock.
1555 */
1556 spin_lock_bh(&net->sctp.addr_wq_lock);
1557 sctp_bh_lock_sock(sk);
1558
1559 /* Hold the sock, since sk_common_release() will put sock_put()
1560 * and we have just a little more cleanup.
1561 */
1562 sock_hold(sk);
1563 sk_common_release(sk);
1564
1565 sctp_bh_unlock_sock(sk);
1566 spin_unlock_bh(&net->sctp.addr_wq_lock);
1567
1568 sock_put(sk);
1569
1570 SCTP_DBG_OBJCNT_DEC(sock);
1571 }
1572
1573 /* Handle EPIPE error. */
1574 static int sctp_error(struct sock *sk, int flags, int err)
1575 {
1576 if (err == -EPIPE)
1577 err = sock_error(sk) ? : -EPIPE;
1578 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1579 send_sig(SIGPIPE, current, 0);
1580 return err;
1581 }
1582
1583 /* API 3.1.3 sendmsg() - UDP Style Syntax
1584 *
1585 * An application uses sendmsg() and recvmsg() calls to transmit data to
1586 * and receive data from its peer.
1587 *
1588 * ssize_t sendmsg(int socket, const struct msghdr *message,
1589 * int flags);
1590 *
1591 * socket - the socket descriptor of the endpoint.
1592 * message - pointer to the msghdr structure which contains a single
1593 * user message and possibly some ancillary data.
1594 *
1595 * See Section 5 for complete description of the data
1596 * structures.
1597 *
1598 * flags - flags sent or received with the user message, see Section
1599 * 5 for complete description of the flags.
1600 *
1601 * Note: This function could use a rewrite especially when explicit
1602 * connect support comes in.
1603 */
1604 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1605
1606 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1607
1608 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1609 struct msghdr *msg, size_t msg_len)
1610 {
1611 struct net *net = sock_net(sk);
1612 struct sctp_sock *sp;
1613 struct sctp_endpoint *ep;
1614 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1615 struct sctp_transport *transport, *chunk_tp;
1616 struct sctp_chunk *chunk;
1617 union sctp_addr to;
1618 struct sockaddr *msg_name = NULL;
1619 struct sctp_sndrcvinfo default_sinfo;
1620 struct sctp_sndrcvinfo *sinfo;
1621 struct sctp_initmsg *sinit;
1622 sctp_assoc_t associd = 0;
1623 sctp_cmsgs_t cmsgs = { NULL };
1624 int err;
1625 sctp_scope_t scope;
1626 long timeo;
1627 __u16 sinfo_flags = 0;
1628 struct sctp_datamsg *datamsg;
1629 int msg_flags = msg->msg_flags;
1630
1631 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1632 sk, msg, msg_len);
1633
1634 err = 0;
1635 sp = sctp_sk(sk);
1636 ep = sp->ep;
1637
1638 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1639
1640 /* We cannot send a message over a TCP-style listening socket. */
1641 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1642 err = -EPIPE;
1643 goto out_nounlock;
1644 }
1645
1646 /* Parse out the SCTP CMSGs. */
1647 err = sctp_msghdr_parse(msg, &cmsgs);
1648
1649 if (err) {
1650 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1651 goto out_nounlock;
1652 }
1653
1654 /* Fetch the destination address for this packet. This
1655 * address only selects the association--it is not necessarily
1656 * the address we will send to.
1657 * For a peeled-off socket, msg_name is ignored.
1658 */
1659 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1660 int msg_namelen = msg->msg_namelen;
1661
1662 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1663 msg_namelen);
1664 if (err)
1665 return err;
1666
1667 if (msg_namelen > sizeof(to))
1668 msg_namelen = sizeof(to);
1669 memcpy(&to, msg->msg_name, msg_namelen);
1670 msg_name = msg->msg_name;
1671 }
1672
1673 sinfo = cmsgs.info;
1674 sinit = cmsgs.init;
1675
1676 /* Did the user specify SNDRCVINFO? */
1677 if (sinfo) {
1678 sinfo_flags = sinfo->sinfo_flags;
1679 associd = sinfo->sinfo_assoc_id;
1680 }
1681
1682 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1683 msg_len, sinfo_flags);
1684
1685 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1686 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1687 err = -EINVAL;
1688 goto out_nounlock;
1689 }
1690
1691 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1692 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1693 * If SCTP_ABORT is set, the message length could be non zero with
1694 * the msg_iov set to the user abort reason.
1695 */
1696 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1697 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1698 err = -EINVAL;
1699 goto out_nounlock;
1700 }
1701
1702 /* If SCTP_ADDR_OVER is set, there must be an address
1703 * specified in msg_name.
1704 */
1705 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1706 err = -EINVAL;
1707 goto out_nounlock;
1708 }
1709
1710 transport = NULL;
1711
1712 SCTP_DEBUG_PRINTK("About to look up association.\n");
1713
1714 sctp_lock_sock(sk);
1715
1716 /* If a msg_name has been specified, assume this is to be used. */
1717 if (msg_name) {
1718 /* Look for a matching association on the endpoint. */
1719 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1720 if (!asoc) {
1721 /* If we could not find a matching association on the
1722 * endpoint, make sure that it is not a TCP-style
1723 * socket that already has an association or there is
1724 * no peeled-off association on another socket.
1725 */
1726 if ((sctp_style(sk, TCP) &&
1727 sctp_sstate(sk, ESTABLISHED)) ||
1728 sctp_endpoint_is_peeled_off(ep, &to)) {
1729 err = -EADDRNOTAVAIL;
1730 goto out_unlock;
1731 }
1732 }
1733 } else {
1734 asoc = sctp_id2assoc(sk, associd);
1735 if (!asoc) {
1736 err = -EPIPE;
1737 goto out_unlock;
1738 }
1739 }
1740
1741 if (asoc) {
1742 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1743
1744 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1745 * socket that has an association in CLOSED state. This can
1746 * happen when an accepted socket has an association that is
1747 * already CLOSED.
1748 */
1749 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1750 err = -EPIPE;
1751 goto out_unlock;
1752 }
1753
1754 if (sinfo_flags & SCTP_EOF) {
1755 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1756 asoc);
1757 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1758 err = 0;
1759 goto out_unlock;
1760 }
1761 if (sinfo_flags & SCTP_ABORT) {
1762
1763 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1764 if (!chunk) {
1765 err = -ENOMEM;
1766 goto out_unlock;
1767 }
1768
1769 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1770 sctp_primitive_ABORT(net, asoc, chunk);
1771 err = 0;
1772 goto out_unlock;
1773 }
1774 }
1775
1776 /* Do we need to create the association? */
1777 if (!asoc) {
1778 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1779
1780 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1781 err = -EINVAL;
1782 goto out_unlock;
1783 }
1784
1785 /* Check for invalid stream against the stream counts,
1786 * either the default or the user specified stream counts.
1787 */
1788 if (sinfo) {
1789 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1790 /* Check against the defaults. */
1791 if (sinfo->sinfo_stream >=
1792 sp->initmsg.sinit_num_ostreams) {
1793 err = -EINVAL;
1794 goto out_unlock;
1795 }
1796 } else {
1797 /* Check against the requested. */
1798 if (sinfo->sinfo_stream >=
1799 sinit->sinit_num_ostreams) {
1800 err = -EINVAL;
1801 goto out_unlock;
1802 }
1803 }
1804 }
1805
1806 /*
1807 * API 3.1.2 bind() - UDP Style Syntax
1808 * If a bind() or sctp_bindx() is not called prior to a
1809 * sendmsg() call that initiates a new association, the
1810 * system picks an ephemeral port and will choose an address
1811 * set equivalent to binding with a wildcard address.
1812 */
1813 if (!ep->base.bind_addr.port) {
1814 if (sctp_autobind(sk)) {
1815 err = -EAGAIN;
1816 goto out_unlock;
1817 }
1818 } else {
1819 /*
1820 * If an unprivileged user inherits a one-to-many
1821 * style socket with open associations on a privileged
1822 * port, it MAY be permitted to accept new associations,
1823 * but it SHOULD NOT be permitted to open new
1824 * associations.
1825 */
1826 if (ep->base.bind_addr.port < PROT_SOCK &&
1827 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1828 err = -EACCES;
1829 goto out_unlock;
1830 }
1831 }
1832
1833 scope = sctp_scope(&to);
1834 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1835 if (!new_asoc) {
1836 err = -ENOMEM;
1837 goto out_unlock;
1838 }
1839 asoc = new_asoc;
1840 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1841 if (err < 0) {
1842 err = -ENOMEM;
1843 goto out_free;
1844 }
1845
1846 /* If the SCTP_INIT ancillary data is specified, set all
1847 * the association init values accordingly.
1848 */
1849 if (sinit) {
1850 if (sinit->sinit_num_ostreams) {
1851 asoc->c.sinit_num_ostreams =
1852 sinit->sinit_num_ostreams;
1853 }
1854 if (sinit->sinit_max_instreams) {
1855 asoc->c.sinit_max_instreams =
1856 sinit->sinit_max_instreams;
1857 }
1858 if (sinit->sinit_max_attempts) {
1859 asoc->max_init_attempts
1860 = sinit->sinit_max_attempts;
1861 }
1862 if (sinit->sinit_max_init_timeo) {
1863 asoc->max_init_timeo =
1864 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1865 }
1866 }
1867
1868 /* Prime the peer's transport structures. */
1869 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1870 if (!transport) {
1871 err = -ENOMEM;
1872 goto out_free;
1873 }
1874 }
1875
1876 /* ASSERT: we have a valid association at this point. */
1877 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1878
1879 if (!sinfo) {
1880 /* If the user didn't specify SNDRCVINFO, make up one with
1881 * some defaults.
1882 */
1883 memset(&default_sinfo, 0, sizeof(default_sinfo));
1884 default_sinfo.sinfo_stream = asoc->default_stream;
1885 default_sinfo.sinfo_flags = asoc->default_flags;
1886 default_sinfo.sinfo_ppid = asoc->default_ppid;
1887 default_sinfo.sinfo_context = asoc->default_context;
1888 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1889 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1890 sinfo = &default_sinfo;
1891 }
1892
1893 /* API 7.1.7, the sndbuf size per association bounds the
1894 * maximum size of data that can be sent in a single send call.
1895 */
1896 if (msg_len > sk->sk_sndbuf) {
1897 err = -EMSGSIZE;
1898 goto out_free;
1899 }
1900
1901 if (asoc->pmtu_pending)
1902 sctp_assoc_pending_pmtu(sk, asoc);
1903
1904 /* If fragmentation is disabled and the message length exceeds the
1905 * association fragmentation point, return EMSGSIZE. The I-D
1906 * does not specify what this error is, but this looks like
1907 * a great fit.
1908 */
1909 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1910 err = -EMSGSIZE;
1911 goto out_free;
1912 }
1913
1914 /* Check for invalid stream. */
1915 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1916 err = -EINVAL;
1917 goto out_free;
1918 }
1919
1920 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1921 if (!sctp_wspace(asoc)) {
1922 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1923 if (err)
1924 goto out_free;
1925 }
1926
1927 /* If an address is passed with the sendto/sendmsg call, it is used
1928 * to override the primary destination address in the TCP model, or
1929 * when SCTP_ADDR_OVER flag is set in the UDP model.
1930 */
1931 if ((sctp_style(sk, TCP) && msg_name) ||
1932 (sinfo_flags & SCTP_ADDR_OVER)) {
1933 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1934 if (!chunk_tp) {
1935 err = -EINVAL;
1936 goto out_free;
1937 }
1938 } else
1939 chunk_tp = NULL;
1940
1941 /* Auto-connect, if we aren't connected already. */
1942 if (sctp_state(asoc, CLOSED)) {
1943 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1944 if (err < 0)
1945 goto out_free;
1946 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1947 }
1948
1949 /* Break the message into multiple chunks of maximum size. */
1950 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1951 if (IS_ERR(datamsg)) {
1952 err = PTR_ERR(datamsg);
1953 goto out_free;
1954 }
1955
1956 /* Now send the (possibly) fragmented message. */
1957 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1958 sctp_chunk_hold(chunk);
1959
1960 /* Do accounting for the write space. */
1961 sctp_set_owner_w(chunk);
1962
1963 chunk->transport = chunk_tp;
1964 }
1965
1966 /* Send it to the lower layers. Note: all chunks
1967 * must either fail or succeed. The lower layer
1968 * works that way today. Keep it that way or this
1969 * breaks.
1970 */
1971 err = sctp_primitive_SEND(net, asoc, datamsg);
1972 /* Did the lower layer accept the chunk? */
1973 if (err)
1974 sctp_datamsg_free(datamsg);
1975 else
1976 sctp_datamsg_put(datamsg);
1977
1978 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1979
1980 if (err)
1981 goto out_free;
1982 else
1983 err = msg_len;
1984
1985 /* If we are already past ASSOCIATE, the lower
1986 * layers are responsible for association cleanup.
1987 */
1988 goto out_unlock;
1989
1990 out_free:
1991 if (new_asoc) {
1992 sctp_unhash_established(asoc);
1993 sctp_association_free(asoc);
1994 }
1995 out_unlock:
1996 sctp_release_sock(sk);
1997
1998 out_nounlock:
1999 return sctp_error(sk, msg_flags, err);
2000
2001 #if 0
2002 do_sock_err:
2003 if (msg_len)
2004 err = msg_len;
2005 else
2006 err = sock_error(sk);
2007 goto out;
2008
2009 do_interrupted:
2010 if (msg_len)
2011 err = msg_len;
2012 goto out;
2013 #endif /* 0 */
2014 }
2015
2016 /* This is an extended version of skb_pull() that removes the data from the
2017 * start of a skb even when data is spread across the list of skb's in the
2018 * frag_list. len specifies the total amount of data that needs to be removed.
2019 * when 'len' bytes could be removed from the skb, it returns 0.
2020 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2021 * could not be removed.
2022 */
2023 static int sctp_skb_pull(struct sk_buff *skb, int len)
2024 {
2025 struct sk_buff *list;
2026 int skb_len = skb_headlen(skb);
2027 int rlen;
2028
2029 if (len <= skb_len) {
2030 __skb_pull(skb, len);
2031 return 0;
2032 }
2033 len -= skb_len;
2034 __skb_pull(skb, skb_len);
2035
2036 skb_walk_frags(skb, list) {
2037 rlen = sctp_skb_pull(list, len);
2038 skb->len -= (len-rlen);
2039 skb->data_len -= (len-rlen);
2040
2041 if (!rlen)
2042 return 0;
2043
2044 len = rlen;
2045 }
2046
2047 return len;
2048 }
2049
2050 /* API 3.1.3 recvmsg() - UDP Style Syntax
2051 *
2052 * ssize_t recvmsg(int socket, struct msghdr *message,
2053 * int flags);
2054 *
2055 * socket - the socket descriptor of the endpoint.
2056 * message - pointer to the msghdr structure which contains a single
2057 * user message and possibly some ancillary data.
2058 *
2059 * See Section 5 for complete description of the data
2060 * structures.
2061 *
2062 * flags - flags sent or received with the user message, see Section
2063 * 5 for complete description of the flags.
2064 */
2065 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2066
2067 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2068 struct msghdr *msg, size_t len, int noblock,
2069 int flags, int *addr_len)
2070 {
2071 struct sctp_ulpevent *event = NULL;
2072 struct sctp_sock *sp = sctp_sk(sk);
2073 struct sk_buff *skb;
2074 int copied;
2075 int err = 0;
2076 int skb_len;
2077
2078 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
2079 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
2080 "len", len, "knoblauch", noblock,
2081 "flags", flags, "addr_len", addr_len);
2082
2083 sctp_lock_sock(sk);
2084
2085 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2086 err = -ENOTCONN;
2087 goto out;
2088 }
2089
2090 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2091 if (!skb)
2092 goto out;
2093
2094 /* Get the total length of the skb including any skb's in the
2095 * frag_list.
2096 */
2097 skb_len = skb->len;
2098
2099 copied = skb_len;
2100 if (copied > len)
2101 copied = len;
2102
2103 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2104
2105 event = sctp_skb2event(skb);
2106
2107 if (err)
2108 goto out_free;
2109
2110 sock_recv_ts_and_drops(msg, sk, skb);
2111 if (sctp_ulpevent_is_notification(event)) {
2112 msg->msg_flags |= MSG_NOTIFICATION;
2113 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2114 } else {
2115 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2116 }
2117
2118 /* Check if we allow SCTP_SNDRCVINFO. */
2119 if (sp->subscribe.sctp_data_io_event)
2120 sctp_ulpevent_read_sndrcvinfo(event, msg);
2121 #if 0
2122 /* FIXME: we should be calling IP/IPv6 layers. */
2123 if (sk->sk_protinfo.af_inet.cmsg_flags)
2124 ip_cmsg_recv(msg, skb);
2125 #endif
2126
2127 err = copied;
2128
2129 /* If skb's length exceeds the user's buffer, update the skb and
2130 * push it back to the receive_queue so that the next call to
2131 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2132 */
2133 if (skb_len > copied) {
2134 msg->msg_flags &= ~MSG_EOR;
2135 if (flags & MSG_PEEK)
2136 goto out_free;
2137 sctp_skb_pull(skb, copied);
2138 skb_queue_head(&sk->sk_receive_queue, skb);
2139
2140 /* When only partial message is copied to the user, increase
2141 * rwnd by that amount. If all the data in the skb is read,
2142 * rwnd is updated when the event is freed.
2143 */
2144 if (!sctp_ulpevent_is_notification(event))
2145 sctp_assoc_rwnd_increase(event->asoc, copied);
2146 goto out;
2147 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2148 (event->msg_flags & MSG_EOR))
2149 msg->msg_flags |= MSG_EOR;
2150 else
2151 msg->msg_flags &= ~MSG_EOR;
2152
2153 out_free:
2154 if (flags & MSG_PEEK) {
2155 /* Release the skb reference acquired after peeking the skb in
2156 * sctp_skb_recv_datagram().
2157 */
2158 kfree_skb(skb);
2159 } else {
2160 /* Free the event which includes releasing the reference to
2161 * the owner of the skb, freeing the skb and updating the
2162 * rwnd.
2163 */
2164 sctp_ulpevent_free(event);
2165 }
2166 out:
2167 sctp_release_sock(sk);
2168 return err;
2169 }
2170
2171 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2172 *
2173 * This option is a on/off flag. If enabled no SCTP message
2174 * fragmentation will be performed. Instead if a message being sent
2175 * exceeds the current PMTU size, the message will NOT be sent and
2176 * instead a error will be indicated to the user.
2177 */
2178 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2179 char __user *optval,
2180 unsigned int optlen)
2181 {
2182 int val;
2183
2184 if (optlen < sizeof(int))
2185 return -EINVAL;
2186
2187 if (get_user(val, (int __user *)optval))
2188 return -EFAULT;
2189
2190 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2191
2192 return 0;
2193 }
2194
2195 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2196 unsigned int optlen)
2197 {
2198 struct sctp_association *asoc;
2199 struct sctp_ulpevent *event;
2200
2201 if (optlen > sizeof(struct sctp_event_subscribe))
2202 return -EINVAL;
2203 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2204 return -EFAULT;
2205
2206 /*
2207 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2208 * if there is no data to be sent or retransmit, the stack will
2209 * immediately send up this notification.
2210 */
2211 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2212 &sctp_sk(sk)->subscribe)) {
2213 asoc = sctp_id2assoc(sk, 0);
2214
2215 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2216 event = sctp_ulpevent_make_sender_dry_event(asoc,
2217 GFP_ATOMIC);
2218 if (!event)
2219 return -ENOMEM;
2220
2221 sctp_ulpq_tail_event(&asoc->ulpq, event);
2222 }
2223 }
2224
2225 return 0;
2226 }
2227
2228 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2229 *
2230 * This socket option is applicable to the UDP-style socket only. When
2231 * set it will cause associations that are idle for more than the
2232 * specified number of seconds to automatically close. An association
2233 * being idle is defined an association that has NOT sent or received
2234 * user data. The special value of '0' indicates that no automatic
2235 * close of any associations should be performed. The option expects an
2236 * integer defining the number of seconds of idle time before an
2237 * association is closed.
2238 */
2239 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2240 unsigned int optlen)
2241 {
2242 struct sctp_sock *sp = sctp_sk(sk);
2243
2244 /* Applicable to UDP-style socket only */
2245 if (sctp_style(sk, TCP))
2246 return -EOPNOTSUPP;
2247 if (optlen != sizeof(int))
2248 return -EINVAL;
2249 if (copy_from_user(&sp->autoclose, optval, optlen))
2250 return -EFAULT;
2251
2252 return 0;
2253 }
2254
2255 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2256 *
2257 * Applications can enable or disable heartbeats for any peer address of
2258 * an association, modify an address's heartbeat interval, force a
2259 * heartbeat to be sent immediately, and adjust the address's maximum
2260 * number of retransmissions sent before an address is considered
2261 * unreachable. The following structure is used to access and modify an
2262 * address's parameters:
2263 *
2264 * struct sctp_paddrparams {
2265 * sctp_assoc_t spp_assoc_id;
2266 * struct sockaddr_storage spp_address;
2267 * uint32_t spp_hbinterval;
2268 * uint16_t spp_pathmaxrxt;
2269 * uint32_t spp_pathmtu;
2270 * uint32_t spp_sackdelay;
2271 * uint32_t spp_flags;
2272 * };
2273 *
2274 * spp_assoc_id - (one-to-many style socket) This is filled in the
2275 * application, and identifies the association for
2276 * this query.
2277 * spp_address - This specifies which address is of interest.
2278 * spp_hbinterval - This contains the value of the heartbeat interval,
2279 * in milliseconds. If a value of zero
2280 * is present in this field then no changes are to
2281 * be made to this parameter.
2282 * spp_pathmaxrxt - This contains the maximum number of
2283 * retransmissions before this address shall be
2284 * considered unreachable. If a value of zero
2285 * is present in this field then no changes are to
2286 * be made to this parameter.
2287 * spp_pathmtu - When Path MTU discovery is disabled the value
2288 * specified here will be the "fixed" path mtu.
2289 * Note that if the spp_address field is empty
2290 * then all associations on this address will
2291 * have this fixed path mtu set upon them.
2292 *
2293 * spp_sackdelay - When delayed sack is enabled, this value specifies
2294 * the number of milliseconds that sacks will be delayed
2295 * for. This value will apply to all addresses of an
2296 * association if the spp_address field is empty. Note
2297 * also, that if delayed sack is enabled and this
2298 * value is set to 0, no change is made to the last
2299 * recorded delayed sack timer value.
2300 *
2301 * spp_flags - These flags are used to control various features
2302 * on an association. The flag field may contain
2303 * zero or more of the following options.
2304 *
2305 * SPP_HB_ENABLE - Enable heartbeats on the
2306 * specified address. Note that if the address
2307 * field is empty all addresses for the association
2308 * have heartbeats enabled upon them.
2309 *
2310 * SPP_HB_DISABLE - Disable heartbeats on the
2311 * speicifed address. Note that if the address
2312 * field is empty all addresses for the association
2313 * will have their heartbeats disabled. Note also
2314 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2315 * mutually exclusive, only one of these two should
2316 * be specified. Enabling both fields will have
2317 * undetermined results.
2318 *
2319 * SPP_HB_DEMAND - Request a user initiated heartbeat
2320 * to be made immediately.
2321 *
2322 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2323 * heartbeat delayis to be set to the value of 0
2324 * milliseconds.
2325 *
2326 * SPP_PMTUD_ENABLE - This field will enable PMTU
2327 * discovery upon the specified address. Note that
2328 * if the address feild is empty then all addresses
2329 * on the association are effected.
2330 *
2331 * SPP_PMTUD_DISABLE - This field will disable PMTU
2332 * discovery upon the specified address. Note that
2333 * if the address feild is empty then all addresses
2334 * on the association are effected. Not also that
2335 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2336 * exclusive. Enabling both will have undetermined
2337 * results.
2338 *
2339 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2340 * on delayed sack. The time specified in spp_sackdelay
2341 * is used to specify the sack delay for this address. Note
2342 * that if spp_address is empty then all addresses will
2343 * enable delayed sack and take on the sack delay
2344 * value specified in spp_sackdelay.
2345 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2346 * off delayed sack. If the spp_address field is blank then
2347 * delayed sack is disabled for the entire association. Note
2348 * also that this field is mutually exclusive to
2349 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2350 * results.
2351 */
2352 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2353 struct sctp_transport *trans,
2354 struct sctp_association *asoc,
2355 struct sctp_sock *sp,
2356 int hb_change,
2357 int pmtud_change,
2358 int sackdelay_change)
2359 {
2360 int error;
2361
2362 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2363 struct net *net = sock_net(trans->asoc->base.sk);
2364
2365 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2366 if (error)
2367 return error;
2368 }
2369
2370 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2371 * this field is ignored. Note also that a value of zero indicates
2372 * the current setting should be left unchanged.
2373 */
2374 if (params->spp_flags & SPP_HB_ENABLE) {
2375
2376 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2377 * set. This lets us use 0 value when this flag
2378 * is set.
2379 */
2380 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2381 params->spp_hbinterval = 0;
2382
2383 if (params->spp_hbinterval ||
2384 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2385 if (trans) {
2386 trans->hbinterval =
2387 msecs_to_jiffies(params->spp_hbinterval);
2388 } else if (asoc) {
2389 asoc->hbinterval =
2390 msecs_to_jiffies(params->spp_hbinterval);
2391 } else {
2392 sp->hbinterval = params->spp_hbinterval;
2393 }
2394 }
2395 }
2396
2397 if (hb_change) {
2398 if (trans) {
2399 trans->param_flags =
2400 (trans->param_flags & ~SPP_HB) | hb_change;
2401 } else if (asoc) {
2402 asoc->param_flags =
2403 (asoc->param_flags & ~SPP_HB) | hb_change;
2404 } else {
2405 sp->param_flags =
2406 (sp->param_flags & ~SPP_HB) | hb_change;
2407 }
2408 }
2409
2410 /* When Path MTU discovery is disabled the value specified here will
2411 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2412 * include the flag SPP_PMTUD_DISABLE for this field to have any
2413 * effect).
2414 */
2415 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2416 if (trans) {
2417 trans->pathmtu = params->spp_pathmtu;
2418 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2419 } else if (asoc) {
2420 asoc->pathmtu = params->spp_pathmtu;
2421 sctp_frag_point(asoc, params->spp_pathmtu);
2422 } else {
2423 sp->pathmtu = params->spp_pathmtu;
2424 }
2425 }
2426
2427 if (pmtud_change) {
2428 if (trans) {
2429 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2430 (params->spp_flags & SPP_PMTUD_ENABLE);
2431 trans->param_flags =
2432 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2433 if (update) {
2434 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2435 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2436 }
2437 } else if (asoc) {
2438 asoc->param_flags =
2439 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2440 } else {
2441 sp->param_flags =
2442 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2443 }
2444 }
2445
2446 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2447 * value of this field is ignored. Note also that a value of zero
2448 * indicates the current setting should be left unchanged.
2449 */
2450 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2451 if (trans) {
2452 trans->sackdelay =
2453 msecs_to_jiffies(params->spp_sackdelay);
2454 } else if (asoc) {
2455 asoc->sackdelay =
2456 msecs_to_jiffies(params->spp_sackdelay);
2457 } else {
2458 sp->sackdelay = params->spp_sackdelay;
2459 }
2460 }
2461
2462 if (sackdelay_change) {
2463 if (trans) {
2464 trans->param_flags =
2465 (trans->param_flags & ~SPP_SACKDELAY) |
2466 sackdelay_change;
2467 } else if (asoc) {
2468 asoc->param_flags =
2469 (asoc->param_flags & ~SPP_SACKDELAY) |
2470 sackdelay_change;
2471 } else {
2472 sp->param_flags =
2473 (sp->param_flags & ~SPP_SACKDELAY) |
2474 sackdelay_change;
2475 }
2476 }
2477
2478 /* Note that a value of zero indicates the current setting should be
2479 left unchanged.
2480 */
2481 if (params->spp_pathmaxrxt) {
2482 if (trans) {
2483 trans->pathmaxrxt = params->spp_pathmaxrxt;
2484 } else if (asoc) {
2485 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2486 } else {
2487 sp->pathmaxrxt = params->spp_pathmaxrxt;
2488 }
2489 }
2490
2491 return 0;
2492 }
2493
2494 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2495 char __user *optval,
2496 unsigned int optlen)
2497 {
2498 struct sctp_paddrparams params;
2499 struct sctp_transport *trans = NULL;
2500 struct sctp_association *asoc = NULL;
2501 struct sctp_sock *sp = sctp_sk(sk);
2502 int error;
2503 int hb_change, pmtud_change, sackdelay_change;
2504
2505 if (optlen != sizeof(struct sctp_paddrparams))
2506 return - EINVAL;
2507
2508 if (copy_from_user(&params, optval, optlen))
2509 return -EFAULT;
2510
2511 /* Validate flags and value parameters. */
2512 hb_change = params.spp_flags & SPP_HB;
2513 pmtud_change = params.spp_flags & SPP_PMTUD;
2514 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2515
2516 if (hb_change == SPP_HB ||
2517 pmtud_change == SPP_PMTUD ||
2518 sackdelay_change == SPP_SACKDELAY ||
2519 params.spp_sackdelay > 500 ||
2520 (params.spp_pathmtu &&
2521 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2522 return -EINVAL;
2523
2524 /* If an address other than INADDR_ANY is specified, and
2525 * no transport is found, then the request is invalid.
2526 */
2527 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
2528 trans = sctp_addr_id2transport(sk, &params.spp_address,
2529 params.spp_assoc_id);
2530 if (!trans)
2531 return -EINVAL;
2532 }
2533
2534 /* Get association, if assoc_id != 0 and the socket is a one
2535 * to many style socket, and an association was not found, then
2536 * the id was invalid.
2537 */
2538 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2539 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2540 return -EINVAL;
2541
2542 /* Heartbeat demand can only be sent on a transport or
2543 * association, but not a socket.
2544 */
2545 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2546 return -EINVAL;
2547
2548 /* Process parameters. */
2549 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2550 hb_change, pmtud_change,
2551 sackdelay_change);
2552
2553 if (error)
2554 return error;
2555
2556 /* If changes are for association, also apply parameters to each
2557 * transport.
2558 */
2559 if (!trans && asoc) {
2560 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2561 transports) {
2562 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2563 hb_change, pmtud_change,
2564 sackdelay_change);
2565 }
2566 }
2567
2568 return 0;
2569 }
2570
2571 /*
2572 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2573 *
2574 * This option will effect the way delayed acks are performed. This
2575 * option allows you to get or set the delayed ack time, in
2576 * milliseconds. It also allows changing the delayed ack frequency.
2577 * Changing the frequency to 1 disables the delayed sack algorithm. If
2578 * the assoc_id is 0, then this sets or gets the endpoints default
2579 * values. If the assoc_id field is non-zero, then the set or get
2580 * effects the specified association for the one to many model (the
2581 * assoc_id field is ignored by the one to one model). Note that if
2582 * sack_delay or sack_freq are 0 when setting this option, then the
2583 * current values will remain unchanged.
2584 *
2585 * struct sctp_sack_info {
2586 * sctp_assoc_t sack_assoc_id;
2587 * uint32_t sack_delay;
2588 * uint32_t sack_freq;
2589 * };
2590 *
2591 * sack_assoc_id - This parameter, indicates which association the user
2592 * is performing an action upon. Note that if this field's value is
2593 * zero then the endpoints default value is changed (effecting future
2594 * associations only).
2595 *
2596 * sack_delay - This parameter contains the number of milliseconds that
2597 * the user is requesting the delayed ACK timer be set to. Note that
2598 * this value is defined in the standard to be between 200 and 500
2599 * milliseconds.
2600 *
2601 * sack_freq - This parameter contains the number of packets that must
2602 * be received before a sack is sent without waiting for the delay
2603 * timer to expire. The default value for this is 2, setting this
2604 * value to 1 will disable the delayed sack algorithm.
2605 */
2606
2607 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2608 char __user *optval, unsigned int optlen)
2609 {
2610 struct sctp_sack_info params;
2611 struct sctp_transport *trans = NULL;
2612 struct sctp_association *asoc = NULL;
2613 struct sctp_sock *sp = sctp_sk(sk);
2614
2615 if (optlen == sizeof(struct sctp_sack_info)) {
2616 if (copy_from_user(&params, optval, optlen))
2617 return -EFAULT;
2618
2619 if (params.sack_delay == 0 && params.sack_freq == 0)
2620 return 0;
2621 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2622 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2623 pr_warn("Use struct sctp_sack_info instead\n");
2624 if (copy_from_user(&params, optval, optlen))
2625 return -EFAULT;
2626
2627 if (params.sack_delay == 0)
2628 params.sack_freq = 1;
2629 else
2630 params.sack_freq = 0;
2631 } else
2632 return - EINVAL;
2633
2634 /* Validate value parameter. */
2635 if (params.sack_delay > 500)
2636 return -EINVAL;
2637
2638 /* Get association, if sack_assoc_id != 0 and the socket is a one
2639 * to many style socket, and an association was not found, then
2640 * the id was invalid.
2641 */
2642 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2643 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2644 return -EINVAL;
2645
2646 if (params.sack_delay) {
2647 if (asoc) {
2648 asoc->sackdelay =
2649 msecs_to_jiffies(params.sack_delay);
2650 asoc->param_flags =
2651 (asoc->param_flags & ~SPP_SACKDELAY) |
2652 SPP_SACKDELAY_ENABLE;
2653 } else {
2654 sp->sackdelay = params.sack_delay;
2655 sp->param_flags =
2656 (sp->param_flags & ~SPP_SACKDELAY) |
2657 SPP_SACKDELAY_ENABLE;
2658 }
2659 }
2660
2661 if (params.sack_freq == 1) {
2662 if (asoc) {
2663 asoc->param_flags =
2664 (asoc->param_flags & ~SPP_SACKDELAY) |
2665 SPP_SACKDELAY_DISABLE;
2666 } else {
2667 sp->param_flags =
2668 (sp->param_flags & ~SPP_SACKDELAY) |
2669 SPP_SACKDELAY_DISABLE;
2670 }
2671 } else if (params.sack_freq > 1) {
2672 if (asoc) {
2673 asoc->sackfreq = params.sack_freq;
2674 asoc->param_flags =
2675 (asoc->param_flags & ~SPP_SACKDELAY) |
2676 SPP_SACKDELAY_ENABLE;
2677 } else {
2678 sp->sackfreq = params.sack_freq;
2679 sp->param_flags =
2680 (sp->param_flags & ~SPP_SACKDELAY) |
2681 SPP_SACKDELAY_ENABLE;
2682 }
2683 }
2684
2685 /* If change is for association, also apply to each transport. */
2686 if (asoc) {
2687 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2688 transports) {
2689 if (params.sack_delay) {
2690 trans->sackdelay =
2691 msecs_to_jiffies(params.sack_delay);
2692 trans->param_flags =
2693 (trans->param_flags & ~SPP_SACKDELAY) |
2694 SPP_SACKDELAY_ENABLE;
2695 }
2696 if (params.sack_freq == 1) {
2697 trans->param_flags =
2698 (trans->param_flags & ~SPP_SACKDELAY) |
2699 SPP_SACKDELAY_DISABLE;
2700 } else if (params.sack_freq > 1) {
2701 trans->sackfreq = params.sack_freq;
2702 trans->param_flags =
2703 (trans->param_flags & ~SPP_SACKDELAY) |
2704 SPP_SACKDELAY_ENABLE;
2705 }
2706 }
2707 }
2708
2709 return 0;
2710 }
2711
2712 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2713 *
2714 * Applications can specify protocol parameters for the default association
2715 * initialization. The option name argument to setsockopt() and getsockopt()
2716 * is SCTP_INITMSG.
2717 *
2718 * Setting initialization parameters is effective only on an unconnected
2719 * socket (for UDP-style sockets only future associations are effected
2720 * by the change). With TCP-style sockets, this option is inherited by
2721 * sockets derived from a listener socket.
2722 */
2723 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2724 {
2725 struct sctp_initmsg sinit;
2726 struct sctp_sock *sp = sctp_sk(sk);
2727
2728 if (optlen != sizeof(struct sctp_initmsg))
2729 return -EINVAL;
2730 if (copy_from_user(&sinit, optval, optlen))
2731 return -EFAULT;
2732
2733 if (sinit.sinit_num_ostreams)
2734 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2735 if (sinit.sinit_max_instreams)
2736 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2737 if (sinit.sinit_max_attempts)
2738 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2739 if (sinit.sinit_max_init_timeo)
2740 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2741
2742 return 0;
2743 }
2744
2745 /*
2746 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2747 *
2748 * Applications that wish to use the sendto() system call may wish to
2749 * specify a default set of parameters that would normally be supplied
2750 * through the inclusion of ancillary data. This socket option allows
2751 * such an application to set the default sctp_sndrcvinfo structure.
2752 * The application that wishes to use this socket option simply passes
2753 * in to this call the sctp_sndrcvinfo structure defined in Section
2754 * 5.2.2) The input parameters accepted by this call include
2755 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2756 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2757 * to this call if the caller is using the UDP model.
2758 */
2759 static int sctp_setsockopt_default_send_param(struct sock *sk,
2760 char __user *optval,
2761 unsigned int optlen)
2762 {
2763 struct sctp_sndrcvinfo info;
2764 struct sctp_association *asoc;
2765 struct sctp_sock *sp = sctp_sk(sk);
2766
2767 if (optlen != sizeof(struct sctp_sndrcvinfo))
2768 return -EINVAL;
2769 if (copy_from_user(&info, optval, optlen))
2770 return -EFAULT;
2771
2772 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2773 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2774 return -EINVAL;
2775
2776 if (asoc) {
2777 asoc->default_stream = info.sinfo_stream;
2778 asoc->default_flags = info.sinfo_flags;
2779 asoc->default_ppid = info.sinfo_ppid;
2780 asoc->default_context = info.sinfo_context;
2781 asoc->default_timetolive = info.sinfo_timetolive;
2782 } else {
2783 sp->default_stream = info.sinfo_stream;
2784 sp->default_flags = info.sinfo_flags;
2785 sp->default_ppid = info.sinfo_ppid;
2786 sp->default_context = info.sinfo_context;
2787 sp->default_timetolive = info.sinfo_timetolive;
2788 }
2789
2790 return 0;
2791 }
2792
2793 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2794 *
2795 * Requests that the local SCTP stack use the enclosed peer address as
2796 * the association primary. The enclosed address must be one of the
2797 * association peer's addresses.
2798 */
2799 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2800 unsigned int optlen)
2801 {
2802 struct sctp_prim prim;
2803 struct sctp_transport *trans;
2804
2805 if (optlen != sizeof(struct sctp_prim))
2806 return -EINVAL;
2807
2808 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2809 return -EFAULT;
2810
2811 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2812 if (!trans)
2813 return -EINVAL;
2814
2815 sctp_assoc_set_primary(trans->asoc, trans);
2816
2817 return 0;
2818 }
2819
2820 /*
2821 * 7.1.5 SCTP_NODELAY
2822 *
2823 * Turn on/off any Nagle-like algorithm. This means that packets are
2824 * generally sent as soon as possible and no unnecessary delays are
2825 * introduced, at the cost of more packets in the network. Expects an
2826 * integer boolean flag.
2827 */
2828 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2829 unsigned int optlen)
2830 {
2831 int val;
2832
2833 if (optlen < sizeof(int))
2834 return -EINVAL;
2835 if (get_user(val, (int __user *)optval))
2836 return -EFAULT;
2837
2838 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2839 return 0;
2840 }
2841
2842 /*
2843 *
2844 * 7.1.1 SCTP_RTOINFO
2845 *
2846 * The protocol parameters used to initialize and bound retransmission
2847 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2848 * and modify these parameters.
2849 * All parameters are time values, in milliseconds. A value of 0, when
2850 * modifying the parameters, indicates that the current value should not
2851 * be changed.
2852 *
2853 */
2854 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2855 {
2856 struct sctp_rtoinfo rtoinfo;
2857 struct sctp_association *asoc;
2858
2859 if (optlen != sizeof (struct sctp_rtoinfo))
2860 return -EINVAL;
2861
2862 if (copy_from_user(&rtoinfo, optval, optlen))
2863 return -EFAULT;
2864
2865 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2866
2867 /* Set the values to the specific association */
2868 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2869 return -EINVAL;
2870
2871 if (asoc) {
2872 if (rtoinfo.srto_initial != 0)
2873 asoc->rto_initial =
2874 msecs_to_jiffies(rtoinfo.srto_initial);
2875 if (rtoinfo.srto_max != 0)
2876 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2877 if (rtoinfo.srto_min != 0)
2878 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2879 } else {
2880 /* If there is no association or the association-id = 0
2881 * set the values to the endpoint.
2882 */
2883 struct sctp_sock *sp = sctp_sk(sk);
2884
2885 if (rtoinfo.srto_initial != 0)
2886 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2887 if (rtoinfo.srto_max != 0)
2888 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2889 if (rtoinfo.srto_min != 0)
2890 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2891 }
2892
2893 return 0;
2894 }
2895
2896 /*
2897 *
2898 * 7.1.2 SCTP_ASSOCINFO
2899 *
2900 * This option is used to tune the maximum retransmission attempts
2901 * of the association.
2902 * Returns an error if the new association retransmission value is
2903 * greater than the sum of the retransmission value of the peer.
2904 * See [SCTP] for more information.
2905 *
2906 */
2907 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2908 {
2909
2910 struct sctp_assocparams assocparams;
2911 struct sctp_association *asoc;
2912
2913 if (optlen != sizeof(struct sctp_assocparams))
2914 return -EINVAL;
2915 if (copy_from_user(&assocparams, optval, optlen))
2916 return -EFAULT;
2917
2918 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2919
2920 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2921 return -EINVAL;
2922
2923 /* Set the values to the specific association */
2924 if (asoc) {
2925 if (assocparams.sasoc_asocmaxrxt != 0) {
2926 __u32 path_sum = 0;
2927 int paths = 0;
2928 struct sctp_transport *peer_addr;
2929
2930 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2931 transports) {
2932 path_sum += peer_addr->pathmaxrxt;
2933 paths++;
2934 }
2935
2936 /* Only validate asocmaxrxt if we have more than
2937 * one path/transport. We do this because path
2938 * retransmissions are only counted when we have more
2939 * then one path.
2940 */
2941 if (paths > 1 &&
2942 assocparams.sasoc_asocmaxrxt > path_sum)
2943 return -EINVAL;
2944
2945 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2946 }
2947
2948 if (assocparams.sasoc_cookie_life != 0) {
2949 asoc->cookie_life.tv_sec =
2950 assocparams.sasoc_cookie_life / 1000;
2951 asoc->cookie_life.tv_usec =
2952 (assocparams.sasoc_cookie_life % 1000)
2953 * 1000;
2954 }
2955 } else {
2956 /* Set the values to the endpoint */
2957 struct sctp_sock *sp = sctp_sk(sk);
2958
2959 if (assocparams.sasoc_asocmaxrxt != 0)
2960 sp->assocparams.sasoc_asocmaxrxt =
2961 assocparams.sasoc_asocmaxrxt;
2962 if (assocparams.sasoc_cookie_life != 0)
2963 sp->assocparams.sasoc_cookie_life =
2964 assocparams.sasoc_cookie_life;
2965 }
2966 return 0;
2967 }
2968
2969 /*
2970 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2971 *
2972 * This socket option is a boolean flag which turns on or off mapped V4
2973 * addresses. If this option is turned on and the socket is type
2974 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2975 * If this option is turned off, then no mapping will be done of V4
2976 * addresses and a user will receive both PF_INET6 and PF_INET type
2977 * addresses on the socket.
2978 */
2979 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2980 {
2981 int val;
2982 struct sctp_sock *sp = sctp_sk(sk);
2983
2984 if (optlen < sizeof(int))
2985 return -EINVAL;
2986 if (get_user(val, (int __user *)optval))
2987 return -EFAULT;
2988 if (val)
2989 sp->v4mapped = 1;
2990 else
2991 sp->v4mapped = 0;
2992
2993 return 0;
2994 }
2995
2996 /*
2997 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2998 * This option will get or set the maximum size to put in any outgoing
2999 * SCTP DATA chunk. If a message is larger than this size it will be
3000 * fragmented by SCTP into the specified size. Note that the underlying
3001 * SCTP implementation may fragment into smaller sized chunks when the
3002 * PMTU of the underlying association is smaller than the value set by
3003 * the user. The default value for this option is '0' which indicates
3004 * the user is NOT limiting fragmentation and only the PMTU will effect
3005 * SCTP's choice of DATA chunk size. Note also that values set larger
3006 * than the maximum size of an IP datagram will effectively let SCTP
3007 * control fragmentation (i.e. the same as setting this option to 0).
3008 *
3009 * The following structure is used to access and modify this parameter:
3010 *
3011 * struct sctp_assoc_value {
3012 * sctp_assoc_t assoc_id;
3013 * uint32_t assoc_value;
3014 * };
3015 *
3016 * assoc_id: This parameter is ignored for one-to-one style sockets.
3017 * For one-to-many style sockets this parameter indicates which
3018 * association the user is performing an action upon. Note that if
3019 * this field's value is zero then the endpoints default value is
3020 * changed (effecting future associations only).
3021 * assoc_value: This parameter specifies the maximum size in bytes.
3022 */
3023 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3024 {
3025 struct sctp_assoc_value params;
3026 struct sctp_association *asoc;
3027 struct sctp_sock *sp = sctp_sk(sk);
3028 int val;
3029
3030 if (optlen == sizeof(int)) {
3031 pr_warn("Use of int in maxseg socket option deprecated\n");
3032 pr_warn("Use struct sctp_assoc_value instead\n");
3033 if (copy_from_user(&val, optval, optlen))
3034 return -EFAULT;
3035 params.assoc_id = 0;
3036 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3037 if (copy_from_user(&params, optval, optlen))
3038 return -EFAULT;
3039 val = params.assoc_value;
3040 } else
3041 return -EINVAL;
3042
3043 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3044 return -EINVAL;
3045
3046 asoc = sctp_id2assoc(sk, params.assoc_id);
3047 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3048 return -EINVAL;
3049
3050 if (asoc) {
3051 if (val == 0) {
3052 val = asoc->pathmtu;
3053 val -= sp->pf->af->net_header_len;
3054 val -= sizeof(struct sctphdr) +
3055 sizeof(struct sctp_data_chunk);
3056 }
3057 asoc->user_frag = val;
3058 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3059 } else {
3060 sp->user_frag = val;
3061 }
3062
3063 return 0;
3064 }
3065
3066
3067 /*
3068 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3069 *
3070 * Requests that the peer mark the enclosed address as the association
3071 * primary. The enclosed address must be one of the association's
3072 * locally bound addresses. The following structure is used to make a
3073 * set primary request:
3074 */
3075 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3076 unsigned int optlen)
3077 {
3078 struct net *net = sock_net(sk);
3079 struct sctp_sock *sp;
3080 struct sctp_association *asoc = NULL;
3081 struct sctp_setpeerprim prim;
3082 struct sctp_chunk *chunk;
3083 struct sctp_af *af;
3084 int err;
3085
3086 sp = sctp_sk(sk);
3087
3088 if (!net->sctp.addip_enable)
3089 return -EPERM;
3090
3091 if (optlen != sizeof(struct sctp_setpeerprim))
3092 return -EINVAL;
3093
3094 if (copy_from_user(&prim, optval, optlen))
3095 return -EFAULT;
3096
3097 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3098 if (!asoc)
3099 return -EINVAL;
3100
3101 if (!asoc->peer.asconf_capable)
3102 return -EPERM;
3103
3104 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3105 return -EPERM;
3106
3107 if (!sctp_state(asoc, ESTABLISHED))
3108 return -ENOTCONN;
3109
3110 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3111 if (!af)
3112 return -EINVAL;
3113
3114 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3115 return -EADDRNOTAVAIL;
3116
3117 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3118 return -EADDRNOTAVAIL;
3119
3120 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3121 chunk = sctp_make_asconf_set_prim(asoc,
3122 (union sctp_addr *)&prim.sspp_addr);
3123 if (!chunk)
3124 return -ENOMEM;
3125
3126 err = sctp_send_asconf(asoc, chunk);
3127
3128 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
3129
3130 return err;
3131 }
3132
3133 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3134 unsigned int optlen)
3135 {
3136 struct sctp_setadaptation adaptation;
3137
3138 if (optlen != sizeof(struct sctp_setadaptation))
3139 return -EINVAL;
3140 if (copy_from_user(&adaptation, optval, optlen))
3141 return -EFAULT;
3142
3143 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3144
3145 return 0;
3146 }
3147
3148 /*
3149 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3150 *
3151 * The context field in the sctp_sndrcvinfo structure is normally only
3152 * used when a failed message is retrieved holding the value that was
3153 * sent down on the actual send call. This option allows the setting of
3154 * a default context on an association basis that will be received on
3155 * reading messages from the peer. This is especially helpful in the
3156 * one-2-many model for an application to keep some reference to an
3157 * internal state machine that is processing messages on the
3158 * association. Note that the setting of this value only effects
3159 * received messages from the peer and does not effect the value that is
3160 * saved with outbound messages.
3161 */
3162 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3163 unsigned int optlen)
3164 {
3165 struct sctp_assoc_value params;
3166 struct sctp_sock *sp;
3167 struct sctp_association *asoc;
3168
3169 if (optlen != sizeof(struct sctp_assoc_value))
3170 return -EINVAL;
3171 if (copy_from_user(&params, optval, optlen))
3172 return -EFAULT;
3173
3174 sp = sctp_sk(sk);
3175
3176 if (params.assoc_id != 0) {
3177 asoc = sctp_id2assoc(sk, params.assoc_id);
3178 if (!asoc)
3179 return -EINVAL;
3180 asoc->default_rcv_context = params.assoc_value;
3181 } else {
3182 sp->default_rcv_context = params.assoc_value;
3183 }
3184
3185 return 0;
3186 }
3187
3188 /*
3189 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3190 *
3191 * This options will at a minimum specify if the implementation is doing
3192 * fragmented interleave. Fragmented interleave, for a one to many
3193 * socket, is when subsequent calls to receive a message may return
3194 * parts of messages from different associations. Some implementations
3195 * may allow you to turn this value on or off. If so, when turned off,
3196 * no fragment interleave will occur (which will cause a head of line
3197 * blocking amongst multiple associations sharing the same one to many
3198 * socket). When this option is turned on, then each receive call may
3199 * come from a different association (thus the user must receive data
3200 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3201 * association each receive belongs to.
3202 *
3203 * This option takes a boolean value. A non-zero value indicates that
3204 * fragmented interleave is on. A value of zero indicates that
3205 * fragmented interleave is off.
3206 *
3207 * Note that it is important that an implementation that allows this
3208 * option to be turned on, have it off by default. Otherwise an unaware
3209 * application using the one to many model may become confused and act
3210 * incorrectly.
3211 */
3212 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3213 char __user *optval,
3214 unsigned int optlen)
3215 {
3216 int val;
3217
3218 if (optlen != sizeof(int))
3219 return -EINVAL;
3220 if (get_user(val, (int __user *)optval))
3221 return -EFAULT;
3222
3223 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3224
3225 return 0;
3226 }
3227
3228 /*
3229 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3230 * (SCTP_PARTIAL_DELIVERY_POINT)
3231 *
3232 * This option will set or get the SCTP partial delivery point. This
3233 * point is the size of a message where the partial delivery API will be
3234 * invoked to help free up rwnd space for the peer. Setting this to a
3235 * lower value will cause partial deliveries to happen more often. The
3236 * calls argument is an integer that sets or gets the partial delivery
3237 * point. Note also that the call will fail if the user attempts to set
3238 * this value larger than the socket receive buffer size.
3239 *
3240 * Note that any single message having a length smaller than or equal to
3241 * the SCTP partial delivery point will be delivered in one single read
3242 * call as long as the user provided buffer is large enough to hold the
3243 * message.
3244 */
3245 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3246 char __user *optval,
3247 unsigned int optlen)
3248 {
3249 u32 val;
3250
3251 if (optlen != sizeof(u32))
3252 return -EINVAL;
3253 if (get_user(val, (int __user *)optval))
3254 return -EFAULT;
3255
3256 /* Note: We double the receive buffer from what the user sets
3257 * it to be, also initial rwnd is based on rcvbuf/2.
3258 */
3259 if (val > (sk->sk_rcvbuf >> 1))
3260 return -EINVAL;
3261
3262 sctp_sk(sk)->pd_point = val;
3263
3264 return 0; /* is this the right error code? */
3265 }
3266
3267 /*
3268 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3269 *
3270 * This option will allow a user to change the maximum burst of packets
3271 * that can be emitted by this association. Note that the default value
3272 * is 4, and some implementations may restrict this setting so that it
3273 * can only be lowered.
3274 *
3275 * NOTE: This text doesn't seem right. Do this on a socket basis with
3276 * future associations inheriting the socket value.
3277 */
3278 static int sctp_setsockopt_maxburst(struct sock *sk,
3279 char __user *optval,
3280 unsigned int optlen)
3281 {
3282 struct sctp_assoc_value params;
3283 struct sctp_sock *sp;
3284 struct sctp_association *asoc;
3285 int val;
3286 int assoc_id = 0;
3287
3288 if (optlen == sizeof(int)) {
3289 pr_warn("Use of int in max_burst socket option deprecated\n");
3290 pr_warn("Use struct sctp_assoc_value instead\n");
3291 if (copy_from_user(&val, optval, optlen))
3292 return -EFAULT;
3293 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3294 if (copy_from_user(&params, optval, optlen))
3295 return -EFAULT;
3296 val = params.assoc_value;
3297 assoc_id = params.assoc_id;
3298 } else
3299 return -EINVAL;
3300
3301 sp = sctp_sk(sk);
3302
3303 if (assoc_id != 0) {
3304 asoc = sctp_id2assoc(sk, assoc_id);
3305 if (!asoc)
3306 return -EINVAL;
3307 asoc->max_burst = val;
3308 } else
3309 sp->max_burst = val;
3310
3311 return 0;
3312 }
3313
3314 /*
3315 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3316 *
3317 * This set option adds a chunk type that the user is requesting to be
3318 * received only in an authenticated way. Changes to the list of chunks
3319 * will only effect future associations on the socket.
3320 */
3321 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3322 char __user *optval,
3323 unsigned int optlen)
3324 {
3325 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3326 struct sctp_authchunk val;
3327
3328 if (!ep->auth_enable)
3329 return -EACCES;
3330
3331 if (optlen != sizeof(struct sctp_authchunk))
3332 return -EINVAL;
3333 if (copy_from_user(&val, optval, optlen))
3334 return -EFAULT;
3335
3336 switch (val.sauth_chunk) {
3337 case SCTP_CID_INIT:
3338 case SCTP_CID_INIT_ACK:
3339 case SCTP_CID_SHUTDOWN_COMPLETE:
3340 case SCTP_CID_AUTH:
3341 return -EINVAL;
3342 }
3343
3344 /* add this chunk id to the endpoint */
3345 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3346 }
3347
3348 /*
3349 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3350 *
3351 * This option gets or sets the list of HMAC algorithms that the local
3352 * endpoint requires the peer to use.
3353 */
3354 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3355 char __user *optval,
3356 unsigned int optlen)
3357 {
3358 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3359 struct sctp_hmacalgo *hmacs;
3360 u32 idents;
3361 int err;
3362
3363 if (!ep->auth_enable)
3364 return -EACCES;
3365
3366 if (optlen < sizeof(struct sctp_hmacalgo))
3367 return -EINVAL;
3368
3369 hmacs= memdup_user(optval, optlen);
3370 if (IS_ERR(hmacs))
3371 return PTR_ERR(hmacs);
3372
3373 idents = hmacs->shmac_num_idents;
3374 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3375 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3376 err = -EINVAL;
3377 goto out;
3378 }
3379
3380 err = sctp_auth_ep_set_hmacs(ep, hmacs);
3381 out:
3382 kfree(hmacs);
3383 return err;
3384 }
3385
3386 /*
3387 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3388 *
3389 * This option will set a shared secret key which is used to build an
3390 * association shared key.
3391 */
3392 static int sctp_setsockopt_auth_key(struct sock *sk,
3393 char __user *optval,
3394 unsigned int optlen)
3395 {
3396 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3397 struct sctp_authkey *authkey;
3398 struct sctp_association *asoc;
3399 int ret;
3400
3401 if (!ep->auth_enable)
3402 return -EACCES;
3403
3404 if (optlen <= sizeof(struct sctp_authkey))
3405 return -EINVAL;
3406
3407 authkey= memdup_user(optval, optlen);
3408 if (IS_ERR(authkey))
3409 return PTR_ERR(authkey);
3410
3411 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3412 ret = -EINVAL;
3413 goto out;
3414 }
3415
3416 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3417 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3418 ret = -EINVAL;
3419 goto out;
3420 }
3421
3422 ret = sctp_auth_set_key(ep, asoc, authkey);
3423 out:
3424 kzfree(authkey);
3425 return ret;
3426 }
3427
3428 /*
3429 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3430 *
3431 * This option will get or set the active shared key to be used to build
3432 * the association shared key.
3433 */
3434 static int sctp_setsockopt_active_key(struct sock *sk,
3435 char __user *optval,
3436 unsigned int optlen)
3437 {
3438 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3439 struct sctp_authkeyid val;
3440 struct sctp_association *asoc;
3441
3442 if (!ep->auth_enable)
3443 return -EACCES;
3444
3445 if (optlen != sizeof(struct sctp_authkeyid))
3446 return -EINVAL;
3447 if (copy_from_user(&val, optval, optlen))
3448 return -EFAULT;
3449
3450 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3451 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3452 return -EINVAL;
3453
3454 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3455 }
3456
3457 /*
3458 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3459 *
3460 * This set option will delete a shared secret key from use.
3461 */
3462 static int sctp_setsockopt_del_key(struct sock *sk,
3463 char __user *optval,
3464 unsigned int optlen)
3465 {
3466 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3467 struct sctp_authkeyid val;
3468 struct sctp_association *asoc;
3469
3470 if (!ep->auth_enable)
3471 return -EACCES;
3472
3473 if (optlen != sizeof(struct sctp_authkeyid))
3474 return -EINVAL;
3475 if (copy_from_user(&val, optval, optlen))
3476 return -EFAULT;
3477
3478 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3479 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3480 return -EINVAL;
3481
3482 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3483
3484 }
3485
3486 /*
3487 * 8.1.23 SCTP_AUTO_ASCONF
3488 *
3489 * This option will enable or disable the use of the automatic generation of
3490 * ASCONF chunks to add and delete addresses to an existing association. Note
3491 * that this option has two caveats namely: a) it only affects sockets that
3492 * are bound to all addresses available to the SCTP stack, and b) the system
3493 * administrator may have an overriding control that turns the ASCONF feature
3494 * off no matter what setting the socket option may have.
3495 * This option expects an integer boolean flag, where a non-zero value turns on
3496 * the option, and a zero value turns off the option.
3497 * Note. In this implementation, socket operation overrides default parameter
3498 * being set by sysctl as well as FreeBSD implementation
3499 */
3500 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3501 unsigned int optlen)
3502 {
3503 int val;
3504 struct sctp_sock *sp = sctp_sk(sk);
3505
3506 if (optlen < sizeof(int))
3507 return -EINVAL;
3508 if (get_user(val, (int __user *)optval))
3509 return -EFAULT;
3510 if (!sctp_is_ep_boundall(sk) && val)
3511 return -EINVAL;
3512 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3513 return 0;
3514
3515 spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3516 if (val == 0 && sp->do_auto_asconf) {
3517 list_del(&sp->auto_asconf_list);
3518 sp->do_auto_asconf = 0;
3519 } else if (val && !sp->do_auto_asconf) {
3520 list_add_tail(&sp->auto_asconf_list,
3521 &sock_net(sk)->sctp.auto_asconf_splist);
3522 sp->do_auto_asconf = 1;
3523 }
3524 spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3525 return 0;
3526 }
3527
3528
3529 /*
3530 * SCTP_PEER_ADDR_THLDS
3531 *
3532 * This option allows us to alter the partially failed threshold for one or all
3533 * transports in an association. See Section 6.1 of:
3534 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3535 */
3536 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3537 char __user *optval,
3538 unsigned int optlen)
3539 {
3540 struct sctp_paddrthlds val;
3541 struct sctp_transport *trans;
3542 struct sctp_association *asoc;
3543
3544 if (optlen < sizeof(struct sctp_paddrthlds))
3545 return -EINVAL;
3546 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3547 sizeof(struct sctp_paddrthlds)))
3548 return -EFAULT;
3549
3550
3551 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3552 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3553 if (!asoc)
3554 return -ENOENT;
3555 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3556 transports) {
3557 if (val.spt_pathmaxrxt)
3558 trans->pathmaxrxt = val.spt_pathmaxrxt;
3559 trans->pf_retrans = val.spt_pathpfthld;
3560 }
3561
3562 if (val.spt_pathmaxrxt)
3563 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3564 asoc->pf_retrans = val.spt_pathpfthld;
3565 } else {
3566 trans = sctp_addr_id2transport(sk, &val.spt_address,
3567 val.spt_assoc_id);
3568 if (!trans)
3569 return -ENOENT;
3570
3571 if (val.spt_pathmaxrxt)
3572 trans->pathmaxrxt = val.spt_pathmaxrxt;
3573 trans->pf_retrans = val.spt_pathpfthld;
3574 }
3575
3576 return 0;
3577 }
3578
3579 /* API 6.2 setsockopt(), getsockopt()
3580 *
3581 * Applications use setsockopt() and getsockopt() to set or retrieve
3582 * socket options. Socket options are used to change the default
3583 * behavior of sockets calls. They are described in Section 7.
3584 *
3585 * The syntax is:
3586 *
3587 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3588 * int __user *optlen);
3589 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3590 * int optlen);
3591 *
3592 * sd - the socket descript.
3593 * level - set to IPPROTO_SCTP for all SCTP options.
3594 * optname - the option name.
3595 * optval - the buffer to store the value of the option.
3596 * optlen - the size of the buffer.
3597 */
3598 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3599 char __user *optval, unsigned int optlen)
3600 {
3601 int retval = 0;
3602
3603 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3604 sk, optname);
3605
3606 /* I can hardly begin to describe how wrong this is. This is
3607 * so broken as to be worse than useless. The API draft
3608 * REALLY is NOT helpful here... I am not convinced that the
3609 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3610 * are at all well-founded.
3611 */
3612 if (level != SOL_SCTP) {
3613 struct sctp_af *af = sctp_sk(sk)->pf->af;
3614 retval = af->setsockopt(sk, level, optname, optval, optlen);
3615 goto out_nounlock;
3616 }
3617
3618 sctp_lock_sock(sk);
3619
3620 switch (optname) {
3621 case SCTP_SOCKOPT_BINDX_ADD:
3622 /* 'optlen' is the size of the addresses buffer. */
3623 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3624 optlen, SCTP_BINDX_ADD_ADDR);
3625 break;
3626
3627 case SCTP_SOCKOPT_BINDX_REM:
3628 /* 'optlen' is the size of the addresses buffer. */
3629 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3630 optlen, SCTP_BINDX_REM_ADDR);
3631 break;
3632
3633 case SCTP_SOCKOPT_CONNECTX_OLD:
3634 /* 'optlen' is the size of the addresses buffer. */
3635 retval = sctp_setsockopt_connectx_old(sk,
3636 (struct sockaddr __user *)optval,
3637 optlen);
3638 break;
3639
3640 case SCTP_SOCKOPT_CONNECTX:
3641 /* 'optlen' is the size of the addresses buffer. */
3642 retval = sctp_setsockopt_connectx(sk,
3643 (struct sockaddr __user *)optval,
3644 optlen);
3645 break;
3646
3647 case SCTP_DISABLE_FRAGMENTS:
3648 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3649 break;
3650
3651 case SCTP_EVENTS:
3652 retval = sctp_setsockopt_events(sk, optval, optlen);
3653 break;
3654
3655 case SCTP_AUTOCLOSE:
3656 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3657 break;
3658
3659 case SCTP_PEER_ADDR_PARAMS:
3660 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3661 break;
3662
3663 case SCTP_DELAYED_SACK:
3664 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3665 break;
3666 case SCTP_PARTIAL_DELIVERY_POINT:
3667 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3668 break;
3669
3670 case SCTP_INITMSG:
3671 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3672 break;
3673 case SCTP_DEFAULT_SEND_PARAM:
3674 retval = sctp_setsockopt_default_send_param(sk, optval,
3675 optlen);
3676 break;
3677 case SCTP_PRIMARY_ADDR:
3678 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3679 break;
3680 case SCTP_SET_PEER_PRIMARY_ADDR:
3681 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3682 break;
3683 case SCTP_NODELAY:
3684 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3685 break;
3686 case SCTP_RTOINFO:
3687 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3688 break;
3689 case SCTP_ASSOCINFO:
3690 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3691 break;
3692 case SCTP_I_WANT_MAPPED_V4_ADDR:
3693 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3694 break;
3695 case SCTP_MAXSEG:
3696 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3697 break;
3698 case SCTP_ADAPTATION_LAYER:
3699 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3700 break;
3701 case SCTP_CONTEXT:
3702 retval = sctp_setsockopt_context(sk, optval, optlen);
3703 break;
3704 case SCTP_FRAGMENT_INTERLEAVE:
3705 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3706 break;
3707 case SCTP_MAX_BURST:
3708 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3709 break;
3710 case SCTP_AUTH_CHUNK:
3711 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3712 break;
3713 case SCTP_HMAC_IDENT:
3714 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3715 break;
3716 case SCTP_AUTH_KEY:
3717 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3718 break;
3719 case SCTP_AUTH_ACTIVE_KEY:
3720 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3721 break;
3722 case SCTP_AUTH_DELETE_KEY:
3723 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3724 break;
3725 case SCTP_AUTO_ASCONF:
3726 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3727 break;
3728 case SCTP_PEER_ADDR_THLDS:
3729 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3730 break;
3731 default:
3732 retval = -ENOPROTOOPT;
3733 break;
3734 }
3735
3736 sctp_release_sock(sk);
3737
3738 out_nounlock:
3739 return retval;
3740 }
3741
3742 /* API 3.1.6 connect() - UDP Style Syntax
3743 *
3744 * An application may use the connect() call in the UDP model to initiate an
3745 * association without sending data.
3746 *
3747 * The syntax is:
3748 *
3749 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3750 *
3751 * sd: the socket descriptor to have a new association added to.
3752 *
3753 * nam: the address structure (either struct sockaddr_in or struct
3754 * sockaddr_in6 defined in RFC2553 [7]).
3755 *
3756 * len: the size of the address.
3757 */
3758 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3759 int addr_len)
3760 {
3761 int err = 0;
3762 struct sctp_af *af;
3763
3764 sctp_lock_sock(sk);
3765
3766 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3767 __func__, sk, addr, addr_len);
3768
3769 /* Validate addr_len before calling common connect/connectx routine. */
3770 af = sctp_get_af_specific(addr->sa_family);
3771 if (!af || addr_len < af->sockaddr_len) {
3772 err = -EINVAL;
3773 } else {
3774 /* Pass correct addr len to common routine (so it knows there
3775 * is only one address being passed.
3776 */
3777 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3778 }
3779
3780 sctp_release_sock(sk);
3781 return err;
3782 }
3783
3784 /* FIXME: Write comments. */
3785 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3786 {
3787 return -EOPNOTSUPP; /* STUB */
3788 }
3789
3790 /* 4.1.4 accept() - TCP Style Syntax
3791 *
3792 * Applications use accept() call to remove an established SCTP
3793 * association from the accept queue of the endpoint. A new socket
3794 * descriptor will be returned from accept() to represent the newly
3795 * formed association.
3796 */
3797 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3798 {
3799 struct sctp_sock *sp;
3800 struct sctp_endpoint *ep;
3801 struct sock *newsk = NULL;
3802 struct sctp_association *asoc;
3803 long timeo;
3804 int error = 0;
3805
3806 sctp_lock_sock(sk);
3807
3808 sp = sctp_sk(sk);
3809 ep = sp->ep;
3810
3811 if (!sctp_style(sk, TCP)) {
3812 error = -EOPNOTSUPP;
3813 goto out;
3814 }
3815
3816 if (!sctp_sstate(sk, LISTENING)) {
3817 error = -EINVAL;
3818 goto out;
3819 }
3820
3821 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3822
3823 error = sctp_wait_for_accept(sk, timeo);
3824 if (error)
3825 goto out;
3826
3827 /* We treat the list of associations on the endpoint as the accept
3828 * queue and pick the first association on the list.
3829 */
3830 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3831
3832 newsk = sp->pf->create_accept_sk(sk, asoc);
3833 if (!newsk) {
3834 error = -ENOMEM;
3835 goto out;
3836 }
3837
3838 /* Populate the fields of the newsk from the oldsk and migrate the
3839 * asoc to the newsk.
3840 */
3841 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3842
3843 out:
3844 sctp_release_sock(sk);
3845 *err = error;
3846 return newsk;
3847 }
3848
3849 /* The SCTP ioctl handler. */
3850 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3851 {
3852 int rc = -ENOTCONN;
3853
3854 sctp_lock_sock(sk);
3855
3856 /*
3857 * SEQPACKET-style sockets in LISTENING state are valid, for
3858 * SCTP, so only discard TCP-style sockets in LISTENING state.
3859 */
3860 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3861 goto out;
3862
3863 switch (cmd) {
3864 case SIOCINQ: {
3865 struct sk_buff *skb;
3866 unsigned int amount = 0;
3867
3868 skb = skb_peek(&sk->sk_receive_queue);
3869 if (skb != NULL) {
3870 /*
3871 * We will only return the amount of this packet since
3872 * that is all that will be read.
3873 */
3874 amount = skb->len;
3875 }
3876 rc = put_user(amount, (int __user *)arg);
3877 break;
3878 }
3879 default:
3880 rc = -ENOIOCTLCMD;
3881 break;
3882 }
3883 out:
3884 sctp_release_sock(sk);
3885 return rc;
3886 }
3887
3888 /* This is the function which gets called during socket creation to
3889 * initialized the SCTP-specific portion of the sock.
3890 * The sock structure should already be zero-filled memory.
3891 */
3892 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3893 {
3894 struct net *net = sock_net(sk);
3895 struct sctp_endpoint *ep;
3896 struct sctp_sock *sp;
3897
3898 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3899
3900 sp = sctp_sk(sk);
3901
3902 /* Initialize the SCTP per socket area. */
3903 switch (sk->sk_type) {
3904 case SOCK_SEQPACKET:
3905 sp->type = SCTP_SOCKET_UDP;
3906 break;
3907 case SOCK_STREAM:
3908 sp->type = SCTP_SOCKET_TCP;
3909 break;
3910 default:
3911 return -ESOCKTNOSUPPORT;
3912 }
3913
3914 /* Initialize default send parameters. These parameters can be
3915 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3916 */
3917 sp->default_stream = 0;
3918 sp->default_ppid = 0;
3919 sp->default_flags = 0;
3920 sp->default_context = 0;
3921 sp->default_timetolive = 0;
3922
3923 sp->default_rcv_context = 0;
3924 sp->max_burst = net->sctp.max_burst;
3925
3926 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
3927
3928 /* Initialize default setup parameters. These parameters
3929 * can be modified with the SCTP_INITMSG socket option or
3930 * overridden by the SCTP_INIT CMSG.
3931 */
3932 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3933 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3934 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
3935 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
3936
3937 /* Initialize default RTO related parameters. These parameters can
3938 * be modified for with the SCTP_RTOINFO socket option.
3939 */
3940 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
3941 sp->rtoinfo.srto_max = net->sctp.rto_max;
3942 sp->rtoinfo.srto_min = net->sctp.rto_min;
3943
3944 /* Initialize default association related parameters. These parameters
3945 * can be modified with the SCTP_ASSOCINFO socket option.
3946 */
3947 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
3948 sp->assocparams.sasoc_number_peer_destinations = 0;
3949 sp->assocparams.sasoc_peer_rwnd = 0;
3950 sp->assocparams.sasoc_local_rwnd = 0;
3951 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
3952
3953 /* Initialize default event subscriptions. By default, all the
3954 * options are off.
3955 */
3956 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3957
3958 /* Default Peer Address Parameters. These defaults can
3959 * be modified via SCTP_PEER_ADDR_PARAMS
3960 */
3961 sp->hbinterval = net->sctp.hb_interval;
3962 sp->pathmaxrxt = net->sctp.max_retrans_path;
3963 sp->pathmtu = 0; // allow default discovery
3964 sp->sackdelay = net->sctp.sack_timeout;
3965 sp->sackfreq = 2;
3966 sp->param_flags = SPP_HB_ENABLE |
3967 SPP_PMTUD_ENABLE |
3968 SPP_SACKDELAY_ENABLE;
3969
3970 /* If enabled no SCTP message fragmentation will be performed.
3971 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3972 */
3973 sp->disable_fragments = 0;
3974
3975 /* Enable Nagle algorithm by default. */
3976 sp->nodelay = 0;
3977
3978 /* Enable by default. */
3979 sp->v4mapped = 1;
3980
3981 /* Auto-close idle associations after the configured
3982 * number of seconds. A value of 0 disables this
3983 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3984 * for UDP-style sockets only.
3985 */
3986 sp->autoclose = 0;
3987
3988 /* User specified fragmentation limit. */
3989 sp->user_frag = 0;
3990
3991 sp->adaptation_ind = 0;
3992
3993 sp->pf = sctp_get_pf_specific(sk->sk_family);
3994
3995 /* Control variables for partial data delivery. */
3996 atomic_set(&sp->pd_mode, 0);
3997 skb_queue_head_init(&sp->pd_lobby);
3998 sp->frag_interleave = 0;
3999
4000 /* Create a per socket endpoint structure. Even if we
4001 * change the data structure relationships, this may still
4002 * be useful for storing pre-connect address information.
4003 */
4004 ep = sctp_endpoint_new(sk, GFP_KERNEL);
4005 if (!ep)
4006 return -ENOMEM;
4007
4008 sp->ep = ep;
4009 sp->hmac = NULL;
4010
4011 SCTP_DBG_OBJCNT_INC(sock);
4012
4013 local_bh_disable();
4014 percpu_counter_inc(&sctp_sockets_allocated);
4015 sock_prot_inuse_add(net, sk->sk_prot, 1);
4016
4017 /* Nothing can fail after this block, otherwise
4018 * sctp_destroy_sock() will be called without addr_wq_lock held
4019 */
4020 if (net->sctp.default_auto_asconf) {
4021 spin_lock(&sock_net(sk)->sctp.addr_wq_lock);
4022 list_add_tail(&sp->auto_asconf_list,
4023 &net->sctp.auto_asconf_splist);
4024 sp->do_auto_asconf = 1;
4025 spin_unlock(&sock_net(sk)->sctp.addr_wq_lock);
4026 } else {
4027 sp->do_auto_asconf = 0;
4028 }
4029
4030 local_bh_enable();
4031
4032 return 0;
4033 }
4034
4035 /* Cleanup any SCTP per socket resources. Must be called with
4036 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4037 */
4038 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
4039 {
4040 struct sctp_sock *sp;
4041
4042 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
4043
4044 /* Release our hold on the endpoint. */
4045 sp = sctp_sk(sk);
4046 /* This could happen during socket init, thus we bail out
4047 * early, since the rest of the below is not setup either.
4048 */
4049 if (sp->ep == NULL)
4050 return;
4051
4052 if (sp->do_auto_asconf) {
4053 sp->do_auto_asconf = 0;
4054 list_del(&sp->auto_asconf_list);
4055 }
4056 sctp_endpoint_free(sp->ep);
4057 local_bh_disable();
4058 percpu_counter_dec(&sctp_sockets_allocated);
4059 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4060 local_bh_enable();
4061 }
4062
4063 /* API 4.1.7 shutdown() - TCP Style Syntax
4064 * int shutdown(int socket, int how);
4065 *
4066 * sd - the socket descriptor of the association to be closed.
4067 * how - Specifies the type of shutdown. The values are
4068 * as follows:
4069 * SHUT_RD
4070 * Disables further receive operations. No SCTP
4071 * protocol action is taken.
4072 * SHUT_WR
4073 * Disables further send operations, and initiates
4074 * the SCTP shutdown sequence.
4075 * SHUT_RDWR
4076 * Disables further send and receive operations
4077 * and initiates the SCTP shutdown sequence.
4078 */
4079 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
4080 {
4081 struct net *net = sock_net(sk);
4082 struct sctp_endpoint *ep;
4083 struct sctp_association *asoc;
4084
4085 if (!sctp_style(sk, TCP))
4086 return;
4087
4088 if (how & SEND_SHUTDOWN) {
4089 ep = sctp_sk(sk)->ep;
4090 if (!list_empty(&ep->asocs)) {
4091 asoc = list_entry(ep->asocs.next,
4092 struct sctp_association, asocs);
4093 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4094 }
4095 }
4096 }
4097
4098 /* 7.2.1 Association Status (SCTP_STATUS)
4099
4100 * Applications can retrieve current status information about an
4101 * association, including association state, peer receiver window size,
4102 * number of unacked data chunks, and number of data chunks pending
4103 * receipt. This information is read-only.
4104 */
4105 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4106 char __user *optval,
4107 int __user *optlen)
4108 {
4109 struct sctp_status status;
4110 struct sctp_association *asoc = NULL;
4111 struct sctp_transport *transport;
4112 sctp_assoc_t associd;
4113 int retval = 0;
4114
4115 if (len < sizeof(status)) {
4116 retval = -EINVAL;
4117 goto out;
4118 }
4119
4120 len = sizeof(status);
4121 if (copy_from_user(&status, optval, len)) {
4122 retval = -EFAULT;
4123 goto out;
4124 }
4125
4126 associd = status.sstat_assoc_id;
4127 asoc = sctp_id2assoc(sk, associd);
4128 if (!asoc) {
4129 retval = -EINVAL;
4130 goto out;
4131 }
4132
4133 transport = asoc->peer.primary_path;
4134
4135 status.sstat_assoc_id = sctp_assoc2id(asoc);
4136 status.sstat_state = asoc->state;
4137 status.sstat_rwnd = asoc->peer.rwnd;
4138 status.sstat_unackdata = asoc->unack_data;
4139
4140 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4141 status.sstat_instrms = asoc->c.sinit_max_instreams;
4142 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4143 status.sstat_fragmentation_point = asoc->frag_point;
4144 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4145 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4146 transport->af_specific->sockaddr_len);
4147 /* Map ipv4 address into v4-mapped-on-v6 address. */
4148 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4149 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4150 status.sstat_primary.spinfo_state = transport->state;
4151 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4152 status.sstat_primary.spinfo_srtt = transport->srtt;
4153 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4154 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4155
4156 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4157 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4158
4159 if (put_user(len, optlen)) {
4160 retval = -EFAULT;
4161 goto out;
4162 }
4163
4164 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
4165 len, status.sstat_state, status.sstat_rwnd,
4166 status.sstat_assoc_id);
4167
4168 if (copy_to_user(optval, &status, len)) {
4169 retval = -EFAULT;
4170 goto out;
4171 }
4172
4173 out:
4174 return retval;
4175 }
4176
4177
4178 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4179 *
4180 * Applications can retrieve information about a specific peer address
4181 * of an association, including its reachability state, congestion
4182 * window, and retransmission timer values. This information is
4183 * read-only.
4184 */
4185 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4186 char __user *optval,
4187 int __user *optlen)
4188 {
4189 struct sctp_paddrinfo pinfo;
4190 struct sctp_transport *transport;
4191 int retval = 0;
4192
4193 if (len < sizeof(pinfo)) {
4194 retval = -EINVAL;
4195 goto out;
4196 }
4197
4198 len = sizeof(pinfo);
4199 if (copy_from_user(&pinfo, optval, len)) {
4200 retval = -EFAULT;
4201 goto out;
4202 }
4203
4204 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4205 pinfo.spinfo_assoc_id);
4206 if (!transport)
4207 return -EINVAL;
4208
4209 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4210 pinfo.spinfo_state = transport->state;
4211 pinfo.spinfo_cwnd = transport->cwnd;
4212 pinfo.spinfo_srtt = transport->srtt;
4213 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4214 pinfo.spinfo_mtu = transport->pathmtu;
4215
4216 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4217 pinfo.spinfo_state = SCTP_ACTIVE;
4218
4219 if (put_user(len, optlen)) {
4220 retval = -EFAULT;
4221 goto out;
4222 }
4223
4224 if (copy_to_user(optval, &pinfo, len)) {
4225 retval = -EFAULT;
4226 goto out;
4227 }
4228
4229 out:
4230 return retval;
4231 }
4232
4233 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4234 *
4235 * This option is a on/off flag. If enabled no SCTP message
4236 * fragmentation will be performed. Instead if a message being sent
4237 * exceeds the current PMTU size, the message will NOT be sent and
4238 * instead a error will be indicated to the user.
4239 */
4240 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4241 char __user *optval, int __user *optlen)
4242 {
4243 int val;
4244
4245 if (len < sizeof(int))
4246 return -EINVAL;
4247
4248 len = sizeof(int);
4249 val = (sctp_sk(sk)->disable_fragments == 1);
4250 if (put_user(len, optlen))
4251 return -EFAULT;
4252 if (copy_to_user(optval, &val, len))
4253 return -EFAULT;
4254 return 0;
4255 }
4256
4257 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4258 *
4259 * This socket option is used to specify various notifications and
4260 * ancillary data the user wishes to receive.
4261 */
4262 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4263 int __user *optlen)
4264 {
4265 if (len == 0)
4266 return -EINVAL;
4267 if (len > sizeof(struct sctp_event_subscribe))
4268 len = sizeof(struct sctp_event_subscribe);
4269 if (put_user(len, optlen))
4270 return -EFAULT;
4271 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4272 return -EFAULT;
4273 return 0;
4274 }
4275
4276 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4277 *
4278 * This socket option is applicable to the UDP-style socket only. When
4279 * set it will cause associations that are idle for more than the
4280 * specified number of seconds to automatically close. An association
4281 * being idle is defined an association that has NOT sent or received
4282 * user data. The special value of '0' indicates that no automatic
4283 * close of any associations should be performed. The option expects an
4284 * integer defining the number of seconds of idle time before an
4285 * association is closed.
4286 */
4287 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4288 {
4289 /* Applicable to UDP-style socket only */
4290 if (sctp_style(sk, TCP))
4291 return -EOPNOTSUPP;
4292 if (len < sizeof(int))
4293 return -EINVAL;
4294 len = sizeof(int);
4295 if (put_user(len, optlen))
4296 return -EFAULT;
4297 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4298 return -EFAULT;
4299 return 0;
4300 }
4301
4302 /* Helper routine to branch off an association to a new socket. */
4303 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4304 {
4305 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4306 struct socket *sock;
4307 struct sctp_af *af;
4308 int err = 0;
4309
4310 if (!asoc)
4311 return -EINVAL;
4312
4313 /* If there is a thread waiting on more sndbuf space for
4314 * sending on this asoc, it cannot be peeled.
4315 */
4316 if (waitqueue_active(&asoc->wait))
4317 return -EBUSY;
4318
4319 /* An association cannot be branched off from an already peeled-off
4320 * socket, nor is this supported for tcp style sockets.
4321 */
4322 if (!sctp_style(sk, UDP))
4323 return -EINVAL;
4324
4325 /* Create a new socket. */
4326 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4327 if (err < 0)
4328 return err;
4329
4330 sctp_copy_sock(sock->sk, sk, asoc);
4331
4332 /* Make peeled-off sockets more like 1-1 accepted sockets.
4333 * Set the daddr and initialize id to something more random
4334 */
4335 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4336 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4337
4338 /* Populate the fields of the newsk from the oldsk and migrate the
4339 * asoc to the newsk.
4340 */
4341 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4342
4343 *sockp = sock;
4344
4345 return err;
4346 }
4347 EXPORT_SYMBOL(sctp_do_peeloff);
4348
4349 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4350 {
4351 sctp_peeloff_arg_t peeloff;
4352 struct socket *newsock;
4353 struct file *newfile;
4354 int retval = 0;
4355
4356 if (len < sizeof(sctp_peeloff_arg_t))
4357 return -EINVAL;
4358 len = sizeof(sctp_peeloff_arg_t);
4359 if (copy_from_user(&peeloff, optval, len))
4360 return -EFAULT;
4361
4362 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4363 if (retval < 0)
4364 goto out;
4365
4366 /* Map the socket to an unused fd that can be returned to the user. */
4367 retval = get_unused_fd();
4368 if (retval < 0) {
4369 sock_release(newsock);
4370 goto out;
4371 }
4372
4373 newfile = sock_alloc_file(newsock, 0, NULL);
4374 if (unlikely(IS_ERR(newfile))) {
4375 put_unused_fd(retval);
4376 sock_release(newsock);
4377 return PTR_ERR(newfile);
4378 }
4379
4380 SCTP_DEBUG_PRINTK("%s: sk: %p newsk: %p sd: %d\n",
4381 __func__, sk, newsock->sk, retval);
4382
4383 /* Return the fd mapped to the new socket. */
4384 if (put_user(len, optlen)) {
4385 fput(newfile);
4386 put_unused_fd(retval);
4387 return -EFAULT;
4388 }
4389 peeloff.sd = retval;
4390 if (copy_to_user(optval, &peeloff, len)) {
4391 fput(newfile);
4392 put_unused_fd(retval);
4393 return -EFAULT;
4394 }
4395 fd_install(retval, newfile);
4396 out:
4397 return retval;
4398 }
4399
4400 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4401 *
4402 * Applications can enable or disable heartbeats for any peer address of
4403 * an association, modify an address's heartbeat interval, force a
4404 * heartbeat to be sent immediately, and adjust the address's maximum
4405 * number of retransmissions sent before an address is considered
4406 * unreachable. The following structure is used to access and modify an
4407 * address's parameters:
4408 *
4409 * struct sctp_paddrparams {
4410 * sctp_assoc_t spp_assoc_id;
4411 * struct sockaddr_storage spp_address;
4412 * uint32_t spp_hbinterval;
4413 * uint16_t spp_pathmaxrxt;
4414 * uint32_t spp_pathmtu;
4415 * uint32_t spp_sackdelay;
4416 * uint32_t spp_flags;
4417 * };
4418 *
4419 * spp_assoc_id - (one-to-many style socket) This is filled in the
4420 * application, and identifies the association for
4421 * this query.
4422 * spp_address - This specifies which address is of interest.
4423 * spp_hbinterval - This contains the value of the heartbeat interval,
4424 * in milliseconds. If a value of zero
4425 * is present in this field then no changes are to
4426 * be made to this parameter.
4427 * spp_pathmaxrxt - This contains the maximum number of
4428 * retransmissions before this address shall be
4429 * considered unreachable. If a value of zero
4430 * is present in this field then no changes are to
4431 * be made to this parameter.
4432 * spp_pathmtu - When Path MTU discovery is disabled the value
4433 * specified here will be the "fixed" path mtu.
4434 * Note that if the spp_address field is empty
4435 * then all associations on this address will
4436 * have this fixed path mtu set upon them.
4437 *
4438 * spp_sackdelay - When delayed sack is enabled, this value specifies
4439 * the number of milliseconds that sacks will be delayed
4440 * for. This value will apply to all addresses of an
4441 * association if the spp_address field is empty. Note
4442 * also, that if delayed sack is enabled and this
4443 * value is set to 0, no change is made to the last
4444 * recorded delayed sack timer value.
4445 *
4446 * spp_flags - These flags are used to control various features
4447 * on an association. The flag field may contain
4448 * zero or more of the following options.
4449 *
4450 * SPP_HB_ENABLE - Enable heartbeats on the
4451 * specified address. Note that if the address
4452 * field is empty all addresses for the association
4453 * have heartbeats enabled upon them.
4454 *
4455 * SPP_HB_DISABLE - Disable heartbeats on the
4456 * speicifed address. Note that if the address
4457 * field is empty all addresses for the association
4458 * will have their heartbeats disabled. Note also
4459 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4460 * mutually exclusive, only one of these two should
4461 * be specified. Enabling both fields will have
4462 * undetermined results.
4463 *
4464 * SPP_HB_DEMAND - Request a user initiated heartbeat
4465 * to be made immediately.
4466 *
4467 * SPP_PMTUD_ENABLE - This field will enable PMTU
4468 * discovery upon the specified address. Note that
4469 * if the address feild is empty then all addresses
4470 * on the association are effected.
4471 *
4472 * SPP_PMTUD_DISABLE - This field will disable PMTU
4473 * discovery upon the specified address. Note that
4474 * if the address feild is empty then all addresses
4475 * on the association are effected. Not also that
4476 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4477 * exclusive. Enabling both will have undetermined
4478 * results.
4479 *
4480 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4481 * on delayed sack. The time specified in spp_sackdelay
4482 * is used to specify the sack delay for this address. Note
4483 * that if spp_address is empty then all addresses will
4484 * enable delayed sack and take on the sack delay
4485 * value specified in spp_sackdelay.
4486 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4487 * off delayed sack. If the spp_address field is blank then
4488 * delayed sack is disabled for the entire association. Note
4489 * also that this field is mutually exclusive to
4490 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4491 * results.
4492 */
4493 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4494 char __user *optval, int __user *optlen)
4495 {
4496 struct sctp_paddrparams params;
4497 struct sctp_transport *trans = NULL;
4498 struct sctp_association *asoc = NULL;
4499 struct sctp_sock *sp = sctp_sk(sk);
4500
4501 if (len < sizeof(struct sctp_paddrparams))
4502 return -EINVAL;
4503 len = sizeof(struct sctp_paddrparams);
4504 if (copy_from_user(&params, optval, len))
4505 return -EFAULT;
4506
4507 /* If an address other than INADDR_ANY is specified, and
4508 * no transport is found, then the request is invalid.
4509 */
4510 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
4511 trans = sctp_addr_id2transport(sk, &params.spp_address,
4512 params.spp_assoc_id);
4513 if (!trans) {
4514 SCTP_DEBUG_PRINTK("Failed no transport\n");
4515 return -EINVAL;
4516 }
4517 }
4518
4519 /* Get association, if assoc_id != 0 and the socket is a one
4520 * to many style socket, and an association was not found, then
4521 * the id was invalid.
4522 */
4523 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4524 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4525 SCTP_DEBUG_PRINTK("Failed no association\n");
4526 return -EINVAL;
4527 }
4528
4529 if (trans) {
4530 /* Fetch transport values. */
4531 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4532 params.spp_pathmtu = trans->pathmtu;
4533 params.spp_pathmaxrxt = trans->pathmaxrxt;
4534 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4535
4536 /*draft-11 doesn't say what to return in spp_flags*/
4537 params.spp_flags = trans->param_flags;
4538 } else if (asoc) {
4539 /* Fetch association values. */
4540 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4541 params.spp_pathmtu = asoc->pathmtu;
4542 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4543 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4544
4545 /*draft-11 doesn't say what to return in spp_flags*/
4546 params.spp_flags = asoc->param_flags;
4547 } else {
4548 /* Fetch socket values. */
4549 params.spp_hbinterval = sp->hbinterval;
4550 params.spp_pathmtu = sp->pathmtu;
4551 params.spp_sackdelay = sp->sackdelay;
4552 params.spp_pathmaxrxt = sp->pathmaxrxt;
4553
4554 /*draft-11 doesn't say what to return in spp_flags*/
4555 params.spp_flags = sp->param_flags;
4556 }
4557
4558 if (copy_to_user(optval, &params, len))
4559 return -EFAULT;
4560
4561 if (put_user(len, optlen))
4562 return -EFAULT;
4563
4564 return 0;
4565 }
4566
4567 /*
4568 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4569 *
4570 * This option will effect the way delayed acks are performed. This
4571 * option allows you to get or set the delayed ack time, in
4572 * milliseconds. It also allows changing the delayed ack frequency.
4573 * Changing the frequency to 1 disables the delayed sack algorithm. If
4574 * the assoc_id is 0, then this sets or gets the endpoints default
4575 * values. If the assoc_id field is non-zero, then the set or get
4576 * effects the specified association for the one to many model (the
4577 * assoc_id field is ignored by the one to one model). Note that if
4578 * sack_delay or sack_freq are 0 when setting this option, then the
4579 * current values will remain unchanged.
4580 *
4581 * struct sctp_sack_info {
4582 * sctp_assoc_t sack_assoc_id;
4583 * uint32_t sack_delay;
4584 * uint32_t sack_freq;
4585 * };
4586 *
4587 * sack_assoc_id - This parameter, indicates which association the user
4588 * is performing an action upon. Note that if this field's value is
4589 * zero then the endpoints default value is changed (effecting future
4590 * associations only).
4591 *
4592 * sack_delay - This parameter contains the number of milliseconds that
4593 * the user is requesting the delayed ACK timer be set to. Note that
4594 * this value is defined in the standard to be between 200 and 500
4595 * milliseconds.
4596 *
4597 * sack_freq - This parameter contains the number of packets that must
4598 * be received before a sack is sent without waiting for the delay
4599 * timer to expire. The default value for this is 2, setting this
4600 * value to 1 will disable the delayed sack algorithm.
4601 */
4602 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4603 char __user *optval,
4604 int __user *optlen)
4605 {
4606 struct sctp_sack_info params;
4607 struct sctp_association *asoc = NULL;
4608 struct sctp_sock *sp = sctp_sk(sk);
4609
4610 if (len >= sizeof(struct sctp_sack_info)) {
4611 len = sizeof(struct sctp_sack_info);
4612
4613 if (copy_from_user(&params, optval, len))
4614 return -EFAULT;
4615 } else if (len == sizeof(struct sctp_assoc_value)) {
4616 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4617 pr_warn("Use struct sctp_sack_info instead\n");
4618 if (copy_from_user(&params, optval, len))
4619 return -EFAULT;
4620 } else
4621 return - EINVAL;
4622
4623 /* Get association, if sack_assoc_id != 0 and the socket is a one
4624 * to many style socket, and an association was not found, then
4625 * the id was invalid.
4626 */
4627 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4628 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4629 return -EINVAL;
4630
4631 if (asoc) {
4632 /* Fetch association values. */
4633 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4634 params.sack_delay = jiffies_to_msecs(
4635 asoc->sackdelay);
4636 params.sack_freq = asoc->sackfreq;
4637
4638 } else {
4639 params.sack_delay = 0;
4640 params.sack_freq = 1;
4641 }
4642 } else {
4643 /* Fetch socket values. */
4644 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4645 params.sack_delay = sp->sackdelay;
4646 params.sack_freq = sp->sackfreq;
4647 } else {
4648 params.sack_delay = 0;
4649 params.sack_freq = 1;
4650 }
4651 }
4652
4653 if (copy_to_user(optval, &params, len))
4654 return -EFAULT;
4655
4656 if (put_user(len, optlen))
4657 return -EFAULT;
4658
4659 return 0;
4660 }
4661
4662 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4663 *
4664 * Applications can specify protocol parameters for the default association
4665 * initialization. The option name argument to setsockopt() and getsockopt()
4666 * is SCTP_INITMSG.
4667 *
4668 * Setting initialization parameters is effective only on an unconnected
4669 * socket (for UDP-style sockets only future associations are effected
4670 * by the change). With TCP-style sockets, this option is inherited by
4671 * sockets derived from a listener socket.
4672 */
4673 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4674 {
4675 if (len < sizeof(struct sctp_initmsg))
4676 return -EINVAL;
4677 len = sizeof(struct sctp_initmsg);
4678 if (put_user(len, optlen))
4679 return -EFAULT;
4680 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4681 return -EFAULT;
4682 return 0;
4683 }
4684
4685
4686 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4687 char __user *optval, int __user *optlen)
4688 {
4689 struct sctp_association *asoc;
4690 int cnt = 0;
4691 struct sctp_getaddrs getaddrs;
4692 struct sctp_transport *from;
4693 void __user *to;
4694 union sctp_addr temp;
4695 struct sctp_sock *sp = sctp_sk(sk);
4696 int addrlen;
4697 size_t space_left;
4698 int bytes_copied;
4699
4700 if (len < sizeof(struct sctp_getaddrs))
4701 return -EINVAL;
4702
4703 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4704 return -EFAULT;
4705
4706 /* For UDP-style sockets, id specifies the association to query. */
4707 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4708 if (!asoc)
4709 return -EINVAL;
4710
4711 to = optval + offsetof(struct sctp_getaddrs,addrs);
4712 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4713
4714 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4715 transports) {
4716 memcpy(&temp, &from->ipaddr, sizeof(temp));
4717 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4718 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4719 if (space_left < addrlen)
4720 return -ENOMEM;
4721 if (copy_to_user(to, &temp, addrlen))
4722 return -EFAULT;
4723 to += addrlen;
4724 cnt++;
4725 space_left -= addrlen;
4726 }
4727
4728 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4729 return -EFAULT;
4730 bytes_copied = ((char __user *)to) - optval;
4731 if (put_user(bytes_copied, optlen))
4732 return -EFAULT;
4733
4734 return 0;
4735 }
4736
4737 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4738 size_t space_left, int *bytes_copied)
4739 {
4740 struct sctp_sockaddr_entry *addr;
4741 union sctp_addr temp;
4742 int cnt = 0;
4743 int addrlen;
4744 struct net *net = sock_net(sk);
4745
4746 rcu_read_lock();
4747 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
4748 if (!addr->valid)
4749 continue;
4750
4751 if ((PF_INET == sk->sk_family) &&
4752 (AF_INET6 == addr->a.sa.sa_family))
4753 continue;
4754 if ((PF_INET6 == sk->sk_family) &&
4755 inet_v6_ipv6only(sk) &&
4756 (AF_INET == addr->a.sa.sa_family))
4757 continue;
4758 memcpy(&temp, &addr->a, sizeof(temp));
4759 if (!temp.v4.sin_port)
4760 temp.v4.sin_port = htons(port);
4761
4762 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4763 &temp);
4764 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4765 if (space_left < addrlen) {
4766 cnt = -ENOMEM;
4767 break;
4768 }
4769 memcpy(to, &temp, addrlen);
4770
4771 to += addrlen;
4772 cnt ++;
4773 space_left -= addrlen;
4774 *bytes_copied += addrlen;
4775 }
4776 rcu_read_unlock();
4777
4778 return cnt;
4779 }
4780
4781
4782 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4783 char __user *optval, int __user *optlen)
4784 {
4785 struct sctp_bind_addr *bp;
4786 struct sctp_association *asoc;
4787 int cnt = 0;
4788 struct sctp_getaddrs getaddrs;
4789 struct sctp_sockaddr_entry *addr;
4790 void __user *to;
4791 union sctp_addr temp;
4792 struct sctp_sock *sp = sctp_sk(sk);
4793 int addrlen;
4794 int err = 0;
4795 size_t space_left;
4796 int bytes_copied = 0;
4797 void *addrs;
4798 void *buf;
4799
4800 if (len < sizeof(struct sctp_getaddrs))
4801 return -EINVAL;
4802
4803 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4804 return -EFAULT;
4805
4806 /*
4807 * For UDP-style sockets, id specifies the association to query.
4808 * If the id field is set to the value '0' then the locally bound
4809 * addresses are returned without regard to any particular
4810 * association.
4811 */
4812 if (0 == getaddrs.assoc_id) {
4813 bp = &sctp_sk(sk)->ep->base.bind_addr;
4814 } else {
4815 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4816 if (!asoc)
4817 return -EINVAL;
4818 bp = &asoc->base.bind_addr;
4819 }
4820
4821 to = optval + offsetof(struct sctp_getaddrs,addrs);
4822 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4823
4824 addrs = kmalloc(space_left, GFP_KERNEL);
4825 if (!addrs)
4826 return -ENOMEM;
4827
4828 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4829 * addresses from the global local address list.
4830 */
4831 if (sctp_list_single_entry(&bp->address_list)) {
4832 addr = list_entry(bp->address_list.next,
4833 struct sctp_sockaddr_entry, list);
4834 if (sctp_is_any(sk, &addr->a)) {
4835 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4836 space_left, &bytes_copied);
4837 if (cnt < 0) {
4838 err = cnt;
4839 goto out;
4840 }
4841 goto copy_getaddrs;
4842 }
4843 }
4844
4845 buf = addrs;
4846 /* Protection on the bound address list is not needed since
4847 * in the socket option context we hold a socket lock and
4848 * thus the bound address list can't change.
4849 */
4850 list_for_each_entry(addr, &bp->address_list, list) {
4851 memcpy(&temp, &addr->a, sizeof(temp));
4852 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4853 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4854 if (space_left < addrlen) {
4855 err = -ENOMEM; /*fixme: right error?*/
4856 goto out;
4857 }
4858 memcpy(buf, &temp, addrlen);
4859 buf += addrlen;
4860 bytes_copied += addrlen;
4861 cnt ++;
4862 space_left -= addrlen;
4863 }
4864
4865 copy_getaddrs:
4866 if (copy_to_user(to, addrs, bytes_copied)) {
4867 err = -EFAULT;
4868 goto out;
4869 }
4870 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4871 err = -EFAULT;
4872 goto out;
4873 }
4874 if (put_user(bytes_copied, optlen))
4875 err = -EFAULT;
4876 out:
4877 kfree(addrs);
4878 return err;
4879 }
4880
4881 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4882 *
4883 * Requests that the local SCTP stack use the enclosed peer address as
4884 * the association primary. The enclosed address must be one of the
4885 * association peer's addresses.
4886 */
4887 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4888 char __user *optval, int __user *optlen)
4889 {
4890 struct sctp_prim prim;
4891 struct sctp_association *asoc;
4892 struct sctp_sock *sp = sctp_sk(sk);
4893
4894 if (len < sizeof(struct sctp_prim))
4895 return -EINVAL;
4896
4897 len = sizeof(struct sctp_prim);
4898
4899 if (copy_from_user(&prim, optval, len))
4900 return -EFAULT;
4901
4902 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4903 if (!asoc)
4904 return -EINVAL;
4905
4906 if (!asoc->peer.primary_path)
4907 return -ENOTCONN;
4908
4909 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4910 asoc->peer.primary_path->af_specific->sockaddr_len);
4911
4912 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4913 (union sctp_addr *)&prim.ssp_addr);
4914
4915 if (put_user(len, optlen))
4916 return -EFAULT;
4917 if (copy_to_user(optval, &prim, len))
4918 return -EFAULT;
4919
4920 return 0;
4921 }
4922
4923 /*
4924 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4925 *
4926 * Requests that the local endpoint set the specified Adaptation Layer
4927 * Indication parameter for all future INIT and INIT-ACK exchanges.
4928 */
4929 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4930 char __user *optval, int __user *optlen)
4931 {
4932 struct sctp_setadaptation adaptation;
4933
4934 if (len < sizeof(struct sctp_setadaptation))
4935 return -EINVAL;
4936
4937 len = sizeof(struct sctp_setadaptation);
4938
4939 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4940
4941 if (put_user(len, optlen))
4942 return -EFAULT;
4943 if (copy_to_user(optval, &adaptation, len))
4944 return -EFAULT;
4945
4946 return 0;
4947 }
4948
4949 /*
4950 *
4951 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4952 *
4953 * Applications that wish to use the sendto() system call may wish to
4954 * specify a default set of parameters that would normally be supplied
4955 * through the inclusion of ancillary data. This socket option allows
4956 * such an application to set the default sctp_sndrcvinfo structure.
4957
4958
4959 * The application that wishes to use this socket option simply passes
4960 * in to this call the sctp_sndrcvinfo structure defined in Section
4961 * 5.2.2) The input parameters accepted by this call include
4962 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4963 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4964 * to this call if the caller is using the UDP model.
4965 *
4966 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4967 */
4968 static int sctp_getsockopt_default_send_param(struct sock *sk,
4969 int len, char __user *optval,
4970 int __user *optlen)
4971 {
4972 struct sctp_sndrcvinfo info;
4973 struct sctp_association *asoc;
4974 struct sctp_sock *sp = sctp_sk(sk);
4975
4976 if (len < sizeof(struct sctp_sndrcvinfo))
4977 return -EINVAL;
4978
4979 len = sizeof(struct sctp_sndrcvinfo);
4980
4981 if (copy_from_user(&info, optval, len))
4982 return -EFAULT;
4983
4984 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4985 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4986 return -EINVAL;
4987
4988 if (asoc) {
4989 info.sinfo_stream = asoc->default_stream;
4990 info.sinfo_flags = asoc->default_flags;
4991 info.sinfo_ppid = asoc->default_ppid;
4992 info.sinfo_context = asoc->default_context;
4993 info.sinfo_timetolive = asoc->default_timetolive;
4994 } else {
4995 info.sinfo_stream = sp->default_stream;
4996 info.sinfo_flags = sp->default_flags;
4997 info.sinfo_ppid = sp->default_ppid;
4998 info.sinfo_context = sp->default_context;
4999 info.sinfo_timetolive = sp->default_timetolive;
5000 }
5001
5002 if (put_user(len, optlen))
5003 return -EFAULT;
5004 if (copy_to_user(optval, &info, len))
5005 return -EFAULT;
5006
5007 return 0;
5008 }
5009
5010 /*
5011 *
5012 * 7.1.5 SCTP_NODELAY
5013 *
5014 * Turn on/off any Nagle-like algorithm. This means that packets are
5015 * generally sent as soon as possible and no unnecessary delays are
5016 * introduced, at the cost of more packets in the network. Expects an
5017 * integer boolean flag.
5018 */
5019
5020 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
5021 char __user *optval, int __user *optlen)
5022 {
5023 int val;
5024
5025 if (len < sizeof(int))
5026 return -EINVAL;
5027
5028 len = sizeof(int);
5029 val = (sctp_sk(sk)->nodelay == 1);
5030 if (put_user(len, optlen))
5031 return -EFAULT;
5032 if (copy_to_user(optval, &val, len))
5033 return -EFAULT;
5034 return 0;
5035 }
5036
5037 /*
5038 *
5039 * 7.1.1 SCTP_RTOINFO
5040 *
5041 * The protocol parameters used to initialize and bound retransmission
5042 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5043 * and modify these parameters.
5044 * All parameters are time values, in milliseconds. A value of 0, when
5045 * modifying the parameters, indicates that the current value should not
5046 * be changed.
5047 *
5048 */
5049 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5050 char __user *optval,
5051 int __user *optlen) {
5052 struct sctp_rtoinfo rtoinfo;
5053 struct sctp_association *asoc;
5054
5055 if (len < sizeof (struct sctp_rtoinfo))
5056 return -EINVAL;
5057
5058 len = sizeof(struct sctp_rtoinfo);
5059
5060 if (copy_from_user(&rtoinfo, optval, len))
5061 return -EFAULT;
5062
5063 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5064
5065 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5066 return -EINVAL;
5067
5068 /* Values corresponding to the specific association. */
5069 if (asoc) {
5070 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5071 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5072 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5073 } else {
5074 /* Values corresponding to the endpoint. */
5075 struct sctp_sock *sp = sctp_sk(sk);
5076
5077 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5078 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5079 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5080 }
5081
5082 if (put_user(len, optlen))
5083 return -EFAULT;
5084
5085 if (copy_to_user(optval, &rtoinfo, len))
5086 return -EFAULT;
5087
5088 return 0;
5089 }
5090
5091 /*
5092 *
5093 * 7.1.2 SCTP_ASSOCINFO
5094 *
5095 * This option is used to tune the maximum retransmission attempts
5096 * of the association.
5097 * Returns an error if the new association retransmission value is
5098 * greater than the sum of the retransmission value of the peer.
5099 * See [SCTP] for more information.
5100 *
5101 */
5102 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5103 char __user *optval,
5104 int __user *optlen)
5105 {
5106
5107 struct sctp_assocparams assocparams;
5108 struct sctp_association *asoc;
5109 struct list_head *pos;
5110 int cnt = 0;
5111
5112 if (len < sizeof (struct sctp_assocparams))
5113 return -EINVAL;
5114
5115 len = sizeof(struct sctp_assocparams);
5116
5117 if (copy_from_user(&assocparams, optval, len))
5118 return -EFAULT;
5119
5120 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5121
5122 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5123 return -EINVAL;
5124
5125 /* Values correspoinding to the specific association */
5126 if (asoc) {
5127 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5128 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5129 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5130 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
5131 * 1000) +
5132 (asoc->cookie_life.tv_usec
5133 / 1000);
5134
5135 list_for_each(pos, &asoc->peer.transport_addr_list) {
5136 cnt ++;
5137 }
5138
5139 assocparams.sasoc_number_peer_destinations = cnt;
5140 } else {
5141 /* Values corresponding to the endpoint */
5142 struct sctp_sock *sp = sctp_sk(sk);
5143
5144 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5145 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5146 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5147 assocparams.sasoc_cookie_life =
5148 sp->assocparams.sasoc_cookie_life;
5149 assocparams.sasoc_number_peer_destinations =
5150 sp->assocparams.
5151 sasoc_number_peer_destinations;
5152 }
5153
5154 if (put_user(len, optlen))
5155 return -EFAULT;
5156
5157 if (copy_to_user(optval, &assocparams, len))
5158 return -EFAULT;
5159
5160 return 0;
5161 }
5162
5163 /*
5164 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5165 *
5166 * This socket option is a boolean flag which turns on or off mapped V4
5167 * addresses. If this option is turned on and the socket is type
5168 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5169 * If this option is turned off, then no mapping will be done of V4
5170 * addresses and a user will receive both PF_INET6 and PF_INET type
5171 * addresses on the socket.
5172 */
5173 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5174 char __user *optval, int __user *optlen)
5175 {
5176 int val;
5177 struct sctp_sock *sp = sctp_sk(sk);
5178
5179 if (len < sizeof(int))
5180 return -EINVAL;
5181
5182 len = sizeof(int);
5183 val = sp->v4mapped;
5184 if (put_user(len, optlen))
5185 return -EFAULT;
5186 if (copy_to_user(optval, &val, len))
5187 return -EFAULT;
5188
5189 return 0;
5190 }
5191
5192 /*
5193 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5194 * (chapter and verse is quoted at sctp_setsockopt_context())
5195 */
5196 static int sctp_getsockopt_context(struct sock *sk, int len,
5197 char __user *optval, int __user *optlen)
5198 {
5199 struct sctp_assoc_value params;
5200 struct sctp_sock *sp;
5201 struct sctp_association *asoc;
5202
5203 if (len < sizeof(struct sctp_assoc_value))
5204 return -EINVAL;
5205
5206 len = sizeof(struct sctp_assoc_value);
5207
5208 if (copy_from_user(&params, optval, len))
5209 return -EFAULT;
5210
5211 sp = sctp_sk(sk);
5212
5213 if (params.assoc_id != 0) {
5214 asoc = sctp_id2assoc(sk, params.assoc_id);
5215 if (!asoc)
5216 return -EINVAL;
5217 params.assoc_value = asoc->default_rcv_context;
5218 } else {
5219 params.assoc_value = sp->default_rcv_context;
5220 }
5221
5222 if (put_user(len, optlen))
5223 return -EFAULT;
5224 if (copy_to_user(optval, &params, len))
5225 return -EFAULT;
5226
5227 return 0;
5228 }
5229
5230 /*
5231 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5232 * This option will get or set the maximum size to put in any outgoing
5233 * SCTP DATA chunk. If a message is larger than this size it will be
5234 * fragmented by SCTP into the specified size. Note that the underlying
5235 * SCTP implementation may fragment into smaller sized chunks when the
5236 * PMTU of the underlying association is smaller than the value set by
5237 * the user. The default value for this option is '0' which indicates
5238 * the user is NOT limiting fragmentation and only the PMTU will effect
5239 * SCTP's choice of DATA chunk size. Note also that values set larger
5240 * than the maximum size of an IP datagram will effectively let SCTP
5241 * control fragmentation (i.e. the same as setting this option to 0).
5242 *
5243 * The following structure is used to access and modify this parameter:
5244 *
5245 * struct sctp_assoc_value {
5246 * sctp_assoc_t assoc_id;
5247 * uint32_t assoc_value;
5248 * };
5249 *
5250 * assoc_id: This parameter is ignored for one-to-one style sockets.
5251 * For one-to-many style sockets this parameter indicates which
5252 * association the user is performing an action upon. Note that if
5253 * this field's value is zero then the endpoints default value is
5254 * changed (effecting future associations only).
5255 * assoc_value: This parameter specifies the maximum size in bytes.
5256 */
5257 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5258 char __user *optval, int __user *optlen)
5259 {
5260 struct sctp_assoc_value params;
5261 struct sctp_association *asoc;
5262
5263 if (len == sizeof(int)) {
5264 pr_warn("Use of int in maxseg socket option deprecated\n");
5265 pr_warn("Use struct sctp_assoc_value instead\n");
5266 params.assoc_id = 0;
5267 } else if (len >= sizeof(struct sctp_assoc_value)) {
5268 len = sizeof(struct sctp_assoc_value);
5269 if (copy_from_user(&params, optval, sizeof(params)))
5270 return -EFAULT;
5271 } else
5272 return -EINVAL;
5273
5274 asoc = sctp_id2assoc(sk, params.assoc_id);
5275 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5276 return -EINVAL;
5277
5278 if (asoc)
5279 params.assoc_value = asoc->frag_point;
5280 else
5281 params.assoc_value = sctp_sk(sk)->user_frag;
5282
5283 if (put_user(len, optlen))
5284 return -EFAULT;
5285 if (len == sizeof(int)) {
5286 if (copy_to_user(optval, &params.assoc_value, len))
5287 return -EFAULT;
5288 } else {
5289 if (copy_to_user(optval, &params, len))
5290 return -EFAULT;
5291 }
5292
5293 return 0;
5294 }
5295
5296 /*
5297 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5298 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5299 */
5300 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5301 char __user *optval, int __user *optlen)
5302 {
5303 int val;
5304
5305 if (len < sizeof(int))
5306 return -EINVAL;
5307
5308 len = sizeof(int);
5309
5310 val = sctp_sk(sk)->frag_interleave;
5311 if (put_user(len, optlen))
5312 return -EFAULT;
5313 if (copy_to_user(optval, &val, len))
5314 return -EFAULT;
5315
5316 return 0;
5317 }
5318
5319 /*
5320 * 7.1.25. Set or Get the sctp partial delivery point
5321 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5322 */
5323 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5324 char __user *optval,
5325 int __user *optlen)
5326 {
5327 u32 val;
5328
5329 if (len < sizeof(u32))
5330 return -EINVAL;
5331
5332 len = sizeof(u32);
5333
5334 val = sctp_sk(sk)->pd_point;
5335 if (put_user(len, optlen))
5336 return -EFAULT;
5337 if (copy_to_user(optval, &val, len))
5338 return -EFAULT;
5339
5340 return 0;
5341 }
5342
5343 /*
5344 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5345 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5346 */
5347 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5348 char __user *optval,
5349 int __user *optlen)
5350 {
5351 struct sctp_assoc_value params;
5352 struct sctp_sock *sp;
5353 struct sctp_association *asoc;
5354
5355 if (len == sizeof(int)) {
5356 pr_warn("Use of int in max_burst socket option deprecated\n");
5357 pr_warn("Use struct sctp_assoc_value instead\n");
5358 params.assoc_id = 0;
5359 } else if (len >= sizeof(struct sctp_assoc_value)) {
5360 len = sizeof(struct sctp_assoc_value);
5361 if (copy_from_user(&params, optval, len))
5362 return -EFAULT;
5363 } else
5364 return -EINVAL;
5365
5366 sp = sctp_sk(sk);
5367
5368 if (params.assoc_id != 0) {
5369 asoc = sctp_id2assoc(sk, params.assoc_id);
5370 if (!asoc)
5371 return -EINVAL;
5372 params.assoc_value = asoc->max_burst;
5373 } else
5374 params.assoc_value = sp->max_burst;
5375
5376 if (len == sizeof(int)) {
5377 if (copy_to_user(optval, &params.assoc_value, len))
5378 return -EFAULT;
5379 } else {
5380 if (copy_to_user(optval, &params, len))
5381 return -EFAULT;
5382 }
5383
5384 return 0;
5385
5386 }
5387
5388 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5389 char __user *optval, int __user *optlen)
5390 {
5391 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5392 struct sctp_hmacalgo __user *p = (void __user *)optval;
5393 struct sctp_hmac_algo_param *hmacs;
5394 __u16 data_len = 0;
5395 u32 num_idents;
5396
5397 if (!ep->auth_enable)
5398 return -EACCES;
5399
5400 hmacs = ep->auth_hmacs_list;
5401 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5402
5403 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5404 return -EINVAL;
5405
5406 len = sizeof(struct sctp_hmacalgo) + data_len;
5407 num_idents = data_len / sizeof(u16);
5408
5409 if (put_user(len, optlen))
5410 return -EFAULT;
5411 if (put_user(num_idents, &p->shmac_num_idents))
5412 return -EFAULT;
5413 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5414 return -EFAULT;
5415 return 0;
5416 }
5417
5418 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5419 char __user *optval, int __user *optlen)
5420 {
5421 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5422 struct sctp_authkeyid val;
5423 struct sctp_association *asoc;
5424
5425 if (!ep->auth_enable)
5426 return -EACCES;
5427
5428 if (len < sizeof(struct sctp_authkeyid))
5429 return -EINVAL;
5430 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5431 return -EFAULT;
5432
5433 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5434 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5435 return -EINVAL;
5436
5437 if (asoc)
5438 val.scact_keynumber = asoc->active_key_id;
5439 else
5440 val.scact_keynumber = ep->active_key_id;
5441
5442 len = sizeof(struct sctp_authkeyid);
5443 if (put_user(len, optlen))
5444 return -EFAULT;
5445 if (copy_to_user(optval, &val, len))
5446 return -EFAULT;
5447
5448 return 0;
5449 }
5450
5451 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5452 char __user *optval, int __user *optlen)
5453 {
5454 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5455 struct sctp_authchunks __user *p = (void __user *)optval;
5456 struct sctp_authchunks val;
5457 struct sctp_association *asoc;
5458 struct sctp_chunks_param *ch;
5459 u32 num_chunks = 0;
5460 char __user *to;
5461
5462 if (!ep->auth_enable)
5463 return -EACCES;
5464
5465 if (len < sizeof(struct sctp_authchunks))
5466 return -EINVAL;
5467
5468 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5469 return -EFAULT;
5470
5471 to = p->gauth_chunks;
5472 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5473 if (!asoc)
5474 return -EINVAL;
5475
5476 ch = asoc->peer.peer_chunks;
5477 if (!ch)
5478 goto num;
5479
5480 /* See if the user provided enough room for all the data */
5481 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5482 if (len < num_chunks)
5483 return -EINVAL;
5484
5485 if (copy_to_user(to, ch->chunks, num_chunks))
5486 return -EFAULT;
5487 num:
5488 len = sizeof(struct sctp_authchunks) + num_chunks;
5489 if (put_user(len, optlen)) return -EFAULT;
5490 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5491 return -EFAULT;
5492 return 0;
5493 }
5494
5495 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5496 char __user *optval, int __user *optlen)
5497 {
5498 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5499 struct sctp_authchunks __user *p = (void __user *)optval;
5500 struct sctp_authchunks val;
5501 struct sctp_association *asoc;
5502 struct sctp_chunks_param *ch;
5503 u32 num_chunks = 0;
5504 char __user *to;
5505
5506 if (!ep->auth_enable)
5507 return -EACCES;
5508
5509 if (len < sizeof(struct sctp_authchunks))
5510 return -EINVAL;
5511
5512 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5513 return -EFAULT;
5514
5515 to = p->gauth_chunks;
5516 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5517 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5518 return -EINVAL;
5519
5520 if (asoc)
5521 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5522 else
5523 ch = ep->auth_chunk_list;
5524
5525 if (!ch)
5526 goto num;
5527
5528 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5529 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5530 return -EINVAL;
5531
5532 if (copy_to_user(to, ch->chunks, num_chunks))
5533 return -EFAULT;
5534 num:
5535 len = sizeof(struct sctp_authchunks) + num_chunks;
5536 if (put_user(len, optlen))
5537 return -EFAULT;
5538 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5539 return -EFAULT;
5540
5541 return 0;
5542 }
5543
5544 /*
5545 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5546 * This option gets the current number of associations that are attached
5547 * to a one-to-many style socket. The option value is an uint32_t.
5548 */
5549 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5550 char __user *optval, int __user *optlen)
5551 {
5552 struct sctp_sock *sp = sctp_sk(sk);
5553 struct sctp_association *asoc;
5554 u32 val = 0;
5555
5556 if (sctp_style(sk, TCP))
5557 return -EOPNOTSUPP;
5558
5559 if (len < sizeof(u32))
5560 return -EINVAL;
5561
5562 len = sizeof(u32);
5563
5564 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5565 val++;
5566 }
5567
5568 if (put_user(len, optlen))
5569 return -EFAULT;
5570 if (copy_to_user(optval, &val, len))
5571 return -EFAULT;
5572
5573 return 0;
5574 }
5575
5576 /*
5577 * 8.1.23 SCTP_AUTO_ASCONF
5578 * See the corresponding setsockopt entry as description
5579 */
5580 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5581 char __user *optval, int __user *optlen)
5582 {
5583 int val = 0;
5584
5585 if (len < sizeof(int))
5586 return -EINVAL;
5587
5588 len = sizeof(int);
5589 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5590 val = 1;
5591 if (put_user(len, optlen))
5592 return -EFAULT;
5593 if (copy_to_user(optval, &val, len))
5594 return -EFAULT;
5595 return 0;
5596 }
5597
5598 /*
5599 * 8.2.6. Get the Current Identifiers of Associations
5600 * (SCTP_GET_ASSOC_ID_LIST)
5601 *
5602 * This option gets the current list of SCTP association identifiers of
5603 * the SCTP associations handled by a one-to-many style socket.
5604 */
5605 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5606 char __user *optval, int __user *optlen)
5607 {
5608 struct sctp_sock *sp = sctp_sk(sk);
5609 struct sctp_association *asoc;
5610 struct sctp_assoc_ids *ids;
5611 u32 num = 0;
5612
5613 if (sctp_style(sk, TCP))
5614 return -EOPNOTSUPP;
5615
5616 if (len < sizeof(struct sctp_assoc_ids))
5617 return -EINVAL;
5618
5619 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5620 num++;
5621 }
5622
5623 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5624 return -EINVAL;
5625
5626 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5627
5628 ids = kmalloc(len, GFP_KERNEL);
5629 if (unlikely(!ids))
5630 return -ENOMEM;
5631
5632 ids->gaids_number_of_ids = num;
5633 num = 0;
5634 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5635 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5636 }
5637
5638 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5639 kfree(ids);
5640 return -EFAULT;
5641 }
5642
5643 kfree(ids);
5644 return 0;
5645 }
5646
5647 /*
5648 * SCTP_PEER_ADDR_THLDS
5649 *
5650 * This option allows us to fetch the partially failed threshold for one or all
5651 * transports in an association. See Section 6.1 of:
5652 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5653 */
5654 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
5655 char __user *optval,
5656 int len,
5657 int __user *optlen)
5658 {
5659 struct sctp_paddrthlds val;
5660 struct sctp_transport *trans;
5661 struct sctp_association *asoc;
5662
5663 if (len < sizeof(struct sctp_paddrthlds))
5664 return -EINVAL;
5665 len = sizeof(struct sctp_paddrthlds);
5666 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
5667 return -EFAULT;
5668
5669 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
5670 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
5671 if (!asoc)
5672 return -ENOENT;
5673
5674 val.spt_pathpfthld = asoc->pf_retrans;
5675 val.spt_pathmaxrxt = asoc->pathmaxrxt;
5676 } else {
5677 trans = sctp_addr_id2transport(sk, &val.spt_address,
5678 val.spt_assoc_id);
5679 if (!trans)
5680 return -ENOENT;
5681
5682 val.spt_pathmaxrxt = trans->pathmaxrxt;
5683 val.spt_pathpfthld = trans->pf_retrans;
5684 }
5685
5686 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
5687 return -EFAULT;
5688
5689 return 0;
5690 }
5691
5692 /*
5693 * SCTP_GET_ASSOC_STATS
5694 *
5695 * This option retrieves local per endpoint statistics. It is modeled
5696 * after OpenSolaris' implementation
5697 */
5698 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
5699 char __user *optval,
5700 int __user *optlen)
5701 {
5702 struct sctp_assoc_stats sas;
5703 struct sctp_association *asoc = NULL;
5704
5705 /* User must provide at least the assoc id */
5706 if (len < sizeof(sctp_assoc_t))
5707 return -EINVAL;
5708
5709 /* Allow the struct to grow and fill in as much as possible */
5710 len = min_t(size_t, len, sizeof(sas));
5711
5712 if (copy_from_user(&sas, optval, len))
5713 return -EFAULT;
5714
5715 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
5716 if (!asoc)
5717 return -EINVAL;
5718
5719 sas.sas_rtxchunks = asoc->stats.rtxchunks;
5720 sas.sas_gapcnt = asoc->stats.gapcnt;
5721 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
5722 sas.sas_osacks = asoc->stats.osacks;
5723 sas.sas_isacks = asoc->stats.isacks;
5724 sas.sas_octrlchunks = asoc->stats.octrlchunks;
5725 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
5726 sas.sas_oodchunks = asoc->stats.oodchunks;
5727 sas.sas_iodchunks = asoc->stats.iodchunks;
5728 sas.sas_ouodchunks = asoc->stats.ouodchunks;
5729 sas.sas_iuodchunks = asoc->stats.iuodchunks;
5730 sas.sas_idupchunks = asoc->stats.idupchunks;
5731 sas.sas_opackets = asoc->stats.opackets;
5732 sas.sas_ipackets = asoc->stats.ipackets;
5733
5734 /* New high max rto observed, will return 0 if not a single
5735 * RTO update took place. obs_rto_ipaddr will be bogus
5736 * in such a case
5737 */
5738 sas.sas_maxrto = asoc->stats.max_obs_rto;
5739 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
5740 sizeof(struct sockaddr_storage));
5741
5742 /* Mark beginning of a new observation period */
5743 asoc->stats.max_obs_rto = asoc->rto_min;
5744
5745 if (put_user(len, optlen))
5746 return -EFAULT;
5747
5748 SCTP_DEBUG_PRINTK("sctp_getsockopt_assoc_stat(%d): %d\n",
5749 len, sas.sas_assoc_id);
5750
5751 if (copy_to_user(optval, &sas, len))
5752 return -EFAULT;
5753
5754 return 0;
5755 }
5756
5757 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5758 char __user *optval, int __user *optlen)
5759 {
5760 int retval = 0;
5761 int len;
5762
5763 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5764 sk, optname);
5765
5766 /* I can hardly begin to describe how wrong this is. This is
5767 * so broken as to be worse than useless. The API draft
5768 * REALLY is NOT helpful here... I am not convinced that the
5769 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5770 * are at all well-founded.
5771 */
5772 if (level != SOL_SCTP) {
5773 struct sctp_af *af = sctp_sk(sk)->pf->af;
5774
5775 retval = af->getsockopt(sk, level, optname, optval, optlen);
5776 return retval;
5777 }
5778
5779 if (get_user(len, optlen))
5780 return -EFAULT;
5781
5782 if (len < 0)
5783 return -EINVAL;
5784
5785 sctp_lock_sock(sk);
5786
5787 switch (optname) {
5788 case SCTP_STATUS:
5789 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5790 break;
5791 case SCTP_DISABLE_FRAGMENTS:
5792 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5793 optlen);
5794 break;
5795 case SCTP_EVENTS:
5796 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5797 break;
5798 case SCTP_AUTOCLOSE:
5799 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5800 break;
5801 case SCTP_SOCKOPT_PEELOFF:
5802 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5803 break;
5804 case SCTP_PEER_ADDR_PARAMS:
5805 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5806 optlen);
5807 break;
5808 case SCTP_DELAYED_SACK:
5809 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5810 optlen);
5811 break;
5812 case SCTP_INITMSG:
5813 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5814 break;
5815 case SCTP_GET_PEER_ADDRS:
5816 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5817 optlen);
5818 break;
5819 case SCTP_GET_LOCAL_ADDRS:
5820 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5821 optlen);
5822 break;
5823 case SCTP_SOCKOPT_CONNECTX3:
5824 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5825 break;
5826 case SCTP_DEFAULT_SEND_PARAM:
5827 retval = sctp_getsockopt_default_send_param(sk, len,
5828 optval, optlen);
5829 break;
5830 case SCTP_PRIMARY_ADDR:
5831 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5832 break;
5833 case SCTP_NODELAY:
5834 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5835 break;
5836 case SCTP_RTOINFO:
5837 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5838 break;
5839 case SCTP_ASSOCINFO:
5840 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5841 break;
5842 case SCTP_I_WANT_MAPPED_V4_ADDR:
5843 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5844 break;
5845 case SCTP_MAXSEG:
5846 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5847 break;
5848 case SCTP_GET_PEER_ADDR_INFO:
5849 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5850 optlen);
5851 break;
5852 case SCTP_ADAPTATION_LAYER:
5853 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5854 optlen);
5855 break;
5856 case SCTP_CONTEXT:
5857 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5858 break;
5859 case SCTP_FRAGMENT_INTERLEAVE:
5860 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5861 optlen);
5862 break;
5863 case SCTP_PARTIAL_DELIVERY_POINT:
5864 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5865 optlen);
5866 break;
5867 case SCTP_MAX_BURST:
5868 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5869 break;
5870 case SCTP_AUTH_KEY:
5871 case SCTP_AUTH_CHUNK:
5872 case SCTP_AUTH_DELETE_KEY:
5873 retval = -EOPNOTSUPP;
5874 break;
5875 case SCTP_HMAC_IDENT:
5876 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5877 break;
5878 case SCTP_AUTH_ACTIVE_KEY:
5879 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5880 break;
5881 case SCTP_PEER_AUTH_CHUNKS:
5882 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5883 optlen);
5884 break;
5885 case SCTP_LOCAL_AUTH_CHUNKS:
5886 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5887 optlen);
5888 break;
5889 case SCTP_GET_ASSOC_NUMBER:
5890 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5891 break;
5892 case SCTP_GET_ASSOC_ID_LIST:
5893 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5894 break;
5895 case SCTP_AUTO_ASCONF:
5896 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5897 break;
5898 case SCTP_PEER_ADDR_THLDS:
5899 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
5900 break;
5901 case SCTP_GET_ASSOC_STATS:
5902 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
5903 break;
5904 default:
5905 retval = -ENOPROTOOPT;
5906 break;
5907 }
5908
5909 sctp_release_sock(sk);
5910 return retval;
5911 }
5912
5913 static void sctp_hash(struct sock *sk)
5914 {
5915 /* STUB */
5916 }
5917
5918 static void sctp_unhash(struct sock *sk)
5919 {
5920 /* STUB */
5921 }
5922
5923 /* Check if port is acceptable. Possibly find first available port.
5924 *
5925 * The port hash table (contained in the 'global' SCTP protocol storage
5926 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5927 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5928 * list (the list number is the port number hashed out, so as you
5929 * would expect from a hash function, all the ports in a given list have
5930 * such a number that hashes out to the same list number; you were
5931 * expecting that, right?); so each list has a set of ports, with a
5932 * link to the socket (struct sock) that uses it, the port number and
5933 * a fastreuse flag (FIXME: NPI ipg).
5934 */
5935 static struct sctp_bind_bucket *sctp_bucket_create(
5936 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
5937
5938 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5939 {
5940 struct sctp_bind_hashbucket *head; /* hash list */
5941 struct sctp_bind_bucket *pp;
5942 unsigned short snum;
5943 int ret;
5944
5945 snum = ntohs(addr->v4.sin_port);
5946
5947 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5948 sctp_local_bh_disable();
5949
5950 if (snum == 0) {
5951 /* Search for an available port. */
5952 int low, high, remaining, index;
5953 unsigned int rover;
5954
5955 inet_get_local_port_range(&low, &high);
5956 remaining = (high - low) + 1;
5957 rover = net_random() % remaining + low;
5958
5959 do {
5960 rover++;
5961 if ((rover < low) || (rover > high))
5962 rover = low;
5963 if (inet_is_reserved_local_port(rover))
5964 continue;
5965 index = sctp_phashfn(sock_net(sk), rover);
5966 head = &sctp_port_hashtable[index];
5967 sctp_spin_lock(&head->lock);
5968 sctp_for_each_hentry(pp, &head->chain)
5969 if ((pp->port == rover) &&
5970 net_eq(sock_net(sk), pp->net))
5971 goto next;
5972 break;
5973 next:
5974 sctp_spin_unlock(&head->lock);
5975 } while (--remaining > 0);
5976
5977 /* Exhausted local port range during search? */
5978 ret = 1;
5979 if (remaining <= 0)
5980 goto fail;
5981
5982 /* OK, here is the one we will use. HEAD (the port
5983 * hash table list entry) is non-NULL and we hold it's
5984 * mutex.
5985 */
5986 snum = rover;
5987 } else {
5988 /* We are given an specific port number; we verify
5989 * that it is not being used. If it is used, we will
5990 * exahust the search in the hash list corresponding
5991 * to the port number (snum) - we detect that with the
5992 * port iterator, pp being NULL.
5993 */
5994 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
5995 sctp_spin_lock(&head->lock);
5996 sctp_for_each_hentry(pp, &head->chain) {
5997 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
5998 goto pp_found;
5999 }
6000 }
6001 pp = NULL;
6002 goto pp_not_found;
6003 pp_found:
6004 if (!hlist_empty(&pp->owner)) {
6005 /* We had a port hash table hit - there is an
6006 * available port (pp != NULL) and it is being
6007 * used by other socket (pp->owner not empty); that other
6008 * socket is going to be sk2.
6009 */
6010 int reuse = sk->sk_reuse;
6011 struct sock *sk2;
6012
6013 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
6014 if (pp->fastreuse && sk->sk_reuse &&
6015 sk->sk_state != SCTP_SS_LISTENING)
6016 goto success;
6017
6018 /* Run through the list of sockets bound to the port
6019 * (pp->port) [via the pointers bind_next and
6020 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
6021 * we get the endpoint they describe and run through
6022 * the endpoint's list of IP (v4 or v6) addresses,
6023 * comparing each of the addresses with the address of
6024 * the socket sk. If we find a match, then that means
6025 * that this port/socket (sk) combination are already
6026 * in an endpoint.
6027 */
6028 sk_for_each_bound(sk2, &pp->owner) {
6029 struct sctp_endpoint *ep2;
6030 ep2 = sctp_sk(sk2)->ep;
6031
6032 if (sk == sk2 ||
6033 (reuse && sk2->sk_reuse &&
6034 sk2->sk_state != SCTP_SS_LISTENING))
6035 continue;
6036
6037 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
6038 sctp_sk(sk2), sctp_sk(sk))) {
6039 ret = (long)sk2;
6040 goto fail_unlock;
6041 }
6042 }
6043 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
6044 }
6045 pp_not_found:
6046 /* If there was a hash table miss, create a new port. */
6047 ret = 1;
6048 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
6049 goto fail_unlock;
6050
6051 /* In either case (hit or miss), make sure fastreuse is 1 only
6052 * if sk->sk_reuse is too (that is, if the caller requested
6053 * SO_REUSEADDR on this socket -sk-).
6054 */
6055 if (hlist_empty(&pp->owner)) {
6056 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
6057 pp->fastreuse = 1;
6058 else
6059 pp->fastreuse = 0;
6060 } else if (pp->fastreuse &&
6061 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6062 pp->fastreuse = 0;
6063
6064 /* We are set, so fill up all the data in the hash table
6065 * entry, tie the socket list information with the rest of the
6066 * sockets FIXME: Blurry, NPI (ipg).
6067 */
6068 success:
6069 if (!sctp_sk(sk)->bind_hash) {
6070 inet_sk(sk)->inet_num = snum;
6071 sk_add_bind_node(sk, &pp->owner);
6072 sctp_sk(sk)->bind_hash = pp;
6073 }
6074 ret = 0;
6075
6076 fail_unlock:
6077 sctp_spin_unlock(&head->lock);
6078
6079 fail:
6080 sctp_local_bh_enable();
6081 return ret;
6082 }
6083
6084 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6085 * port is requested.
6086 */
6087 static int sctp_get_port(struct sock *sk, unsigned short snum)
6088 {
6089 long ret;
6090 union sctp_addr addr;
6091 struct sctp_af *af = sctp_sk(sk)->pf->af;
6092
6093 /* Set up a dummy address struct from the sk. */
6094 af->from_sk(&addr, sk);
6095 addr.v4.sin_port = htons(snum);
6096
6097 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6098 ret = sctp_get_port_local(sk, &addr);
6099
6100 return ret ? 1 : 0;
6101 }
6102
6103 /*
6104 * Move a socket to LISTENING state.
6105 */
6106 SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
6107 {
6108 struct sctp_sock *sp = sctp_sk(sk);
6109 struct sctp_endpoint *ep = sp->ep;
6110 struct crypto_hash *tfm = NULL;
6111 char alg[32];
6112
6113 /* Allocate HMAC for generating cookie. */
6114 if (!sp->hmac && sp->sctp_hmac_alg) {
6115 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6116 tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC);
6117 if (IS_ERR(tfm)) {
6118 net_info_ratelimited("failed to load transform for %s: %ld\n",
6119 sp->sctp_hmac_alg, PTR_ERR(tfm));
6120 return -ENOSYS;
6121 }
6122 sctp_sk(sk)->hmac = tfm;
6123 }
6124
6125 /*
6126 * If a bind() or sctp_bindx() is not called prior to a listen()
6127 * call that allows new associations to be accepted, the system
6128 * picks an ephemeral port and will choose an address set equivalent
6129 * to binding with a wildcard address.
6130 *
6131 * This is not currently spelled out in the SCTP sockets
6132 * extensions draft, but follows the practice as seen in TCP
6133 * sockets.
6134 *
6135 */
6136 sk->sk_state = SCTP_SS_LISTENING;
6137 if (!ep->base.bind_addr.port) {
6138 if (sctp_autobind(sk))
6139 return -EAGAIN;
6140 } else {
6141 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6142 sk->sk_state = SCTP_SS_CLOSED;
6143 return -EADDRINUSE;
6144 }
6145 }
6146
6147 sk->sk_max_ack_backlog = backlog;
6148 sctp_hash_endpoint(ep);
6149 return 0;
6150 }
6151
6152 /*
6153 * 4.1.3 / 5.1.3 listen()
6154 *
6155 * By default, new associations are not accepted for UDP style sockets.
6156 * An application uses listen() to mark a socket as being able to
6157 * accept new associations.
6158 *
6159 * On TCP style sockets, applications use listen() to ready the SCTP
6160 * endpoint for accepting inbound associations.
6161 *
6162 * On both types of endpoints a backlog of '0' disables listening.
6163 *
6164 * Move a socket to LISTENING state.
6165 */
6166 int sctp_inet_listen(struct socket *sock, int backlog)
6167 {
6168 struct sock *sk = sock->sk;
6169 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6170 int err = -EINVAL;
6171
6172 if (unlikely(backlog < 0))
6173 return err;
6174
6175 sctp_lock_sock(sk);
6176
6177 /* Peeled-off sockets are not allowed to listen(). */
6178 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6179 goto out;
6180
6181 if (sock->state != SS_UNCONNECTED)
6182 goto out;
6183
6184 if (!sctp_sstate(sk, LISTENING) && !sctp_sstate(sk, CLOSED))
6185 goto out;
6186
6187 /* If backlog is zero, disable listening. */
6188 if (!backlog) {
6189 if (sctp_sstate(sk, CLOSED))
6190 goto out;
6191
6192 err = 0;
6193 sctp_unhash_endpoint(ep);
6194 sk->sk_state = SCTP_SS_CLOSED;
6195 if (sk->sk_reuse)
6196 sctp_sk(sk)->bind_hash->fastreuse = 1;
6197 goto out;
6198 }
6199
6200 /* If we are already listening, just update the backlog */
6201 if (sctp_sstate(sk, LISTENING))
6202 sk->sk_max_ack_backlog = backlog;
6203 else {
6204 err = sctp_listen_start(sk, backlog);
6205 if (err)
6206 goto out;
6207 }
6208
6209 err = 0;
6210 out:
6211 sctp_release_sock(sk);
6212 return err;
6213 }
6214
6215 /*
6216 * This function is done by modeling the current datagram_poll() and the
6217 * tcp_poll(). Note that, based on these implementations, we don't
6218 * lock the socket in this function, even though it seems that,
6219 * ideally, locking or some other mechanisms can be used to ensure
6220 * the integrity of the counters (sndbuf and wmem_alloc) used
6221 * in this place. We assume that we don't need locks either until proven
6222 * otherwise.
6223 *
6224 * Another thing to note is that we include the Async I/O support
6225 * here, again, by modeling the current TCP/UDP code. We don't have
6226 * a good way to test with it yet.
6227 */
6228 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6229 {
6230 struct sock *sk = sock->sk;
6231 struct sctp_sock *sp = sctp_sk(sk);
6232 unsigned int mask;
6233
6234 poll_wait(file, sk_sleep(sk), wait);
6235
6236 /* A TCP-style listening socket becomes readable when the accept queue
6237 * is not empty.
6238 */
6239 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6240 return (!list_empty(&sp->ep->asocs)) ?
6241 (POLLIN | POLLRDNORM) : 0;
6242
6243 mask = 0;
6244
6245 /* Is there any exceptional events? */
6246 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6247 mask |= POLLERR |
6248 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
6249 if (sk->sk_shutdown & RCV_SHUTDOWN)
6250 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6251 if (sk->sk_shutdown == SHUTDOWN_MASK)
6252 mask |= POLLHUP;
6253
6254 /* Is it readable? Reconsider this code with TCP-style support. */
6255 if (!skb_queue_empty(&sk->sk_receive_queue))
6256 mask |= POLLIN | POLLRDNORM;
6257
6258 /* The association is either gone or not ready. */
6259 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6260 return mask;
6261
6262 /* Is it writable? */
6263 if (sctp_writeable(sk)) {
6264 mask |= POLLOUT | POLLWRNORM;
6265 } else {
6266 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6267 /*
6268 * Since the socket is not locked, the buffer
6269 * might be made available after the writeable check and
6270 * before the bit is set. This could cause a lost I/O
6271 * signal. tcp_poll() has a race breaker for this race
6272 * condition. Based on their implementation, we put
6273 * in the following code to cover it as well.
6274 */
6275 if (sctp_writeable(sk))
6276 mask |= POLLOUT | POLLWRNORM;
6277 }
6278 return mask;
6279 }
6280
6281 /********************************************************************
6282 * 2nd Level Abstractions
6283 ********************************************************************/
6284
6285 static struct sctp_bind_bucket *sctp_bucket_create(
6286 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6287 {
6288 struct sctp_bind_bucket *pp;
6289
6290 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6291 if (pp) {
6292 SCTP_DBG_OBJCNT_INC(bind_bucket);
6293 pp->port = snum;
6294 pp->fastreuse = 0;
6295 INIT_HLIST_HEAD(&pp->owner);
6296 pp->net = net;
6297 hlist_add_head(&pp->node, &head->chain);
6298 }
6299 return pp;
6300 }
6301
6302 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6303 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6304 {
6305 if (pp && hlist_empty(&pp->owner)) {
6306 __hlist_del(&pp->node);
6307 kmem_cache_free(sctp_bucket_cachep, pp);
6308 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6309 }
6310 }
6311
6312 /* Release this socket's reference to a local port. */
6313 static inline void __sctp_put_port(struct sock *sk)
6314 {
6315 struct sctp_bind_hashbucket *head =
6316 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6317 inet_sk(sk)->inet_num)];
6318 struct sctp_bind_bucket *pp;
6319
6320 sctp_spin_lock(&head->lock);
6321 pp = sctp_sk(sk)->bind_hash;
6322 __sk_del_bind_node(sk);
6323 sctp_sk(sk)->bind_hash = NULL;
6324 inet_sk(sk)->inet_num = 0;
6325 sctp_bucket_destroy(pp);
6326 sctp_spin_unlock(&head->lock);
6327 }
6328
6329 void sctp_put_port(struct sock *sk)
6330 {
6331 sctp_local_bh_disable();
6332 __sctp_put_port(sk);
6333 sctp_local_bh_enable();
6334 }
6335
6336 /*
6337 * The system picks an ephemeral port and choose an address set equivalent
6338 * to binding with a wildcard address.
6339 * One of those addresses will be the primary address for the association.
6340 * This automatically enables the multihoming capability of SCTP.
6341 */
6342 static int sctp_autobind(struct sock *sk)
6343 {
6344 union sctp_addr autoaddr;
6345 struct sctp_af *af;
6346 __be16 port;
6347
6348 /* Initialize a local sockaddr structure to INADDR_ANY. */
6349 af = sctp_sk(sk)->pf->af;
6350
6351 port = htons(inet_sk(sk)->inet_num);
6352 af->inaddr_any(&autoaddr, port);
6353
6354 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6355 }
6356
6357 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6358 *
6359 * From RFC 2292
6360 * 4.2 The cmsghdr Structure *
6361 *
6362 * When ancillary data is sent or received, any number of ancillary data
6363 * objects can be specified by the msg_control and msg_controllen members of
6364 * the msghdr structure, because each object is preceded by
6365 * a cmsghdr structure defining the object's length (the cmsg_len member).
6366 * Historically Berkeley-derived implementations have passed only one object
6367 * at a time, but this API allows multiple objects to be
6368 * passed in a single call to sendmsg() or recvmsg(). The following example
6369 * shows two ancillary data objects in a control buffer.
6370 *
6371 * |<--------------------------- msg_controllen -------------------------->|
6372 * | |
6373 *
6374 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6375 *
6376 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6377 * | | |
6378 *
6379 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6380 *
6381 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6382 * | | | | |
6383 *
6384 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6385 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6386 *
6387 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6388 *
6389 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6390 * ^
6391 * |
6392 *
6393 * msg_control
6394 * points here
6395 */
6396 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
6397 sctp_cmsgs_t *cmsgs)
6398 {
6399 struct cmsghdr *cmsg;
6400 struct msghdr *my_msg = (struct msghdr *)msg;
6401
6402 for (cmsg = CMSG_FIRSTHDR(msg);
6403 cmsg != NULL;
6404 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6405 if (!CMSG_OK(my_msg, cmsg))
6406 return -EINVAL;
6407
6408 /* Should we parse this header or ignore? */
6409 if (cmsg->cmsg_level != IPPROTO_SCTP)
6410 continue;
6411
6412 /* Strictly check lengths following example in SCM code. */
6413 switch (cmsg->cmsg_type) {
6414 case SCTP_INIT:
6415 /* SCTP Socket API Extension
6416 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6417 *
6418 * This cmsghdr structure provides information for
6419 * initializing new SCTP associations with sendmsg().
6420 * The SCTP_INITMSG socket option uses this same data
6421 * structure. This structure is not used for
6422 * recvmsg().
6423 *
6424 * cmsg_level cmsg_type cmsg_data[]
6425 * ------------ ------------ ----------------------
6426 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6427 */
6428 if (cmsg->cmsg_len !=
6429 CMSG_LEN(sizeof(struct sctp_initmsg)))
6430 return -EINVAL;
6431 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6432 break;
6433
6434 case SCTP_SNDRCV:
6435 /* SCTP Socket API Extension
6436 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6437 *
6438 * This cmsghdr structure specifies SCTP options for
6439 * sendmsg() and describes SCTP header information
6440 * about a received message through recvmsg().
6441 *
6442 * cmsg_level cmsg_type cmsg_data[]
6443 * ------------ ------------ ----------------------
6444 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6445 */
6446 if (cmsg->cmsg_len !=
6447 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6448 return -EINVAL;
6449
6450 cmsgs->info =
6451 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6452
6453 /* Minimally, validate the sinfo_flags. */
6454 if (cmsgs->info->sinfo_flags &
6455 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6456 SCTP_ABORT | SCTP_EOF))
6457 return -EINVAL;
6458 break;
6459
6460 default:
6461 return -EINVAL;
6462 }
6463 }
6464 return 0;
6465 }
6466
6467 /*
6468 * Wait for a packet..
6469 * Note: This function is the same function as in core/datagram.c
6470 * with a few modifications to make lksctp work.
6471 */
6472 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6473 {
6474 int error;
6475 DEFINE_WAIT(wait);
6476
6477 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6478
6479 /* Socket errors? */
6480 error = sock_error(sk);
6481 if (error)
6482 goto out;
6483
6484 if (!skb_queue_empty(&sk->sk_receive_queue))
6485 goto ready;
6486
6487 /* Socket shut down? */
6488 if (sk->sk_shutdown & RCV_SHUTDOWN)
6489 goto out;
6490
6491 /* Sequenced packets can come disconnected. If so we report the
6492 * problem.
6493 */
6494 error = -ENOTCONN;
6495
6496 /* Is there a good reason to think that we may receive some data? */
6497 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6498 goto out;
6499
6500 /* Handle signals. */
6501 if (signal_pending(current))
6502 goto interrupted;
6503
6504 /* Let another process have a go. Since we are going to sleep
6505 * anyway. Note: This may cause odd behaviors if the message
6506 * does not fit in the user's buffer, but this seems to be the
6507 * only way to honor MSG_DONTWAIT realistically.
6508 */
6509 sctp_release_sock(sk);
6510 *timeo_p = schedule_timeout(*timeo_p);
6511 sctp_lock_sock(sk);
6512
6513 ready:
6514 finish_wait(sk_sleep(sk), &wait);
6515 return 0;
6516
6517 interrupted:
6518 error = sock_intr_errno(*timeo_p);
6519
6520 out:
6521 finish_wait(sk_sleep(sk), &wait);
6522 *err = error;
6523 return error;
6524 }
6525
6526 /* Receive a datagram.
6527 * Note: This is pretty much the same routine as in core/datagram.c
6528 * with a few changes to make lksctp work.
6529 */
6530 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6531 int noblock, int *err)
6532 {
6533 int error;
6534 struct sk_buff *skb;
6535 long timeo;
6536
6537 timeo = sock_rcvtimeo(sk, noblock);
6538
6539 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6540 timeo, MAX_SCHEDULE_TIMEOUT);
6541
6542 do {
6543 /* Again only user level code calls this function,
6544 * so nothing interrupt level
6545 * will suddenly eat the receive_queue.
6546 *
6547 * Look at current nfs client by the way...
6548 * However, this function was correct in any case. 8)
6549 */
6550 if (flags & MSG_PEEK) {
6551 spin_lock_bh(&sk->sk_receive_queue.lock);
6552 skb = skb_peek(&sk->sk_receive_queue);
6553 if (skb)
6554 atomic_inc(&skb->users);
6555 spin_unlock_bh(&sk->sk_receive_queue.lock);
6556 } else {
6557 skb = skb_dequeue(&sk->sk_receive_queue);
6558 }
6559
6560 if (skb)
6561 return skb;
6562
6563 /* Caller is allowed not to check sk->sk_err before calling. */
6564 error = sock_error(sk);
6565 if (error)
6566 goto no_packet;
6567
6568 if (sk->sk_shutdown & RCV_SHUTDOWN)
6569 break;
6570
6571 /* User doesn't want to wait. */
6572 error = -EAGAIN;
6573 if (!timeo)
6574 goto no_packet;
6575 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6576
6577 return NULL;
6578
6579 no_packet:
6580 *err = error;
6581 return NULL;
6582 }
6583
6584 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6585 static void __sctp_write_space(struct sctp_association *asoc)
6586 {
6587 struct sock *sk = asoc->base.sk;
6588 struct socket *sock = sk->sk_socket;
6589
6590 if ((sctp_wspace(asoc) > 0) && sock) {
6591 if (waitqueue_active(&asoc->wait))
6592 wake_up_interruptible(&asoc->wait);
6593
6594 if (sctp_writeable(sk)) {
6595 wait_queue_head_t *wq = sk_sleep(sk);
6596
6597 if (wq && waitqueue_active(wq))
6598 wake_up_interruptible(wq);
6599
6600 /* Note that we try to include the Async I/O support
6601 * here by modeling from the current TCP/UDP code.
6602 * We have not tested with it yet.
6603 */
6604 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6605 sock_wake_async(sock,
6606 SOCK_WAKE_SPACE, POLL_OUT);
6607 }
6608 }
6609 }
6610
6611 static void sctp_wake_up_waiters(struct sock *sk,
6612 struct sctp_association *asoc)
6613 {
6614 struct sctp_association *tmp = asoc;
6615
6616 /* We do accounting for the sndbuf space per association,
6617 * so we only need to wake our own association.
6618 */
6619 if (asoc->ep->sndbuf_policy)
6620 return __sctp_write_space(asoc);
6621
6622 /* If association goes down and is just flushing its
6623 * outq, then just normally notify others.
6624 */
6625 if (asoc->base.dead)
6626 return sctp_write_space(sk);
6627
6628 /* Accounting for the sndbuf space is per socket, so we
6629 * need to wake up others, try to be fair and in case of
6630 * other associations, let them have a go first instead
6631 * of just doing a sctp_write_space() call.
6632 *
6633 * Note that we reach sctp_wake_up_waiters() only when
6634 * associations free up queued chunks, thus we are under
6635 * lock and the list of associations on a socket is
6636 * guaranteed not to change.
6637 */
6638 for (tmp = list_next_entry(tmp, asocs); 1;
6639 tmp = list_next_entry(tmp, asocs)) {
6640 /* Manually skip the head element. */
6641 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
6642 continue;
6643 /* Wake up association. */
6644 __sctp_write_space(tmp);
6645 /* We've reached the end. */
6646 if (tmp == asoc)
6647 break;
6648 }
6649 }
6650
6651 /* Do accounting for the sndbuf space.
6652 * Decrement the used sndbuf space of the corresponding association by the
6653 * data size which was just transmitted(freed).
6654 */
6655 static void sctp_wfree(struct sk_buff *skb)
6656 {
6657 struct sctp_association *asoc;
6658 struct sctp_chunk *chunk;
6659 struct sock *sk;
6660
6661 /* Get the saved chunk pointer. */
6662 chunk = *((struct sctp_chunk **)(skb->cb));
6663 asoc = chunk->asoc;
6664 sk = asoc->base.sk;
6665 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6666 sizeof(struct sk_buff) +
6667 sizeof(struct sctp_chunk);
6668
6669 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6670
6671 /*
6672 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6673 */
6674 sk->sk_wmem_queued -= skb->truesize;
6675 sk_mem_uncharge(sk, skb->truesize);
6676
6677 sock_wfree(skb);
6678 sctp_wake_up_waiters(sk, asoc);
6679
6680 sctp_association_put(asoc);
6681 }
6682
6683 /* Do accounting for the receive space on the socket.
6684 * Accounting for the association is done in ulpevent.c
6685 * We set this as a destructor for the cloned data skbs so that
6686 * accounting is done at the correct time.
6687 */
6688 void sctp_sock_rfree(struct sk_buff *skb)
6689 {
6690 struct sock *sk = skb->sk;
6691 struct sctp_ulpevent *event = sctp_skb2event(skb);
6692
6693 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6694
6695 /*
6696 * Mimic the behavior of sock_rfree
6697 */
6698 sk_mem_uncharge(sk, event->rmem_len);
6699 }
6700
6701
6702 /* Helper function to wait for space in the sndbuf. */
6703 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6704 size_t msg_len)
6705 {
6706 struct sock *sk = asoc->base.sk;
6707 int err = 0;
6708 long current_timeo = *timeo_p;
6709 DEFINE_WAIT(wait);
6710
6711 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6712 asoc, (long)(*timeo_p), msg_len);
6713
6714 /* Increment the association's refcnt. */
6715 sctp_association_hold(asoc);
6716
6717 /* Wait on the association specific sndbuf space. */
6718 for (;;) {
6719 prepare_to_wait_exclusive(&asoc->wait, &wait,
6720 TASK_INTERRUPTIBLE);
6721 if (!*timeo_p)
6722 goto do_nonblock;
6723 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6724 asoc->base.dead)
6725 goto do_error;
6726 if (signal_pending(current))
6727 goto do_interrupted;
6728 if (msg_len <= sctp_wspace(asoc))
6729 break;
6730
6731 /* Let another process have a go. Since we are going
6732 * to sleep anyway.
6733 */
6734 sctp_release_sock(sk);
6735 current_timeo = schedule_timeout(current_timeo);
6736 sctp_lock_sock(sk);
6737
6738 *timeo_p = current_timeo;
6739 }
6740
6741 out:
6742 finish_wait(&asoc->wait, &wait);
6743
6744 /* Release the association's refcnt. */
6745 sctp_association_put(asoc);
6746
6747 return err;
6748
6749 do_error:
6750 err = -EPIPE;
6751 goto out;
6752
6753 do_interrupted:
6754 err = sock_intr_errno(*timeo_p);
6755 goto out;
6756
6757 do_nonblock:
6758 err = -EAGAIN;
6759 goto out;
6760 }
6761
6762 void sctp_data_ready(struct sock *sk, int len)
6763 {
6764 struct socket_wq *wq;
6765
6766 rcu_read_lock();
6767 wq = rcu_dereference(sk->sk_wq);
6768 if (wq_has_sleeper(wq))
6769 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6770 POLLRDNORM | POLLRDBAND);
6771 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6772 rcu_read_unlock();
6773 }
6774
6775 /* If socket sndbuf has changed, wake up all per association waiters. */
6776 void sctp_write_space(struct sock *sk)
6777 {
6778 struct sctp_association *asoc;
6779
6780 /* Wake up the tasks in each wait queue. */
6781 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6782 __sctp_write_space(asoc);
6783 }
6784 }
6785
6786 /* Is there any sndbuf space available on the socket?
6787 *
6788 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6789 * associations on the same socket. For a UDP-style socket with
6790 * multiple associations, it is possible for it to be "unwriteable"
6791 * prematurely. I assume that this is acceptable because
6792 * a premature "unwriteable" is better than an accidental "writeable" which
6793 * would cause an unwanted block under certain circumstances. For the 1-1
6794 * UDP-style sockets or TCP-style sockets, this code should work.
6795 * - Daisy
6796 */
6797 static int sctp_writeable(struct sock *sk)
6798 {
6799 int amt = 0;
6800
6801 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6802 if (amt < 0)
6803 amt = 0;
6804 return amt;
6805 }
6806
6807 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6808 * returns immediately with EINPROGRESS.
6809 */
6810 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6811 {
6812 struct sock *sk = asoc->base.sk;
6813 int err = 0;
6814 long current_timeo = *timeo_p;
6815 DEFINE_WAIT(wait);
6816
6817 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6818 (long)(*timeo_p));
6819
6820 /* Increment the association's refcnt. */
6821 sctp_association_hold(asoc);
6822
6823 for (;;) {
6824 prepare_to_wait_exclusive(&asoc->wait, &wait,
6825 TASK_INTERRUPTIBLE);
6826 if (!*timeo_p)
6827 goto do_nonblock;
6828 if (sk->sk_shutdown & RCV_SHUTDOWN)
6829 break;
6830 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6831 asoc->base.dead)
6832 goto do_error;
6833 if (signal_pending(current))
6834 goto do_interrupted;
6835
6836 if (sctp_state(asoc, ESTABLISHED))
6837 break;
6838
6839 /* Let another process have a go. Since we are going
6840 * to sleep anyway.
6841 */
6842 sctp_release_sock(sk);
6843 current_timeo = schedule_timeout(current_timeo);
6844 sctp_lock_sock(sk);
6845
6846 *timeo_p = current_timeo;
6847 }
6848
6849 out:
6850 finish_wait(&asoc->wait, &wait);
6851
6852 /* Release the association's refcnt. */
6853 sctp_association_put(asoc);
6854
6855 return err;
6856
6857 do_error:
6858 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6859 err = -ETIMEDOUT;
6860 else
6861 err = -ECONNREFUSED;
6862 goto out;
6863
6864 do_interrupted:
6865 err = sock_intr_errno(*timeo_p);
6866 goto out;
6867
6868 do_nonblock:
6869 err = -EINPROGRESS;
6870 goto out;
6871 }
6872
6873 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6874 {
6875 struct sctp_endpoint *ep;
6876 int err = 0;
6877 DEFINE_WAIT(wait);
6878
6879 ep = sctp_sk(sk)->ep;
6880
6881
6882 for (;;) {
6883 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6884 TASK_INTERRUPTIBLE);
6885
6886 if (list_empty(&ep->asocs)) {
6887 sctp_release_sock(sk);
6888 timeo = schedule_timeout(timeo);
6889 sctp_lock_sock(sk);
6890 }
6891
6892 err = -EINVAL;
6893 if (!sctp_sstate(sk, LISTENING))
6894 break;
6895
6896 err = 0;
6897 if (!list_empty(&ep->asocs))
6898 break;
6899
6900 err = sock_intr_errno(timeo);
6901 if (signal_pending(current))
6902 break;
6903
6904 err = -EAGAIN;
6905 if (!timeo)
6906 break;
6907 }
6908
6909 finish_wait(sk_sleep(sk), &wait);
6910
6911 return err;
6912 }
6913
6914 static void sctp_wait_for_close(struct sock *sk, long timeout)
6915 {
6916 DEFINE_WAIT(wait);
6917
6918 do {
6919 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6920 if (list_empty(&sctp_sk(sk)->ep->asocs))
6921 break;
6922 sctp_release_sock(sk);
6923 timeout = schedule_timeout(timeout);
6924 sctp_lock_sock(sk);
6925 } while (!signal_pending(current) && timeout);
6926
6927 finish_wait(sk_sleep(sk), &wait);
6928 }
6929
6930 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6931 {
6932 struct sk_buff *frag;
6933
6934 if (!skb->data_len)
6935 goto done;
6936
6937 /* Don't forget the fragments. */
6938 skb_walk_frags(skb, frag)
6939 sctp_skb_set_owner_r_frag(frag, sk);
6940
6941 done:
6942 sctp_skb_set_owner_r(skb, sk);
6943 }
6944
6945 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6946 struct sctp_association *asoc)
6947 {
6948 struct inet_sock *inet = inet_sk(sk);
6949 struct inet_sock *newinet;
6950
6951 newsk->sk_type = sk->sk_type;
6952 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6953 newsk->sk_flags = sk->sk_flags;
6954 newsk->sk_no_check = sk->sk_no_check;
6955 newsk->sk_reuse = sk->sk_reuse;
6956
6957 newsk->sk_shutdown = sk->sk_shutdown;
6958 newsk->sk_destruct = inet_sock_destruct;
6959 newsk->sk_family = sk->sk_family;
6960 newsk->sk_protocol = IPPROTO_SCTP;
6961 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6962 newsk->sk_sndbuf = sk->sk_sndbuf;
6963 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6964 newsk->sk_lingertime = sk->sk_lingertime;
6965 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6966 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6967
6968 newinet = inet_sk(newsk);
6969
6970 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6971 * getsockname() and getpeername()
6972 */
6973 newinet->inet_sport = inet->inet_sport;
6974 newinet->inet_saddr = inet->inet_saddr;
6975 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6976 newinet->inet_dport = htons(asoc->peer.port);
6977 newinet->pmtudisc = inet->pmtudisc;
6978 newinet->inet_id = asoc->next_tsn ^ jiffies;
6979
6980 newinet->uc_ttl = inet->uc_ttl;
6981 newinet->mc_loop = 1;
6982 newinet->mc_ttl = 1;
6983 newinet->mc_index = 0;
6984 newinet->mc_list = NULL;
6985
6986 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
6987 net_enable_timestamp();
6988 }
6989
6990 static inline void sctp_copy_descendant(struct sock *sk_to,
6991 const struct sock *sk_from)
6992 {
6993 int ancestor_size = sizeof(struct inet_sock) +
6994 sizeof(struct sctp_sock) -
6995 offsetof(struct sctp_sock, auto_asconf_list);
6996
6997 if (sk_from->sk_family == PF_INET6)
6998 ancestor_size += sizeof(struct ipv6_pinfo);
6999
7000 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
7001 }
7002
7003 /* Populate the fields of the newsk from the oldsk and migrate the assoc
7004 * and its messages to the newsk.
7005 */
7006 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
7007 struct sctp_association *assoc,
7008 sctp_socket_type_t type)
7009 {
7010 struct sctp_sock *oldsp = sctp_sk(oldsk);
7011 struct sctp_sock *newsp = sctp_sk(newsk);
7012 struct sctp_bind_bucket *pp; /* hash list port iterator */
7013 struct sctp_endpoint *newep = newsp->ep;
7014 struct sk_buff *skb, *tmp;
7015 struct sctp_ulpevent *event;
7016 struct sctp_bind_hashbucket *head;
7017
7018 /* Migrate socket buffer sizes and all the socket level options to the
7019 * new socket.
7020 */
7021 newsk->sk_sndbuf = oldsk->sk_sndbuf;
7022 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
7023 /* Brute force copy old sctp opt. */
7024 sctp_copy_descendant(newsk, oldsk);
7025
7026 /* Restore the ep value that was overwritten with the above structure
7027 * copy.
7028 */
7029 newsp->ep = newep;
7030 newsp->hmac = NULL;
7031
7032 /* Hook this new socket in to the bind_hash list. */
7033 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
7034 inet_sk(oldsk)->inet_num)];
7035 sctp_local_bh_disable();
7036 sctp_spin_lock(&head->lock);
7037 pp = sctp_sk(oldsk)->bind_hash;
7038 sk_add_bind_node(newsk, &pp->owner);
7039 sctp_sk(newsk)->bind_hash = pp;
7040 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
7041 sctp_spin_unlock(&head->lock);
7042 sctp_local_bh_enable();
7043
7044 /* Copy the bind_addr list from the original endpoint to the new
7045 * endpoint so that we can handle restarts properly
7046 */
7047 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
7048 &oldsp->ep->base.bind_addr, GFP_KERNEL);
7049
7050 /* Move any messages in the old socket's receive queue that are for the
7051 * peeled off association to the new socket's receive queue.
7052 */
7053 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
7054 event = sctp_skb2event(skb);
7055 if (event->asoc == assoc) {
7056 __skb_unlink(skb, &oldsk->sk_receive_queue);
7057 __skb_queue_tail(&newsk->sk_receive_queue, skb);
7058 sctp_skb_set_owner_r_frag(skb, newsk);
7059 }
7060 }
7061
7062 /* Clean up any messages pending delivery due to partial
7063 * delivery. Three cases:
7064 * 1) No partial deliver; no work.
7065 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
7066 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
7067 */
7068 skb_queue_head_init(&newsp->pd_lobby);
7069 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
7070
7071 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
7072 struct sk_buff_head *queue;
7073
7074 /* Decide which queue to move pd_lobby skbs to. */
7075 if (assoc->ulpq.pd_mode) {
7076 queue = &newsp->pd_lobby;
7077 } else
7078 queue = &newsk->sk_receive_queue;
7079
7080 /* Walk through the pd_lobby, looking for skbs that
7081 * need moved to the new socket.
7082 */
7083 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
7084 event = sctp_skb2event(skb);
7085 if (event->asoc == assoc) {
7086 __skb_unlink(skb, &oldsp->pd_lobby);
7087 __skb_queue_tail(queue, skb);
7088 sctp_skb_set_owner_r_frag(skb, newsk);
7089 }
7090 }
7091
7092 /* Clear up any skbs waiting for the partial
7093 * delivery to finish.
7094 */
7095 if (assoc->ulpq.pd_mode)
7096 sctp_clear_pd(oldsk, NULL);
7097
7098 }
7099
7100 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
7101 sctp_skb_set_owner_r_frag(skb, newsk);
7102
7103 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
7104 sctp_skb_set_owner_r_frag(skb, newsk);
7105
7106 /* Set the type of socket to indicate that it is peeled off from the
7107 * original UDP-style socket or created with the accept() call on a
7108 * TCP-style socket..
7109 */
7110 newsp->type = type;
7111
7112 /* Mark the new socket "in-use" by the user so that any packets
7113 * that may arrive on the association after we've moved it are
7114 * queued to the backlog. This prevents a potential race between
7115 * backlog processing on the old socket and new-packet processing
7116 * on the new socket.
7117 *
7118 * The caller has just allocated newsk so we can guarantee that other
7119 * paths won't try to lock it and then oldsk.
7120 */
7121 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7122 sctp_assoc_migrate(assoc, newsk);
7123
7124 /* If the association on the newsk is already closed before accept()
7125 * is called, set RCV_SHUTDOWN flag.
7126 */
7127 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
7128 newsk->sk_shutdown |= RCV_SHUTDOWN;
7129
7130 newsk->sk_state = SCTP_SS_ESTABLISHED;
7131 sctp_release_sock(newsk);
7132 }
7133
7134
7135 /* This proto struct describes the ULP interface for SCTP. */
7136 struct proto sctp_prot = {
7137 .name = "SCTP",
7138 .owner = THIS_MODULE,
7139 .close = sctp_close,
7140 .connect = sctp_connect,
7141 .disconnect = sctp_disconnect,
7142 .accept = sctp_accept,
7143 .ioctl = sctp_ioctl,
7144 .init = sctp_init_sock,
7145 .destroy = sctp_destroy_sock,
7146 .shutdown = sctp_shutdown,
7147 .setsockopt = sctp_setsockopt,
7148 .getsockopt = sctp_getsockopt,
7149 .sendmsg = sctp_sendmsg,
7150 .recvmsg = sctp_recvmsg,
7151 .bind = sctp_bind,
7152 .backlog_rcv = sctp_backlog_rcv,
7153 .hash = sctp_hash,
7154 .unhash = sctp_unhash,
7155 .get_port = sctp_get_port,
7156 .obj_size = sizeof(struct sctp_sock),
7157 .sysctl_mem = sysctl_sctp_mem,
7158 .sysctl_rmem = sysctl_sctp_rmem,
7159 .sysctl_wmem = sysctl_sctp_wmem,
7160 .memory_pressure = &sctp_memory_pressure,
7161 .enter_memory_pressure = sctp_enter_memory_pressure,
7162 .memory_allocated = &sctp_memory_allocated,
7163 .sockets_allocated = &sctp_sockets_allocated,
7164 };
7165
7166 #if IS_ENABLED(CONFIG_IPV6)
7167
7168 #include <net/transp_v6.h>
7169 static void sctp_v6_destroy_sock(struct sock *sk)
7170 {
7171 sctp_destroy_sock(sk);
7172 inet6_destroy_sock(sk);
7173 }
7174
7175 struct proto sctpv6_prot = {
7176 .name = "SCTPv6",
7177 .owner = THIS_MODULE,
7178 .close = sctp_close,
7179 .connect = sctp_connect,
7180 .disconnect = sctp_disconnect,
7181 .accept = sctp_accept,
7182 .ioctl = sctp_ioctl,
7183 .init = sctp_init_sock,
7184 .destroy = sctp_v6_destroy_sock,
7185 .shutdown = sctp_shutdown,
7186 .setsockopt = sctp_setsockopt,
7187 .getsockopt = sctp_getsockopt,
7188 .sendmsg = sctp_sendmsg,
7189 .recvmsg = sctp_recvmsg,
7190 .bind = sctp_bind,
7191 .backlog_rcv = sctp_backlog_rcv,
7192 .hash = sctp_hash,
7193 .unhash = sctp_unhash,
7194 .get_port = sctp_get_port,
7195 .obj_size = sizeof(struct sctp6_sock),
7196 .sysctl_mem = sysctl_sctp_mem,
7197 .sysctl_rmem = sysctl_sctp_rmem,
7198 .sysctl_wmem = sysctl_sctp_wmem,
7199 .memory_pressure = &sctp_memory_pressure,
7200 .enter_memory_pressure = sctp_enter_memory_pressure,
7201 .memory_allocated = &sctp_memory_allocated,
7202 .sockets_allocated = &sctp_sockets_allocated,
7203 };
7204 #endif /* IS_ENABLED(CONFIG_IPV6) */