]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* SCTP kernel reference 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 reference 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 | * The SCTP reference 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 | * The SCTP reference 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 | #include <linux/config.h> | |
61 | #include <linux/types.h> | |
62 | #include <linux/kernel.h> | |
63 | #include <linux/wait.h> | |
64 | #include <linux/time.h> | |
65 | #include <linux/ip.h> | |
66 | #include <linux/fcntl.h> | |
67 | #include <linux/poll.h> | |
68 | #include <linux/init.h> | |
69 | #include <linux/crypto.h> | |
70 | ||
71 | #include <net/ip.h> | |
72 | #include <net/icmp.h> | |
73 | #include <net/route.h> | |
74 | #include <net/ipv6.h> | |
75 | #include <net/inet_common.h> | |
76 | ||
77 | #include <linux/socket.h> /* for sa_family_t */ | |
78 | #include <net/sock.h> | |
79 | #include <net/sctp/sctp.h> | |
80 | #include <net/sctp/sm.h> | |
81 | ||
82 | /* WARNING: Please do not remove the SCTP_STATIC attribute to | |
83 | * any of the functions below as they are used to export functions | |
84 | * used by a project regression testsuite. | |
85 | */ | |
86 | ||
87 | /* Forward declarations for internal helper functions. */ | |
88 | static int sctp_writeable(struct sock *sk); | |
89 | static void sctp_wfree(struct sk_buff *skb); | |
90 | static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p, | |
91 | size_t msg_len); | |
92 | static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p); | |
93 | static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p); | |
94 | static int sctp_wait_for_accept(struct sock *sk, long timeo); | |
95 | static void sctp_wait_for_close(struct sock *sk, long timeo); | |
96 | static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt, | |
97 | union sctp_addr *addr, int len); | |
98 | static int sctp_bindx_add(struct sock *, struct sockaddr *, int); | |
99 | static int sctp_bindx_rem(struct sock *, struct sockaddr *, int); | |
100 | static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int); | |
101 | static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int); | |
102 | static int sctp_send_asconf(struct sctp_association *asoc, | |
103 | struct sctp_chunk *chunk); | |
104 | static int sctp_do_bind(struct sock *, union sctp_addr *, int); | |
105 | static int sctp_autobind(struct sock *sk); | |
106 | static void sctp_sock_migrate(struct sock *, struct sock *, | |
107 | struct sctp_association *, sctp_socket_type_t); | |
108 | static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG; | |
109 | ||
110 | extern kmem_cache_t *sctp_bucket_cachep; | |
111 | ||
112 | /* Get the sndbuf space available at the time on the association. */ | |
113 | static inline int sctp_wspace(struct sctp_association *asoc) | |
114 | { | |
115 | struct sock *sk = asoc->base.sk; | |
116 | int amt = 0; | |
117 | ||
4eb701df NH |
118 | if (asoc->ep->sndbuf_policy) { |
119 | /* make sure that no association uses more than sk_sndbuf */ | |
120 | amt = sk->sk_sndbuf - asoc->sndbuf_used; | |
121 | } else { | |
122 | /* do socket level accounting */ | |
123 | amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc); | |
124 | } | |
125 | ||
1da177e4 LT |
126 | if (amt < 0) |
127 | amt = 0; | |
4eb701df | 128 | |
1da177e4 LT |
129 | return amt; |
130 | } | |
131 | ||
132 | /* Increment the used sndbuf space count of the corresponding association by | |
133 | * the size of the outgoing data chunk. | |
134 | * Also, set the skb destructor for sndbuf accounting later. | |
135 | * | |
136 | * Since it is always 1-1 between chunk and skb, and also a new skb is always | |
137 | * allocated for chunk bundling in sctp_packet_transmit(), we can use the | |
138 | * destructor in the data chunk skb for the purpose of the sndbuf space | |
139 | * tracking. | |
140 | */ | |
141 | static inline void sctp_set_owner_w(struct sctp_chunk *chunk) | |
142 | { | |
143 | struct sctp_association *asoc = chunk->asoc; | |
144 | struct sock *sk = asoc->base.sk; | |
145 | ||
146 | /* The sndbuf space is tracked per association. */ | |
147 | sctp_association_hold(asoc); | |
148 | ||
4eb701df NH |
149 | skb_set_owner_w(chunk->skb, sk); |
150 | ||
1da177e4 LT |
151 | chunk->skb->destructor = sctp_wfree; |
152 | /* Save the chunk pointer in skb for sctp_wfree to use later. */ | |
153 | *((struct sctp_chunk **)(chunk->skb->cb)) = chunk; | |
154 | ||
4eb701df NH |
155 | asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) + |
156 | sizeof(struct sk_buff) + | |
157 | sizeof(struct sctp_chunk); | |
158 | ||
4eb701df | 159 | atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc); |
1da177e4 LT |
160 | } |
161 | ||
162 | /* Verify that this is a valid address. */ | |
163 | static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr, | |
164 | int len) | |
165 | { | |
166 | struct sctp_af *af; | |
167 | ||
168 | /* Verify basic sockaddr. */ | |
169 | af = sctp_sockaddr_af(sctp_sk(sk), addr, len); | |
170 | if (!af) | |
171 | return -EINVAL; | |
172 | ||
173 | /* Is this a valid SCTP address? */ | |
174 | if (!af->addr_valid(addr, sctp_sk(sk))) | |
175 | return -EINVAL; | |
176 | ||
177 | if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr))) | |
178 | return -EINVAL; | |
179 | ||
180 | return 0; | |
181 | } | |
182 | ||
183 | /* Look up the association by its id. If this is not a UDP-style | |
184 | * socket, the ID field is always ignored. | |
185 | */ | |
186 | struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id) | |
187 | { | |
188 | struct sctp_association *asoc = NULL; | |
189 | ||
190 | /* If this is not a UDP-style socket, assoc id should be ignored. */ | |
191 | if (!sctp_style(sk, UDP)) { | |
192 | /* Return NULL if the socket state is not ESTABLISHED. It | |
193 | * could be a TCP-style listening socket or a socket which | |
194 | * hasn't yet called connect() to establish an association. | |
195 | */ | |
196 | if (!sctp_sstate(sk, ESTABLISHED)) | |
197 | return NULL; | |
198 | ||
199 | /* Get the first and the only association from the list. */ | |
200 | if (!list_empty(&sctp_sk(sk)->ep->asocs)) | |
201 | asoc = list_entry(sctp_sk(sk)->ep->asocs.next, | |
202 | struct sctp_association, asocs); | |
203 | return asoc; | |
204 | } | |
205 | ||
206 | /* Otherwise this is a UDP-style socket. */ | |
207 | if (!id || (id == (sctp_assoc_t)-1)) | |
208 | return NULL; | |
209 | ||
210 | spin_lock_bh(&sctp_assocs_id_lock); | |
211 | asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id); | |
212 | spin_unlock_bh(&sctp_assocs_id_lock); | |
213 | ||
214 | if (!asoc || (asoc->base.sk != sk) || asoc->base.dead) | |
215 | return NULL; | |
216 | ||
217 | return asoc; | |
218 | } | |
219 | ||
220 | /* Look up the transport from an address and an assoc id. If both address and | |
221 | * id are specified, the associations matching the address and the id should be | |
222 | * the same. | |
223 | */ | |
224 | static struct sctp_transport *sctp_addr_id2transport(struct sock *sk, | |
225 | struct sockaddr_storage *addr, | |
226 | sctp_assoc_t id) | |
227 | { | |
228 | struct sctp_association *addr_asoc = NULL, *id_asoc = NULL; | |
229 | struct sctp_transport *transport; | |
230 | union sctp_addr *laddr = (union sctp_addr *)addr; | |
231 | ||
232 | laddr->v4.sin_port = ntohs(laddr->v4.sin_port); | |
233 | addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep, | |
234 | (union sctp_addr *)addr, | |
235 | &transport); | |
236 | laddr->v4.sin_port = htons(laddr->v4.sin_port); | |
237 | ||
238 | if (!addr_asoc) | |
239 | return NULL; | |
240 | ||
241 | id_asoc = sctp_id2assoc(sk, id); | |
242 | if (id_asoc && (id_asoc != addr_asoc)) | |
243 | return NULL; | |
244 | ||
245 | sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk), | |
246 | (union sctp_addr *)addr); | |
247 | ||
248 | return transport; | |
249 | } | |
250 | ||
251 | /* API 3.1.2 bind() - UDP Style Syntax | |
252 | * The syntax of bind() is, | |
253 | * | |
254 | * ret = bind(int sd, struct sockaddr *addr, int addrlen); | |
255 | * | |
256 | * sd - the socket descriptor returned by socket(). | |
257 | * addr - the address structure (struct sockaddr_in or struct | |
258 | * sockaddr_in6 [RFC 2553]), | |
259 | * addr_len - the size of the address structure. | |
260 | */ | |
3f7a87d2 | 261 | SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len) |
1da177e4 LT |
262 | { |
263 | int retval = 0; | |
264 | ||
265 | sctp_lock_sock(sk); | |
266 | ||
3f7a87d2 FF |
267 | SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n", |
268 | sk, addr, addr_len); | |
1da177e4 LT |
269 | |
270 | /* Disallow binding twice. */ | |
271 | if (!sctp_sk(sk)->ep->base.bind_addr.port) | |
3f7a87d2 | 272 | retval = sctp_do_bind(sk, (union sctp_addr *)addr, |
1da177e4 LT |
273 | addr_len); |
274 | else | |
275 | retval = -EINVAL; | |
276 | ||
277 | sctp_release_sock(sk); | |
278 | ||
279 | return retval; | |
280 | } | |
281 | ||
282 | static long sctp_get_port_local(struct sock *, union sctp_addr *); | |
283 | ||
284 | /* Verify this is a valid sockaddr. */ | |
285 | static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt, | |
286 | union sctp_addr *addr, int len) | |
287 | { | |
288 | struct sctp_af *af; | |
289 | ||
290 | /* Check minimum size. */ | |
291 | if (len < sizeof (struct sockaddr)) | |
292 | return NULL; | |
293 | ||
294 | /* Does this PF support this AF? */ | |
295 | if (!opt->pf->af_supported(addr->sa.sa_family, opt)) | |
296 | return NULL; | |
297 | ||
298 | /* If we get this far, af is valid. */ | |
299 | af = sctp_get_af_specific(addr->sa.sa_family); | |
300 | ||
301 | if (len < af->sockaddr_len) | |
302 | return NULL; | |
303 | ||
304 | return af; | |
305 | } | |
306 | ||
307 | /* Bind a local address either to an endpoint or to an association. */ | |
308 | SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len) | |
309 | { | |
310 | struct sctp_sock *sp = sctp_sk(sk); | |
311 | struct sctp_endpoint *ep = sp->ep; | |
312 | struct sctp_bind_addr *bp = &ep->base.bind_addr; | |
313 | struct sctp_af *af; | |
314 | unsigned short snum; | |
315 | int ret = 0; | |
316 | ||
1da177e4 LT |
317 | /* Common sockaddr verification. */ |
318 | af = sctp_sockaddr_af(sp, addr, len); | |
3f7a87d2 FF |
319 | if (!af) { |
320 | SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n", | |
321 | sk, addr, len); | |
1da177e4 | 322 | return -EINVAL; |
3f7a87d2 FF |
323 | } |
324 | ||
325 | snum = ntohs(addr->v4.sin_port); | |
326 | ||
327 | SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ", | |
328 | ", port: %d, new port: %d, len: %d)\n", | |
329 | sk, | |
330 | addr, | |
331 | bp->port, snum, | |
332 | len); | |
1da177e4 LT |
333 | |
334 | /* PF specific bind() address verification. */ | |
335 | if (!sp->pf->bind_verify(sp, addr)) | |
336 | return -EADDRNOTAVAIL; | |
337 | ||
1da177e4 LT |
338 | /* We must either be unbound, or bind to the same port. */ |
339 | if (bp->port && (snum != bp->port)) { | |
340 | SCTP_DEBUG_PRINTK("sctp_do_bind:" | |
341 | " New port %d does not match existing port " | |
342 | "%d.\n", snum, bp->port); | |
343 | return -EINVAL; | |
344 | } | |
345 | ||
346 | if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE)) | |
347 | return -EACCES; | |
348 | ||
349 | /* Make sure we are allowed to bind here. | |
350 | * The function sctp_get_port_local() does duplicate address | |
351 | * detection. | |
352 | */ | |
353 | if ((ret = sctp_get_port_local(sk, addr))) { | |
354 | if (ret == (long) sk) { | |
355 | /* This endpoint has a conflicting address. */ | |
356 | return -EINVAL; | |
357 | } else { | |
358 | return -EADDRINUSE; | |
359 | } | |
360 | } | |
361 | ||
362 | /* Refresh ephemeral port. */ | |
363 | if (!bp->port) | |
364 | bp->port = inet_sk(sk)->num; | |
365 | ||
366 | /* Add the address to the bind address list. */ | |
367 | sctp_local_bh_disable(); | |
368 | sctp_write_lock(&ep->base.addr_lock); | |
369 | ||
370 | /* Use GFP_ATOMIC since BHs are disabled. */ | |
371 | addr->v4.sin_port = ntohs(addr->v4.sin_port); | |
372 | ret = sctp_add_bind_addr(bp, addr, GFP_ATOMIC); | |
373 | addr->v4.sin_port = htons(addr->v4.sin_port); | |
374 | sctp_write_unlock(&ep->base.addr_lock); | |
375 | sctp_local_bh_enable(); | |
376 | ||
377 | /* Copy back into socket for getsockname() use. */ | |
378 | if (!ret) { | |
379 | inet_sk(sk)->sport = htons(inet_sk(sk)->num); | |
380 | af->to_sk_saddr(addr, sk); | |
381 | } | |
382 | ||
383 | return ret; | |
384 | } | |
385 | ||
386 | /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks | |
387 | * | |
388 | * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged | |
389 | * at any one time. If a sender, after sending an ASCONF chunk, decides | |
390 | * it needs to transfer another ASCONF Chunk, it MUST wait until the | |
391 | * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a | |
392 | * subsequent ASCONF. Note this restriction binds each side, so at any | |
393 | * time two ASCONF may be in-transit on any given association (one sent | |
394 | * from each endpoint). | |
395 | */ | |
396 | static int sctp_send_asconf(struct sctp_association *asoc, | |
397 | struct sctp_chunk *chunk) | |
398 | { | |
399 | int retval = 0; | |
400 | ||
401 | /* If there is an outstanding ASCONF chunk, queue it for later | |
402 | * transmission. | |
403 | */ | |
404 | if (asoc->addip_last_asconf) { | |
79af02c2 | 405 | list_add_tail(&chunk->list, &asoc->addip_chunk_list); |
1da177e4 LT |
406 | goto out; |
407 | } | |
408 | ||
409 | /* Hold the chunk until an ASCONF_ACK is received. */ | |
410 | sctp_chunk_hold(chunk); | |
411 | retval = sctp_primitive_ASCONF(asoc, chunk); | |
412 | if (retval) | |
413 | sctp_chunk_free(chunk); | |
414 | else | |
415 | asoc->addip_last_asconf = chunk; | |
416 | ||
417 | out: | |
418 | return retval; | |
419 | } | |
420 | ||
421 | /* Add a list of addresses as bind addresses to local endpoint or | |
422 | * association. | |
423 | * | |
424 | * Basically run through each address specified in the addrs/addrcnt | |
425 | * array/length pair, determine if it is IPv6 or IPv4 and call | |
426 | * sctp_do_bind() on it. | |
427 | * | |
428 | * If any of them fails, then the operation will be reversed and the | |
429 | * ones that were added will be removed. | |
430 | * | |
431 | * Only sctp_setsockopt_bindx() is supposed to call this function. | |
432 | */ | |
433 | int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt) | |
434 | { | |
435 | int cnt; | |
436 | int retval = 0; | |
437 | void *addr_buf; | |
438 | struct sockaddr *sa_addr; | |
439 | struct sctp_af *af; | |
440 | ||
441 | SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n", | |
442 | sk, addrs, addrcnt); | |
443 | ||
444 | addr_buf = addrs; | |
445 | for (cnt = 0; cnt < addrcnt; cnt++) { | |
446 | /* The list may contain either IPv4 or IPv6 address; | |
447 | * determine the address length for walking thru the list. | |
448 | */ | |
449 | sa_addr = (struct sockaddr *)addr_buf; | |
450 | af = sctp_get_af_specific(sa_addr->sa_family); | |
451 | if (!af) { | |
452 | retval = -EINVAL; | |
453 | goto err_bindx_add; | |
454 | } | |
455 | ||
456 | retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr, | |
457 | af->sockaddr_len); | |
458 | ||
459 | addr_buf += af->sockaddr_len; | |
460 | ||
461 | err_bindx_add: | |
462 | if (retval < 0) { | |
463 | /* Failed. Cleanup the ones that have been added */ | |
464 | if (cnt > 0) | |
465 | sctp_bindx_rem(sk, addrs, cnt); | |
466 | return retval; | |
467 | } | |
468 | } | |
469 | ||
470 | return retval; | |
471 | } | |
472 | ||
473 | /* Send an ASCONF chunk with Add IP address parameters to all the peers of the | |
474 | * associations that are part of the endpoint indicating that a list of local | |
475 | * addresses are added to the endpoint. | |
476 | * | |
477 | * If any of the addresses is already in the bind address list of the | |
478 | * association, we do not send the chunk for that association. But it will not | |
479 | * affect other associations. | |
480 | * | |
481 | * Only sctp_setsockopt_bindx() is supposed to call this function. | |
482 | */ | |
483 | static int sctp_send_asconf_add_ip(struct sock *sk, | |
484 | struct sockaddr *addrs, | |
485 | int addrcnt) | |
486 | { | |
487 | struct sctp_sock *sp; | |
488 | struct sctp_endpoint *ep; | |
489 | struct sctp_association *asoc; | |
490 | struct sctp_bind_addr *bp; | |
491 | struct sctp_chunk *chunk; | |
492 | struct sctp_sockaddr_entry *laddr; | |
493 | union sctp_addr *addr; | |
494 | void *addr_buf; | |
495 | struct sctp_af *af; | |
496 | struct list_head *pos; | |
497 | struct list_head *p; | |
498 | int i; | |
499 | int retval = 0; | |
500 | ||
501 | if (!sctp_addip_enable) | |
502 | return retval; | |
503 | ||
504 | sp = sctp_sk(sk); | |
505 | ep = sp->ep; | |
506 | ||
507 | SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n", | |
508 | __FUNCTION__, sk, addrs, addrcnt); | |
509 | ||
510 | list_for_each(pos, &ep->asocs) { | |
511 | asoc = list_entry(pos, struct sctp_association, asocs); | |
512 | ||
513 | if (!asoc->peer.asconf_capable) | |
514 | continue; | |
515 | ||
516 | if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP) | |
517 | continue; | |
518 | ||
519 | if (!sctp_state(asoc, ESTABLISHED)) | |
520 | continue; | |
521 | ||
522 | /* Check if any address in the packed array of addresses is | |
523 | * in the bind address list of the association. If so, | |
524 | * do not send the asconf chunk to its peer, but continue with | |
525 | * other associations. | |
526 | */ | |
527 | addr_buf = addrs; | |
528 | for (i = 0; i < addrcnt; i++) { | |
529 | addr = (union sctp_addr *)addr_buf; | |
530 | af = sctp_get_af_specific(addr->v4.sin_family); | |
531 | if (!af) { | |
532 | retval = -EINVAL; | |
533 | goto out; | |
534 | } | |
535 | ||
536 | if (sctp_assoc_lookup_laddr(asoc, addr)) | |
537 | break; | |
538 | ||
539 | addr_buf += af->sockaddr_len; | |
540 | } | |
541 | if (i < addrcnt) | |
542 | continue; | |
543 | ||
544 | /* Use the first address in bind addr list of association as | |
545 | * Address Parameter of ASCONF CHUNK. | |
546 | */ | |
547 | sctp_read_lock(&asoc->base.addr_lock); | |
548 | bp = &asoc->base.bind_addr; | |
549 | p = bp->address_list.next; | |
550 | laddr = list_entry(p, struct sctp_sockaddr_entry, list); | |
551 | sctp_read_unlock(&asoc->base.addr_lock); | |
552 | ||
553 | chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs, | |
554 | addrcnt, SCTP_PARAM_ADD_IP); | |
555 | if (!chunk) { | |
556 | retval = -ENOMEM; | |
557 | goto out; | |
558 | } | |
559 | ||
560 | retval = sctp_send_asconf(asoc, chunk); | |
561 | ||
562 | /* FIXME: After sending the add address ASCONF chunk, we | |
563 | * cannot append the address to the association's binding | |
564 | * address list, because the new address may be used as the | |
565 | * source of a message sent to the peer before the ASCONF | |
566 | * chunk is received by the peer. So we should wait until | |
567 | * ASCONF_ACK is received. | |
568 | */ | |
569 | } | |
570 | ||
571 | out: | |
572 | return retval; | |
573 | } | |
574 | ||
575 | /* Remove a list of addresses from bind addresses list. Do not remove the | |
576 | * last address. | |
577 | * | |
578 | * Basically run through each address specified in the addrs/addrcnt | |
579 | * array/length pair, determine if it is IPv6 or IPv4 and call | |
580 | * sctp_del_bind() on it. | |
581 | * | |
582 | * If any of them fails, then the operation will be reversed and the | |
583 | * ones that were removed will be added back. | |
584 | * | |
585 | * At least one address has to be left; if only one address is | |
586 | * available, the operation will return -EBUSY. | |
587 | * | |
588 | * Only sctp_setsockopt_bindx() is supposed to call this function. | |
589 | */ | |
590 | int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt) | |
591 | { | |
592 | struct sctp_sock *sp = sctp_sk(sk); | |
593 | struct sctp_endpoint *ep = sp->ep; | |
594 | int cnt; | |
595 | struct sctp_bind_addr *bp = &ep->base.bind_addr; | |
596 | int retval = 0; | |
597 | union sctp_addr saveaddr; | |
598 | void *addr_buf; | |
599 | struct sockaddr *sa_addr; | |
600 | struct sctp_af *af; | |
601 | ||
602 | SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n", | |
603 | sk, addrs, addrcnt); | |
604 | ||
605 | addr_buf = addrs; | |
606 | for (cnt = 0; cnt < addrcnt; cnt++) { | |
607 | /* If the bind address list is empty or if there is only one | |
608 | * bind address, there is nothing more to be removed (we need | |
609 | * at least one address here). | |
610 | */ | |
611 | if (list_empty(&bp->address_list) || | |
612 | (sctp_list_single_entry(&bp->address_list))) { | |
613 | retval = -EBUSY; | |
614 | goto err_bindx_rem; | |
615 | } | |
616 | ||
617 | /* The list may contain either IPv4 or IPv6 address; | |
618 | * determine the address length to copy the address to | |
619 | * saveaddr. | |
620 | */ | |
621 | sa_addr = (struct sockaddr *)addr_buf; | |
622 | af = sctp_get_af_specific(sa_addr->sa_family); | |
623 | if (!af) { | |
624 | retval = -EINVAL; | |
625 | goto err_bindx_rem; | |
626 | } | |
627 | memcpy(&saveaddr, sa_addr, af->sockaddr_len); | |
628 | saveaddr.v4.sin_port = ntohs(saveaddr.v4.sin_port); | |
629 | if (saveaddr.v4.sin_port != bp->port) { | |
630 | retval = -EINVAL; | |
631 | goto err_bindx_rem; | |
632 | } | |
633 | ||
634 | /* FIXME - There is probably a need to check if sk->sk_saddr and | |
635 | * sk->sk_rcv_addr are currently set to one of the addresses to | |
636 | * be removed. This is something which needs to be looked into | |
637 | * when we are fixing the outstanding issues with multi-homing | |
638 | * socket routing and failover schemes. Refer to comments in | |
639 | * sctp_do_bind(). -daisy | |
640 | */ | |
641 | sctp_local_bh_disable(); | |
642 | sctp_write_lock(&ep->base.addr_lock); | |
643 | ||
644 | retval = sctp_del_bind_addr(bp, &saveaddr); | |
645 | ||
646 | sctp_write_unlock(&ep->base.addr_lock); | |
647 | sctp_local_bh_enable(); | |
648 | ||
649 | addr_buf += af->sockaddr_len; | |
650 | err_bindx_rem: | |
651 | if (retval < 0) { | |
652 | /* Failed. Add the ones that has been removed back */ | |
653 | if (cnt > 0) | |
654 | sctp_bindx_add(sk, addrs, cnt); | |
655 | return retval; | |
656 | } | |
657 | } | |
658 | ||
659 | return retval; | |
660 | } | |
661 | ||
662 | /* Send an ASCONF chunk with Delete IP address parameters to all the peers of | |
663 | * the associations that are part of the endpoint indicating that a list of | |
664 | * local addresses are removed from the endpoint. | |
665 | * | |
666 | * If any of the addresses is already in the bind address list of the | |
667 | * association, we do not send the chunk for that association. But it will not | |
668 | * affect other associations. | |
669 | * | |
670 | * Only sctp_setsockopt_bindx() is supposed to call this function. | |
671 | */ | |
672 | static int sctp_send_asconf_del_ip(struct sock *sk, | |
673 | struct sockaddr *addrs, | |
674 | int addrcnt) | |
675 | { | |
676 | struct sctp_sock *sp; | |
677 | struct sctp_endpoint *ep; | |
678 | struct sctp_association *asoc; | |
679 | struct sctp_bind_addr *bp; | |
680 | struct sctp_chunk *chunk; | |
681 | union sctp_addr *laddr; | |
682 | void *addr_buf; | |
683 | struct sctp_af *af; | |
684 | struct list_head *pos; | |
685 | int i; | |
686 | int retval = 0; | |
687 | ||
688 | if (!sctp_addip_enable) | |
689 | return retval; | |
690 | ||
691 | sp = sctp_sk(sk); | |
692 | ep = sp->ep; | |
693 | ||
694 | SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n", | |
695 | __FUNCTION__, sk, addrs, addrcnt); | |
696 | ||
697 | list_for_each(pos, &ep->asocs) { | |
698 | asoc = list_entry(pos, struct sctp_association, asocs); | |
699 | ||
700 | if (!asoc->peer.asconf_capable) | |
701 | continue; | |
702 | ||
703 | if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP) | |
704 | continue; | |
705 | ||
706 | if (!sctp_state(asoc, ESTABLISHED)) | |
707 | continue; | |
708 | ||
709 | /* Check if any address in the packed array of addresses is | |
710 | * not present in the bind address list of the association. | |
711 | * If so, do not send the asconf chunk to its peer, but | |
712 | * continue with other associations. | |
713 | */ | |
714 | addr_buf = addrs; | |
715 | for (i = 0; i < addrcnt; i++) { | |
716 | laddr = (union sctp_addr *)addr_buf; | |
717 | af = sctp_get_af_specific(laddr->v4.sin_family); | |
718 | if (!af) { | |
719 | retval = -EINVAL; | |
720 | goto out; | |
721 | } | |
722 | ||
723 | if (!sctp_assoc_lookup_laddr(asoc, laddr)) | |
724 | break; | |
725 | ||
726 | addr_buf += af->sockaddr_len; | |
727 | } | |
728 | if (i < addrcnt) | |
729 | continue; | |
730 | ||
731 | /* Find one address in the association's bind address list | |
732 | * that is not in the packed array of addresses. This is to | |
733 | * make sure that we do not delete all the addresses in the | |
734 | * association. | |
735 | */ | |
736 | sctp_read_lock(&asoc->base.addr_lock); | |
737 | bp = &asoc->base.bind_addr; | |
738 | laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs, | |
739 | addrcnt, sp); | |
740 | sctp_read_unlock(&asoc->base.addr_lock); | |
741 | if (!laddr) | |
742 | continue; | |
743 | ||
744 | chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt, | |
745 | SCTP_PARAM_DEL_IP); | |
746 | if (!chunk) { | |
747 | retval = -ENOMEM; | |
748 | goto out; | |
749 | } | |
750 | ||
751 | retval = sctp_send_asconf(asoc, chunk); | |
752 | ||
753 | /* FIXME: After sending the delete address ASCONF chunk, we | |
754 | * cannot remove the addresses from the association's bind | |
755 | * address list, because there maybe some packet send to | |
756 | * the delete addresses, so we should wait until ASCONF_ACK | |
757 | * packet is received. | |
758 | */ | |
759 | } | |
760 | out: | |
761 | return retval; | |
762 | } | |
763 | ||
764 | /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt() | |
765 | * | |
766 | * API 8.1 | |
767 | * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt, | |
768 | * int flags); | |
769 | * | |
770 | * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses. | |
771 | * If the sd is an IPv6 socket, the addresses passed can either be IPv4 | |
772 | * or IPv6 addresses. | |
773 | * | |
774 | * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see | |
775 | * Section 3.1.2 for this usage. | |
776 | * | |
777 | * addrs is a pointer to an array of one or more socket addresses. Each | |
778 | * address is contained in its appropriate structure (i.e. struct | |
779 | * sockaddr_in or struct sockaddr_in6) the family of the address type | |
780 | * must be used to distengish the address length (note that this | |
781 | * representation is termed a "packed array" of addresses). The caller | |
782 | * specifies the number of addresses in the array with addrcnt. | |
783 | * | |
784 | * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns | |
785 | * -1, and sets errno to the appropriate error code. | |
786 | * | |
787 | * For SCTP, the port given in each socket address must be the same, or | |
788 | * sctp_bindx() will fail, setting errno to EINVAL. | |
789 | * | |
790 | * The flags parameter is formed from the bitwise OR of zero or more of | |
791 | * the following currently defined flags: | |
792 | * | |
793 | * SCTP_BINDX_ADD_ADDR | |
794 | * | |
795 | * SCTP_BINDX_REM_ADDR | |
796 | * | |
797 | * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the | |
798 | * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given | |
799 | * addresses from the association. The two flags are mutually exclusive; | |
800 | * if both are given, sctp_bindx() will fail with EINVAL. A caller may | |
801 | * not remove all addresses from an association; sctp_bindx() will | |
802 | * reject such an attempt with EINVAL. | |
803 | * | |
804 | * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate | |
805 | * additional addresses with an endpoint after calling bind(). Or use | |
806 | * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening | |
807 | * socket is associated with so that no new association accepted will be | |
808 | * associated with those addresses. If the endpoint supports dynamic | |
809 | * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a | |
810 | * endpoint to send the appropriate message to the peer to change the | |
811 | * peers address lists. | |
812 | * | |
813 | * Adding and removing addresses from a connected association is | |
814 | * optional functionality. Implementations that do not support this | |
815 | * functionality should return EOPNOTSUPP. | |
816 | * | |
817 | * Basically do nothing but copying the addresses from user to kernel | |
818 | * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk. | |
3f7a87d2 FF |
819 | * This is used for tunneling the sctp_bindx() request through sctp_setsockopt() |
820 | * from userspace. | |
1da177e4 LT |
821 | * |
822 | * We don't use copy_from_user() for optimization: we first do the | |
823 | * sanity checks (buffer size -fast- and access check-healthy | |
824 | * pointer); if all of those succeed, then we can alloc the memory | |
825 | * (expensive operation) needed to copy the data to kernel. Then we do | |
826 | * the copying without checking the user space area | |
827 | * (__copy_from_user()). | |
828 | * | |
829 | * On exit there is no need to do sockfd_put(), sys_setsockopt() does | |
830 | * it. | |
831 | * | |
832 | * sk The sk of the socket | |
833 | * addrs The pointer to the addresses in user land | |
834 | * addrssize Size of the addrs buffer | |
835 | * op Operation to perform (add or remove, see the flags of | |
836 | * sctp_bindx) | |
837 | * | |
838 | * Returns 0 if ok, <0 errno code on error. | |
839 | */ | |
840 | SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk, | |
841 | struct sockaddr __user *addrs, | |
842 | int addrs_size, int op) | |
843 | { | |
844 | struct sockaddr *kaddrs; | |
845 | int err; | |
846 | int addrcnt = 0; | |
847 | int walk_size = 0; | |
848 | struct sockaddr *sa_addr; | |
849 | void *addr_buf; | |
850 | struct sctp_af *af; | |
851 | ||
852 | SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p" | |
853 | " addrs_size %d opt %d\n", sk, addrs, addrs_size, op); | |
854 | ||
855 | if (unlikely(addrs_size <= 0)) | |
856 | return -EINVAL; | |
857 | ||
858 | /* Check the user passed a healthy pointer. */ | |
859 | if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size))) | |
860 | return -EFAULT; | |
861 | ||
862 | /* Alloc space for the address array in kernel memory. */ | |
863 | kaddrs = (struct sockaddr *)kmalloc(addrs_size, GFP_KERNEL); | |
864 | if (unlikely(!kaddrs)) | |
865 | return -ENOMEM; | |
866 | ||
867 | if (__copy_from_user(kaddrs, addrs, addrs_size)) { | |
868 | kfree(kaddrs); | |
869 | return -EFAULT; | |
870 | } | |
871 | ||
872 | /* Walk through the addrs buffer and count the number of addresses. */ | |
873 | addr_buf = kaddrs; | |
874 | while (walk_size < addrs_size) { | |
875 | sa_addr = (struct sockaddr *)addr_buf; | |
876 | af = sctp_get_af_specific(sa_addr->sa_family); | |
877 | ||
878 | /* If the address family is not supported or if this address | |
879 | * causes the address buffer to overflow return EINVAL. | |
880 | */ | |
881 | if (!af || (walk_size + af->sockaddr_len) > addrs_size) { | |
882 | kfree(kaddrs); | |
883 | return -EINVAL; | |
884 | } | |
885 | addrcnt++; | |
886 | addr_buf += af->sockaddr_len; | |
887 | walk_size += af->sockaddr_len; | |
888 | } | |
889 | ||
890 | /* Do the work. */ | |
891 | switch (op) { | |
892 | case SCTP_BINDX_ADD_ADDR: | |
893 | err = sctp_bindx_add(sk, kaddrs, addrcnt); | |
894 | if (err) | |
895 | goto out; | |
896 | err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt); | |
897 | break; | |
898 | ||
899 | case SCTP_BINDX_REM_ADDR: | |
900 | err = sctp_bindx_rem(sk, kaddrs, addrcnt); | |
901 | if (err) | |
902 | goto out; | |
903 | err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt); | |
904 | break; | |
905 | ||
906 | default: | |
907 | err = -EINVAL; | |
908 | break; | |
909 | }; | |
910 | ||
911 | out: | |
912 | kfree(kaddrs); | |
913 | ||
914 | return err; | |
915 | } | |
916 | ||
3f7a87d2 FF |
917 | /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size) |
918 | * | |
919 | * Common routine for handling connect() and sctp_connectx(). | |
920 | * Connect will come in with just a single address. | |
921 | */ | |
922 | static int __sctp_connect(struct sock* sk, | |
923 | struct sockaddr *kaddrs, | |
924 | int addrs_size) | |
925 | { | |
926 | struct sctp_sock *sp; | |
927 | struct sctp_endpoint *ep; | |
928 | struct sctp_association *asoc = NULL; | |
929 | struct sctp_association *asoc2; | |
930 | struct sctp_transport *transport; | |
931 | union sctp_addr to; | |
932 | struct sctp_af *af; | |
933 | sctp_scope_t scope; | |
934 | long timeo; | |
935 | int err = 0; | |
936 | int addrcnt = 0; | |
937 | int walk_size = 0; | |
938 | struct sockaddr *sa_addr; | |
939 | void *addr_buf; | |
940 | ||
941 | sp = sctp_sk(sk); | |
942 | ep = sp->ep; | |
943 | ||
944 | /* connect() cannot be done on a socket that is already in ESTABLISHED | |
945 | * state - UDP-style peeled off socket or a TCP-style socket that | |
946 | * is already connected. | |
947 | * It cannot be done even on a TCP-style listening socket. | |
948 | */ | |
949 | if (sctp_sstate(sk, ESTABLISHED) || | |
950 | (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) { | |
951 | err = -EISCONN; | |
952 | goto out_free; | |
953 | } | |
954 | ||
955 | /* Walk through the addrs buffer and count the number of addresses. */ | |
956 | addr_buf = kaddrs; | |
957 | while (walk_size < addrs_size) { | |
958 | sa_addr = (struct sockaddr *)addr_buf; | |
959 | af = sctp_get_af_specific(sa_addr->sa_family); | |
960 | ||
961 | /* If the address family is not supported or if this address | |
962 | * causes the address buffer to overflow return EINVAL. | |
963 | */ | |
964 | if (!af || (walk_size + af->sockaddr_len) > addrs_size) { | |
965 | err = -EINVAL; | |
966 | goto out_free; | |
967 | } | |
968 | ||
969 | err = sctp_verify_addr(sk, (union sctp_addr *)sa_addr, | |
970 | af->sockaddr_len); | |
971 | if (err) | |
972 | goto out_free; | |
973 | ||
974 | memcpy(&to, sa_addr, af->sockaddr_len); | |
975 | to.v4.sin_port = ntohs(to.v4.sin_port); | |
976 | ||
977 | /* Check if there already is a matching association on the | |
978 | * endpoint (other than the one created here). | |
979 | */ | |
980 | asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport); | |
981 | if (asoc2 && asoc2 != asoc) { | |
982 | if (asoc2->state >= SCTP_STATE_ESTABLISHED) | |
983 | err = -EISCONN; | |
984 | else | |
985 | err = -EALREADY; | |
986 | goto out_free; | |
987 | } | |
988 | ||
989 | /* If we could not find a matching association on the endpoint, | |
990 | * make sure that there is no peeled-off association matching | |
991 | * the peer address even on another socket. | |
992 | */ | |
993 | if (sctp_endpoint_is_peeled_off(ep, &to)) { | |
994 | err = -EADDRNOTAVAIL; | |
995 | goto out_free; | |
996 | } | |
997 | ||
998 | if (!asoc) { | |
999 | /* If a bind() or sctp_bindx() is not called prior to | |
1000 | * an sctp_connectx() call, the system picks an | |
1001 | * ephemeral port and will choose an address set | |
1002 | * equivalent to binding with a wildcard address. | |
1003 | */ | |
1004 | if (!ep->base.bind_addr.port) { | |
1005 | if (sctp_autobind(sk)) { | |
1006 | err = -EAGAIN; | |
1007 | goto out_free; | |
1008 | } | |
64a0c1c8 ISJ |
1009 | } else { |
1010 | /* | |
1011 | * If an unprivileged user inherits a 1-many | |
1012 | * style socket with open associations on a | |
1013 | * privileged port, it MAY be permitted to | |
1014 | * accept new associations, but it SHOULD NOT | |
1015 | * be permitted to open new associations. | |
1016 | */ | |
1017 | if (ep->base.bind_addr.port < PROT_SOCK && | |
1018 | !capable(CAP_NET_BIND_SERVICE)) { | |
1019 | err = -EACCES; | |
1020 | goto out_free; | |
1021 | } | |
3f7a87d2 FF |
1022 | } |
1023 | ||
1024 | scope = sctp_scope(&to); | |
1025 | asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL); | |
1026 | if (!asoc) { | |
1027 | err = -ENOMEM; | |
1028 | goto out_free; | |
1029 | } | |
1030 | } | |
1031 | ||
1032 | /* Prime the peer's transport structures. */ | |
1033 | transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, | |
1034 | SCTP_UNKNOWN); | |
1035 | if (!transport) { | |
1036 | err = -ENOMEM; | |
1037 | goto out_free; | |
1038 | } | |
1039 | ||
1040 | addrcnt++; | |
1041 | addr_buf += af->sockaddr_len; | |
1042 | walk_size += af->sockaddr_len; | |
1043 | } | |
1044 | ||
1045 | err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL); | |
1046 | if (err < 0) { | |
1047 | goto out_free; | |
1048 | } | |
1049 | ||
1050 | err = sctp_primitive_ASSOCIATE(asoc, NULL); | |
1051 | if (err < 0) { | |
1052 | goto out_free; | |
1053 | } | |
1054 | ||
1055 | /* Initialize sk's dport and daddr for getpeername() */ | |
1056 | inet_sk(sk)->dport = htons(asoc->peer.port); | |
1057 | af = sctp_get_af_specific(to.sa.sa_family); | |
1058 | af->to_sk_daddr(&to, sk); | |
1059 | ||
1060 | timeo = sock_sndtimeo(sk, sk->sk_socket->file->f_flags & O_NONBLOCK); | |
1061 | err = sctp_wait_for_connect(asoc, &timeo); | |
1062 | ||
1063 | /* Don't free association on exit. */ | |
1064 | asoc = NULL; | |
1065 | ||
1066 | out_free: | |
1067 | ||
1068 | SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p" | |
1069 | " kaddrs: %p err: %d\n", | |
1070 | asoc, kaddrs, err); | |
1071 | if (asoc) | |
1072 | sctp_association_free(asoc); | |
1073 | return err; | |
1074 | } | |
1075 | ||
1076 | /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt() | |
1077 | * | |
1078 | * API 8.9 | |
1079 | * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt); | |
1080 | * | |
1081 | * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses. | |
1082 | * If the sd is an IPv6 socket, the addresses passed can either be IPv4 | |
1083 | * or IPv6 addresses. | |
1084 | * | |
1085 | * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see | |
1086 | * Section 3.1.2 for this usage. | |
1087 | * | |
1088 | * addrs is a pointer to an array of one or more socket addresses. Each | |
1089 | * address is contained in its appropriate structure (i.e. struct | |
1090 | * sockaddr_in or struct sockaddr_in6) the family of the address type | |
1091 | * must be used to distengish the address length (note that this | |
1092 | * representation is termed a "packed array" of addresses). The caller | |
1093 | * specifies the number of addresses in the array with addrcnt. | |
1094 | * | |
1095 | * On success, sctp_connectx() returns 0. On failure, sctp_connectx() returns | |
1096 | * -1, and sets errno to the appropriate error code. | |
1097 | * | |
1098 | * For SCTP, the port given in each socket address must be the same, or | |
1099 | * sctp_connectx() will fail, setting errno to EINVAL. | |
1100 | * | |
1101 | * An application can use sctp_connectx to initiate an association with | |
1102 | * an endpoint that is multi-homed. Much like sctp_bindx() this call | |
1103 | * allows a caller to specify multiple addresses at which a peer can be | |
1104 | * reached. The way the SCTP stack uses the list of addresses to set up | |
1105 | * the association is implementation dependant. This function only | |
1106 | * specifies that the stack will try to make use of all the addresses in | |
1107 | * the list when needed. | |
1108 | * | |
1109 | * Note that the list of addresses passed in is only used for setting up | |
1110 | * the association. It does not necessarily equal the set of addresses | |
1111 | * the peer uses for the resulting association. If the caller wants to | |
1112 | * find out the set of peer addresses, it must use sctp_getpaddrs() to | |
1113 | * retrieve them after the association has been set up. | |
1114 | * | |
1115 | * Basically do nothing but copying the addresses from user to kernel | |
1116 | * land and invoking either sctp_connectx(). This is used for tunneling | |
1117 | * the sctp_connectx() request through sctp_setsockopt() from userspace. | |
1118 | * | |
1119 | * We don't use copy_from_user() for optimization: we first do the | |
1120 | * sanity checks (buffer size -fast- and access check-healthy | |
1121 | * pointer); if all of those succeed, then we can alloc the memory | |
1122 | * (expensive operation) needed to copy the data to kernel. Then we do | |
1123 | * the copying without checking the user space area | |
1124 | * (__copy_from_user()). | |
1125 | * | |
1126 | * On exit there is no need to do sockfd_put(), sys_setsockopt() does | |
1127 | * it. | |
1128 | * | |
1129 | * sk The sk of the socket | |
1130 | * addrs The pointer to the addresses in user land | |
1131 | * addrssize Size of the addrs buffer | |
1132 | * | |
1133 | * Returns 0 if ok, <0 errno code on error. | |
1134 | */ | |
1135 | SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk, | |
1136 | struct sockaddr __user *addrs, | |
1137 | int addrs_size) | |
1138 | { | |
1139 | int err = 0; | |
1140 | struct sockaddr *kaddrs; | |
1141 | ||
1142 | SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n", | |
1143 | __FUNCTION__, sk, addrs, addrs_size); | |
1144 | ||
1145 | if (unlikely(addrs_size <= 0)) | |
1146 | return -EINVAL; | |
1147 | ||
1148 | /* Check the user passed a healthy pointer. */ | |
1149 | if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size))) | |
1150 | return -EFAULT; | |
1151 | ||
1152 | /* Alloc space for the address array in kernel memory. */ | |
1153 | kaddrs = (struct sockaddr *)kmalloc(addrs_size, GFP_KERNEL); | |
1154 | if (unlikely(!kaddrs)) | |
1155 | return -ENOMEM; | |
1156 | ||
1157 | if (__copy_from_user(kaddrs, addrs, addrs_size)) { | |
1158 | err = -EFAULT; | |
1159 | } else { | |
1160 | err = __sctp_connect(sk, kaddrs, addrs_size); | |
1161 | } | |
1162 | ||
1163 | kfree(kaddrs); | |
1164 | return err; | |
1165 | } | |
1166 | ||
1da177e4 LT |
1167 | /* API 3.1.4 close() - UDP Style Syntax |
1168 | * Applications use close() to perform graceful shutdown (as described in | |
1169 | * Section 10.1 of [SCTP]) on ALL the associations currently represented | |
1170 | * by a UDP-style socket. | |
1171 | * | |
1172 | * The syntax is | |
1173 | * | |
1174 | * ret = close(int sd); | |
1175 | * | |
1176 | * sd - the socket descriptor of the associations to be closed. | |
1177 | * | |
1178 | * To gracefully shutdown a specific association represented by the | |
1179 | * UDP-style socket, an application should use the sendmsg() call, | |
1180 | * passing no user data, but including the appropriate flag in the | |
1181 | * ancillary data (see Section xxxx). | |
1182 | * | |
1183 | * If sd in the close() call is a branched-off socket representing only | |
1184 | * one association, the shutdown is performed on that association only. | |
1185 | * | |
1186 | * 4.1.6 close() - TCP Style Syntax | |
1187 | * | |
1188 | * Applications use close() to gracefully close down an association. | |
1189 | * | |
1190 | * The syntax is: | |
1191 | * | |
1192 | * int close(int sd); | |
1193 | * | |
1194 | * sd - the socket descriptor of the association to be closed. | |
1195 | * | |
1196 | * After an application calls close() on a socket descriptor, no further | |
1197 | * socket operations will succeed on that descriptor. | |
1198 | * | |
1199 | * API 7.1.4 SO_LINGER | |
1200 | * | |
1201 | * An application using the TCP-style socket can use this option to | |
1202 | * perform the SCTP ABORT primitive. The linger option structure is: | |
1203 | * | |
1204 | * struct linger { | |
1205 | * int l_onoff; // option on/off | |
1206 | * int l_linger; // linger time | |
1207 | * }; | |
1208 | * | |
1209 | * To enable the option, set l_onoff to 1. If the l_linger value is set | |
1210 | * to 0, calling close() is the same as the ABORT primitive. If the | |
1211 | * value is set to a negative value, the setsockopt() call will return | |
1212 | * an error. If the value is set to a positive value linger_time, the | |
1213 | * close() can be blocked for at most linger_time ms. If the graceful | |
1214 | * shutdown phase does not finish during this period, close() will | |
1215 | * return but the graceful shutdown phase continues in the system. | |
1216 | */ | |
1217 | SCTP_STATIC void sctp_close(struct sock *sk, long timeout) | |
1218 | { | |
1219 | struct sctp_endpoint *ep; | |
1220 | struct sctp_association *asoc; | |
1221 | struct list_head *pos, *temp; | |
1222 | ||
1223 | SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout); | |
1224 | ||
1225 | sctp_lock_sock(sk); | |
1226 | sk->sk_shutdown = SHUTDOWN_MASK; | |
1227 | ||
1228 | ep = sctp_sk(sk)->ep; | |
1229 | ||
1230 | /* Walk all associations on a socket, not on an endpoint. */ | |
1231 | list_for_each_safe(pos, temp, &ep->asocs) { | |
1232 | asoc = list_entry(pos, struct sctp_association, asocs); | |
1233 | ||
1234 | if (sctp_style(sk, TCP)) { | |
1235 | /* A closed association can still be in the list if | |
1236 | * it belongs to a TCP-style listening socket that is | |
1237 | * not yet accepted. If so, free it. If not, send an | |
1238 | * ABORT or SHUTDOWN based on the linger options. | |
1239 | */ | |
1240 | if (sctp_state(asoc, CLOSED)) { | |
1241 | sctp_unhash_established(asoc); | |
1242 | sctp_association_free(asoc); | |
1243 | ||
1244 | } else if (sock_flag(sk, SOCK_LINGER) && | |
1245 | !sk->sk_lingertime) | |
1246 | sctp_primitive_ABORT(asoc, NULL); | |
1247 | else | |
1248 | sctp_primitive_SHUTDOWN(asoc, NULL); | |
1249 | } else | |
1250 | sctp_primitive_SHUTDOWN(asoc, NULL); | |
1251 | } | |
1252 | ||
1253 | /* Clean up any skbs sitting on the receive queue. */ | |
1254 | sctp_queue_purge_ulpevents(&sk->sk_receive_queue); | |
1255 | sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby); | |
1256 | ||
1257 | /* On a TCP-style socket, block for at most linger_time if set. */ | |
1258 | if (sctp_style(sk, TCP) && timeout) | |
1259 | sctp_wait_for_close(sk, timeout); | |
1260 | ||
1261 | /* This will run the backlog queue. */ | |
1262 | sctp_release_sock(sk); | |
1263 | ||
1264 | /* Supposedly, no process has access to the socket, but | |
1265 | * the net layers still may. | |
1266 | */ | |
1267 | sctp_local_bh_disable(); | |
1268 | sctp_bh_lock_sock(sk); | |
1269 | ||
1270 | /* Hold the sock, since sk_common_release() will put sock_put() | |
1271 | * and we have just a little more cleanup. | |
1272 | */ | |
1273 | sock_hold(sk); | |
1274 | sk_common_release(sk); | |
1275 | ||
1276 | sctp_bh_unlock_sock(sk); | |
1277 | sctp_local_bh_enable(); | |
1278 | ||
1279 | sock_put(sk); | |
1280 | ||
1281 | SCTP_DBG_OBJCNT_DEC(sock); | |
1282 | } | |
1283 | ||
1284 | /* Handle EPIPE error. */ | |
1285 | static int sctp_error(struct sock *sk, int flags, int err) | |
1286 | { | |
1287 | if (err == -EPIPE) | |
1288 | err = sock_error(sk) ? : -EPIPE; | |
1289 | if (err == -EPIPE && !(flags & MSG_NOSIGNAL)) | |
1290 | send_sig(SIGPIPE, current, 0); | |
1291 | return err; | |
1292 | } | |
1293 | ||
1294 | /* API 3.1.3 sendmsg() - UDP Style Syntax | |
1295 | * | |
1296 | * An application uses sendmsg() and recvmsg() calls to transmit data to | |
1297 | * and receive data from its peer. | |
1298 | * | |
1299 | * ssize_t sendmsg(int socket, const struct msghdr *message, | |
1300 | * int flags); | |
1301 | * | |
1302 | * socket - the socket descriptor of the endpoint. | |
1303 | * message - pointer to the msghdr structure which contains a single | |
1304 | * user message and possibly some ancillary data. | |
1305 | * | |
1306 | * See Section 5 for complete description of the data | |
1307 | * structures. | |
1308 | * | |
1309 | * flags - flags sent or received with the user message, see Section | |
1310 | * 5 for complete description of the flags. | |
1311 | * | |
1312 | * Note: This function could use a rewrite especially when explicit | |
1313 | * connect support comes in. | |
1314 | */ | |
1315 | /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */ | |
1316 | ||
1317 | SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *); | |
1318 | ||
1319 | SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk, | |
1320 | struct msghdr *msg, size_t msg_len) | |
1321 | { | |
1322 | struct sctp_sock *sp; | |
1323 | struct sctp_endpoint *ep; | |
1324 | struct sctp_association *new_asoc=NULL, *asoc=NULL; | |
1325 | struct sctp_transport *transport, *chunk_tp; | |
1326 | struct sctp_chunk *chunk; | |
1327 | union sctp_addr to; | |
1328 | struct sockaddr *msg_name = NULL; | |
1329 | struct sctp_sndrcvinfo default_sinfo = { 0 }; | |
1330 | struct sctp_sndrcvinfo *sinfo; | |
1331 | struct sctp_initmsg *sinit; | |
1332 | sctp_assoc_t associd = 0; | |
1333 | sctp_cmsgs_t cmsgs = { NULL }; | |
1334 | int err; | |
1335 | sctp_scope_t scope; | |
1336 | long timeo; | |
1337 | __u16 sinfo_flags = 0; | |
1338 | struct sctp_datamsg *datamsg; | |
1339 | struct list_head *pos; | |
1340 | int msg_flags = msg->msg_flags; | |
1341 | ||
1342 | SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n", | |
1343 | sk, msg, msg_len); | |
1344 | ||
1345 | err = 0; | |
1346 | sp = sctp_sk(sk); | |
1347 | ep = sp->ep; | |
1348 | ||
3f7a87d2 | 1349 | SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep); |
1da177e4 LT |
1350 | |
1351 | /* We cannot send a message over a TCP-style listening socket. */ | |
1352 | if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) { | |
1353 | err = -EPIPE; | |
1354 | goto out_nounlock; | |
1355 | } | |
1356 | ||
1357 | /* Parse out the SCTP CMSGs. */ | |
1358 | err = sctp_msghdr_parse(msg, &cmsgs); | |
1359 | ||
1360 | if (err) { | |
1361 | SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err); | |
1362 | goto out_nounlock; | |
1363 | } | |
1364 | ||
1365 | /* Fetch the destination address for this packet. This | |
1366 | * address only selects the association--it is not necessarily | |
1367 | * the address we will send to. | |
1368 | * For a peeled-off socket, msg_name is ignored. | |
1369 | */ | |
1370 | if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) { | |
1371 | int msg_namelen = msg->msg_namelen; | |
1372 | ||
1373 | err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name, | |
1374 | msg_namelen); | |
1375 | if (err) | |
1376 | return err; | |
1377 | ||
1378 | if (msg_namelen > sizeof(to)) | |
1379 | msg_namelen = sizeof(to); | |
1380 | memcpy(&to, msg->msg_name, msg_namelen); | |
1381 | SCTP_DEBUG_PRINTK("Just memcpy'd. msg_name is " | |
1382 | "0x%x:%u.\n", | |
1383 | to.v4.sin_addr.s_addr, to.v4.sin_port); | |
1384 | ||
1385 | to.v4.sin_port = ntohs(to.v4.sin_port); | |
1386 | msg_name = msg->msg_name; | |
1387 | } | |
1388 | ||
1389 | sinfo = cmsgs.info; | |
1390 | sinit = cmsgs.init; | |
1391 | ||
1392 | /* Did the user specify SNDRCVINFO? */ | |
1393 | if (sinfo) { | |
1394 | sinfo_flags = sinfo->sinfo_flags; | |
1395 | associd = sinfo->sinfo_assoc_id; | |
1396 | } | |
1397 | ||
1398 | SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n", | |
1399 | msg_len, sinfo_flags); | |
1400 | ||
eaa5c54d ISJ |
1401 | /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */ |
1402 | if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) { | |
1da177e4 LT |
1403 | err = -EINVAL; |
1404 | goto out_nounlock; | |
1405 | } | |
1406 | ||
eaa5c54d ISJ |
1407 | /* If SCTP_EOF is set, no data can be sent. Disallow sending zero |
1408 | * length messages when SCTP_EOF|SCTP_ABORT is not set. | |
1409 | * If SCTP_ABORT is set, the message length could be non zero with | |
1da177e4 LT |
1410 | * the msg_iov set to the user abort reason. |
1411 | */ | |
eaa5c54d ISJ |
1412 | if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) || |
1413 | (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) { | |
1da177e4 LT |
1414 | err = -EINVAL; |
1415 | goto out_nounlock; | |
1416 | } | |
1417 | ||
eaa5c54d | 1418 | /* If SCTP_ADDR_OVER is set, there must be an address |
1da177e4 LT |
1419 | * specified in msg_name. |
1420 | */ | |
eaa5c54d | 1421 | if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) { |
1da177e4 LT |
1422 | err = -EINVAL; |
1423 | goto out_nounlock; | |
1424 | } | |
1425 | ||
1426 | transport = NULL; | |
1427 | ||
1428 | SCTP_DEBUG_PRINTK("About to look up association.\n"); | |
1429 | ||
1430 | sctp_lock_sock(sk); | |
1431 | ||
1432 | /* If a msg_name has been specified, assume this is to be used. */ | |
1433 | if (msg_name) { | |
1434 | /* Look for a matching association on the endpoint. */ | |
1435 | asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport); | |
1436 | if (!asoc) { | |
1437 | /* If we could not find a matching association on the | |
1438 | * endpoint, make sure that it is not a TCP-style | |
1439 | * socket that already has an association or there is | |
1440 | * no peeled-off association on another socket. | |
1441 | */ | |
1442 | if ((sctp_style(sk, TCP) && | |
1443 | sctp_sstate(sk, ESTABLISHED)) || | |
1444 | sctp_endpoint_is_peeled_off(ep, &to)) { | |
1445 | err = -EADDRNOTAVAIL; | |
1446 | goto out_unlock; | |
1447 | } | |
1448 | } | |
1449 | } else { | |
1450 | asoc = sctp_id2assoc(sk, associd); | |
1451 | if (!asoc) { | |
1452 | err = -EPIPE; | |
1453 | goto out_unlock; | |
1454 | } | |
1455 | } | |
1456 | ||
1457 | if (asoc) { | |
1458 | SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc); | |
1459 | ||
1460 | /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED | |
1461 | * socket that has an association in CLOSED state. This can | |
1462 | * happen when an accepted socket has an association that is | |
1463 | * already CLOSED. | |
1464 | */ | |
1465 | if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) { | |
1466 | err = -EPIPE; | |
1467 | goto out_unlock; | |
1468 | } | |
1469 | ||
eaa5c54d | 1470 | if (sinfo_flags & SCTP_EOF) { |
1da177e4 LT |
1471 | SCTP_DEBUG_PRINTK("Shutting down association: %p\n", |
1472 | asoc); | |
1473 | sctp_primitive_SHUTDOWN(asoc, NULL); | |
1474 | err = 0; | |
1475 | goto out_unlock; | |
1476 | } | |
eaa5c54d | 1477 | if (sinfo_flags & SCTP_ABORT) { |
1da177e4 LT |
1478 | SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc); |
1479 | sctp_primitive_ABORT(asoc, msg); | |
1480 | err = 0; | |
1481 | goto out_unlock; | |
1482 | } | |
1483 | } | |
1484 | ||
1485 | /* Do we need to create the association? */ | |
1486 | if (!asoc) { | |
1487 | SCTP_DEBUG_PRINTK("There is no association yet.\n"); | |
1488 | ||
eaa5c54d | 1489 | if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) { |
1da177e4 LT |
1490 | err = -EINVAL; |
1491 | goto out_unlock; | |
1492 | } | |
1493 | ||
1494 | /* Check for invalid stream against the stream counts, | |
1495 | * either the default or the user specified stream counts. | |
1496 | */ | |
1497 | if (sinfo) { | |
1498 | if (!sinit || (sinit && !sinit->sinit_num_ostreams)) { | |
1499 | /* Check against the defaults. */ | |
1500 | if (sinfo->sinfo_stream >= | |
1501 | sp->initmsg.sinit_num_ostreams) { | |
1502 | err = -EINVAL; | |
1503 | goto out_unlock; | |
1504 | } | |
1505 | } else { | |
1506 | /* Check against the requested. */ | |
1507 | if (sinfo->sinfo_stream >= | |
1508 | sinit->sinit_num_ostreams) { | |
1509 | err = -EINVAL; | |
1510 | goto out_unlock; | |
1511 | } | |
1512 | } | |
1513 | } | |
1514 | ||
1515 | /* | |
1516 | * API 3.1.2 bind() - UDP Style Syntax | |
1517 | * If a bind() or sctp_bindx() is not called prior to a | |
1518 | * sendmsg() call that initiates a new association, the | |
1519 | * system picks an ephemeral port and will choose an address | |
1520 | * set equivalent to binding with a wildcard address. | |
1521 | */ | |
1522 | if (!ep->base.bind_addr.port) { | |
1523 | if (sctp_autobind(sk)) { | |
1524 | err = -EAGAIN; | |
1525 | goto out_unlock; | |
1526 | } | |
64a0c1c8 ISJ |
1527 | } else { |
1528 | /* | |
1529 | * If an unprivileged user inherits a one-to-many | |
1530 | * style socket with open associations on a privileged | |
1531 | * port, it MAY be permitted to accept new associations, | |
1532 | * but it SHOULD NOT be permitted to open new | |
1533 | * associations. | |
1534 | */ | |
1535 | if (ep->base.bind_addr.port < PROT_SOCK && | |
1536 | !capable(CAP_NET_BIND_SERVICE)) { | |
1537 | err = -EACCES; | |
1538 | goto out_unlock; | |
1539 | } | |
1da177e4 LT |
1540 | } |
1541 | ||
1542 | scope = sctp_scope(&to); | |
1543 | new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL); | |
1544 | if (!new_asoc) { | |
1545 | err = -ENOMEM; | |
1546 | goto out_unlock; | |
1547 | } | |
1548 | asoc = new_asoc; | |
1549 | ||
1550 | /* If the SCTP_INIT ancillary data is specified, set all | |
1551 | * the association init values accordingly. | |
1552 | */ | |
1553 | if (sinit) { | |
1554 | if (sinit->sinit_num_ostreams) { | |
1555 | asoc->c.sinit_num_ostreams = | |
1556 | sinit->sinit_num_ostreams; | |
1557 | } | |
1558 | if (sinit->sinit_max_instreams) { | |
1559 | asoc->c.sinit_max_instreams = | |
1560 | sinit->sinit_max_instreams; | |
1561 | } | |
1562 | if (sinit->sinit_max_attempts) { | |
1563 | asoc->max_init_attempts | |
1564 | = sinit->sinit_max_attempts; | |
1565 | } | |
1566 | if (sinit->sinit_max_init_timeo) { | |
1567 | asoc->max_init_timeo = | |
1568 | msecs_to_jiffies(sinit->sinit_max_init_timeo); | |
1569 | } | |
1570 | } | |
1571 | ||
1572 | /* Prime the peer's transport structures. */ | |
3f7a87d2 | 1573 | transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN); |
1da177e4 LT |
1574 | if (!transport) { |
1575 | err = -ENOMEM; | |
1576 | goto out_free; | |
1577 | } | |
1578 | err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL); | |
1579 | if (err < 0) { | |
1580 | err = -ENOMEM; | |
1581 | goto out_free; | |
1582 | } | |
1583 | } | |
1584 | ||
1585 | /* ASSERT: we have a valid association at this point. */ | |
1586 | SCTP_DEBUG_PRINTK("We have a valid association.\n"); | |
1587 | ||
1588 | if (!sinfo) { | |
1589 | /* If the user didn't specify SNDRCVINFO, make up one with | |
1590 | * some defaults. | |
1591 | */ | |
1592 | default_sinfo.sinfo_stream = asoc->default_stream; | |
1593 | default_sinfo.sinfo_flags = asoc->default_flags; | |
1594 | default_sinfo.sinfo_ppid = asoc->default_ppid; | |
1595 | default_sinfo.sinfo_context = asoc->default_context; | |
1596 | default_sinfo.sinfo_timetolive = asoc->default_timetolive; | |
1597 | default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc); | |
1598 | sinfo = &default_sinfo; | |
1599 | } | |
1600 | ||
1601 | /* API 7.1.7, the sndbuf size per association bounds the | |
1602 | * maximum size of data that can be sent in a single send call. | |
1603 | */ | |
1604 | if (msg_len > sk->sk_sndbuf) { | |
1605 | err = -EMSGSIZE; | |
1606 | goto out_free; | |
1607 | } | |
1608 | ||
1609 | /* If fragmentation is disabled and the message length exceeds the | |
1610 | * association fragmentation point, return EMSGSIZE. The I-D | |
1611 | * does not specify what this error is, but this looks like | |
1612 | * a great fit. | |
1613 | */ | |
1614 | if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) { | |
1615 | err = -EMSGSIZE; | |
1616 | goto out_free; | |
1617 | } | |
1618 | ||
1619 | if (sinfo) { | |
1620 | /* Check for invalid stream. */ | |
1621 | if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) { | |
1622 | err = -EINVAL; | |
1623 | goto out_free; | |
1624 | } | |
1625 | } | |
1626 | ||
1627 | timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT); | |
1628 | if (!sctp_wspace(asoc)) { | |
1629 | err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len); | |
1630 | if (err) | |
1631 | goto out_free; | |
1632 | } | |
1633 | ||
1634 | /* If an address is passed with the sendto/sendmsg call, it is used | |
1635 | * to override the primary destination address in the TCP model, or | |
eaa5c54d | 1636 | * when SCTP_ADDR_OVER flag is set in the UDP model. |
1da177e4 LT |
1637 | */ |
1638 | if ((sctp_style(sk, TCP) && msg_name) || | |
eaa5c54d | 1639 | (sinfo_flags & SCTP_ADDR_OVER)) { |
1da177e4 LT |
1640 | chunk_tp = sctp_assoc_lookup_paddr(asoc, &to); |
1641 | if (!chunk_tp) { | |
1642 | err = -EINVAL; | |
1643 | goto out_free; | |
1644 | } | |
1645 | } else | |
1646 | chunk_tp = NULL; | |
1647 | ||
1648 | /* Auto-connect, if we aren't connected already. */ | |
1649 | if (sctp_state(asoc, CLOSED)) { | |
1650 | err = sctp_primitive_ASSOCIATE(asoc, NULL); | |
1651 | if (err < 0) | |
1652 | goto out_free; | |
1653 | SCTP_DEBUG_PRINTK("We associated primitively.\n"); | |
1654 | } | |
1655 | ||
1656 | /* Break the message into multiple chunks of maximum size. */ | |
1657 | datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len); | |
1658 | if (!datamsg) { | |
1659 | err = -ENOMEM; | |
1660 | goto out_free; | |
1661 | } | |
1662 | ||
1663 | /* Now send the (possibly) fragmented message. */ | |
1664 | list_for_each(pos, &datamsg->chunks) { | |
1665 | chunk = list_entry(pos, struct sctp_chunk, frag_list); | |
1666 | sctp_datamsg_track(chunk); | |
1667 | ||
1668 | /* Do accounting for the write space. */ | |
1669 | sctp_set_owner_w(chunk); | |
1670 | ||
1671 | chunk->transport = chunk_tp; | |
1672 | ||
1673 | /* Send it to the lower layers. Note: all chunks | |
1674 | * must either fail or succeed. The lower layer | |
1675 | * works that way today. Keep it that way or this | |
1676 | * breaks. | |
1677 | */ | |
1678 | err = sctp_primitive_SEND(asoc, chunk); | |
1679 | /* Did the lower layer accept the chunk? */ | |
1680 | if (err) | |
1681 | sctp_chunk_free(chunk); | |
1682 | SCTP_DEBUG_PRINTK("We sent primitively.\n"); | |
1683 | } | |
1684 | ||
1685 | sctp_datamsg_free(datamsg); | |
1686 | if (err) | |
1687 | goto out_free; | |
1688 | else | |
1689 | err = msg_len; | |
1690 | ||
1691 | /* If we are already past ASSOCIATE, the lower | |
1692 | * layers are responsible for association cleanup. | |
1693 | */ | |
1694 | goto out_unlock; | |
1695 | ||
1696 | out_free: | |
1697 | if (new_asoc) | |
1698 | sctp_association_free(asoc); | |
1699 | out_unlock: | |
1700 | sctp_release_sock(sk); | |
1701 | ||
1702 | out_nounlock: | |
1703 | return sctp_error(sk, msg_flags, err); | |
1704 | ||
1705 | #if 0 | |
1706 | do_sock_err: | |
1707 | if (msg_len) | |
1708 | err = msg_len; | |
1709 | else | |
1710 | err = sock_error(sk); | |
1711 | goto out; | |
1712 | ||
1713 | do_interrupted: | |
1714 | if (msg_len) | |
1715 | err = msg_len; | |
1716 | goto out; | |
1717 | #endif /* 0 */ | |
1718 | } | |
1719 | ||
1720 | /* This is an extended version of skb_pull() that removes the data from the | |
1721 | * start of a skb even when data is spread across the list of skb's in the | |
1722 | * frag_list. len specifies the total amount of data that needs to be removed. | |
1723 | * when 'len' bytes could be removed from the skb, it returns 0. | |
1724 | * If 'len' exceeds the total skb length, it returns the no. of bytes that | |
1725 | * could not be removed. | |
1726 | */ | |
1727 | static int sctp_skb_pull(struct sk_buff *skb, int len) | |
1728 | { | |
1729 | struct sk_buff *list; | |
1730 | int skb_len = skb_headlen(skb); | |
1731 | int rlen; | |
1732 | ||
1733 | if (len <= skb_len) { | |
1734 | __skb_pull(skb, len); | |
1735 | return 0; | |
1736 | } | |
1737 | len -= skb_len; | |
1738 | __skb_pull(skb, skb_len); | |
1739 | ||
1740 | for (list = skb_shinfo(skb)->frag_list; list; list = list->next) { | |
1741 | rlen = sctp_skb_pull(list, len); | |
1742 | skb->len -= (len-rlen); | |
1743 | skb->data_len -= (len-rlen); | |
1744 | ||
1745 | if (!rlen) | |
1746 | return 0; | |
1747 | ||
1748 | len = rlen; | |
1749 | } | |
1750 | ||
1751 | return len; | |
1752 | } | |
1753 | ||
1754 | /* API 3.1.3 recvmsg() - UDP Style Syntax | |
1755 | * | |
1756 | * ssize_t recvmsg(int socket, struct msghdr *message, | |
1757 | * int flags); | |
1758 | * | |
1759 | * socket - the socket descriptor of the endpoint. | |
1760 | * message - pointer to the msghdr structure which contains a single | |
1761 | * user message and possibly some ancillary data. | |
1762 | * | |
1763 | * See Section 5 for complete description of the data | |
1764 | * structures. | |
1765 | * | |
1766 | * flags - flags sent or received with the user message, see Section | |
1767 | * 5 for complete description of the flags. | |
1768 | */ | |
1769 | static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *); | |
1770 | ||
1771 | SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk, | |
1772 | struct msghdr *msg, size_t len, int noblock, | |
1773 | int flags, int *addr_len) | |
1774 | { | |
1775 | struct sctp_ulpevent *event = NULL; | |
1776 | struct sctp_sock *sp = sctp_sk(sk); | |
1777 | struct sk_buff *skb; | |
1778 | int copied; | |
1779 | int err = 0; | |
1780 | int skb_len; | |
1781 | ||
1782 | SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: " | |
1783 | "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg, | |
1784 | "len", len, "knoblauch", noblock, | |
1785 | "flags", flags, "addr_len", addr_len); | |
1786 | ||
1787 | sctp_lock_sock(sk); | |
1788 | ||
1789 | if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) { | |
1790 | err = -ENOTCONN; | |
1791 | goto out; | |
1792 | } | |
1793 | ||
1794 | skb = sctp_skb_recv_datagram(sk, flags, noblock, &err); | |
1795 | if (!skb) | |
1796 | goto out; | |
1797 | ||
1798 | /* Get the total length of the skb including any skb's in the | |
1799 | * frag_list. | |
1800 | */ | |
1801 | skb_len = skb->len; | |
1802 | ||
1803 | copied = skb_len; | |
1804 | if (copied > len) | |
1805 | copied = len; | |
1806 | ||
1807 | err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); | |
1808 | ||
1809 | event = sctp_skb2event(skb); | |
1810 | ||
1811 | if (err) | |
1812 | goto out_free; | |
1813 | ||
1814 | sock_recv_timestamp(msg, sk, skb); | |
1815 | if (sctp_ulpevent_is_notification(event)) { | |
1816 | msg->msg_flags |= MSG_NOTIFICATION; | |
1817 | sp->pf->event_msgname(event, msg->msg_name, addr_len); | |
1818 | } else { | |
1819 | sp->pf->skb_msgname(skb, msg->msg_name, addr_len); | |
1820 | } | |
1821 | ||
1822 | /* Check if we allow SCTP_SNDRCVINFO. */ | |
1823 | if (sp->subscribe.sctp_data_io_event) | |
1824 | sctp_ulpevent_read_sndrcvinfo(event, msg); | |
1825 | #if 0 | |
1826 | /* FIXME: we should be calling IP/IPv6 layers. */ | |
1827 | if (sk->sk_protinfo.af_inet.cmsg_flags) | |
1828 | ip_cmsg_recv(msg, skb); | |
1829 | #endif | |
1830 | ||
1831 | err = copied; | |
1832 | ||
1833 | /* If skb's length exceeds the user's buffer, update the skb and | |
1834 | * push it back to the receive_queue so that the next call to | |
1835 | * recvmsg() will return the remaining data. Don't set MSG_EOR. | |
1836 | */ | |
1837 | if (skb_len > copied) { | |
1838 | msg->msg_flags &= ~MSG_EOR; | |
1839 | if (flags & MSG_PEEK) | |
1840 | goto out_free; | |
1841 | sctp_skb_pull(skb, copied); | |
1842 | skb_queue_head(&sk->sk_receive_queue, skb); | |
1843 | ||
1844 | /* When only partial message is copied to the user, increase | |
1845 | * rwnd by that amount. If all the data in the skb is read, | |
1846 | * rwnd is updated when the event is freed. | |
1847 | */ | |
1848 | sctp_assoc_rwnd_increase(event->asoc, copied); | |
1849 | goto out; | |
1850 | } else if ((event->msg_flags & MSG_NOTIFICATION) || | |
1851 | (event->msg_flags & MSG_EOR)) | |
1852 | msg->msg_flags |= MSG_EOR; | |
1853 | else | |
1854 | msg->msg_flags &= ~MSG_EOR; | |
1855 | ||
1856 | out_free: | |
1857 | if (flags & MSG_PEEK) { | |
1858 | /* Release the skb reference acquired after peeking the skb in | |
1859 | * sctp_skb_recv_datagram(). | |
1860 | */ | |
1861 | kfree_skb(skb); | |
1862 | } else { | |
1863 | /* Free the event which includes releasing the reference to | |
1864 | * the owner of the skb, freeing the skb and updating the | |
1865 | * rwnd. | |
1866 | */ | |
1867 | sctp_ulpevent_free(event); | |
1868 | } | |
1869 | out: | |
1870 | sctp_release_sock(sk); | |
1871 | return err; | |
1872 | } | |
1873 | ||
1874 | /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS) | |
1875 | * | |
1876 | * This option is a on/off flag. If enabled no SCTP message | |
1877 | * fragmentation will be performed. Instead if a message being sent | |
1878 | * exceeds the current PMTU size, the message will NOT be sent and | |
1879 | * instead a error will be indicated to the user. | |
1880 | */ | |
1881 | static int sctp_setsockopt_disable_fragments(struct sock *sk, | |
1882 | char __user *optval, int optlen) | |
1883 | { | |
1884 | int val; | |
1885 | ||
1886 | if (optlen < sizeof(int)) | |
1887 | return -EINVAL; | |
1888 | ||
1889 | if (get_user(val, (int __user *)optval)) | |
1890 | return -EFAULT; | |
1891 | ||
1892 | sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1; | |
1893 | ||
1894 | return 0; | |
1895 | } | |
1896 | ||
1897 | static int sctp_setsockopt_events(struct sock *sk, char __user *optval, | |
1898 | int optlen) | |
1899 | { | |
1900 | if (optlen != sizeof(struct sctp_event_subscribe)) | |
1901 | return -EINVAL; | |
1902 | if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen)) | |
1903 | return -EFAULT; | |
1904 | return 0; | |
1905 | } | |
1906 | ||
1907 | /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE) | |
1908 | * | |
1909 | * This socket option is applicable to the UDP-style socket only. When | |
1910 | * set it will cause associations that are idle for more than the | |
1911 | * specified number of seconds to automatically close. An association | |
1912 | * being idle is defined an association that has NOT sent or received | |
1913 | * user data. The special value of '0' indicates that no automatic | |
1914 | * close of any associations should be performed. The option expects an | |
1915 | * integer defining the number of seconds of idle time before an | |
1916 | * association is closed. | |
1917 | */ | |
1918 | static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval, | |
1919 | int optlen) | |
1920 | { | |
1921 | struct sctp_sock *sp = sctp_sk(sk); | |
1922 | ||
1923 | /* Applicable to UDP-style socket only */ | |
1924 | if (sctp_style(sk, TCP)) | |
1925 | return -EOPNOTSUPP; | |
1926 | if (optlen != sizeof(int)) | |
1927 | return -EINVAL; | |
1928 | if (copy_from_user(&sp->autoclose, optval, optlen)) | |
1929 | return -EFAULT; | |
1930 | ||
1da177e4 LT |
1931 | return 0; |
1932 | } | |
1933 | ||
1934 | /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS) | |
1935 | * | |
1936 | * Applications can enable or disable heartbeats for any peer address of | |
1937 | * an association, modify an address's heartbeat interval, force a | |
1938 | * heartbeat to be sent immediately, and adjust the address's maximum | |
1939 | * number of retransmissions sent before an address is considered | |
1940 | * unreachable. The following structure is used to access and modify an | |
1941 | * address's parameters: | |
1942 | * | |
1943 | * struct sctp_paddrparams { | |
1944 | * sctp_assoc_t spp_assoc_id; | |
1945 | * struct sockaddr_storage spp_address; | |
1946 | * uint32_t spp_hbinterval; | |
1947 | * uint16_t spp_pathmaxrxt; | |
1948 | * }; | |
1949 | * | |
1950 | * spp_assoc_id - (UDP style socket) This is filled in the application, | |
1951 | * and identifies the association for this query. | |
1952 | * spp_address - This specifies which address is of interest. | |
1953 | * spp_hbinterval - This contains the value of the heartbeat interval, | |
1954 | * in milliseconds. A value of 0, when modifying the | |
1955 | * parameter, specifies that the heartbeat on this | |
1956 | * address should be disabled. A value of UINT32_MAX | |
1957 | * (4294967295), when modifying the parameter, | |
1958 | * specifies that a heartbeat should be sent | |
1959 | * immediately to the peer address, and the current | |
1960 | * interval should remain unchanged. | |
1961 | * spp_pathmaxrxt - This contains the maximum number of | |
1962 | * retransmissions before this address shall be | |
1963 | * considered unreachable. | |
1964 | */ | |
1965 | static int sctp_setsockopt_peer_addr_params(struct sock *sk, | |
1966 | char __user *optval, int optlen) | |
1967 | { | |
1968 | struct sctp_paddrparams params; | |
1969 | struct sctp_transport *trans; | |
1970 | int error; | |
1971 | ||
1972 | if (optlen != sizeof(struct sctp_paddrparams)) | |
1973 | return -EINVAL; | |
1974 | if (copy_from_user(¶ms, optval, optlen)) | |
1975 | return -EFAULT; | |
1976 | ||
1977 | /* | |
1978 | * API 7. Socket Options (setting the default value for the endpoint) | |
1979 | * All options that support specific settings on an association by | |
1980 | * filling in either an association id variable or a sockaddr_storage | |
1981 | * SHOULD also support setting of the same value for the entire endpoint | |
1982 | * (i.e. future associations). To accomplish this the following logic is | |
1983 | * used when setting one of these options: | |
1984 | ||
1985 | * c) If neither the sockaddr_storage or association identification is | |
1986 | * set i.e. the sockaddr_storage is set to all 0's (INADDR_ANY) and | |
1987 | * the association identification is 0, the settings are a default | |
1988 | * and to be applied to the endpoint (all future associations). | |
1989 | */ | |
1990 | ||
1991 | /* update default value for endpoint (all future associations) */ | |
1992 | if (!params.spp_assoc_id && | |
1993 | sctp_is_any(( union sctp_addr *)¶ms.spp_address)) { | |
1994 | /* Manual heartbeat on an endpoint is invalid. */ | |
1995 | if (0xffffffff == params.spp_hbinterval) | |
1996 | return -EINVAL; | |
1997 | else if (params.spp_hbinterval) | |
1998 | sctp_sk(sk)->paddrparam.spp_hbinterval = | |
1999 | params.spp_hbinterval; | |
2000 | if (params.spp_pathmaxrxt) | |
2001 | sctp_sk(sk)->paddrparam.spp_pathmaxrxt = | |
2002 | params.spp_pathmaxrxt; | |
2003 | return 0; | |
2004 | } | |
2005 | ||
2006 | trans = sctp_addr_id2transport(sk, ¶ms.spp_address, | |
2007 | params.spp_assoc_id); | |
2008 | if (!trans) | |
2009 | return -EINVAL; | |
2010 | ||
2011 | /* Applications can enable or disable heartbeats for any peer address | |
2012 | * of an association, modify an address's heartbeat interval, force a | |
2013 | * heartbeat to be sent immediately, and adjust the address's maximum | |
2014 | * number of retransmissions sent before an address is considered | |
2015 | * unreachable. | |
2016 | * | |
2017 | * The value of the heartbeat interval, in milliseconds. A value of | |
2018 | * UINT32_MAX (4294967295), when modifying the parameter, specifies | |
2019 | * that a heartbeat should be sent immediately to the peer address, | |
2020 | * and the current interval should remain unchanged. | |
2021 | */ | |
2022 | if (0xffffffff == params.spp_hbinterval) { | |
2023 | error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans); | |
2024 | if (error) | |
2025 | return error; | |
2026 | } else { | |
2027 | /* The value of the heartbeat interval, in milliseconds. A value of 0, | |
2028 | * when modifying the parameter, specifies that the heartbeat on this | |
2029 | * address should be disabled. | |
2030 | */ | |
2031 | if (params.spp_hbinterval) { | |
2032 | trans->hb_allowed = 1; | |
2033 | trans->hb_interval = | |
2034 | msecs_to_jiffies(params.spp_hbinterval); | |
2035 | } else | |
2036 | trans->hb_allowed = 0; | |
2037 | } | |
2038 | ||
2039 | /* spp_pathmaxrxt contains the maximum number of retransmissions | |
2040 | * before this address shall be considered unreachable. | |
2041 | */ | |
2042 | if (params.spp_pathmaxrxt) | |
2043 | trans->max_retrans = params.spp_pathmaxrxt; | |
2044 | ||
2045 | return 0; | |
2046 | } | |
2047 | ||
2048 | /* 7.1.3 Initialization Parameters (SCTP_INITMSG) | |
2049 | * | |
2050 | * Applications can specify protocol parameters for the default association | |
2051 | * initialization. The option name argument to setsockopt() and getsockopt() | |
2052 | * is SCTP_INITMSG. | |
2053 | * | |
2054 | * Setting initialization parameters is effective only on an unconnected | |
2055 | * socket (for UDP-style sockets only future associations are effected | |
2056 | * by the change). With TCP-style sockets, this option is inherited by | |
2057 | * sockets derived from a listener socket. | |
2058 | */ | |
2059 | static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, int optlen) | |
2060 | { | |
2061 | struct sctp_initmsg sinit; | |
2062 | struct sctp_sock *sp = sctp_sk(sk); | |
2063 | ||
2064 | if (optlen != sizeof(struct sctp_initmsg)) | |
2065 | return -EINVAL; | |
2066 | if (copy_from_user(&sinit, optval, optlen)) | |
2067 | return -EFAULT; | |
2068 | ||
2069 | if (sinit.sinit_num_ostreams) | |
2070 | sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams; | |
2071 | if (sinit.sinit_max_instreams) | |
2072 | sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams; | |
2073 | if (sinit.sinit_max_attempts) | |
2074 | sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts; | |
2075 | if (sinit.sinit_max_init_timeo) | |
2076 | sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo; | |
2077 | ||
2078 | return 0; | |
2079 | } | |
2080 | ||
2081 | /* | |
2082 | * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM) | |
2083 | * | |
2084 | * Applications that wish to use the sendto() system call may wish to | |
2085 | * specify a default set of parameters that would normally be supplied | |
2086 | * through the inclusion of ancillary data. This socket option allows | |
2087 | * such an application to set the default sctp_sndrcvinfo structure. | |
2088 | * The application that wishes to use this socket option simply passes | |
2089 | * in to this call the sctp_sndrcvinfo structure defined in Section | |
2090 | * 5.2.2) The input parameters accepted by this call include | |
2091 | * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context, | |
2092 | * sinfo_timetolive. The user must provide the sinfo_assoc_id field in | |
2093 | * to this call if the caller is using the UDP model. | |
2094 | */ | |
2095 | static int sctp_setsockopt_default_send_param(struct sock *sk, | |
2096 | char __user *optval, int optlen) | |
2097 | { | |
2098 | struct sctp_sndrcvinfo info; | |
2099 | struct sctp_association *asoc; | |
2100 | struct sctp_sock *sp = sctp_sk(sk); | |
2101 | ||
2102 | if (optlen != sizeof(struct sctp_sndrcvinfo)) | |
2103 | return -EINVAL; | |
2104 | if (copy_from_user(&info, optval, optlen)) | |
2105 | return -EFAULT; | |
2106 | ||
2107 | asoc = sctp_id2assoc(sk, info.sinfo_assoc_id); | |
2108 | if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP)) | |
2109 | return -EINVAL; | |
2110 | ||
2111 | if (asoc) { | |
2112 | asoc->default_stream = info.sinfo_stream; | |
2113 | asoc->default_flags = info.sinfo_flags; | |
2114 | asoc->default_ppid = info.sinfo_ppid; | |
2115 | asoc->default_context = info.sinfo_context; | |
2116 | asoc->default_timetolive = info.sinfo_timetolive; | |
2117 | } else { | |
2118 | sp->default_stream = info.sinfo_stream; | |
2119 | sp->default_flags = info.sinfo_flags; | |
2120 | sp->default_ppid = info.sinfo_ppid; | |
2121 | sp->default_context = info.sinfo_context; | |
2122 | sp->default_timetolive = info.sinfo_timetolive; | |
2123 | } | |
2124 | ||
2125 | return 0; | |
2126 | } | |
2127 | ||
2128 | /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR) | |
2129 | * | |
2130 | * Requests that the local SCTP stack use the enclosed peer address as | |
2131 | * the association primary. The enclosed address must be one of the | |
2132 | * association peer's addresses. | |
2133 | */ | |
2134 | static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval, | |
2135 | int optlen) | |
2136 | { | |
2137 | struct sctp_prim prim; | |
2138 | struct sctp_transport *trans; | |
2139 | ||
2140 | if (optlen != sizeof(struct sctp_prim)) | |
2141 | return -EINVAL; | |
2142 | ||
2143 | if (copy_from_user(&prim, optval, sizeof(struct sctp_prim))) | |
2144 | return -EFAULT; | |
2145 | ||
2146 | trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id); | |
2147 | if (!trans) | |
2148 | return -EINVAL; | |
2149 | ||
2150 | sctp_assoc_set_primary(trans->asoc, trans); | |
2151 | ||
2152 | return 0; | |
2153 | } | |
2154 | ||
2155 | /* | |
2156 | * 7.1.5 SCTP_NODELAY | |
2157 | * | |
2158 | * Turn on/off any Nagle-like algorithm. This means that packets are | |
2159 | * generally sent as soon as possible and no unnecessary delays are | |
2160 | * introduced, at the cost of more packets in the network. Expects an | |
2161 | * integer boolean flag. | |
2162 | */ | |
2163 | static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval, | |
2164 | int optlen) | |
2165 | { | |
2166 | int val; | |
2167 | ||
2168 | if (optlen < sizeof(int)) | |
2169 | return -EINVAL; | |
2170 | if (get_user(val, (int __user *)optval)) | |
2171 | return -EFAULT; | |
2172 | ||
2173 | sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1; | |
2174 | return 0; | |
2175 | } | |
2176 | ||
2177 | /* | |
2178 | * | |
2179 | * 7.1.1 SCTP_RTOINFO | |
2180 | * | |
2181 | * The protocol parameters used to initialize and bound retransmission | |
2182 | * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access | |
2183 | * and modify these parameters. | |
2184 | * All parameters are time values, in milliseconds. A value of 0, when | |
2185 | * modifying the parameters, indicates that the current value should not | |
2186 | * be changed. | |
2187 | * | |
2188 | */ | |
2189 | static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, int optlen) { | |
2190 | struct sctp_rtoinfo rtoinfo; | |
2191 | struct sctp_association *asoc; | |
2192 | ||
2193 | if (optlen != sizeof (struct sctp_rtoinfo)) | |
2194 | return -EINVAL; | |
2195 | ||
2196 | if (copy_from_user(&rtoinfo, optval, optlen)) | |
2197 | return -EFAULT; | |
2198 | ||
2199 | asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id); | |
2200 | ||
2201 | /* Set the values to the specific association */ | |
2202 | if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP)) | |
2203 | return -EINVAL; | |
2204 | ||
2205 | if (asoc) { | |
2206 | if (rtoinfo.srto_initial != 0) | |
2207 | asoc->rto_initial = | |
2208 | msecs_to_jiffies(rtoinfo.srto_initial); | |
2209 | if (rtoinfo.srto_max != 0) | |
2210 | asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max); | |
2211 | if (rtoinfo.srto_min != 0) | |
2212 | asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min); | |
2213 | } else { | |
2214 | /* If there is no association or the association-id = 0 | |
2215 | * set the values to the endpoint. | |
2216 | */ | |
2217 | struct sctp_sock *sp = sctp_sk(sk); | |
2218 | ||
2219 | if (rtoinfo.srto_initial != 0) | |
2220 | sp->rtoinfo.srto_initial = rtoinfo.srto_initial; | |
2221 | if (rtoinfo.srto_max != 0) | |
2222 | sp->rtoinfo.srto_max = rtoinfo.srto_max; | |
2223 | if (rtoinfo.srto_min != 0) | |
2224 | sp->rtoinfo.srto_min = rtoinfo.srto_min; | |
2225 | } | |
2226 | ||
2227 | return 0; | |
2228 | } | |
2229 | ||
2230 | /* | |
2231 | * | |
2232 | * 7.1.2 SCTP_ASSOCINFO | |
2233 | * | |
2234 | * This option is used to tune the the maximum retransmission attempts | |
2235 | * of the association. | |
2236 | * Returns an error if the new association retransmission value is | |
2237 | * greater than the sum of the retransmission value of the peer. | |
2238 | * See [SCTP] for more information. | |
2239 | * | |
2240 | */ | |
2241 | static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, int optlen) | |
2242 | { | |
2243 | ||
2244 | struct sctp_assocparams assocparams; | |
2245 | struct sctp_association *asoc; | |
2246 | ||
2247 | if (optlen != sizeof(struct sctp_assocparams)) | |
2248 | return -EINVAL; | |
2249 | if (copy_from_user(&assocparams, optval, optlen)) | |
2250 | return -EFAULT; | |
2251 | ||
2252 | asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id); | |
2253 | ||
2254 | if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP)) | |
2255 | return -EINVAL; | |
2256 | ||
2257 | /* Set the values to the specific association */ | |
2258 | if (asoc) { | |
2259 | if (assocparams.sasoc_asocmaxrxt != 0) | |
2260 | asoc->max_retrans = assocparams.sasoc_asocmaxrxt; | |
2261 | if (assocparams.sasoc_cookie_life != 0) { | |
2262 | asoc->cookie_life.tv_sec = | |
2263 | assocparams.sasoc_cookie_life / 1000; | |
2264 | asoc->cookie_life.tv_usec = | |
2265 | (assocparams.sasoc_cookie_life % 1000) | |
2266 | * 1000; | |
2267 | } | |
2268 | } else { | |
2269 | /* Set the values to the endpoint */ | |
2270 | struct sctp_sock *sp = sctp_sk(sk); | |
2271 | ||
2272 | if (assocparams.sasoc_asocmaxrxt != 0) | |
2273 | sp->assocparams.sasoc_asocmaxrxt = | |
2274 | assocparams.sasoc_asocmaxrxt; | |
2275 | if (assocparams.sasoc_cookie_life != 0) | |
2276 | sp->assocparams.sasoc_cookie_life = | |
2277 | assocparams.sasoc_cookie_life; | |
2278 | } | |
2279 | return 0; | |
2280 | } | |
2281 | ||
2282 | /* | |
2283 | * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR) | |
2284 | * | |
2285 | * This socket option is a boolean flag which turns on or off mapped V4 | |
2286 | * addresses. If this option is turned on and the socket is type | |
2287 | * PF_INET6, then IPv4 addresses will be mapped to V6 representation. | |
2288 | * If this option is turned off, then no mapping will be done of V4 | |
2289 | * addresses and a user will receive both PF_INET6 and PF_INET type | |
2290 | * addresses on the socket. | |
2291 | */ | |
2292 | static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, int optlen) | |
2293 | { | |
2294 | int val; | |
2295 | struct sctp_sock *sp = sctp_sk(sk); | |
2296 | ||
2297 | if (optlen < sizeof(int)) | |
2298 | return -EINVAL; | |
2299 | if (get_user(val, (int __user *)optval)) | |
2300 | return -EFAULT; | |
2301 | if (val) | |
2302 | sp->v4mapped = 1; | |
2303 | else | |
2304 | sp->v4mapped = 0; | |
2305 | ||
2306 | return 0; | |
2307 | } | |
2308 | ||
2309 | /* | |
2310 | * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG) | |
2311 | * | |
2312 | * This socket option specifies the maximum size to put in any outgoing | |
2313 | * SCTP chunk. If a message is larger than this size it will be | |
2314 | * fragmented by SCTP into the specified size. Note that the underlying | |
2315 | * SCTP implementation may fragment into smaller sized chunks when the | |
2316 | * PMTU of the underlying association is smaller than the value set by | |
2317 | * the user. | |
2318 | */ | |
2319 | static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, int optlen) | |
2320 | { | |
2321 | struct sctp_association *asoc; | |
2322 | struct list_head *pos; | |
2323 | struct sctp_sock *sp = sctp_sk(sk); | |
2324 | int val; | |
2325 | ||
2326 | if (optlen < sizeof(int)) | |
2327 | return -EINVAL; | |
2328 | if (get_user(val, (int __user *)optval)) | |
2329 | return -EFAULT; | |
96a33998 | 2330 | if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN))) |
1da177e4 LT |
2331 | return -EINVAL; |
2332 | sp->user_frag = val; | |
2333 | ||
96a33998 ISJ |
2334 | /* Update the frag_point of the existing associations. */ |
2335 | list_for_each(pos, &(sp->ep->asocs)) { | |
2336 | asoc = list_entry(pos, struct sctp_association, asocs); | |
2337 | asoc->frag_point = sctp_frag_point(sp, asoc->pmtu); | |
1da177e4 LT |
2338 | } |
2339 | ||
2340 | return 0; | |
2341 | } | |
2342 | ||
2343 | ||
2344 | /* | |
2345 | * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR) | |
2346 | * | |
2347 | * Requests that the peer mark the enclosed address as the association | |
2348 | * primary. The enclosed address must be one of the association's | |
2349 | * locally bound addresses. The following structure is used to make a | |
2350 | * set primary request: | |
2351 | */ | |
2352 | static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval, | |
2353 | int optlen) | |
2354 | { | |
2355 | struct sctp_sock *sp; | |
2356 | struct sctp_endpoint *ep; | |
2357 | struct sctp_association *asoc = NULL; | |
2358 | struct sctp_setpeerprim prim; | |
2359 | struct sctp_chunk *chunk; | |
2360 | int err; | |
2361 | ||
2362 | sp = sctp_sk(sk); | |
2363 | ep = sp->ep; | |
2364 | ||
2365 | if (!sctp_addip_enable) | |
2366 | return -EPERM; | |
2367 | ||
2368 | if (optlen != sizeof(struct sctp_setpeerprim)) | |
2369 | return -EINVAL; | |
2370 | ||
2371 | if (copy_from_user(&prim, optval, optlen)) | |
2372 | return -EFAULT; | |
2373 | ||
2374 | asoc = sctp_id2assoc(sk, prim.sspp_assoc_id); | |
2375 | if (!asoc) | |
2376 | return -EINVAL; | |
2377 | ||
2378 | if (!asoc->peer.asconf_capable) | |
2379 | return -EPERM; | |
2380 | ||
2381 | if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY) | |
2382 | return -EPERM; | |
2383 | ||
2384 | if (!sctp_state(asoc, ESTABLISHED)) | |
2385 | return -ENOTCONN; | |
2386 | ||
2387 | if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr)) | |
2388 | return -EADDRNOTAVAIL; | |
2389 | ||
2390 | /* Create an ASCONF chunk with SET_PRIMARY parameter */ | |
2391 | chunk = sctp_make_asconf_set_prim(asoc, | |
2392 | (union sctp_addr *)&prim.sspp_addr); | |
2393 | if (!chunk) | |
2394 | return -ENOMEM; | |
2395 | ||
2396 | err = sctp_send_asconf(asoc, chunk); | |
2397 | ||
2398 | SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n"); | |
2399 | ||
2400 | return err; | |
2401 | } | |
2402 | ||
2403 | static int sctp_setsockopt_adaption_layer(struct sock *sk, char __user *optval, | |
2404 | int optlen) | |
2405 | { | |
a1ab3582 | 2406 | struct sctp_setadaption adaption; |
1da177e4 | 2407 | |
a1ab3582 | 2408 | if (optlen != sizeof(struct sctp_setadaption)) |
1da177e4 | 2409 | return -EINVAL; |
a1ab3582 | 2410 | if (copy_from_user(&adaption, optval, optlen)) |
1da177e4 LT |
2411 | return -EFAULT; |
2412 | ||
a1ab3582 | 2413 | sctp_sk(sk)->adaption_ind = adaption.ssb_adaption_ind; |
1da177e4 LT |
2414 | |
2415 | return 0; | |
2416 | } | |
2417 | ||
2418 | /* API 6.2 setsockopt(), getsockopt() | |
2419 | * | |
2420 | * Applications use setsockopt() and getsockopt() to set or retrieve | |
2421 | * socket options. Socket options are used to change the default | |
2422 | * behavior of sockets calls. They are described in Section 7. | |
2423 | * | |
2424 | * The syntax is: | |
2425 | * | |
2426 | * ret = getsockopt(int sd, int level, int optname, void __user *optval, | |
2427 | * int __user *optlen); | |
2428 | * ret = setsockopt(int sd, int level, int optname, const void __user *optval, | |
2429 | * int optlen); | |
2430 | * | |
2431 | * sd - the socket descript. | |
2432 | * level - set to IPPROTO_SCTP for all SCTP options. | |
2433 | * optname - the option name. | |
2434 | * optval - the buffer to store the value of the option. | |
2435 | * optlen - the size of the buffer. | |
2436 | */ | |
2437 | SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname, | |
2438 | char __user *optval, int optlen) | |
2439 | { | |
2440 | int retval = 0; | |
2441 | ||
2442 | SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n", | |
2443 | sk, optname); | |
2444 | ||
2445 | /* I can hardly begin to describe how wrong this is. This is | |
2446 | * so broken as to be worse than useless. The API draft | |
2447 | * REALLY is NOT helpful here... I am not convinced that the | |
2448 | * semantics of setsockopt() with a level OTHER THAN SOL_SCTP | |
2449 | * are at all well-founded. | |
2450 | */ | |
2451 | if (level != SOL_SCTP) { | |
2452 | struct sctp_af *af = sctp_sk(sk)->pf->af; | |
2453 | retval = af->setsockopt(sk, level, optname, optval, optlen); | |
2454 | goto out_nounlock; | |
2455 | } | |
2456 | ||
2457 | sctp_lock_sock(sk); | |
2458 | ||
2459 | switch (optname) { | |
2460 | case SCTP_SOCKOPT_BINDX_ADD: | |
2461 | /* 'optlen' is the size of the addresses buffer. */ | |
2462 | retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval, | |
2463 | optlen, SCTP_BINDX_ADD_ADDR); | |
2464 | break; | |
2465 | ||
2466 | case SCTP_SOCKOPT_BINDX_REM: | |
2467 | /* 'optlen' is the size of the addresses buffer. */ | |
2468 | retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval, | |
2469 | optlen, SCTP_BINDX_REM_ADDR); | |
2470 | break; | |
2471 | ||
3f7a87d2 FF |
2472 | case SCTP_SOCKOPT_CONNECTX: |
2473 | /* 'optlen' is the size of the addresses buffer. */ | |
2474 | retval = sctp_setsockopt_connectx(sk, (struct sockaddr __user *)optval, | |
2475 | optlen); | |
2476 | break; | |
2477 | ||
1da177e4 LT |
2478 | case SCTP_DISABLE_FRAGMENTS: |
2479 | retval = sctp_setsockopt_disable_fragments(sk, optval, optlen); | |
2480 | break; | |
2481 | ||
2482 | case SCTP_EVENTS: | |
2483 | retval = sctp_setsockopt_events(sk, optval, optlen); | |
2484 | break; | |
2485 | ||
2486 | case SCTP_AUTOCLOSE: | |
2487 | retval = sctp_setsockopt_autoclose(sk, optval, optlen); | |
2488 | break; | |
2489 | ||
2490 | case SCTP_PEER_ADDR_PARAMS: | |
2491 | retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen); | |
2492 | break; | |
2493 | ||
2494 | case SCTP_INITMSG: | |
2495 | retval = sctp_setsockopt_initmsg(sk, optval, optlen); | |
2496 | break; | |
2497 | case SCTP_DEFAULT_SEND_PARAM: | |
2498 | retval = sctp_setsockopt_default_send_param(sk, optval, | |
2499 | optlen); | |
2500 | break; | |
2501 | case SCTP_PRIMARY_ADDR: | |
2502 | retval = sctp_setsockopt_primary_addr(sk, optval, optlen); | |
2503 | break; | |
2504 | case SCTP_SET_PEER_PRIMARY_ADDR: | |
2505 | retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen); | |
2506 | break; | |
2507 | case SCTP_NODELAY: | |
2508 | retval = sctp_setsockopt_nodelay(sk, optval, optlen); | |
2509 | break; | |
2510 | case SCTP_RTOINFO: | |
2511 | retval = sctp_setsockopt_rtoinfo(sk, optval, optlen); | |
2512 | break; | |
2513 | case SCTP_ASSOCINFO: | |
2514 | retval = sctp_setsockopt_associnfo(sk, optval, optlen); | |
2515 | break; | |
2516 | case SCTP_I_WANT_MAPPED_V4_ADDR: | |
2517 | retval = sctp_setsockopt_mappedv4(sk, optval, optlen); | |
2518 | break; | |
2519 | case SCTP_MAXSEG: | |
2520 | retval = sctp_setsockopt_maxseg(sk, optval, optlen); | |
2521 | break; | |
2522 | case SCTP_ADAPTION_LAYER: | |
2523 | retval = sctp_setsockopt_adaption_layer(sk, optval, optlen); | |
2524 | break; | |
2525 | ||
2526 | default: | |
2527 | retval = -ENOPROTOOPT; | |
2528 | break; | |
2529 | }; | |
2530 | ||
2531 | sctp_release_sock(sk); | |
2532 | ||
2533 | out_nounlock: | |
2534 | return retval; | |
2535 | } | |
2536 | ||
2537 | /* API 3.1.6 connect() - UDP Style Syntax | |
2538 | * | |
2539 | * An application may use the connect() call in the UDP model to initiate an | |
2540 | * association without sending data. | |
2541 | * | |
2542 | * The syntax is: | |
2543 | * | |
2544 | * ret = connect(int sd, const struct sockaddr *nam, socklen_t len); | |
2545 | * | |
2546 | * sd: the socket descriptor to have a new association added to. | |
2547 | * | |
2548 | * nam: the address structure (either struct sockaddr_in or struct | |
2549 | * sockaddr_in6 defined in RFC2553 [7]). | |
2550 | * | |
2551 | * len: the size of the address. | |
2552 | */ | |
3f7a87d2 | 2553 | SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr, |
1da177e4 LT |
2554 | int addr_len) |
2555 | { | |
1da177e4 | 2556 | int err = 0; |
3f7a87d2 | 2557 | struct sctp_af *af; |
1da177e4 LT |
2558 | |
2559 | sctp_lock_sock(sk); | |
2560 | ||
3f7a87d2 FF |
2561 | SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n", |
2562 | __FUNCTION__, sk, addr, addr_len); | |
1da177e4 | 2563 | |
3f7a87d2 FF |
2564 | /* Validate addr_len before calling common connect/connectx routine. */ |
2565 | af = sctp_get_af_specific(addr->sa_family); | |
2566 | if (!af || addr_len < af->sockaddr_len) { | |
2567 | err = -EINVAL; | |
2568 | } else { | |
2569 | /* Pass correct addr len to common routine (so it knows there | |
2570 | * is only one address being passed. | |
2571 | */ | |
2572 | err = __sctp_connect(sk, addr, af->sockaddr_len); | |
1da177e4 LT |
2573 | } |
2574 | ||
1da177e4 | 2575 | sctp_release_sock(sk); |
1da177e4 LT |
2576 | return err; |
2577 | } | |
2578 | ||
2579 | /* FIXME: Write comments. */ | |
2580 | SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags) | |
2581 | { | |
2582 | return -EOPNOTSUPP; /* STUB */ | |
2583 | } | |
2584 | ||
2585 | /* 4.1.4 accept() - TCP Style Syntax | |
2586 | * | |
2587 | * Applications use accept() call to remove an established SCTP | |
2588 | * association from the accept queue of the endpoint. A new socket | |
2589 | * descriptor will be returned from accept() to represent the newly | |
2590 | * formed association. | |
2591 | */ | |
2592 | SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err) | |
2593 | { | |
2594 | struct sctp_sock *sp; | |
2595 | struct sctp_endpoint *ep; | |
2596 | struct sock *newsk = NULL; | |
2597 | struct sctp_association *asoc; | |
2598 | long timeo; | |
2599 | int error = 0; | |
2600 | ||
2601 | sctp_lock_sock(sk); | |
2602 | ||
2603 | sp = sctp_sk(sk); | |
2604 | ep = sp->ep; | |
2605 | ||
2606 | if (!sctp_style(sk, TCP)) { | |
2607 | error = -EOPNOTSUPP; | |
2608 | goto out; | |
2609 | } | |
2610 | ||
2611 | if (!sctp_sstate(sk, LISTENING)) { | |
2612 | error = -EINVAL; | |
2613 | goto out; | |
2614 | } | |
2615 | ||
2616 | timeo = sock_rcvtimeo(sk, sk->sk_socket->file->f_flags & O_NONBLOCK); | |
2617 | ||
2618 | error = sctp_wait_for_accept(sk, timeo); | |
2619 | if (error) | |
2620 | goto out; | |
2621 | ||
2622 | /* We treat the list of associations on the endpoint as the accept | |
2623 | * queue and pick the first association on the list. | |
2624 | */ | |
2625 | asoc = list_entry(ep->asocs.next, struct sctp_association, asocs); | |
2626 | ||
2627 | newsk = sp->pf->create_accept_sk(sk, asoc); | |
2628 | if (!newsk) { | |
2629 | error = -ENOMEM; | |
2630 | goto out; | |
2631 | } | |
2632 | ||
2633 | /* Populate the fields of the newsk from the oldsk and migrate the | |
2634 | * asoc to the newsk. | |
2635 | */ | |
2636 | sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP); | |
2637 | ||
2638 | out: | |
2639 | sctp_release_sock(sk); | |
2640 | *err = error; | |
2641 | return newsk; | |
2642 | } | |
2643 | ||
2644 | /* The SCTP ioctl handler. */ | |
2645 | SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg) | |
2646 | { | |
2647 | return -ENOIOCTLCMD; | |
2648 | } | |
2649 | ||
2650 | /* This is the function which gets called during socket creation to | |
2651 | * initialized the SCTP-specific portion of the sock. | |
2652 | * The sock structure should already be zero-filled memory. | |
2653 | */ | |
2654 | SCTP_STATIC int sctp_init_sock(struct sock *sk) | |
2655 | { | |
2656 | struct sctp_endpoint *ep; | |
2657 | struct sctp_sock *sp; | |
2658 | ||
2659 | SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk); | |
2660 | ||
2661 | sp = sctp_sk(sk); | |
2662 | ||
2663 | /* Initialize the SCTP per socket area. */ | |
2664 | switch (sk->sk_type) { | |
2665 | case SOCK_SEQPACKET: | |
2666 | sp->type = SCTP_SOCKET_UDP; | |
2667 | break; | |
2668 | case SOCK_STREAM: | |
2669 | sp->type = SCTP_SOCKET_TCP; | |
2670 | break; | |
2671 | default: | |
2672 | return -ESOCKTNOSUPPORT; | |
2673 | } | |
2674 | ||
2675 | /* Initialize default send parameters. These parameters can be | |
2676 | * modified with the SCTP_DEFAULT_SEND_PARAM socket option. | |
2677 | */ | |
2678 | sp->default_stream = 0; | |
2679 | sp->default_ppid = 0; | |
2680 | sp->default_flags = 0; | |
2681 | sp->default_context = 0; | |
2682 | sp->default_timetolive = 0; | |
2683 | ||
2684 | /* Initialize default setup parameters. These parameters | |
2685 | * can be modified with the SCTP_INITMSG socket option or | |
2686 | * overridden by the SCTP_INIT CMSG. | |
2687 | */ | |
2688 | sp->initmsg.sinit_num_ostreams = sctp_max_outstreams; | |
2689 | sp->initmsg.sinit_max_instreams = sctp_max_instreams; | |
2690 | sp->initmsg.sinit_max_attempts = sctp_max_retrans_init; | |
2691 | sp->initmsg.sinit_max_init_timeo = jiffies_to_msecs(sctp_rto_max); | |
2692 | ||
2693 | /* Initialize default RTO related parameters. These parameters can | |
2694 | * be modified for with the SCTP_RTOINFO socket option. | |
2695 | */ | |
2696 | sp->rtoinfo.srto_initial = jiffies_to_msecs(sctp_rto_initial); | |
2697 | sp->rtoinfo.srto_max = jiffies_to_msecs(sctp_rto_max); | |
2698 | sp->rtoinfo.srto_min = jiffies_to_msecs(sctp_rto_min); | |
2699 | ||
2700 | /* Initialize default association related parameters. These parameters | |
2701 | * can be modified with the SCTP_ASSOCINFO socket option. | |
2702 | */ | |
2703 | sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association; | |
2704 | sp->assocparams.sasoc_number_peer_destinations = 0; | |
2705 | sp->assocparams.sasoc_peer_rwnd = 0; | |
2706 | sp->assocparams.sasoc_local_rwnd = 0; | |
2707 | sp->assocparams.sasoc_cookie_life = | |
2708 | jiffies_to_msecs(sctp_valid_cookie_life); | |
2709 | ||
2710 | /* Initialize default event subscriptions. By default, all the | |
2711 | * options are off. | |
2712 | */ | |
2713 | memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe)); | |
2714 | ||
2715 | /* Default Peer Address Parameters. These defaults can | |
2716 | * be modified via SCTP_PEER_ADDR_PARAMS | |
2717 | */ | |
2718 | sp->paddrparam.spp_hbinterval = jiffies_to_msecs(sctp_hb_interval); | |
2719 | sp->paddrparam.spp_pathmaxrxt = sctp_max_retrans_path; | |
2720 | ||
2721 | /* If enabled no SCTP message fragmentation will be performed. | |
2722 | * Configure through SCTP_DISABLE_FRAGMENTS socket option. | |
2723 | */ | |
2724 | sp->disable_fragments = 0; | |
2725 | ||
2726 | /* Turn on/off any Nagle-like algorithm. */ | |
2727 | sp->nodelay = 1; | |
2728 | ||
2729 | /* Enable by default. */ | |
2730 | sp->v4mapped = 1; | |
2731 | ||
2732 | /* Auto-close idle associations after the configured | |
2733 | * number of seconds. A value of 0 disables this | |
2734 | * feature. Configure through the SCTP_AUTOCLOSE socket option, | |
2735 | * for UDP-style sockets only. | |
2736 | */ | |
2737 | sp->autoclose = 0; | |
2738 | ||
2739 | /* User specified fragmentation limit. */ | |
2740 | sp->user_frag = 0; | |
2741 | ||
2742 | sp->adaption_ind = 0; | |
2743 | ||
2744 | sp->pf = sctp_get_pf_specific(sk->sk_family); | |
2745 | ||
2746 | /* Control variables for partial data delivery. */ | |
2747 | sp->pd_mode = 0; | |
2748 | skb_queue_head_init(&sp->pd_lobby); | |
2749 | ||
2750 | /* Create a per socket endpoint structure. Even if we | |
2751 | * change the data structure relationships, this may still | |
2752 | * be useful for storing pre-connect address information. | |
2753 | */ | |
2754 | ep = sctp_endpoint_new(sk, GFP_KERNEL); | |
2755 | if (!ep) | |
2756 | return -ENOMEM; | |
2757 | ||
2758 | sp->ep = ep; | |
2759 | sp->hmac = NULL; | |
2760 | ||
2761 | SCTP_DBG_OBJCNT_INC(sock); | |
2762 | return 0; | |
2763 | } | |
2764 | ||
2765 | /* Cleanup any SCTP per socket resources. */ | |
2766 | SCTP_STATIC int sctp_destroy_sock(struct sock *sk) | |
2767 | { | |
2768 | struct sctp_endpoint *ep; | |
2769 | ||
2770 | SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk); | |
2771 | ||
2772 | /* Release our hold on the endpoint. */ | |
2773 | ep = sctp_sk(sk)->ep; | |
2774 | sctp_endpoint_free(ep); | |
2775 | ||
2776 | return 0; | |
2777 | } | |
2778 | ||
2779 | /* API 4.1.7 shutdown() - TCP Style Syntax | |
2780 | * int shutdown(int socket, int how); | |
2781 | * | |
2782 | * sd - the socket descriptor of the association to be closed. | |
2783 | * how - Specifies the type of shutdown. The values are | |
2784 | * as follows: | |
2785 | * SHUT_RD | |
2786 | * Disables further receive operations. No SCTP | |
2787 | * protocol action is taken. | |
2788 | * SHUT_WR | |
2789 | * Disables further send operations, and initiates | |
2790 | * the SCTP shutdown sequence. | |
2791 | * SHUT_RDWR | |
2792 | * Disables further send and receive operations | |
2793 | * and initiates the SCTP shutdown sequence. | |
2794 | */ | |
2795 | SCTP_STATIC void sctp_shutdown(struct sock *sk, int how) | |
2796 | { | |
2797 | struct sctp_endpoint *ep; | |
2798 | struct sctp_association *asoc; | |
2799 | ||
2800 | if (!sctp_style(sk, TCP)) | |
2801 | return; | |
2802 | ||
2803 | if (how & SEND_SHUTDOWN) { | |
2804 | ep = sctp_sk(sk)->ep; | |
2805 | if (!list_empty(&ep->asocs)) { | |
2806 | asoc = list_entry(ep->asocs.next, | |
2807 | struct sctp_association, asocs); | |
2808 | sctp_primitive_SHUTDOWN(asoc, NULL); | |
2809 | } | |
2810 | } | |
2811 | } | |
2812 | ||
2813 | /* 7.2.1 Association Status (SCTP_STATUS) | |
2814 | ||
2815 | * Applications can retrieve current status information about an | |
2816 | * association, including association state, peer receiver window size, | |
2817 | * number of unacked data chunks, and number of data chunks pending | |
2818 | * receipt. This information is read-only. | |
2819 | */ | |
2820 | static int sctp_getsockopt_sctp_status(struct sock *sk, int len, | |
2821 | char __user *optval, | |
2822 | int __user *optlen) | |
2823 | { | |
2824 | struct sctp_status status; | |
2825 | struct sctp_association *asoc = NULL; | |
2826 | struct sctp_transport *transport; | |
2827 | sctp_assoc_t associd; | |
2828 | int retval = 0; | |
2829 | ||
2830 | if (len != sizeof(status)) { | |
2831 | retval = -EINVAL; | |
2832 | goto out; | |
2833 | } | |
2834 | ||
2835 | if (copy_from_user(&status, optval, sizeof(status))) { | |
2836 | retval = -EFAULT; | |
2837 | goto out; | |
2838 | } | |
2839 | ||
2840 | associd = status.sstat_assoc_id; | |
2841 | asoc = sctp_id2assoc(sk, associd); | |
2842 | if (!asoc) { | |
2843 | retval = -EINVAL; | |
2844 | goto out; | |
2845 | } | |
2846 | ||
2847 | transport = asoc->peer.primary_path; | |
2848 | ||
2849 | status.sstat_assoc_id = sctp_assoc2id(asoc); | |
2850 | status.sstat_state = asoc->state; | |
2851 | status.sstat_rwnd = asoc->peer.rwnd; | |
2852 | status.sstat_unackdata = asoc->unack_data; | |
2853 | ||
2854 | status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map); | |
2855 | status.sstat_instrms = asoc->c.sinit_max_instreams; | |
2856 | status.sstat_outstrms = asoc->c.sinit_num_ostreams; | |
2857 | status.sstat_fragmentation_point = asoc->frag_point; | |
2858 | status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc); | |
2859 | memcpy(&status.sstat_primary.spinfo_address, | |
2860 | &(transport->ipaddr), sizeof(union sctp_addr)); | |
2861 | /* Map ipv4 address into v4-mapped-on-v6 address. */ | |
2862 | sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk), | |
2863 | (union sctp_addr *)&status.sstat_primary.spinfo_address); | |
3f7a87d2 | 2864 | status.sstat_primary.spinfo_state = transport->state; |
1da177e4 LT |
2865 | status.sstat_primary.spinfo_cwnd = transport->cwnd; |
2866 | status.sstat_primary.spinfo_srtt = transport->srtt; | |
2867 | status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto); | |
2868 | status.sstat_primary.spinfo_mtu = transport->pmtu; | |
2869 | ||
3f7a87d2 FF |
2870 | if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN) |
2871 | status.sstat_primary.spinfo_state = SCTP_ACTIVE; | |
2872 | ||
1da177e4 LT |
2873 | if (put_user(len, optlen)) { |
2874 | retval = -EFAULT; | |
2875 | goto out; | |
2876 | } | |
2877 | ||
2878 | SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n", | |
2879 | len, status.sstat_state, status.sstat_rwnd, | |
2880 | status.sstat_assoc_id); | |
2881 | ||
2882 | if (copy_to_user(optval, &status, len)) { | |
2883 | retval = -EFAULT; | |
2884 | goto out; | |
2885 | } | |
2886 | ||
2887 | out: | |
2888 | return (retval); | |
2889 | } | |
2890 | ||
2891 | ||
2892 | /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO) | |
2893 | * | |
2894 | * Applications can retrieve information about a specific peer address | |
2895 | * of an association, including its reachability state, congestion | |
2896 | * window, and retransmission timer values. This information is | |
2897 | * read-only. | |
2898 | */ | |
2899 | static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len, | |
2900 | char __user *optval, | |
2901 | int __user *optlen) | |
2902 | { | |
2903 | struct sctp_paddrinfo pinfo; | |
2904 | struct sctp_transport *transport; | |
2905 | int retval = 0; | |
2906 | ||
2907 | if (len != sizeof(pinfo)) { | |
2908 | retval = -EINVAL; | |
2909 | goto out; | |
2910 | } | |
2911 | ||
2912 | if (copy_from_user(&pinfo, optval, sizeof(pinfo))) { | |
2913 | retval = -EFAULT; | |
2914 | goto out; | |
2915 | } | |
2916 | ||
2917 | transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address, | |
2918 | pinfo.spinfo_assoc_id); | |
2919 | if (!transport) | |
2920 | return -EINVAL; | |
2921 | ||
2922 | pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc); | |
3f7a87d2 | 2923 | pinfo.spinfo_state = transport->state; |
1da177e4 LT |
2924 | pinfo.spinfo_cwnd = transport->cwnd; |
2925 | pinfo.spinfo_srtt = transport->srtt; | |
2926 | pinfo.spinfo_rto = jiffies_to_msecs(transport->rto); | |
2927 | pinfo.spinfo_mtu = transport->pmtu; | |
2928 | ||
3f7a87d2 FF |
2929 | if (pinfo.spinfo_state == SCTP_UNKNOWN) |
2930 | pinfo.spinfo_state = SCTP_ACTIVE; | |
2931 | ||
1da177e4 LT |
2932 | if (put_user(len, optlen)) { |
2933 | retval = -EFAULT; | |
2934 | goto out; | |
2935 | } | |
2936 | ||
2937 | if (copy_to_user(optval, &pinfo, len)) { | |
2938 | retval = -EFAULT; | |
2939 | goto out; | |
2940 | } | |
2941 | ||
2942 | out: | |
2943 | return (retval); | |
2944 | } | |
2945 | ||
2946 | /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS) | |
2947 | * | |
2948 | * This option is a on/off flag. If enabled no SCTP message | |
2949 | * fragmentation will be performed. Instead if a message being sent | |
2950 | * exceeds the current PMTU size, the message will NOT be sent and | |
2951 | * instead a error will be indicated to the user. | |
2952 | */ | |
2953 | static int sctp_getsockopt_disable_fragments(struct sock *sk, int len, | |
2954 | char __user *optval, int __user *optlen) | |
2955 | { | |
2956 | int val; | |
2957 | ||
2958 | if (len < sizeof(int)) | |
2959 | return -EINVAL; | |
2960 | ||
2961 | len = sizeof(int); | |
2962 | val = (sctp_sk(sk)->disable_fragments == 1); | |
2963 | if (put_user(len, optlen)) | |
2964 | return -EFAULT; | |
2965 | if (copy_to_user(optval, &val, len)) | |
2966 | return -EFAULT; | |
2967 | return 0; | |
2968 | } | |
2969 | ||
2970 | /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS) | |
2971 | * | |
2972 | * This socket option is used to specify various notifications and | |
2973 | * ancillary data the user wishes to receive. | |
2974 | */ | |
2975 | static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval, | |
2976 | int __user *optlen) | |
2977 | { | |
2978 | if (len != sizeof(struct sctp_event_subscribe)) | |
2979 | return -EINVAL; | |
2980 | if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len)) | |
2981 | return -EFAULT; | |
2982 | return 0; | |
2983 | } | |
2984 | ||
2985 | /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE) | |
2986 | * | |
2987 | * This socket option is applicable to the UDP-style socket only. When | |
2988 | * set it will cause associations that are idle for more than the | |
2989 | * specified number of seconds to automatically close. An association | |
2990 | * being idle is defined an association that has NOT sent or received | |
2991 | * user data. The special value of '0' indicates that no automatic | |
2992 | * close of any associations should be performed. The option expects an | |
2993 | * integer defining the number of seconds of idle time before an | |
2994 | * association is closed. | |
2995 | */ | |
2996 | static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen) | |
2997 | { | |
2998 | /* Applicable to UDP-style socket only */ | |
2999 | if (sctp_style(sk, TCP)) | |
3000 | return -EOPNOTSUPP; | |
3001 | if (len != sizeof(int)) | |
3002 | return -EINVAL; | |
3003 | if (copy_to_user(optval, &sctp_sk(sk)->autoclose, len)) | |
3004 | return -EFAULT; | |
3005 | return 0; | |
3006 | } | |
3007 | ||
3008 | /* Helper routine to branch off an association to a new socket. */ | |
3009 | SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc, | |
3010 | struct socket **sockp) | |
3011 | { | |
3012 | struct sock *sk = asoc->base.sk; | |
3013 | struct socket *sock; | |
3014 | int err = 0; | |
3015 | ||
3016 | /* An association cannot be branched off from an already peeled-off | |
3017 | * socket, nor is this supported for tcp style sockets. | |
3018 | */ | |
3019 | if (!sctp_style(sk, UDP)) | |
3020 | return -EINVAL; | |
3021 | ||
3022 | /* Create a new socket. */ | |
3023 | err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock); | |
3024 | if (err < 0) | |
3025 | return err; | |
3026 | ||
3027 | /* Populate the fields of the newsk from the oldsk and migrate the | |
3028 | * asoc to the newsk. | |
3029 | */ | |
3030 | sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH); | |
3031 | *sockp = sock; | |
3032 | ||
3033 | return err; | |
3034 | } | |
3035 | ||
3036 | static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen) | |
3037 | { | |
3038 | sctp_peeloff_arg_t peeloff; | |
3039 | struct socket *newsock; | |
3040 | int retval = 0; | |
3041 | struct sctp_association *asoc; | |
3042 | ||
3043 | if (len != sizeof(sctp_peeloff_arg_t)) | |
3044 | return -EINVAL; | |
3045 | if (copy_from_user(&peeloff, optval, len)) | |
3046 | return -EFAULT; | |
3047 | ||
3048 | asoc = sctp_id2assoc(sk, peeloff.associd); | |
3049 | if (!asoc) { | |
3050 | retval = -EINVAL; | |
3051 | goto out; | |
3052 | } | |
3053 | ||
3054 | SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __FUNCTION__, sk, asoc); | |
3055 | ||
3056 | retval = sctp_do_peeloff(asoc, &newsock); | |
3057 | if (retval < 0) | |
3058 | goto out; | |
3059 | ||
3060 | /* Map the socket to an unused fd that can be returned to the user. */ | |
3061 | retval = sock_map_fd(newsock); | |
3062 | if (retval < 0) { | |
3063 | sock_release(newsock); | |
3064 | goto out; | |
3065 | } | |
3066 | ||
3067 | SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n", | |
3068 | __FUNCTION__, sk, asoc, newsock->sk, retval); | |
3069 | ||
3070 | /* Return the fd mapped to the new socket. */ | |
3071 | peeloff.sd = retval; | |
3072 | if (copy_to_user(optval, &peeloff, len)) | |
3073 | retval = -EFAULT; | |
3074 | ||
3075 | out: | |
3076 | return retval; | |
3077 | } | |
3078 | ||
3079 | /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS) | |
3080 | * | |
3081 | * Applications can enable or disable heartbeats for any peer address of | |
3082 | * an association, modify an address's heartbeat interval, force a | |
3083 | * heartbeat to be sent immediately, and adjust the address's maximum | |
3084 | * number of retransmissions sent before an address is considered | |
3085 | * unreachable. The following structure is used to access and modify an | |
3086 | * address's parameters: | |
3087 | * | |
3088 | * struct sctp_paddrparams { | |
3089 | * sctp_assoc_t spp_assoc_id; | |
3090 | * struct sockaddr_storage spp_address; | |
3091 | * uint32_t spp_hbinterval; | |
3092 | * uint16_t spp_pathmaxrxt; | |
3093 | * }; | |
3094 | * | |
3095 | * spp_assoc_id - (UDP style socket) This is filled in the application, | |
3096 | * and identifies the association for this query. | |
3097 | * spp_address - This specifies which address is of interest. | |
3098 | * spp_hbinterval - This contains the value of the heartbeat interval, | |
3099 | * in milliseconds. A value of 0, when modifying the | |
3100 | * parameter, specifies that the heartbeat on this | |
3101 | * address should be disabled. A value of UINT32_MAX | |
3102 | * (4294967295), when modifying the parameter, | |
3103 | * specifies that a heartbeat should be sent | |
3104 | * immediately to the peer address, and the current | |
3105 | * interval should remain unchanged. | |
3106 | * spp_pathmaxrxt - This contains the maximum number of | |
3107 | * retransmissions before this address shall be | |
3108 | * considered unreachable. | |
3109 | */ | |
3110 | static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len, | |
3111 | char __user *optval, int __user *optlen) | |
3112 | { | |
3113 | struct sctp_paddrparams params; | |
3114 | struct sctp_transport *trans; | |
3115 | ||
3116 | if (len != sizeof(struct sctp_paddrparams)) | |
3117 | return -EINVAL; | |
3118 | if (copy_from_user(¶ms, optval, len)) | |
3119 | return -EFAULT; | |
3120 | ||
3121 | /* If no association id is specified retrieve the default value | |
3122 | * for the endpoint that will be used for all future associations | |
3123 | */ | |
3124 | if (!params.spp_assoc_id && | |
3125 | sctp_is_any(( union sctp_addr *)¶ms.spp_address)) { | |
3126 | params.spp_hbinterval = sctp_sk(sk)->paddrparam.spp_hbinterval; | |
3127 | params.spp_pathmaxrxt = sctp_sk(sk)->paddrparam.spp_pathmaxrxt; | |
3128 | ||
3129 | goto done; | |
3130 | } | |
3131 | ||
3132 | trans = sctp_addr_id2transport(sk, ¶ms.spp_address, | |
3133 | params.spp_assoc_id); | |
3134 | if (!trans) | |
3135 | return -EINVAL; | |
3136 | ||
3137 | /* The value of the heartbeat interval, in milliseconds. A value of 0, | |
3138 | * when modifying the parameter, specifies that the heartbeat on this | |
3139 | * address should be disabled. | |
3140 | */ | |
3141 | if (!trans->hb_allowed) | |
3142 | params.spp_hbinterval = 0; | |
3143 | else | |
3144 | params.spp_hbinterval = jiffies_to_msecs(trans->hb_interval); | |
3145 | ||
3146 | /* spp_pathmaxrxt contains the maximum number of retransmissions | |
3147 | * before this address shall be considered unreachable. | |
3148 | */ | |
3149 | params.spp_pathmaxrxt = trans->max_retrans; | |
3150 | ||
3151 | done: | |
3152 | if (copy_to_user(optval, ¶ms, len)) | |
3153 | return -EFAULT; | |
3154 | ||
3155 | if (put_user(len, optlen)) | |
3156 | return -EFAULT; | |
3157 | ||
3158 | return 0; | |
3159 | } | |
3160 | ||
3161 | /* 7.1.3 Initialization Parameters (SCTP_INITMSG) | |
3162 | * | |
3163 | * Applications can specify protocol parameters for the default association | |
3164 | * initialization. The option name argument to setsockopt() and getsockopt() | |
3165 | * is SCTP_INITMSG. | |
3166 | * | |
3167 | * Setting initialization parameters is effective only on an unconnected | |
3168 | * socket (for UDP-style sockets only future associations are effected | |
3169 | * by the change). With TCP-style sockets, this option is inherited by | |
3170 | * sockets derived from a listener socket. | |
3171 | */ | |
3172 | static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen) | |
3173 | { | |
3174 | if (len != sizeof(struct sctp_initmsg)) | |
3175 | return -EINVAL; | |
3176 | if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len)) | |
3177 | return -EFAULT; | |
3178 | return 0; | |
3179 | } | |
3180 | ||