Commit | Line | Data |
---|---|---|
d021c344 AK |
1 | /* |
2 | * VMware vSockets Driver | |
3 | * | |
4 | * Copyright (C) 2007-2013 VMware, Inc. All rights reserved. | |
5 | * | |
6 | * This program is free software; you can redistribute it and/or modify it | |
7 | * under the terms of the GNU General Public License as published by the Free | |
8 | * Software Foundation version 2 and no later version. | |
9 | * | |
10 | * This program is distributed in the hope that it will be useful, but WITHOUT | |
11 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
12 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for | |
13 | * more details. | |
14 | */ | |
15 | ||
16 | /* Implementation notes: | |
17 | * | |
18 | * - There are two kinds of sockets: those created by user action (such as | |
19 | * calling socket(2)) and those created by incoming connection request packets. | |
20 | * | |
21 | * - There are two "global" tables, one for bound sockets (sockets that have | |
22 | * specified an address that they are responsible for) and one for connected | |
23 | * sockets (sockets that have established a connection with another socket). | |
24 | * These tables are "global" in that all sockets on the system are placed | |
25 | * within them. - Note, though, that the bound table contains an extra entry | |
26 | * for a list of unbound sockets and SOCK_DGRAM sockets will always remain in | |
27 | * that list. The bound table is used solely for lookup of sockets when packets | |
28 | * are received and that's not necessary for SOCK_DGRAM sockets since we create | |
29 | * a datagram handle for each and need not perform a lookup. Keeping SOCK_DGRAM | |
30 | * sockets out of the bound hash buckets will reduce the chance of collisions | |
31 | * when looking for SOCK_STREAM sockets and prevents us from having to check the | |
32 | * socket type in the hash table lookups. | |
33 | * | |
34 | * - Sockets created by user action will either be "client" sockets that | |
35 | * initiate a connection or "server" sockets that listen for connections; we do | |
36 | * not support simultaneous connects (two "client" sockets connecting). | |
37 | * | |
38 | * - "Server" sockets are referred to as listener sockets throughout this | |
ea3803c1 SH |
39 | * implementation because they are in the VSOCK_SS_LISTEN state. When a |
40 | * connection request is received (the second kind of socket mentioned above), | |
41 | * we create a new socket and refer to it as a pending socket. These pending | |
42 | * sockets are placed on the pending connection list of the listener socket. | |
43 | * When future packets are received for the address the listener socket is | |
44 | * bound to, we check if the source of the packet is from one that has an | |
45 | * existing pending connection. If it does, we process the packet for the | |
46 | * pending socket. When that socket reaches the connected state, it is removed | |
47 | * from the listener socket's pending list and enqueued in the listener | |
48 | * socket's accept queue. Callers of accept(2) will accept connected sockets | |
49 | * from the listener socket's accept queue. If the socket cannot be accepted | |
50 | * for some reason then it is marked rejected. Once the connection is | |
51 | * accepted, it is owned by the user process and the responsibility for cleanup | |
52 | * falls with that user process. | |
d021c344 AK |
53 | * |
54 | * - It is possible that these pending sockets will never reach the connected | |
55 | * state; in fact, we may never receive another packet after the connection | |
56 | * request. Because of this, we must schedule a cleanup function to run in the | |
57 | * future, after some amount of time passes where a connection should have been | |
58 | * established. This function ensures that the socket is off all lists so it | |
59 | * cannot be retrieved, then drops all references to the socket so it is cleaned | |
60 | * up (sock_put() -> sk_free() -> our sk_destruct implementation). Note this | |
61 | * function will also cleanup rejected sockets, those that reach the connected | |
62 | * state but leave it before they have been accepted. | |
63 | * | |
64 | * - Sockets created by user action will be cleaned up when the user process | |
65 | * calls close(2), causing our release implementation to be called. Our release | |
66 | * implementation will perform some cleanup then drop the last reference so our | |
67 | * sk_destruct implementation is invoked. Our sk_destruct implementation will | |
68 | * perform additional cleanup that's common for both types of sockets. | |
69 | * | |
70 | * - A socket's reference count is what ensures that the structure won't be | |
71 | * freed. Each entry in a list (such as the "global" bound and connected tables | |
72 | * and the listener socket's pending list and connected queue) ensures a | |
73 | * reference. When we defer work until process context and pass a socket as our | |
74 | * argument, we must ensure the reference count is increased to ensure the | |
75 | * socket isn't freed before the function is run; the deferred function will | |
76 | * then drop the reference. | |
77 | */ | |
78 | ||
79 | #include <linux/types.h> | |
d021c344 AK |
80 | #include <linux/bitops.h> |
81 | #include <linux/cred.h> | |
82 | #include <linux/init.h> | |
83 | #include <linux/io.h> | |
84 | #include <linux/kernel.h> | |
85 | #include <linux/kmod.h> | |
86 | #include <linux/list.h> | |
87 | #include <linux/miscdevice.h> | |
88 | #include <linux/module.h> | |
89 | #include <linux/mutex.h> | |
90 | #include <linux/net.h> | |
91 | #include <linux/poll.h> | |
92 | #include <linux/skbuff.h> | |
93 | #include <linux/smp.h> | |
94 | #include <linux/socket.h> | |
95 | #include <linux/stddef.h> | |
96 | #include <linux/unistd.h> | |
97 | #include <linux/wait.h> | |
98 | #include <linux/workqueue.h> | |
99 | #include <net/sock.h> | |
82a54d0e | 100 | #include <net/af_vsock.h> |
d021c344 AK |
101 | |
102 | static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr); | |
103 | static void vsock_sk_destruct(struct sock *sk); | |
104 | static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb); | |
105 | ||
106 | /* Protocol family. */ | |
107 | static struct proto vsock_proto = { | |
108 | .name = "AF_VSOCK", | |
109 | .owner = THIS_MODULE, | |
110 | .obj_size = sizeof(struct vsock_sock), | |
111 | }; | |
112 | ||
113 | /* The default peer timeout indicates how long we will wait for a peer response | |
114 | * to a control message. | |
115 | */ | |
116 | #define VSOCK_DEFAULT_CONNECT_TIMEOUT (2 * HZ) | |
117 | ||
d021c344 AK |
118 | static const struct vsock_transport *transport; |
119 | static DEFINE_MUTEX(vsock_register_mutex); | |
120 | ||
121 | /**** EXPORTS ****/ | |
122 | ||
123 | /* Get the ID of the local context. This is transport dependent. */ | |
124 | ||
125 | int vm_sockets_get_local_cid(void) | |
126 | { | |
127 | return transport->get_local_cid(); | |
128 | } | |
129 | EXPORT_SYMBOL_GPL(vm_sockets_get_local_cid); | |
130 | ||
131 | /**** UTILS ****/ | |
132 | ||
133 | /* Each bound VSocket is stored in the bind hash table and each connected | |
134 | * VSocket is stored in the connected hash table. | |
135 | * | |
136 | * Unbound sockets are all put on the same list attached to the end of the hash | |
137 | * table (vsock_unbound_sockets). Bound sockets are added to the hash table in | |
138 | * the bucket that their local address hashes to (vsock_bound_sockets(addr) | |
139 | * represents the list that addr hashes to). | |
140 | * | |
141 | * Specifically, we initialize the vsock_bind_table array to a size of | |
142 | * VSOCK_HASH_SIZE + 1 so that vsock_bind_table[0] through | |
143 | * vsock_bind_table[VSOCK_HASH_SIZE - 1] are for bound sockets and | |
144 | * vsock_bind_table[VSOCK_HASH_SIZE] is for unbound sockets. The hash function | |
a49dd9dc | 145 | * mods with VSOCK_HASH_SIZE to ensure this. |
d021c344 AK |
146 | */ |
147 | #define VSOCK_HASH_SIZE 251 | |
148 | #define MAX_PORT_RETRIES 24 | |
149 | ||
a49dd9dc | 150 | #define VSOCK_HASH(addr) ((addr)->svm_port % VSOCK_HASH_SIZE) |
d021c344 AK |
151 | #define vsock_bound_sockets(addr) (&vsock_bind_table[VSOCK_HASH(addr)]) |
152 | #define vsock_unbound_sockets (&vsock_bind_table[VSOCK_HASH_SIZE]) | |
153 | ||
154 | /* XXX This can probably be implemented in a better way. */ | |
155 | #define VSOCK_CONN_HASH(src, dst) \ | |
a49dd9dc | 156 | (((src)->svm_cid ^ (dst)->svm_port) % VSOCK_HASH_SIZE) |
d021c344 AK |
157 | #define vsock_connected_sockets(src, dst) \ |
158 | (&vsock_connected_table[VSOCK_CONN_HASH(src, dst)]) | |
159 | #define vsock_connected_sockets_vsk(vsk) \ | |
160 | vsock_connected_sockets(&(vsk)->remote_addr, &(vsk)->local_addr) | |
161 | ||
162 | static struct list_head vsock_bind_table[VSOCK_HASH_SIZE + 1]; | |
163 | static struct list_head vsock_connected_table[VSOCK_HASH_SIZE]; | |
164 | static DEFINE_SPINLOCK(vsock_table_lock); | |
165 | ||
b3a6dfe8 AH |
166 | /* Autobind this socket to the local address if necessary. */ |
167 | static int vsock_auto_bind(struct vsock_sock *vsk) | |
168 | { | |
169 | struct sock *sk = sk_vsock(vsk); | |
170 | struct sockaddr_vm local_addr; | |
171 | ||
172 | if (vsock_addr_bound(&vsk->local_addr)) | |
173 | return 0; | |
174 | vsock_addr_init(&local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); | |
175 | return __vsock_bind(sk, &local_addr); | |
176 | } | |
177 | ||
22ee3b57 | 178 | static void vsock_init_tables(void) |
d021c344 AK |
179 | { |
180 | int i; | |
181 | ||
182 | for (i = 0; i < ARRAY_SIZE(vsock_bind_table); i++) | |
183 | INIT_LIST_HEAD(&vsock_bind_table[i]); | |
184 | ||
185 | for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) | |
186 | INIT_LIST_HEAD(&vsock_connected_table[i]); | |
187 | } | |
188 | ||
189 | static void __vsock_insert_bound(struct list_head *list, | |
190 | struct vsock_sock *vsk) | |
191 | { | |
192 | sock_hold(&vsk->sk); | |
193 | list_add(&vsk->bound_table, list); | |
194 | } | |
195 | ||
196 | static void __vsock_insert_connected(struct list_head *list, | |
197 | struct vsock_sock *vsk) | |
198 | { | |
199 | sock_hold(&vsk->sk); | |
200 | list_add(&vsk->connected_table, list); | |
201 | } | |
202 | ||
203 | static void __vsock_remove_bound(struct vsock_sock *vsk) | |
204 | { | |
205 | list_del_init(&vsk->bound_table); | |
206 | sock_put(&vsk->sk); | |
207 | } | |
208 | ||
209 | static void __vsock_remove_connected(struct vsock_sock *vsk) | |
210 | { | |
211 | list_del_init(&vsk->connected_table); | |
212 | sock_put(&vsk->sk); | |
213 | } | |
214 | ||
215 | static struct sock *__vsock_find_bound_socket(struct sockaddr_vm *addr) | |
216 | { | |
217 | struct vsock_sock *vsk; | |
218 | ||
219 | list_for_each_entry(vsk, vsock_bound_sockets(addr), bound_table) | |
990454b5 | 220 | if (addr->svm_port == vsk->local_addr.svm_port) |
d021c344 AK |
221 | return sk_vsock(vsk); |
222 | ||
223 | return NULL; | |
224 | } | |
225 | ||
226 | static struct sock *__vsock_find_connected_socket(struct sockaddr_vm *src, | |
227 | struct sockaddr_vm *dst) | |
228 | { | |
229 | struct vsock_sock *vsk; | |
230 | ||
231 | list_for_each_entry(vsk, vsock_connected_sockets(src, dst), | |
232 | connected_table) { | |
990454b5 RG |
233 | if (vsock_addr_equals_addr(src, &vsk->remote_addr) && |
234 | dst->svm_port == vsk->local_addr.svm_port) { | |
d021c344 AK |
235 | return sk_vsock(vsk); |
236 | } | |
237 | } | |
238 | ||
239 | return NULL; | |
240 | } | |
241 | ||
242 | static bool __vsock_in_bound_table(struct vsock_sock *vsk) | |
243 | { | |
244 | return !list_empty(&vsk->bound_table); | |
245 | } | |
246 | ||
247 | static bool __vsock_in_connected_table(struct vsock_sock *vsk) | |
248 | { | |
249 | return !list_empty(&vsk->connected_table); | |
250 | } | |
251 | ||
252 | static void vsock_insert_unbound(struct vsock_sock *vsk) | |
253 | { | |
254 | spin_lock_bh(&vsock_table_lock); | |
255 | __vsock_insert_bound(vsock_unbound_sockets, vsk); | |
256 | spin_unlock_bh(&vsock_table_lock); | |
257 | } | |
258 | ||
259 | void vsock_insert_connected(struct vsock_sock *vsk) | |
260 | { | |
261 | struct list_head *list = vsock_connected_sockets( | |
262 | &vsk->remote_addr, &vsk->local_addr); | |
263 | ||
264 | spin_lock_bh(&vsock_table_lock); | |
265 | __vsock_insert_connected(list, vsk); | |
266 | spin_unlock_bh(&vsock_table_lock); | |
267 | } | |
268 | EXPORT_SYMBOL_GPL(vsock_insert_connected); | |
269 | ||
270 | void vsock_remove_bound(struct vsock_sock *vsk) | |
271 | { | |
272 | spin_lock_bh(&vsock_table_lock); | |
273 | __vsock_remove_bound(vsk); | |
274 | spin_unlock_bh(&vsock_table_lock); | |
275 | } | |
276 | EXPORT_SYMBOL_GPL(vsock_remove_bound); | |
277 | ||
278 | void vsock_remove_connected(struct vsock_sock *vsk) | |
279 | { | |
280 | spin_lock_bh(&vsock_table_lock); | |
281 | __vsock_remove_connected(vsk); | |
282 | spin_unlock_bh(&vsock_table_lock); | |
283 | } | |
284 | EXPORT_SYMBOL_GPL(vsock_remove_connected); | |
285 | ||
286 | struct sock *vsock_find_bound_socket(struct sockaddr_vm *addr) | |
287 | { | |
288 | struct sock *sk; | |
289 | ||
290 | spin_lock_bh(&vsock_table_lock); | |
291 | sk = __vsock_find_bound_socket(addr); | |
292 | if (sk) | |
293 | sock_hold(sk); | |
294 | ||
295 | spin_unlock_bh(&vsock_table_lock); | |
296 | ||
297 | return sk; | |
298 | } | |
299 | EXPORT_SYMBOL_GPL(vsock_find_bound_socket); | |
300 | ||
301 | struct sock *vsock_find_connected_socket(struct sockaddr_vm *src, | |
302 | struct sockaddr_vm *dst) | |
303 | { | |
304 | struct sock *sk; | |
305 | ||
306 | spin_lock_bh(&vsock_table_lock); | |
307 | sk = __vsock_find_connected_socket(src, dst); | |
308 | if (sk) | |
309 | sock_hold(sk); | |
310 | ||
311 | spin_unlock_bh(&vsock_table_lock); | |
312 | ||
313 | return sk; | |
314 | } | |
315 | EXPORT_SYMBOL_GPL(vsock_find_connected_socket); | |
316 | ||
317 | static bool vsock_in_bound_table(struct vsock_sock *vsk) | |
318 | { | |
319 | bool ret; | |
320 | ||
321 | spin_lock_bh(&vsock_table_lock); | |
322 | ret = __vsock_in_bound_table(vsk); | |
323 | spin_unlock_bh(&vsock_table_lock); | |
324 | ||
325 | return ret; | |
326 | } | |
327 | ||
328 | static bool vsock_in_connected_table(struct vsock_sock *vsk) | |
329 | { | |
330 | bool ret; | |
331 | ||
332 | spin_lock_bh(&vsock_table_lock); | |
333 | ret = __vsock_in_connected_table(vsk); | |
334 | spin_unlock_bh(&vsock_table_lock); | |
335 | ||
336 | return ret; | |
337 | } | |
338 | ||
339 | void vsock_for_each_connected_socket(void (*fn)(struct sock *sk)) | |
340 | { | |
341 | int i; | |
342 | ||
343 | spin_lock_bh(&vsock_table_lock); | |
344 | ||
345 | for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) { | |
346 | struct vsock_sock *vsk; | |
347 | list_for_each_entry(vsk, &vsock_connected_table[i], | |
d9af2d67 | 348 | connected_table) |
d021c344 AK |
349 | fn(sk_vsock(vsk)); |
350 | } | |
351 | ||
352 | spin_unlock_bh(&vsock_table_lock); | |
353 | } | |
354 | EXPORT_SYMBOL_GPL(vsock_for_each_connected_socket); | |
355 | ||
356 | void vsock_add_pending(struct sock *listener, struct sock *pending) | |
357 | { | |
358 | struct vsock_sock *vlistener; | |
359 | struct vsock_sock *vpending; | |
360 | ||
361 | vlistener = vsock_sk(listener); | |
362 | vpending = vsock_sk(pending); | |
363 | ||
364 | sock_hold(pending); | |
365 | sock_hold(listener); | |
366 | list_add_tail(&vpending->pending_links, &vlistener->pending_links); | |
367 | } | |
368 | EXPORT_SYMBOL_GPL(vsock_add_pending); | |
369 | ||
370 | void vsock_remove_pending(struct sock *listener, struct sock *pending) | |
371 | { | |
372 | struct vsock_sock *vpending = vsock_sk(pending); | |
373 | ||
374 | list_del_init(&vpending->pending_links); | |
375 | sock_put(listener); | |
376 | sock_put(pending); | |
377 | } | |
378 | EXPORT_SYMBOL_GPL(vsock_remove_pending); | |
379 | ||
380 | void vsock_enqueue_accept(struct sock *listener, struct sock *connected) | |
381 | { | |
382 | struct vsock_sock *vlistener; | |
383 | struct vsock_sock *vconnected; | |
384 | ||
385 | vlistener = vsock_sk(listener); | |
386 | vconnected = vsock_sk(connected); | |
387 | ||
388 | sock_hold(connected); | |
389 | sock_hold(listener); | |
390 | list_add_tail(&vconnected->accept_queue, &vlistener->accept_queue); | |
391 | } | |
392 | EXPORT_SYMBOL_GPL(vsock_enqueue_accept); | |
393 | ||
394 | static struct sock *vsock_dequeue_accept(struct sock *listener) | |
395 | { | |
396 | struct vsock_sock *vlistener; | |
397 | struct vsock_sock *vconnected; | |
398 | ||
399 | vlistener = vsock_sk(listener); | |
400 | ||
401 | if (list_empty(&vlistener->accept_queue)) | |
402 | return NULL; | |
403 | ||
404 | vconnected = list_entry(vlistener->accept_queue.next, | |
405 | struct vsock_sock, accept_queue); | |
406 | ||
407 | list_del_init(&vconnected->accept_queue); | |
408 | sock_put(listener); | |
409 | /* The caller will need a reference on the connected socket so we let | |
410 | * it call sock_put(). | |
411 | */ | |
412 | ||
413 | return sk_vsock(vconnected); | |
414 | } | |
415 | ||
416 | static bool vsock_is_accept_queue_empty(struct sock *sk) | |
417 | { | |
418 | struct vsock_sock *vsk = vsock_sk(sk); | |
419 | return list_empty(&vsk->accept_queue); | |
420 | } | |
421 | ||
422 | static bool vsock_is_pending(struct sock *sk) | |
423 | { | |
424 | struct vsock_sock *vsk = vsock_sk(sk); | |
425 | return !list_empty(&vsk->pending_links); | |
426 | } | |
427 | ||
428 | static int vsock_send_shutdown(struct sock *sk, int mode) | |
429 | { | |
430 | return transport->shutdown(vsock_sk(sk), mode); | |
431 | } | |
432 | ||
433 | void vsock_pending_work(struct work_struct *work) | |
434 | { | |
435 | struct sock *sk; | |
436 | struct sock *listener; | |
437 | struct vsock_sock *vsk; | |
438 | bool cleanup; | |
439 | ||
440 | vsk = container_of(work, struct vsock_sock, dwork.work); | |
441 | sk = sk_vsock(vsk); | |
442 | listener = vsk->listener; | |
443 | cleanup = true; | |
444 | ||
445 | lock_sock(listener); | |
446 | lock_sock(sk); | |
447 | ||
448 | if (vsock_is_pending(sk)) { | |
449 | vsock_remove_pending(listener, sk); | |
450 | } else if (!vsk->rejected) { | |
451 | /* We are not on the pending list and accept() did not reject | |
452 | * us, so we must have been accepted by our user process. We | |
453 | * just need to drop our references to the sockets and be on | |
454 | * our way. | |
455 | */ | |
456 | cleanup = false; | |
457 | goto out; | |
458 | } | |
459 | ||
460 | listener->sk_ack_backlog--; | |
461 | ||
462 | /* We need to remove ourself from the global connected sockets list so | |
463 | * incoming packets can't find this socket, and to reduce the reference | |
464 | * count. | |
465 | */ | |
466 | if (vsock_in_connected_table(vsk)) | |
467 | vsock_remove_connected(vsk); | |
468 | ||
469 | sk->sk_state = SS_FREE; | |
470 | ||
471 | out: | |
472 | release_sock(sk); | |
473 | release_sock(listener); | |
474 | if (cleanup) | |
475 | sock_put(sk); | |
476 | ||
477 | sock_put(sk); | |
478 | sock_put(listener); | |
479 | } | |
480 | EXPORT_SYMBOL_GPL(vsock_pending_work); | |
481 | ||
482 | /**** SOCKET OPERATIONS ****/ | |
483 | ||
484 | static int __vsock_bind_stream(struct vsock_sock *vsk, | |
485 | struct sockaddr_vm *addr) | |
486 | { | |
487 | static u32 port = LAST_RESERVED_PORT + 1; | |
488 | struct sockaddr_vm new_addr; | |
489 | ||
490 | vsock_addr_init(&new_addr, addr->svm_cid, addr->svm_port); | |
491 | ||
492 | if (addr->svm_port == VMADDR_PORT_ANY) { | |
493 | bool found = false; | |
494 | unsigned int i; | |
495 | ||
496 | for (i = 0; i < MAX_PORT_RETRIES; i++) { | |
497 | if (port <= LAST_RESERVED_PORT) | |
498 | port = LAST_RESERVED_PORT + 1; | |
499 | ||
500 | new_addr.svm_port = port++; | |
501 | ||
502 | if (!__vsock_find_bound_socket(&new_addr)) { | |
503 | found = true; | |
504 | break; | |
505 | } | |
506 | } | |
507 | ||
508 | if (!found) | |
509 | return -EADDRNOTAVAIL; | |
510 | } else { | |
511 | /* If port is in reserved range, ensure caller | |
512 | * has necessary privileges. | |
513 | */ | |
514 | if (addr->svm_port <= LAST_RESERVED_PORT && | |
515 | !capable(CAP_NET_BIND_SERVICE)) { | |
516 | return -EACCES; | |
517 | } | |
518 | ||
519 | if (__vsock_find_bound_socket(&new_addr)) | |
520 | return -EADDRINUSE; | |
521 | } | |
522 | ||
523 | vsock_addr_init(&vsk->local_addr, new_addr.svm_cid, new_addr.svm_port); | |
524 | ||
525 | /* Remove stream sockets from the unbound list and add them to the hash | |
526 | * table for easy lookup by its address. The unbound list is simply an | |
527 | * extra entry at the end of the hash table, a trick used by AF_UNIX. | |
528 | */ | |
529 | __vsock_remove_bound(vsk); | |
530 | __vsock_insert_bound(vsock_bound_sockets(&vsk->local_addr), vsk); | |
531 | ||
532 | return 0; | |
533 | } | |
534 | ||
535 | static int __vsock_bind_dgram(struct vsock_sock *vsk, | |
536 | struct sockaddr_vm *addr) | |
537 | { | |
538 | return transport->dgram_bind(vsk, addr); | |
539 | } | |
540 | ||
541 | static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr) | |
542 | { | |
543 | struct vsock_sock *vsk = vsock_sk(sk); | |
544 | u32 cid; | |
545 | int retval; | |
546 | ||
547 | /* First ensure this socket isn't already bound. */ | |
548 | if (vsock_addr_bound(&vsk->local_addr)) | |
549 | return -EINVAL; | |
550 | ||
551 | /* Now bind to the provided address or select appropriate values if | |
552 | * none are provided (VMADDR_CID_ANY and VMADDR_PORT_ANY). Note that | |
553 | * like AF_INET prevents binding to a non-local IP address (in most | |
554 | * cases), we only allow binding to the local CID. | |
555 | */ | |
556 | cid = transport->get_local_cid(); | |
557 | if (addr->svm_cid != cid && addr->svm_cid != VMADDR_CID_ANY) | |
558 | return -EADDRNOTAVAIL; | |
559 | ||
560 | switch (sk->sk_socket->type) { | |
561 | case SOCK_STREAM: | |
562 | spin_lock_bh(&vsock_table_lock); | |
563 | retval = __vsock_bind_stream(vsk, addr); | |
564 | spin_unlock_bh(&vsock_table_lock); | |
565 | break; | |
566 | ||
567 | case SOCK_DGRAM: | |
568 | retval = __vsock_bind_dgram(vsk, addr); | |
569 | break; | |
570 | ||
571 | default: | |
572 | retval = -EINVAL; | |
573 | break; | |
574 | } | |
575 | ||
576 | return retval; | |
577 | } | |
578 | ||
579 | struct sock *__vsock_create(struct net *net, | |
580 | struct socket *sock, | |
581 | struct sock *parent, | |
582 | gfp_t priority, | |
11aa9c28 EB |
583 | unsigned short type, |
584 | int kern) | |
d021c344 AK |
585 | { |
586 | struct sock *sk; | |
587 | struct vsock_sock *psk; | |
588 | struct vsock_sock *vsk; | |
589 | ||
11aa9c28 | 590 | sk = sk_alloc(net, AF_VSOCK, priority, &vsock_proto, kern); |
d021c344 AK |
591 | if (!sk) |
592 | return NULL; | |
593 | ||
594 | sock_init_data(sock, sk); | |
595 | ||
596 | /* sk->sk_type is normally set in sock_init_data, but only if sock is | |
597 | * non-NULL. We make sure that our sockets always have a type by | |
598 | * setting it here if needed. | |
599 | */ | |
600 | if (!sock) | |
601 | sk->sk_type = type; | |
602 | ||
603 | vsk = vsock_sk(sk); | |
604 | vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); | |
605 | vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); | |
606 | ||
607 | sk->sk_destruct = vsock_sk_destruct; | |
608 | sk->sk_backlog_rcv = vsock_queue_rcv_skb; | |
609 | sk->sk_state = 0; | |
610 | sock_reset_flag(sk, SOCK_DONE); | |
611 | ||
612 | INIT_LIST_HEAD(&vsk->bound_table); | |
613 | INIT_LIST_HEAD(&vsk->connected_table); | |
614 | vsk->listener = NULL; | |
615 | INIT_LIST_HEAD(&vsk->pending_links); | |
616 | INIT_LIST_HEAD(&vsk->accept_queue); | |
617 | vsk->rejected = false; | |
618 | vsk->sent_request = false; | |
619 | vsk->ignore_connecting_rst = false; | |
620 | vsk->peer_shutdown = 0; | |
621 | ||
622 | psk = parent ? vsock_sk(parent) : NULL; | |
623 | if (parent) { | |
624 | vsk->trusted = psk->trusted; | |
625 | vsk->owner = get_cred(psk->owner); | |
626 | vsk->connect_timeout = psk->connect_timeout; | |
627 | } else { | |
628 | vsk->trusted = capable(CAP_NET_ADMIN); | |
629 | vsk->owner = get_current_cred(); | |
630 | vsk->connect_timeout = VSOCK_DEFAULT_CONNECT_TIMEOUT; | |
631 | } | |
632 | ||
633 | if (transport->init(vsk, psk) < 0) { | |
634 | sk_free(sk); | |
635 | return NULL; | |
636 | } | |
637 | ||
638 | if (sock) | |
639 | vsock_insert_unbound(vsk); | |
640 | ||
641 | return sk; | |
642 | } | |
643 | EXPORT_SYMBOL_GPL(__vsock_create); | |
644 | ||
645 | static void __vsock_release(struct sock *sk) | |
646 | { | |
647 | if (sk) { | |
648 | struct sk_buff *skb; | |
649 | struct sock *pending; | |
650 | struct vsock_sock *vsk; | |
651 | ||
652 | vsk = vsock_sk(sk); | |
653 | pending = NULL; /* Compiler warning. */ | |
654 | ||
655 | if (vsock_in_bound_table(vsk)) | |
656 | vsock_remove_bound(vsk); | |
657 | ||
658 | if (vsock_in_connected_table(vsk)) | |
659 | vsock_remove_connected(vsk); | |
660 | ||
661 | transport->release(vsk); | |
662 | ||
663 | lock_sock(sk); | |
664 | sock_orphan(sk); | |
665 | sk->sk_shutdown = SHUTDOWN_MASK; | |
666 | ||
667 | while ((skb = skb_dequeue(&sk->sk_receive_queue))) | |
668 | kfree_skb(skb); | |
669 | ||
670 | /* Clean up any sockets that never were accepted. */ | |
671 | while ((pending = vsock_dequeue_accept(sk)) != NULL) { | |
672 | __vsock_release(pending); | |
673 | sock_put(pending); | |
674 | } | |
675 | ||
676 | release_sock(sk); | |
677 | sock_put(sk); | |
678 | } | |
679 | } | |
680 | ||
681 | static void vsock_sk_destruct(struct sock *sk) | |
682 | { | |
683 | struct vsock_sock *vsk = vsock_sk(sk); | |
684 | ||
685 | transport->destruct(vsk); | |
686 | ||
687 | /* When clearing these addresses, there's no need to set the family and | |
688 | * possibly register the address family with the kernel. | |
689 | */ | |
690 | vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); | |
691 | vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); | |
692 | ||
693 | put_cred(vsk->owner); | |
694 | } | |
695 | ||
696 | static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) | |
697 | { | |
698 | int err; | |
699 | ||
700 | err = sock_queue_rcv_skb(sk, skb); | |
701 | if (err) | |
702 | kfree_skb(skb); | |
703 | ||
704 | return err; | |
705 | } | |
706 | ||
707 | s64 vsock_stream_has_data(struct vsock_sock *vsk) | |
708 | { | |
709 | return transport->stream_has_data(vsk); | |
710 | } | |
711 | EXPORT_SYMBOL_GPL(vsock_stream_has_data); | |
712 | ||
713 | s64 vsock_stream_has_space(struct vsock_sock *vsk) | |
714 | { | |
715 | return transport->stream_has_space(vsk); | |
716 | } | |
717 | EXPORT_SYMBOL_GPL(vsock_stream_has_space); | |
718 | ||
719 | static int vsock_release(struct socket *sock) | |
720 | { | |
721 | __vsock_release(sock->sk); | |
722 | sock->sk = NULL; | |
723 | sock->state = SS_FREE; | |
724 | ||
725 | return 0; | |
726 | } | |
727 | ||
728 | static int | |
729 | vsock_bind(struct socket *sock, struct sockaddr *addr, int addr_len) | |
730 | { | |
731 | int err; | |
732 | struct sock *sk; | |
733 | struct sockaddr_vm *vm_addr; | |
734 | ||
735 | sk = sock->sk; | |
736 | ||
737 | if (vsock_addr_cast(addr, addr_len, &vm_addr) != 0) | |
738 | return -EINVAL; | |
739 | ||
740 | lock_sock(sk); | |
741 | err = __vsock_bind(sk, vm_addr); | |
742 | release_sock(sk); | |
743 | ||
744 | return err; | |
745 | } | |
746 | ||
747 | static int vsock_getname(struct socket *sock, | |
748 | struct sockaddr *addr, int *addr_len, int peer) | |
749 | { | |
750 | int err; | |
751 | struct sock *sk; | |
752 | struct vsock_sock *vsk; | |
753 | struct sockaddr_vm *vm_addr; | |
754 | ||
755 | sk = sock->sk; | |
756 | vsk = vsock_sk(sk); | |
757 | err = 0; | |
758 | ||
759 | lock_sock(sk); | |
760 | ||
761 | if (peer) { | |
762 | if (sock->state != SS_CONNECTED) { | |
763 | err = -ENOTCONN; | |
764 | goto out; | |
765 | } | |
766 | vm_addr = &vsk->remote_addr; | |
767 | } else { | |
768 | vm_addr = &vsk->local_addr; | |
769 | } | |
770 | ||
771 | if (!vm_addr) { | |
772 | err = -EINVAL; | |
773 | goto out; | |
774 | } | |
775 | ||
776 | /* sys_getsockname() and sys_getpeername() pass us a | |
777 | * MAX_SOCK_ADDR-sized buffer and don't set addr_len. Unfortunately | |
778 | * that macro is defined in socket.c instead of .h, so we hardcode its | |
779 | * value here. | |
780 | */ | |
781 | BUILD_BUG_ON(sizeof(*vm_addr) > 128); | |
782 | memcpy(addr, vm_addr, sizeof(*vm_addr)); | |
783 | *addr_len = sizeof(*vm_addr); | |
784 | ||
785 | out: | |
786 | release_sock(sk); | |
787 | return err; | |
788 | } | |
789 | ||
790 | static int vsock_shutdown(struct socket *sock, int mode) | |
791 | { | |
792 | int err; | |
793 | struct sock *sk; | |
794 | ||
795 | /* User level uses SHUT_RD (0) and SHUT_WR (1), but the kernel uses | |
796 | * RCV_SHUTDOWN (1) and SEND_SHUTDOWN (2), so we must increment mode | |
797 | * here like the other address families do. Note also that the | |
798 | * increment makes SHUT_RDWR (2) into RCV_SHUTDOWN | SEND_SHUTDOWN (3), | |
799 | * which is what we want. | |
800 | */ | |
801 | mode++; | |
802 | ||
803 | if ((mode & ~SHUTDOWN_MASK) || !mode) | |
804 | return -EINVAL; | |
805 | ||
806 | /* If this is a STREAM socket and it is not connected then bail out | |
807 | * immediately. If it is a DGRAM socket then we must first kick the | |
808 | * socket so that it wakes up from any sleeping calls, for example | |
809 | * recv(), and then afterwards return the error. | |
810 | */ | |
811 | ||
812 | sk = sock->sk; | |
813 | if (sock->state == SS_UNCONNECTED) { | |
814 | err = -ENOTCONN; | |
815 | if (sk->sk_type == SOCK_STREAM) | |
816 | return err; | |
817 | } else { | |
818 | sock->state = SS_DISCONNECTING; | |
819 | err = 0; | |
820 | } | |
821 | ||
822 | /* Receive and send shutdowns are treated alike. */ | |
823 | mode = mode & (RCV_SHUTDOWN | SEND_SHUTDOWN); | |
824 | if (mode) { | |
825 | lock_sock(sk); | |
826 | sk->sk_shutdown |= mode; | |
827 | sk->sk_state_change(sk); | |
828 | release_sock(sk); | |
829 | ||
830 | if (sk->sk_type == SOCK_STREAM) { | |
831 | sock_reset_flag(sk, SOCK_DONE); | |
832 | vsock_send_shutdown(sk, mode); | |
833 | } | |
834 | } | |
835 | ||
836 | return err; | |
837 | } | |
838 | ||
839 | static unsigned int vsock_poll(struct file *file, struct socket *sock, | |
840 | poll_table *wait) | |
841 | { | |
842 | struct sock *sk; | |
843 | unsigned int mask; | |
844 | struct vsock_sock *vsk; | |
845 | ||
846 | sk = sock->sk; | |
847 | vsk = vsock_sk(sk); | |
848 | ||
849 | poll_wait(file, sk_sleep(sk), wait); | |
850 | mask = 0; | |
851 | ||
852 | if (sk->sk_err) | |
853 | /* Signify that there has been an error on this socket. */ | |
854 | mask |= POLLERR; | |
855 | ||
856 | /* INET sockets treat local write shutdown and peer write shutdown as a | |
857 | * case of POLLHUP set. | |
858 | */ | |
859 | if ((sk->sk_shutdown == SHUTDOWN_MASK) || | |
860 | ((sk->sk_shutdown & SEND_SHUTDOWN) && | |
861 | (vsk->peer_shutdown & SEND_SHUTDOWN))) { | |
862 | mask |= POLLHUP; | |
863 | } | |
864 | ||
865 | if (sk->sk_shutdown & RCV_SHUTDOWN || | |
866 | vsk->peer_shutdown & SEND_SHUTDOWN) { | |
867 | mask |= POLLRDHUP; | |
868 | } | |
869 | ||
870 | if (sock->type == SOCK_DGRAM) { | |
871 | /* For datagram sockets we can read if there is something in | |
872 | * the queue and write as long as the socket isn't shutdown for | |
873 | * sending. | |
874 | */ | |
875 | if (!skb_queue_empty(&sk->sk_receive_queue) || | |
876 | (sk->sk_shutdown & RCV_SHUTDOWN)) { | |
877 | mask |= POLLIN | POLLRDNORM; | |
878 | } | |
879 | ||
880 | if (!(sk->sk_shutdown & SEND_SHUTDOWN)) | |
881 | mask |= POLLOUT | POLLWRNORM | POLLWRBAND; | |
882 | ||
883 | } else if (sock->type == SOCK_STREAM) { | |
884 | lock_sock(sk); | |
885 | ||
886 | /* Listening sockets that have connections in their accept | |
887 | * queue can be read. | |
888 | */ | |
ea3803c1 | 889 | if (sk->sk_state == VSOCK_SS_LISTEN |
d021c344 AK |
890 | && !vsock_is_accept_queue_empty(sk)) |
891 | mask |= POLLIN | POLLRDNORM; | |
892 | ||
893 | /* If there is something in the queue then we can read. */ | |
894 | if (transport->stream_is_active(vsk) && | |
895 | !(sk->sk_shutdown & RCV_SHUTDOWN)) { | |
896 | bool data_ready_now = false; | |
897 | int ret = transport->notify_poll_in( | |
898 | vsk, 1, &data_ready_now); | |
899 | if (ret < 0) { | |
900 | mask |= POLLERR; | |
901 | } else { | |
902 | if (data_ready_now) | |
903 | mask |= POLLIN | POLLRDNORM; | |
904 | ||
905 | } | |
906 | } | |
907 | ||
908 | /* Sockets whose connections have been closed, reset, or | |
909 | * terminated should also be considered read, and we check the | |
910 | * shutdown flag for that. | |
911 | */ | |
912 | if (sk->sk_shutdown & RCV_SHUTDOWN || | |
913 | vsk->peer_shutdown & SEND_SHUTDOWN) { | |
914 | mask |= POLLIN | POLLRDNORM; | |
915 | } | |
916 | ||
917 | /* Connected sockets that can produce data can be written. */ | |
918 | if (sk->sk_state == SS_CONNECTED) { | |
919 | if (!(sk->sk_shutdown & SEND_SHUTDOWN)) { | |
920 | bool space_avail_now = false; | |
921 | int ret = transport->notify_poll_out( | |
922 | vsk, 1, &space_avail_now); | |
923 | if (ret < 0) { | |
924 | mask |= POLLERR; | |
925 | } else { | |
926 | if (space_avail_now) | |
927 | /* Remove POLLWRBAND since INET | |
928 | * sockets are not setting it. | |
929 | */ | |
930 | mask |= POLLOUT | POLLWRNORM; | |
931 | ||
932 | } | |
933 | } | |
934 | } | |
935 | ||
936 | /* Simulate INET socket poll behaviors, which sets | |
937 | * POLLOUT|POLLWRNORM when peer is closed and nothing to read, | |
938 | * but local send is not shutdown. | |
939 | */ | |
940 | if (sk->sk_state == SS_UNCONNECTED) { | |
941 | if (!(sk->sk_shutdown & SEND_SHUTDOWN)) | |
942 | mask |= POLLOUT | POLLWRNORM; | |
943 | ||
944 | } | |
945 | ||
946 | release_sock(sk); | |
947 | } | |
948 | ||
949 | return mask; | |
950 | } | |
951 | ||
1b784140 YX |
952 | static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg, |
953 | size_t len) | |
d021c344 AK |
954 | { |
955 | int err; | |
956 | struct sock *sk; | |
957 | struct vsock_sock *vsk; | |
958 | struct sockaddr_vm *remote_addr; | |
959 | ||
960 | if (msg->msg_flags & MSG_OOB) | |
961 | return -EOPNOTSUPP; | |
962 | ||
963 | /* For now, MSG_DONTWAIT is always assumed... */ | |
964 | err = 0; | |
965 | sk = sock->sk; | |
966 | vsk = vsock_sk(sk); | |
967 | ||
968 | lock_sock(sk); | |
969 | ||
b3a6dfe8 AH |
970 | err = vsock_auto_bind(vsk); |
971 | if (err) | |
972 | goto out; | |
d021c344 | 973 | |
d021c344 AK |
974 | |
975 | /* If the provided message contains an address, use that. Otherwise | |
976 | * fall back on the socket's remote handle (if it has been connected). | |
977 | */ | |
978 | if (msg->msg_name && | |
979 | vsock_addr_cast(msg->msg_name, msg->msg_namelen, | |
980 | &remote_addr) == 0) { | |
981 | /* Ensure this address is of the right type and is a valid | |
982 | * destination. | |
983 | */ | |
984 | ||
985 | if (remote_addr->svm_cid == VMADDR_CID_ANY) | |
986 | remote_addr->svm_cid = transport->get_local_cid(); | |
987 | ||
988 | if (!vsock_addr_bound(remote_addr)) { | |
989 | err = -EINVAL; | |
990 | goto out; | |
991 | } | |
992 | } else if (sock->state == SS_CONNECTED) { | |
993 | remote_addr = &vsk->remote_addr; | |
994 | ||
995 | if (remote_addr->svm_cid == VMADDR_CID_ANY) | |
996 | remote_addr->svm_cid = transport->get_local_cid(); | |
997 | ||
998 | /* XXX Should connect() or this function ensure remote_addr is | |
999 | * bound? | |
1000 | */ | |
1001 | if (!vsock_addr_bound(&vsk->remote_addr)) { | |
1002 | err = -EINVAL; | |
1003 | goto out; | |
1004 | } | |
1005 | } else { | |
1006 | err = -EINVAL; | |
1007 | goto out; | |
1008 | } | |
1009 | ||
1010 | if (!transport->dgram_allow(remote_addr->svm_cid, | |
1011 | remote_addr->svm_port)) { | |
1012 | err = -EINVAL; | |
1013 | goto out; | |
1014 | } | |
1015 | ||
0f7db23a | 1016 | err = transport->dgram_enqueue(vsk, remote_addr, msg, len); |
d021c344 AK |
1017 | |
1018 | out: | |
1019 | release_sock(sk); | |
1020 | return err; | |
1021 | } | |
1022 | ||
1023 | static int vsock_dgram_connect(struct socket *sock, | |
1024 | struct sockaddr *addr, int addr_len, int flags) | |
1025 | { | |
1026 | int err; | |
1027 | struct sock *sk; | |
1028 | struct vsock_sock *vsk; | |
1029 | struct sockaddr_vm *remote_addr; | |
1030 | ||
1031 | sk = sock->sk; | |
1032 | vsk = vsock_sk(sk); | |
1033 | ||
1034 | err = vsock_addr_cast(addr, addr_len, &remote_addr); | |
1035 | if (err == -EAFNOSUPPORT && remote_addr->svm_family == AF_UNSPEC) { | |
1036 | lock_sock(sk); | |
1037 | vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, | |
1038 | VMADDR_PORT_ANY); | |
1039 | sock->state = SS_UNCONNECTED; | |
1040 | release_sock(sk); | |
1041 | return 0; | |
1042 | } else if (err != 0) | |
1043 | return -EINVAL; | |
1044 | ||
1045 | lock_sock(sk); | |
1046 | ||
b3a6dfe8 AH |
1047 | err = vsock_auto_bind(vsk); |
1048 | if (err) | |
1049 | goto out; | |
d021c344 AK |
1050 | |
1051 | if (!transport->dgram_allow(remote_addr->svm_cid, | |
1052 | remote_addr->svm_port)) { | |
1053 | err = -EINVAL; | |
1054 | goto out; | |
1055 | } | |
1056 | ||
1057 | memcpy(&vsk->remote_addr, remote_addr, sizeof(vsk->remote_addr)); | |
1058 | sock->state = SS_CONNECTED; | |
1059 | ||
1060 | out: | |
1061 | release_sock(sk); | |
1062 | return err; | |
1063 | } | |
1064 | ||
1b784140 YX |
1065 | static int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg, |
1066 | size_t len, int flags) | |
d021c344 | 1067 | { |
1b784140 | 1068 | return transport->dgram_dequeue(vsock_sk(sock->sk), msg, len, flags); |
d021c344 AK |
1069 | } |
1070 | ||
1071 | static const struct proto_ops vsock_dgram_ops = { | |
1072 | .family = PF_VSOCK, | |
1073 | .owner = THIS_MODULE, | |
1074 | .release = vsock_release, | |
1075 | .bind = vsock_bind, | |
1076 | .connect = vsock_dgram_connect, | |
1077 | .socketpair = sock_no_socketpair, | |
1078 | .accept = sock_no_accept, | |
1079 | .getname = vsock_getname, | |
1080 | .poll = vsock_poll, | |
1081 | .ioctl = sock_no_ioctl, | |
1082 | .listen = sock_no_listen, | |
1083 | .shutdown = vsock_shutdown, | |
1084 | .setsockopt = sock_no_setsockopt, | |
1085 | .getsockopt = sock_no_getsockopt, | |
1086 | .sendmsg = vsock_dgram_sendmsg, | |
1087 | .recvmsg = vsock_dgram_recvmsg, | |
1088 | .mmap = sock_no_mmap, | |
1089 | .sendpage = sock_no_sendpage, | |
1090 | }; | |
1091 | ||
1092 | static void vsock_connect_timeout(struct work_struct *work) | |
1093 | { | |
1094 | struct sock *sk; | |
1095 | struct vsock_sock *vsk; | |
1096 | ||
1097 | vsk = container_of(work, struct vsock_sock, dwork.work); | |
1098 | sk = sk_vsock(vsk); | |
1099 | ||
1100 | lock_sock(sk); | |
1101 | if (sk->sk_state == SS_CONNECTING && | |
1102 | (sk->sk_shutdown != SHUTDOWN_MASK)) { | |
1103 | sk->sk_state = SS_UNCONNECTED; | |
1104 | sk->sk_err = ETIMEDOUT; | |
1105 | sk->sk_error_report(sk); | |
1106 | } | |
1107 | release_sock(sk); | |
1108 | ||
1109 | sock_put(sk); | |
1110 | } | |
1111 | ||
1112 | static int vsock_stream_connect(struct socket *sock, struct sockaddr *addr, | |
1113 | int addr_len, int flags) | |
1114 | { | |
1115 | int err; | |
1116 | struct sock *sk; | |
1117 | struct vsock_sock *vsk; | |
1118 | struct sockaddr_vm *remote_addr; | |
1119 | long timeout; | |
1120 | DEFINE_WAIT(wait); | |
1121 | ||
1122 | err = 0; | |
1123 | sk = sock->sk; | |
1124 | vsk = vsock_sk(sk); | |
1125 | ||
1126 | lock_sock(sk); | |
1127 | ||
1128 | /* XXX AF_UNSPEC should make us disconnect like AF_INET. */ | |
1129 | switch (sock->state) { | |
1130 | case SS_CONNECTED: | |
1131 | err = -EISCONN; | |
1132 | goto out; | |
1133 | case SS_DISCONNECTING: | |
1134 | err = -EINVAL; | |
1135 | goto out; | |
1136 | case SS_CONNECTING: | |
1137 | /* This continues on so we can move sock into the SS_CONNECTED | |
1138 | * state once the connection has completed (at which point err | |
1139 | * will be set to zero also). Otherwise, we will either wait | |
1140 | * for the connection or return -EALREADY should this be a | |
1141 | * non-blocking call. | |
1142 | */ | |
1143 | err = -EALREADY; | |
1144 | break; | |
1145 | default: | |
ea3803c1 | 1146 | if ((sk->sk_state == VSOCK_SS_LISTEN) || |
d021c344 AK |
1147 | vsock_addr_cast(addr, addr_len, &remote_addr) != 0) { |
1148 | err = -EINVAL; | |
1149 | goto out; | |
1150 | } | |
1151 | ||
1152 | /* The hypervisor and well-known contexts do not have socket | |
1153 | * endpoints. | |
1154 | */ | |
1155 | if (!transport->stream_allow(remote_addr->svm_cid, | |
1156 | remote_addr->svm_port)) { | |
1157 | err = -ENETUNREACH; | |
1158 | goto out; | |
1159 | } | |
1160 | ||
1161 | /* Set the remote address that we are connecting to. */ | |
1162 | memcpy(&vsk->remote_addr, remote_addr, | |
1163 | sizeof(vsk->remote_addr)); | |
1164 | ||
b3a6dfe8 AH |
1165 | err = vsock_auto_bind(vsk); |
1166 | if (err) | |
1167 | goto out; | |
d021c344 AK |
1168 | |
1169 | sk->sk_state = SS_CONNECTING; | |
1170 | ||
1171 | err = transport->connect(vsk); | |
1172 | if (err < 0) | |
1173 | goto out; | |
1174 | ||
1175 | /* Mark sock as connecting and set the error code to in | |
1176 | * progress in case this is a non-blocking connect. | |
1177 | */ | |
1178 | sock->state = SS_CONNECTING; | |
1179 | err = -EINPROGRESS; | |
1180 | } | |
1181 | ||
1182 | /* The receive path will handle all communication until we are able to | |
1183 | * enter the connected state. Here we wait for the connection to be | |
1184 | * completed or a notification of an error. | |
1185 | */ | |
1186 | timeout = vsk->connect_timeout; | |
1187 | prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); | |
1188 | ||
1189 | while (sk->sk_state != SS_CONNECTED && sk->sk_err == 0) { | |
1190 | if (flags & O_NONBLOCK) { | |
1191 | /* If we're not going to block, we schedule a timeout | |
1192 | * function to generate a timeout on the connection | |
1193 | * attempt, in case the peer doesn't respond in a | |
1194 | * timely manner. We hold on to the socket until the | |
1195 | * timeout fires. | |
1196 | */ | |
1197 | sock_hold(sk); | |
1198 | INIT_DELAYED_WORK(&vsk->dwork, | |
1199 | vsock_connect_timeout); | |
1200 | schedule_delayed_work(&vsk->dwork, timeout); | |
1201 | ||
1202 | /* Skip ahead to preserve error code set above. */ | |
1203 | goto out_wait; | |
1204 | } | |
1205 | ||
1206 | release_sock(sk); | |
1207 | timeout = schedule_timeout(timeout); | |
1208 | lock_sock(sk); | |
1209 | ||
1210 | if (signal_pending(current)) { | |
1211 | err = sock_intr_errno(timeout); | |
f7f9b5e7 CI |
1212 | sk->sk_state = SS_UNCONNECTED; |
1213 | sock->state = SS_UNCONNECTED; | |
1214 | goto out_wait; | |
d021c344 AK |
1215 | } else if (timeout == 0) { |
1216 | err = -ETIMEDOUT; | |
f7f9b5e7 CI |
1217 | sk->sk_state = SS_UNCONNECTED; |
1218 | sock->state = SS_UNCONNECTED; | |
1219 | goto out_wait; | |
d021c344 AK |
1220 | } |
1221 | ||
1222 | prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); | |
1223 | } | |
1224 | ||
1225 | if (sk->sk_err) { | |
1226 | err = -sk->sk_err; | |
f7f9b5e7 CI |
1227 | sk->sk_state = SS_UNCONNECTED; |
1228 | sock->state = SS_UNCONNECTED; | |
1229 | } else { | |
d021c344 | 1230 | err = 0; |
f7f9b5e7 | 1231 | } |
d021c344 AK |
1232 | |
1233 | out_wait: | |
1234 | finish_wait(sk_sleep(sk), &wait); | |
1235 | out: | |
1236 | release_sock(sk); | |
1237 | return err; | |
d021c344 AK |
1238 | } |
1239 | ||
1240 | static int vsock_accept(struct socket *sock, struct socket *newsock, int flags) | |
1241 | { | |
1242 | struct sock *listener; | |
1243 | int err; | |
1244 | struct sock *connected; | |
1245 | struct vsock_sock *vconnected; | |
1246 | long timeout; | |
1247 | DEFINE_WAIT(wait); | |
1248 | ||
1249 | err = 0; | |
1250 | listener = sock->sk; | |
1251 | ||
1252 | lock_sock(listener); | |
1253 | ||
1254 | if (sock->type != SOCK_STREAM) { | |
1255 | err = -EOPNOTSUPP; | |
1256 | goto out; | |
1257 | } | |
1258 | ||
ea3803c1 | 1259 | if (listener->sk_state != VSOCK_SS_LISTEN) { |
d021c344 AK |
1260 | err = -EINVAL; |
1261 | goto out; | |
1262 | } | |
1263 | ||
1264 | /* Wait for children sockets to appear; these are the new sockets | |
1265 | * created upon connection establishment. | |
1266 | */ | |
1267 | timeout = sock_sndtimeo(listener, flags & O_NONBLOCK); | |
1268 | prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE); | |
1269 | ||
1270 | while ((connected = vsock_dequeue_accept(listener)) == NULL && | |
1271 | listener->sk_err == 0) { | |
1272 | release_sock(listener); | |
1273 | timeout = schedule_timeout(timeout); | |
f7f9b5e7 | 1274 | finish_wait(sk_sleep(listener), &wait); |
d021c344 AK |
1275 | lock_sock(listener); |
1276 | ||
1277 | if (signal_pending(current)) { | |
1278 | err = sock_intr_errno(timeout); | |
f7f9b5e7 | 1279 | goto out; |
d021c344 AK |
1280 | } else if (timeout == 0) { |
1281 | err = -EAGAIN; | |
f7f9b5e7 | 1282 | goto out; |
d021c344 AK |
1283 | } |
1284 | ||
1285 | prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE); | |
1286 | } | |
f7f9b5e7 | 1287 | finish_wait(sk_sleep(listener), &wait); |
d021c344 AK |
1288 | |
1289 | if (listener->sk_err) | |
1290 | err = -listener->sk_err; | |
1291 | ||
1292 | if (connected) { | |
1293 | listener->sk_ack_backlog--; | |
1294 | ||
1295 | lock_sock(connected); | |
1296 | vconnected = vsock_sk(connected); | |
1297 | ||
1298 | /* If the listener socket has received an error, then we should | |
1299 | * reject this socket and return. Note that we simply mark the | |
1300 | * socket rejected, drop our reference, and let the cleanup | |
1301 | * function handle the cleanup; the fact that we found it in | |
1302 | * the listener's accept queue guarantees that the cleanup | |
1303 | * function hasn't run yet. | |
1304 | */ | |
1305 | if (err) { | |
1306 | vconnected->rejected = true; | |
f7f9b5e7 CI |
1307 | } else { |
1308 | newsock->state = SS_CONNECTED; | |
1309 | sock_graft(connected, newsock); | |
d021c344 AK |
1310 | } |
1311 | ||
d021c344 AK |
1312 | release_sock(connected); |
1313 | sock_put(connected); | |
1314 | } | |
1315 | ||
d021c344 AK |
1316 | out: |
1317 | release_sock(listener); | |
1318 | return err; | |
1319 | } | |
1320 | ||
1321 | static int vsock_listen(struct socket *sock, int backlog) | |
1322 | { | |
1323 | int err; | |
1324 | struct sock *sk; | |
1325 | struct vsock_sock *vsk; | |
1326 | ||
1327 | sk = sock->sk; | |
1328 | ||
1329 | lock_sock(sk); | |
1330 | ||
1331 | if (sock->type != SOCK_STREAM) { | |
1332 | err = -EOPNOTSUPP; | |
1333 | goto out; | |
1334 | } | |
1335 | ||
1336 | if (sock->state != SS_UNCONNECTED) { | |
1337 | err = -EINVAL; | |
1338 | goto out; | |
1339 | } | |
1340 | ||
1341 | vsk = vsock_sk(sk); | |
1342 | ||
1343 | if (!vsock_addr_bound(&vsk->local_addr)) { | |
1344 | err = -EINVAL; | |
1345 | goto out; | |
1346 | } | |
1347 | ||
1348 | sk->sk_max_ack_backlog = backlog; | |
ea3803c1 | 1349 | sk->sk_state = VSOCK_SS_LISTEN; |
d021c344 AK |
1350 | |
1351 | err = 0; | |
1352 | ||
1353 | out: | |
1354 | release_sock(sk); | |
1355 | return err; | |
1356 | } | |
1357 | ||
1358 | static int vsock_stream_setsockopt(struct socket *sock, | |
1359 | int level, | |
1360 | int optname, | |
1361 | char __user *optval, | |
1362 | unsigned int optlen) | |
1363 | { | |
1364 | int err; | |
1365 | struct sock *sk; | |
1366 | struct vsock_sock *vsk; | |
1367 | u64 val; | |
1368 | ||
1369 | if (level != AF_VSOCK) | |
1370 | return -ENOPROTOOPT; | |
1371 | ||
1372 | #define COPY_IN(_v) \ | |
1373 | do { \ | |
1374 | if (optlen < sizeof(_v)) { \ | |
1375 | err = -EINVAL; \ | |
1376 | goto exit; \ | |
1377 | } \ | |
1378 | if (copy_from_user(&_v, optval, sizeof(_v)) != 0) { \ | |
1379 | err = -EFAULT; \ | |
1380 | goto exit; \ | |
1381 | } \ | |
1382 | } while (0) | |
1383 | ||
1384 | err = 0; | |
1385 | sk = sock->sk; | |
1386 | vsk = vsock_sk(sk); | |
1387 | ||
1388 | lock_sock(sk); | |
1389 | ||
1390 | switch (optname) { | |
1391 | case SO_VM_SOCKETS_BUFFER_SIZE: | |
1392 | COPY_IN(val); | |
1393 | transport->set_buffer_size(vsk, val); | |
1394 | break; | |
1395 | ||
1396 | case SO_VM_SOCKETS_BUFFER_MAX_SIZE: | |
1397 | COPY_IN(val); | |
1398 | transport->set_max_buffer_size(vsk, val); | |
1399 | break; | |
1400 | ||
1401 | case SO_VM_SOCKETS_BUFFER_MIN_SIZE: | |
1402 | COPY_IN(val); | |
1403 | transport->set_min_buffer_size(vsk, val); | |
1404 | break; | |
1405 | ||
1406 | case SO_VM_SOCKETS_CONNECT_TIMEOUT: { | |
1407 | struct timeval tv; | |
1408 | COPY_IN(tv); | |
1409 | if (tv.tv_sec >= 0 && tv.tv_usec < USEC_PER_SEC && | |
1410 | tv.tv_sec < (MAX_SCHEDULE_TIMEOUT / HZ - 1)) { | |
1411 | vsk->connect_timeout = tv.tv_sec * HZ + | |
1412 | DIV_ROUND_UP(tv.tv_usec, (1000000 / HZ)); | |
1413 | if (vsk->connect_timeout == 0) | |
1414 | vsk->connect_timeout = | |
1415 | VSOCK_DEFAULT_CONNECT_TIMEOUT; | |
1416 | ||
1417 | } else { | |
1418 | err = -ERANGE; | |
1419 | } | |
1420 | break; | |
1421 | } | |
1422 | ||
1423 | default: | |
1424 | err = -ENOPROTOOPT; | |
1425 | break; | |
1426 | } | |
1427 | ||
1428 | #undef COPY_IN | |
1429 | ||
1430 | exit: | |
1431 | release_sock(sk); | |
1432 | return err; | |
1433 | } | |
1434 | ||
1435 | static int vsock_stream_getsockopt(struct socket *sock, | |
1436 | int level, int optname, | |
1437 | char __user *optval, | |
1438 | int __user *optlen) | |
1439 | { | |
1440 | int err; | |
1441 | int len; | |
1442 | struct sock *sk; | |
1443 | struct vsock_sock *vsk; | |
1444 | u64 val; | |
1445 | ||
1446 | if (level != AF_VSOCK) | |
1447 | return -ENOPROTOOPT; | |
1448 | ||
1449 | err = get_user(len, optlen); | |
1450 | if (err != 0) | |
1451 | return err; | |
1452 | ||
1453 | #define COPY_OUT(_v) \ | |
1454 | do { \ | |
1455 | if (len < sizeof(_v)) \ | |
1456 | return -EINVAL; \ | |
1457 | \ | |
1458 | len = sizeof(_v); \ | |
1459 | if (copy_to_user(optval, &_v, len) != 0) \ | |
1460 | return -EFAULT; \ | |
1461 | \ | |
1462 | } while (0) | |
1463 | ||
1464 | err = 0; | |
1465 | sk = sock->sk; | |
1466 | vsk = vsock_sk(sk); | |
1467 | ||
1468 | switch (optname) { | |
1469 | case SO_VM_SOCKETS_BUFFER_SIZE: | |
1470 | val = transport->get_buffer_size(vsk); | |
1471 | COPY_OUT(val); | |
1472 | break; | |
1473 | ||
1474 | case SO_VM_SOCKETS_BUFFER_MAX_SIZE: | |
1475 | val = transport->get_max_buffer_size(vsk); | |
1476 | COPY_OUT(val); | |
1477 | break; | |
1478 | ||
1479 | case SO_VM_SOCKETS_BUFFER_MIN_SIZE: | |
1480 | val = transport->get_min_buffer_size(vsk); | |
1481 | COPY_OUT(val); | |
1482 | break; | |
1483 | ||
1484 | case SO_VM_SOCKETS_CONNECT_TIMEOUT: { | |
1485 | struct timeval tv; | |
1486 | tv.tv_sec = vsk->connect_timeout / HZ; | |
1487 | tv.tv_usec = | |
1488 | (vsk->connect_timeout - | |
1489 | tv.tv_sec * HZ) * (1000000 / HZ); | |
1490 | COPY_OUT(tv); | |
1491 | break; | |
1492 | } | |
1493 | default: | |
1494 | return -ENOPROTOOPT; | |
1495 | } | |
1496 | ||
1497 | err = put_user(len, optlen); | |
1498 | if (err != 0) | |
1499 | return -EFAULT; | |
1500 | ||
1501 | #undef COPY_OUT | |
1502 | ||
1503 | return 0; | |
1504 | } | |
1505 | ||
1b784140 YX |
1506 | static int vsock_stream_sendmsg(struct socket *sock, struct msghdr *msg, |
1507 | size_t len) | |
d021c344 AK |
1508 | { |
1509 | struct sock *sk; | |
1510 | struct vsock_sock *vsk; | |
1511 | ssize_t total_written; | |
1512 | long timeout; | |
1513 | int err; | |
1514 | struct vsock_transport_send_notify_data send_data; | |
1515 | ||
1516 | DEFINE_WAIT(wait); | |
1517 | ||
1518 | sk = sock->sk; | |
1519 | vsk = vsock_sk(sk); | |
1520 | total_written = 0; | |
1521 | err = 0; | |
1522 | ||
1523 | if (msg->msg_flags & MSG_OOB) | |
1524 | return -EOPNOTSUPP; | |
1525 | ||
1526 | lock_sock(sk); | |
1527 | ||
1528 | /* Callers should not provide a destination with stream sockets. */ | |
1529 | if (msg->msg_namelen) { | |
1530 | err = sk->sk_state == SS_CONNECTED ? -EISCONN : -EOPNOTSUPP; | |
1531 | goto out; | |
1532 | } | |
1533 | ||
1534 | /* Send data only if both sides are not shutdown in the direction. */ | |
1535 | if (sk->sk_shutdown & SEND_SHUTDOWN || | |
1536 | vsk->peer_shutdown & RCV_SHUTDOWN) { | |
1537 | err = -EPIPE; | |
1538 | goto out; | |
1539 | } | |
1540 | ||
1541 | if (sk->sk_state != SS_CONNECTED || | |
1542 | !vsock_addr_bound(&vsk->local_addr)) { | |
1543 | err = -ENOTCONN; | |
1544 | goto out; | |
1545 | } | |
1546 | ||
1547 | if (!vsock_addr_bound(&vsk->remote_addr)) { | |
1548 | err = -EDESTADDRREQ; | |
1549 | goto out; | |
1550 | } | |
1551 | ||
1552 | /* Wait for room in the produce queue to enqueue our user's data. */ | |
1553 | timeout = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT); | |
1554 | ||
1555 | err = transport->notify_send_init(vsk, &send_data); | |
1556 | if (err < 0) | |
1557 | goto out; | |
1558 | ||
6f57e56a | 1559 | |
d021c344 AK |
1560 | while (total_written < len) { |
1561 | ssize_t written; | |
1562 | ||
f7f9b5e7 | 1563 | prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); |
d021c344 AK |
1564 | while (vsock_stream_has_space(vsk) == 0 && |
1565 | sk->sk_err == 0 && | |
1566 | !(sk->sk_shutdown & SEND_SHUTDOWN) && | |
1567 | !(vsk->peer_shutdown & RCV_SHUTDOWN)) { | |
1568 | ||
1569 | /* Don't wait for non-blocking sockets. */ | |
1570 | if (timeout == 0) { | |
1571 | err = -EAGAIN; | |
f7f9b5e7 CI |
1572 | finish_wait(sk_sleep(sk), &wait); |
1573 | goto out_err; | |
d021c344 AK |
1574 | } |
1575 | ||
1576 | err = transport->notify_send_pre_block(vsk, &send_data); | |
f7f9b5e7 CI |
1577 | if (err < 0) { |
1578 | finish_wait(sk_sleep(sk), &wait); | |
1579 | goto out_err; | |
1580 | } | |
d021c344 AK |
1581 | |
1582 | release_sock(sk); | |
1583 | timeout = schedule_timeout(timeout); | |
1584 | lock_sock(sk); | |
1585 | if (signal_pending(current)) { | |
1586 | err = sock_intr_errno(timeout); | |
f7f9b5e7 CI |
1587 | finish_wait(sk_sleep(sk), &wait); |
1588 | goto out_err; | |
d021c344 AK |
1589 | } else if (timeout == 0) { |
1590 | err = -EAGAIN; | |
f7f9b5e7 CI |
1591 | finish_wait(sk_sleep(sk), &wait); |
1592 | goto out_err; | |
d021c344 AK |
1593 | } |
1594 | ||
6f57e56a CI |
1595 | prepare_to_wait(sk_sleep(sk), &wait, |
1596 | TASK_INTERRUPTIBLE); | |
d021c344 | 1597 | } |
f7f9b5e7 | 1598 | finish_wait(sk_sleep(sk), &wait); |
d021c344 AK |
1599 | |
1600 | /* These checks occur both as part of and after the loop | |
1601 | * conditional since we need to check before and after | |
1602 | * sleeping. | |
1603 | */ | |
1604 | if (sk->sk_err) { | |
1605 | err = -sk->sk_err; | |
f7f9b5e7 | 1606 | goto out_err; |
d021c344 AK |
1607 | } else if ((sk->sk_shutdown & SEND_SHUTDOWN) || |
1608 | (vsk->peer_shutdown & RCV_SHUTDOWN)) { | |
1609 | err = -EPIPE; | |
f7f9b5e7 | 1610 | goto out_err; |
d021c344 AK |
1611 | } |
1612 | ||
1613 | err = transport->notify_send_pre_enqueue(vsk, &send_data); | |
1614 | if (err < 0) | |
f7f9b5e7 | 1615 | goto out_err; |
d021c344 AK |
1616 | |
1617 | /* Note that enqueue will only write as many bytes as are free | |
1618 | * in the produce queue, so we don't need to ensure len is | |
1619 | * smaller than the queue size. It is the caller's | |
1620 | * responsibility to check how many bytes we were able to send. | |
1621 | */ | |
1622 | ||
1623 | written = transport->stream_enqueue( | |
0f7db23a | 1624 | vsk, msg, |
d021c344 AK |
1625 | len - total_written); |
1626 | if (written < 0) { | |
1627 | err = -ENOMEM; | |
f7f9b5e7 | 1628 | goto out_err; |
d021c344 AK |
1629 | } |
1630 | ||
1631 | total_written += written; | |
1632 | ||
1633 | err = transport->notify_send_post_enqueue( | |
1634 | vsk, written, &send_data); | |
1635 | if (err < 0) | |
f7f9b5e7 | 1636 | goto out_err; |
d021c344 AK |
1637 | |
1638 | } | |
1639 | ||
f7f9b5e7 | 1640 | out_err: |
d021c344 AK |
1641 | if (total_written > 0) |
1642 | err = total_written; | |
d021c344 AK |
1643 | out: |
1644 | release_sock(sk); | |
1645 | return err; | |
1646 | } | |
1647 | ||
1648 | ||
1649 | static int | |
1b784140 YX |
1650 | vsock_stream_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, |
1651 | int flags) | |
d021c344 AK |
1652 | { |
1653 | struct sock *sk; | |
1654 | struct vsock_sock *vsk; | |
1655 | int err; | |
1656 | size_t target; | |
1657 | ssize_t copied; | |
1658 | long timeout; | |
1659 | struct vsock_transport_recv_notify_data recv_data; | |
1660 | ||
1661 | DEFINE_WAIT(wait); | |
1662 | ||
1663 | sk = sock->sk; | |
1664 | vsk = vsock_sk(sk); | |
1665 | err = 0; | |
1666 | ||
1667 | lock_sock(sk); | |
1668 | ||
1669 | if (sk->sk_state != SS_CONNECTED) { | |
1670 | /* Recvmsg is supposed to return 0 if a peer performs an | |
1671 | * orderly shutdown. Differentiate between that case and when a | |
1672 | * peer has not connected or a local shutdown occured with the | |
1673 | * SOCK_DONE flag. | |
1674 | */ | |
1675 | if (sock_flag(sk, SOCK_DONE)) | |
1676 | err = 0; | |
1677 | else | |
1678 | err = -ENOTCONN; | |
1679 | ||
1680 | goto out; | |
1681 | } | |
1682 | ||
1683 | if (flags & MSG_OOB) { | |
1684 | err = -EOPNOTSUPP; | |
1685 | goto out; | |
1686 | } | |
1687 | ||
1688 | /* We don't check peer_shutdown flag here since peer may actually shut | |
1689 | * down, but there can be data in the queue that a local socket can | |
1690 | * receive. | |
1691 | */ | |
1692 | if (sk->sk_shutdown & RCV_SHUTDOWN) { | |
1693 | err = 0; | |
1694 | goto out; | |
1695 | } | |
1696 | ||
1697 | /* It is valid on Linux to pass in a zero-length receive buffer. This | |
1698 | * is not an error. We may as well bail out now. | |
1699 | */ | |
1700 | if (!len) { | |
1701 | err = 0; | |
1702 | goto out; | |
1703 | } | |
1704 | ||
1705 | /* We must not copy less than target bytes into the user's buffer | |
1706 | * before returning successfully, so we wait for the consume queue to | |
1707 | * have that much data to consume before dequeueing. Note that this | |
1708 | * makes it impossible to handle cases where target is greater than the | |
1709 | * queue size. | |
1710 | */ | |
1711 | target = sock_rcvlowat(sk, flags & MSG_WAITALL, len); | |
1712 | if (target >= transport->stream_rcvhiwat(vsk)) { | |
1713 | err = -ENOMEM; | |
1714 | goto out; | |
1715 | } | |
1716 | timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT); | |
1717 | copied = 0; | |
1718 | ||
1719 | err = transport->notify_recv_init(vsk, target, &recv_data); | |
1720 | if (err < 0) | |
1721 | goto out; | |
1722 | ||
d021c344 AK |
1723 | |
1724 | while (1) { | |
f7f9b5e7 | 1725 | s64 ready; |
d021c344 | 1726 | |
f7f9b5e7 CI |
1727 | prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); |
1728 | ready = vsock_stream_has_data(vsk); | |
d021c344 | 1729 | |
f7f9b5e7 CI |
1730 | if (ready == 0) { |
1731 | if (sk->sk_err != 0 || | |
1732 | (sk->sk_shutdown & RCV_SHUTDOWN) || | |
1733 | (vsk->peer_shutdown & SEND_SHUTDOWN)) { | |
1734 | finish_wait(sk_sleep(sk), &wait); | |
1735 | break; | |
1736 | } | |
1737 | /* Don't wait for non-blocking sockets. */ | |
1738 | if (timeout == 0) { | |
1739 | err = -EAGAIN; | |
1740 | finish_wait(sk_sleep(sk), &wait); | |
1741 | break; | |
1742 | } | |
1743 | ||
1744 | err = transport->notify_recv_pre_block( | |
1745 | vsk, target, &recv_data); | |
1746 | if (err < 0) { | |
1747 | finish_wait(sk_sleep(sk), &wait); | |
1748 | break; | |
1749 | } | |
1750 | release_sock(sk); | |
1751 | timeout = schedule_timeout(timeout); | |
1752 | lock_sock(sk); | |
1753 | ||
1754 | if (signal_pending(current)) { | |
1755 | err = sock_intr_errno(timeout); | |
1756 | finish_wait(sk_sleep(sk), &wait); | |
1757 | break; | |
1758 | } else if (timeout == 0) { | |
1759 | err = -EAGAIN; | |
1760 | finish_wait(sk_sleep(sk), &wait); | |
1761 | break; | |
1762 | } | |
1763 | } else { | |
d021c344 AK |
1764 | ssize_t read; |
1765 | ||
f7f9b5e7 CI |
1766 | finish_wait(sk_sleep(sk), &wait); |
1767 | ||
1768 | if (ready < 0) { | |
1769 | /* Invalid queue pair content. XXX This should | |
1770 | * be changed to a connection reset in a later | |
1771 | * change. | |
1772 | */ | |
1773 | ||
1774 | err = -ENOMEM; | |
1775 | goto out; | |
1776 | } | |
1777 | ||
d021c344 AK |
1778 | err = transport->notify_recv_pre_dequeue( |
1779 | vsk, target, &recv_data); | |
1780 | if (err < 0) | |
1781 | break; | |
1782 | ||
1783 | read = transport->stream_dequeue( | |
0f7db23a | 1784 | vsk, msg, |
d021c344 AK |
1785 | len - copied, flags); |
1786 | if (read < 0) { | |
1787 | err = -ENOMEM; | |
1788 | break; | |
1789 | } | |
1790 | ||
1791 | copied += read; | |
1792 | ||
1793 | err = transport->notify_recv_post_dequeue( | |
1794 | vsk, target, read, | |
1795 | !(flags & MSG_PEEK), &recv_data); | |
1796 | if (err < 0) | |
f7f9b5e7 | 1797 | goto out; |
d021c344 AK |
1798 | |
1799 | if (read >= target || flags & MSG_PEEK) | |
1800 | break; | |
1801 | ||
1802 | target -= read; | |
d021c344 AK |
1803 | } |
1804 | } | |
1805 | ||
1806 | if (sk->sk_err) | |
1807 | err = -sk->sk_err; | |
1808 | else if (sk->sk_shutdown & RCV_SHUTDOWN) | |
1809 | err = 0; | |
1810 | ||
1811 | if (copied > 0) { | |
1812 | /* We only do these additional bookkeeping/notification steps | |
1813 | * if we actually copied something out of the queue pair | |
1814 | * instead of just peeking ahead. | |
1815 | */ | |
1816 | ||
1817 | if (!(flags & MSG_PEEK)) { | |
1818 | /* If the other side has shutdown for sending and there | |
1819 | * is nothing more to read, then modify the socket | |
1820 | * state. | |
1821 | */ | |
1822 | if (vsk->peer_shutdown & SEND_SHUTDOWN) { | |
1823 | if (vsock_stream_has_data(vsk) <= 0) { | |
1824 | sk->sk_state = SS_UNCONNECTED; | |
1825 | sock_set_flag(sk, SOCK_DONE); | |
1826 | sk->sk_state_change(sk); | |
1827 | } | |
1828 | } | |
1829 | } | |
1830 | err = copied; | |
1831 | } | |
1832 | ||
d021c344 AK |
1833 | out: |
1834 | release_sock(sk); | |
1835 | return err; | |
1836 | } | |
1837 | ||
1838 | static const struct proto_ops vsock_stream_ops = { | |
1839 | .family = PF_VSOCK, | |
1840 | .owner = THIS_MODULE, | |
1841 | .release = vsock_release, | |
1842 | .bind = vsock_bind, | |
1843 | .connect = vsock_stream_connect, | |
1844 | .socketpair = sock_no_socketpair, | |
1845 | .accept = vsock_accept, | |
1846 | .getname = vsock_getname, | |
1847 | .poll = vsock_poll, | |
1848 | .ioctl = sock_no_ioctl, | |
1849 | .listen = vsock_listen, | |
1850 | .shutdown = vsock_shutdown, | |
1851 | .setsockopt = vsock_stream_setsockopt, | |
1852 | .getsockopt = vsock_stream_getsockopt, | |
1853 | .sendmsg = vsock_stream_sendmsg, | |
1854 | .recvmsg = vsock_stream_recvmsg, | |
1855 | .mmap = sock_no_mmap, | |
1856 | .sendpage = sock_no_sendpage, | |
1857 | }; | |
1858 | ||
1859 | static int vsock_create(struct net *net, struct socket *sock, | |
1860 | int protocol, int kern) | |
1861 | { | |
1862 | if (!sock) | |
1863 | return -EINVAL; | |
1864 | ||
6cf1c5fc | 1865 | if (protocol && protocol != PF_VSOCK) |
d021c344 AK |
1866 | return -EPROTONOSUPPORT; |
1867 | ||
1868 | switch (sock->type) { | |
1869 | case SOCK_DGRAM: | |
1870 | sock->ops = &vsock_dgram_ops; | |
1871 | break; | |
1872 | case SOCK_STREAM: | |
1873 | sock->ops = &vsock_stream_ops; | |
1874 | break; | |
1875 | default: | |
1876 | return -ESOCKTNOSUPPORT; | |
1877 | } | |
1878 | ||
1879 | sock->state = SS_UNCONNECTED; | |
1880 | ||
11aa9c28 | 1881 | return __vsock_create(net, sock, NULL, GFP_KERNEL, 0, kern) ? 0 : -ENOMEM; |
d021c344 AK |
1882 | } |
1883 | ||
1884 | static const struct net_proto_family vsock_family_ops = { | |
1885 | .family = AF_VSOCK, | |
1886 | .create = vsock_create, | |
1887 | .owner = THIS_MODULE, | |
1888 | }; | |
1889 | ||
1890 | static long vsock_dev_do_ioctl(struct file *filp, | |
1891 | unsigned int cmd, void __user *ptr) | |
1892 | { | |
1893 | u32 __user *p = ptr; | |
1894 | int retval = 0; | |
1895 | ||
1896 | switch (cmd) { | |
1897 | case IOCTL_VM_SOCKETS_GET_LOCAL_CID: | |
1898 | if (put_user(transport->get_local_cid(), p) != 0) | |
1899 | retval = -EFAULT; | |
1900 | break; | |
1901 | ||
1902 | default: | |
1903 | pr_err("Unknown ioctl %d\n", cmd); | |
1904 | retval = -EINVAL; | |
1905 | } | |
1906 | ||
1907 | return retval; | |
1908 | } | |
1909 | ||
1910 | static long vsock_dev_ioctl(struct file *filp, | |
1911 | unsigned int cmd, unsigned long arg) | |
1912 | { | |
1913 | return vsock_dev_do_ioctl(filp, cmd, (void __user *)arg); | |
1914 | } | |
1915 | ||
1916 | #ifdef CONFIG_COMPAT | |
1917 | static long vsock_dev_compat_ioctl(struct file *filp, | |
1918 | unsigned int cmd, unsigned long arg) | |
1919 | { | |
1920 | return vsock_dev_do_ioctl(filp, cmd, compat_ptr(arg)); | |
1921 | } | |
1922 | #endif | |
1923 | ||
1924 | static const struct file_operations vsock_device_ops = { | |
1925 | .owner = THIS_MODULE, | |
1926 | .unlocked_ioctl = vsock_dev_ioctl, | |
1927 | #ifdef CONFIG_COMPAT | |
1928 | .compat_ioctl = vsock_dev_compat_ioctl, | |
1929 | #endif | |
1930 | .open = nonseekable_open, | |
1931 | }; | |
1932 | ||
1933 | static struct miscdevice vsock_device = { | |
1934 | .name = "vsock", | |
d021c344 AK |
1935 | .fops = &vsock_device_ops, |
1936 | }; | |
1937 | ||
2c4a336e | 1938 | int __vsock_core_init(const struct vsock_transport *t, struct module *owner) |
d021c344 | 1939 | { |
2c4a336e AK |
1940 | int err = mutex_lock_interruptible(&vsock_register_mutex); |
1941 | ||
1942 | if (err) | |
1943 | return err; | |
1944 | ||
1945 | if (transport) { | |
1946 | err = -EBUSY; | |
1947 | goto err_busy; | |
1948 | } | |
1949 | ||
1950 | /* Transport must be the owner of the protocol so that it can't | |
1951 | * unload while there are open sockets. | |
1952 | */ | |
1953 | vsock_proto.owner = owner; | |
1954 | transport = t; | |
d021c344 AK |
1955 | |
1956 | vsock_init_tables(); | |
1957 | ||
6ad0b2f7 | 1958 | vsock_device.minor = MISC_DYNAMIC_MINOR; |
d021c344 AK |
1959 | err = misc_register(&vsock_device); |
1960 | if (err) { | |
1961 | pr_err("Failed to register misc device\n"); | |
f6a835bb | 1962 | goto err_reset_transport; |
d021c344 AK |
1963 | } |
1964 | ||
1965 | err = proto_register(&vsock_proto, 1); /* we want our slab */ | |
1966 | if (err) { | |
1967 | pr_err("Cannot register vsock protocol\n"); | |
f6a835bb | 1968 | goto err_deregister_misc; |
d021c344 AK |
1969 | } |
1970 | ||
1971 | err = sock_register(&vsock_family_ops); | |
1972 | if (err) { | |
1973 | pr_err("could not register af_vsock (%d) address family: %d\n", | |
1974 | AF_VSOCK, err); | |
1975 | goto err_unregister_proto; | |
1976 | } | |
1977 | ||
2c4a336e | 1978 | mutex_unlock(&vsock_register_mutex); |
d021c344 AK |
1979 | return 0; |
1980 | ||
1981 | err_unregister_proto: | |
1982 | proto_unregister(&vsock_proto); | |
f6a835bb | 1983 | err_deregister_misc: |
d021c344 | 1984 | misc_deregister(&vsock_device); |
f6a835bb | 1985 | err_reset_transport: |
2c4a336e AK |
1986 | transport = NULL; |
1987 | err_busy: | |
d021c344 | 1988 | mutex_unlock(&vsock_register_mutex); |
2c4a336e | 1989 | return err; |
d021c344 | 1990 | } |
2c4a336e | 1991 | EXPORT_SYMBOL_GPL(__vsock_core_init); |
d021c344 AK |
1992 | |
1993 | void vsock_core_exit(void) | |
1994 | { | |
1995 | mutex_lock(&vsock_register_mutex); | |
1996 | ||
1997 | misc_deregister(&vsock_device); | |
1998 | sock_unregister(AF_VSOCK); | |
1999 | proto_unregister(&vsock_proto); | |
2000 | ||
2001 | /* We do not want the assignment below re-ordered. */ | |
2002 | mb(); | |
2003 | transport = NULL; | |
2004 | ||
2005 | mutex_unlock(&vsock_register_mutex); | |
2006 | } | |
2007 | EXPORT_SYMBOL_GPL(vsock_core_exit); | |
2008 | ||
2009 | MODULE_AUTHOR("VMware, Inc."); | |
2010 | MODULE_DESCRIPTION("VMware Virtual Socket Family"); | |
2c4a336e | 2011 | MODULE_VERSION("1.0.1.0-k"); |
d021c344 | 2012 | MODULE_LICENSE("GPL v2"); |