blob: df5fc6b340f1bbde621fbe84fa44e0af6e2e00af [file] [log] [blame]
Andy Kingd021c342013-02-06 14:23:56 +00001/*
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
39 * implementation because they are in the SS_LISTEN state. When a connection
40 * request is received (the second kind of socket mentioned above), we create a
41 * new socket and refer to it as a pending socket. These pending sockets are
42 * placed on the pending connection list of the listener socket. When future
43 * packets are received for the address the listener socket is bound to, we
44 * check if the source of the packet is from one that has an existing pending
45 * connection. If it does, we process the packet for the pending socket. When
46 * that socket reaches the connected state, it is removed from the listener
47 * socket's pending list and enqueued in the listener socket's accept queue.
48 * Callers of accept(2) will accept connected sockets from the listener socket's
49 * accept queue. If the socket cannot be accepted for some reason then it is
50 * marked rejected. Once the connection is accepted, it is owned by the user
51 * process and the responsibility for cleanup falls with that user process.
52 *
53 * - It is possible that these pending sockets will never reach the connected
54 * state; in fact, we may never receive another packet after the connection
55 * request. Because of this, we must schedule a cleanup function to run in the
56 * future, after some amount of time passes where a connection should have been
57 * established. This function ensures that the socket is off all lists so it
58 * cannot be retrieved, then drops all references to the socket so it is cleaned
59 * up (sock_put() -> sk_free() -> our sk_destruct implementation). Note this
60 * function will also cleanup rejected sockets, those that reach the connected
61 * state but leave it before they have been accepted.
62 *
63 * - Sockets created by user action will be cleaned up when the user process
64 * calls close(2), causing our release implementation to be called. Our release
65 * implementation will perform some cleanup then drop the last reference so our
66 * sk_destruct implementation is invoked. Our sk_destruct implementation will
67 * perform additional cleanup that's common for both types of sockets.
68 *
69 * - A socket's reference count is what ensures that the structure won't be
70 * freed. Each entry in a list (such as the "global" bound and connected tables
71 * and the listener socket's pending list and connected queue) ensures a
72 * reference. When we defer work until process context and pass a socket as our
73 * argument, we must ensure the reference count is increased to ensure the
74 * socket isn't freed before the function is run; the deferred function will
75 * then drop the reference.
76 */
77
78#include <linux/types.h>
Andy Kingd021c342013-02-06 14:23:56 +000079#include <linux/bitops.h>
80#include <linux/cred.h>
81#include <linux/init.h>
82#include <linux/io.h>
83#include <linux/kernel.h>
84#include <linux/kmod.h>
85#include <linux/list.h>
86#include <linux/miscdevice.h>
87#include <linux/module.h>
88#include <linux/mutex.h>
89#include <linux/net.h>
90#include <linux/poll.h>
91#include <linux/skbuff.h>
92#include <linux/smp.h>
93#include <linux/socket.h>
94#include <linux/stddef.h>
95#include <linux/unistd.h>
96#include <linux/wait.h>
97#include <linux/workqueue.h>
98#include <net/sock.h>
Asias He82a54d02013-07-25 17:39:34 +080099#include <net/af_vsock.h>
Andy Kingd021c342013-02-06 14:23:56 +0000100
101static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr);
102static void vsock_sk_destruct(struct sock *sk);
103static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
104
105/* Protocol family. */
106static struct proto vsock_proto = {
107 .name = "AF_VSOCK",
108 .owner = THIS_MODULE,
109 .obj_size = sizeof(struct vsock_sock),
110};
111
112/* The default peer timeout indicates how long we will wait for a peer response
113 * to a control message.
114 */
115#define VSOCK_DEFAULT_CONNECT_TIMEOUT (2 * HZ)
116
117#define SS_LISTEN 255
118
119static const struct vsock_transport *transport;
120static DEFINE_MUTEX(vsock_register_mutex);
121
122/**** EXPORTS ****/
123
124/* Get the ID of the local context. This is transport dependent. */
125
126int vm_sockets_get_local_cid(void)
127{
128 return transport->get_local_cid();
129}
130EXPORT_SYMBOL_GPL(vm_sockets_get_local_cid);
131
132/**** UTILS ****/
133
134/* Each bound VSocket is stored in the bind hash table and each connected
135 * VSocket is stored in the connected hash table.
136 *
137 * Unbound sockets are all put on the same list attached to the end of the hash
138 * table (vsock_unbound_sockets). Bound sockets are added to the hash table in
139 * the bucket that their local address hashes to (vsock_bound_sockets(addr)
140 * represents the list that addr hashes to).
141 *
142 * Specifically, we initialize the vsock_bind_table array to a size of
143 * VSOCK_HASH_SIZE + 1 so that vsock_bind_table[0] through
144 * vsock_bind_table[VSOCK_HASH_SIZE - 1] are for bound sockets and
145 * vsock_bind_table[VSOCK_HASH_SIZE] is for unbound sockets. The hash function
Asias Hea49dd9d2013-06-20 17:20:33 +0800146 * mods with VSOCK_HASH_SIZE to ensure this.
Andy Kingd021c342013-02-06 14:23:56 +0000147 */
148#define VSOCK_HASH_SIZE 251
149#define MAX_PORT_RETRIES 24
150
Asias Hea49dd9d2013-06-20 17:20:33 +0800151#define VSOCK_HASH(addr) ((addr)->svm_port % VSOCK_HASH_SIZE)
Andy Kingd021c342013-02-06 14:23:56 +0000152#define vsock_bound_sockets(addr) (&vsock_bind_table[VSOCK_HASH(addr)])
153#define vsock_unbound_sockets (&vsock_bind_table[VSOCK_HASH_SIZE])
154
155/* XXX This can probably be implemented in a better way. */
156#define VSOCK_CONN_HASH(src, dst) \
Asias Hea49dd9d2013-06-20 17:20:33 +0800157 (((src)->svm_cid ^ (dst)->svm_port) % VSOCK_HASH_SIZE)
Andy Kingd021c342013-02-06 14:23:56 +0000158#define vsock_connected_sockets(src, dst) \
159 (&vsock_connected_table[VSOCK_CONN_HASH(src, dst)])
160#define vsock_connected_sockets_vsk(vsk) \
161 vsock_connected_sockets(&(vsk)->remote_addr, &(vsk)->local_addr)
162
163static struct list_head vsock_bind_table[VSOCK_HASH_SIZE + 1];
164static struct list_head vsock_connected_table[VSOCK_HASH_SIZE];
165static DEFINE_SPINLOCK(vsock_table_lock);
166
Asias Heb3a6dfe2013-06-20 17:20:30 +0800167/* Autobind this socket to the local address if necessary. */
168static int vsock_auto_bind(struct vsock_sock *vsk)
169{
170 struct sock *sk = sk_vsock(vsk);
171 struct sockaddr_vm local_addr;
172
173 if (vsock_addr_bound(&vsk->local_addr))
174 return 0;
175 vsock_addr_init(&local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
176 return __vsock_bind(sk, &local_addr);
177}
178
Geert Uytterhoeven22ee3b52013-04-23 23:40:55 +0000179static void vsock_init_tables(void)
Andy Kingd021c342013-02-06 14:23:56 +0000180{
181 int i;
182
183 for (i = 0; i < ARRAY_SIZE(vsock_bind_table); i++)
184 INIT_LIST_HEAD(&vsock_bind_table[i]);
185
186 for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++)
187 INIT_LIST_HEAD(&vsock_connected_table[i]);
188}
189
190static void __vsock_insert_bound(struct list_head *list,
191 struct vsock_sock *vsk)
192{
193 sock_hold(&vsk->sk);
194 list_add(&vsk->bound_table, list);
195}
196
197static void __vsock_insert_connected(struct list_head *list,
198 struct vsock_sock *vsk)
199{
200 sock_hold(&vsk->sk);
201 list_add(&vsk->connected_table, list);
202}
203
204static void __vsock_remove_bound(struct vsock_sock *vsk)
205{
206 list_del_init(&vsk->bound_table);
207 sock_put(&vsk->sk);
208}
209
210static void __vsock_remove_connected(struct vsock_sock *vsk)
211{
212 list_del_init(&vsk->connected_table);
213 sock_put(&vsk->sk);
214}
215
216static struct sock *__vsock_find_bound_socket(struct sockaddr_vm *addr)
217{
218 struct vsock_sock *vsk;
219
220 list_for_each_entry(vsk, vsock_bound_sockets(addr), bound_table)
Reilly Grant990454b2013-04-01 11:41:52 -0700221 if (addr->svm_port == vsk->local_addr.svm_port)
Andy Kingd021c342013-02-06 14:23:56 +0000222 return sk_vsock(vsk);
223
224 return NULL;
225}
226
227static struct sock *__vsock_find_connected_socket(struct sockaddr_vm *src,
228 struct sockaddr_vm *dst)
229{
230 struct vsock_sock *vsk;
231
232 list_for_each_entry(vsk, vsock_connected_sockets(src, dst),
233 connected_table) {
Reilly Grant990454b2013-04-01 11:41:52 -0700234 if (vsock_addr_equals_addr(src, &vsk->remote_addr) &&
235 dst->svm_port == vsk->local_addr.svm_port) {
Andy Kingd021c342013-02-06 14:23:56 +0000236 return sk_vsock(vsk);
237 }
238 }
239
240 return NULL;
241}
242
243static bool __vsock_in_bound_table(struct vsock_sock *vsk)
244{
245 return !list_empty(&vsk->bound_table);
246}
247
248static bool __vsock_in_connected_table(struct vsock_sock *vsk)
249{
250 return !list_empty(&vsk->connected_table);
251}
252
253static void vsock_insert_unbound(struct vsock_sock *vsk)
254{
255 spin_lock_bh(&vsock_table_lock);
256 __vsock_insert_bound(vsock_unbound_sockets, vsk);
257 spin_unlock_bh(&vsock_table_lock);
258}
259
260void vsock_insert_connected(struct vsock_sock *vsk)
261{
262 struct list_head *list = vsock_connected_sockets(
263 &vsk->remote_addr, &vsk->local_addr);
264
265 spin_lock_bh(&vsock_table_lock);
266 __vsock_insert_connected(list, vsk);
267 spin_unlock_bh(&vsock_table_lock);
268}
269EXPORT_SYMBOL_GPL(vsock_insert_connected);
270
271void vsock_remove_bound(struct vsock_sock *vsk)
272{
273 spin_lock_bh(&vsock_table_lock);
274 __vsock_remove_bound(vsk);
275 spin_unlock_bh(&vsock_table_lock);
276}
277EXPORT_SYMBOL_GPL(vsock_remove_bound);
278
279void vsock_remove_connected(struct vsock_sock *vsk)
280{
281 spin_lock_bh(&vsock_table_lock);
282 __vsock_remove_connected(vsk);
283 spin_unlock_bh(&vsock_table_lock);
284}
285EXPORT_SYMBOL_GPL(vsock_remove_connected);
286
287struct sock *vsock_find_bound_socket(struct sockaddr_vm *addr)
288{
289 struct sock *sk;
290
291 spin_lock_bh(&vsock_table_lock);
292 sk = __vsock_find_bound_socket(addr);
293 if (sk)
294 sock_hold(sk);
295
296 spin_unlock_bh(&vsock_table_lock);
297
298 return sk;
299}
300EXPORT_SYMBOL_GPL(vsock_find_bound_socket);
301
302struct sock *vsock_find_connected_socket(struct sockaddr_vm *src,
303 struct sockaddr_vm *dst)
304{
305 struct sock *sk;
306
307 spin_lock_bh(&vsock_table_lock);
308 sk = __vsock_find_connected_socket(src, dst);
309 if (sk)
310 sock_hold(sk);
311
312 spin_unlock_bh(&vsock_table_lock);
313
314 return sk;
315}
316EXPORT_SYMBOL_GPL(vsock_find_connected_socket);
317
318static bool vsock_in_bound_table(struct vsock_sock *vsk)
319{
320 bool ret;
321
322 spin_lock_bh(&vsock_table_lock);
323 ret = __vsock_in_bound_table(vsk);
324 spin_unlock_bh(&vsock_table_lock);
325
326 return ret;
327}
328
329static bool vsock_in_connected_table(struct vsock_sock *vsk)
330{
331 bool ret;
332
333 spin_lock_bh(&vsock_table_lock);
334 ret = __vsock_in_connected_table(vsk);
335 spin_unlock_bh(&vsock_table_lock);
336
337 return ret;
338}
339
340void vsock_for_each_connected_socket(void (*fn)(struct sock *sk))
341{
342 int i;
343
344 spin_lock_bh(&vsock_table_lock);
345
346 for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) {
347 struct vsock_sock *vsk;
348 list_for_each_entry(vsk, &vsock_connected_table[i],
Julia Lawalld9af2d62013-08-05 16:47:38 +0200349 connected_table)
Andy Kingd021c342013-02-06 14:23:56 +0000350 fn(sk_vsock(vsk));
351 }
352
353 spin_unlock_bh(&vsock_table_lock);
354}
355EXPORT_SYMBOL_GPL(vsock_for_each_connected_socket);
356
357void vsock_add_pending(struct sock *listener, struct sock *pending)
358{
359 struct vsock_sock *vlistener;
360 struct vsock_sock *vpending;
361
362 vlistener = vsock_sk(listener);
363 vpending = vsock_sk(pending);
364
365 sock_hold(pending);
366 sock_hold(listener);
367 list_add_tail(&vpending->pending_links, &vlistener->pending_links);
368}
369EXPORT_SYMBOL_GPL(vsock_add_pending);
370
371void vsock_remove_pending(struct sock *listener, struct sock *pending)
372{
373 struct vsock_sock *vpending = vsock_sk(pending);
374
375 list_del_init(&vpending->pending_links);
376 sock_put(listener);
377 sock_put(pending);
378}
379EXPORT_SYMBOL_GPL(vsock_remove_pending);
380
381void vsock_enqueue_accept(struct sock *listener, struct sock *connected)
382{
383 struct vsock_sock *vlistener;
384 struct vsock_sock *vconnected;
385
386 vlistener = vsock_sk(listener);
387 vconnected = vsock_sk(connected);
388
389 sock_hold(connected);
390 sock_hold(listener);
391 list_add_tail(&vconnected->accept_queue, &vlistener->accept_queue);
392}
393EXPORT_SYMBOL_GPL(vsock_enqueue_accept);
394
395static struct sock *vsock_dequeue_accept(struct sock *listener)
396{
397 struct vsock_sock *vlistener;
398 struct vsock_sock *vconnected;
399
400 vlistener = vsock_sk(listener);
401
402 if (list_empty(&vlistener->accept_queue))
403 return NULL;
404
405 vconnected = list_entry(vlistener->accept_queue.next,
406 struct vsock_sock, accept_queue);
407
408 list_del_init(&vconnected->accept_queue);
409 sock_put(listener);
410 /* The caller will need a reference on the connected socket so we let
411 * it call sock_put().
412 */
413
414 return sk_vsock(vconnected);
415}
416
417static bool vsock_is_accept_queue_empty(struct sock *sk)
418{
419 struct vsock_sock *vsk = vsock_sk(sk);
420 return list_empty(&vsk->accept_queue);
421}
422
423static bool vsock_is_pending(struct sock *sk)
424{
425 struct vsock_sock *vsk = vsock_sk(sk);
426 return !list_empty(&vsk->pending_links);
427}
428
429static int vsock_send_shutdown(struct sock *sk, int mode)
430{
431 return transport->shutdown(vsock_sk(sk), mode);
432}
433
434void vsock_pending_work(struct work_struct *work)
435{
436 struct sock *sk;
437 struct sock *listener;
438 struct vsock_sock *vsk;
439 bool cleanup;
440
441 vsk = container_of(work, struct vsock_sock, dwork.work);
442 sk = sk_vsock(vsk);
443 listener = vsk->listener;
444 cleanup = true;
445
446 lock_sock(listener);
447 lock_sock(sk);
448
449 if (vsock_is_pending(sk)) {
450 vsock_remove_pending(listener, sk);
451 } else if (!vsk->rejected) {
452 /* We are not on the pending list and accept() did not reject
453 * us, so we must have been accepted by our user process. We
454 * just need to drop our references to the sockets and be on
455 * our way.
456 */
457 cleanup = false;
458 goto out;
459 }
460
461 listener->sk_ack_backlog--;
462
463 /* We need to remove ourself from the global connected sockets list so
464 * incoming packets can't find this socket, and to reduce the reference
465 * count.
466 */
467 if (vsock_in_connected_table(vsk))
468 vsock_remove_connected(vsk);
469
470 sk->sk_state = SS_FREE;
471
472out:
473 release_sock(sk);
474 release_sock(listener);
475 if (cleanup)
476 sock_put(sk);
477
478 sock_put(sk);
479 sock_put(listener);
480}
481EXPORT_SYMBOL_GPL(vsock_pending_work);
482
483/**** SOCKET OPERATIONS ****/
484
485static int __vsock_bind_stream(struct vsock_sock *vsk,
486 struct sockaddr_vm *addr)
487{
488 static u32 port = LAST_RESERVED_PORT + 1;
489 struct sockaddr_vm new_addr;
490
491 vsock_addr_init(&new_addr, addr->svm_cid, addr->svm_port);
492
493 if (addr->svm_port == VMADDR_PORT_ANY) {
494 bool found = false;
495 unsigned int i;
496
497 for (i = 0; i < MAX_PORT_RETRIES; i++) {
498 if (port <= LAST_RESERVED_PORT)
499 port = LAST_RESERVED_PORT + 1;
500
501 new_addr.svm_port = port++;
502
503 if (!__vsock_find_bound_socket(&new_addr)) {
504 found = true;
505 break;
506 }
507 }
508
509 if (!found)
510 return -EADDRNOTAVAIL;
511 } else {
512 /* If port is in reserved range, ensure caller
513 * has necessary privileges.
514 */
515 if (addr->svm_port <= LAST_RESERVED_PORT &&
516 !capable(CAP_NET_BIND_SERVICE)) {
517 return -EACCES;
518 }
519
520 if (__vsock_find_bound_socket(&new_addr))
521 return -EADDRINUSE;
522 }
523
524 vsock_addr_init(&vsk->local_addr, new_addr.svm_cid, new_addr.svm_port);
525
526 /* Remove stream sockets from the unbound list and add them to the hash
527 * table for easy lookup by its address. The unbound list is simply an
528 * extra entry at the end of the hash table, a trick used by AF_UNIX.
529 */
530 __vsock_remove_bound(vsk);
531 __vsock_insert_bound(vsock_bound_sockets(&vsk->local_addr), vsk);
532
533 return 0;
534}
535
536static int __vsock_bind_dgram(struct vsock_sock *vsk,
537 struct sockaddr_vm *addr)
538{
539 return transport->dgram_bind(vsk, addr);
540}
541
542static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr)
543{
544 struct vsock_sock *vsk = vsock_sk(sk);
545 u32 cid;
546 int retval;
547
548 /* First ensure this socket isn't already bound. */
549 if (vsock_addr_bound(&vsk->local_addr))
550 return -EINVAL;
551
552 /* Now bind to the provided address or select appropriate values if
553 * none are provided (VMADDR_CID_ANY and VMADDR_PORT_ANY). Note that
554 * like AF_INET prevents binding to a non-local IP address (in most
555 * cases), we only allow binding to the local CID.
556 */
557 cid = transport->get_local_cid();
558 if (addr->svm_cid != cid && addr->svm_cid != VMADDR_CID_ANY)
559 return -EADDRNOTAVAIL;
560
561 switch (sk->sk_socket->type) {
562 case SOCK_STREAM:
563 spin_lock_bh(&vsock_table_lock);
564 retval = __vsock_bind_stream(vsk, addr);
565 spin_unlock_bh(&vsock_table_lock);
566 break;
567
568 case SOCK_DGRAM:
569 retval = __vsock_bind_dgram(vsk, addr);
570 break;
571
572 default:
573 retval = -EINVAL;
574 break;
575 }
576
577 return retval;
578}
579
580struct sock *__vsock_create(struct net *net,
581 struct socket *sock,
582 struct sock *parent,
583 gfp_t priority,
Eric W. Biederman11aa9c22015-05-08 21:09:13 -0500584 unsigned short type,
585 int kern)
Andy Kingd021c342013-02-06 14:23:56 +0000586{
587 struct sock *sk;
588 struct vsock_sock *psk;
589 struct vsock_sock *vsk;
590
Eric W. Biederman11aa9c22015-05-08 21:09:13 -0500591 sk = sk_alloc(net, AF_VSOCK, priority, &vsock_proto, kern);
Andy Kingd021c342013-02-06 14:23:56 +0000592 if (!sk)
593 return NULL;
594
595 sock_init_data(sock, sk);
596
597 /* sk->sk_type is normally set in sock_init_data, but only if sock is
598 * non-NULL. We make sure that our sockets always have a type by
599 * setting it here if needed.
600 */
601 if (!sock)
602 sk->sk_type = type;
603
604 vsk = vsock_sk(sk);
605 vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
606 vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
607
608 sk->sk_destruct = vsock_sk_destruct;
609 sk->sk_backlog_rcv = vsock_queue_rcv_skb;
610 sk->sk_state = 0;
611 sock_reset_flag(sk, SOCK_DONE);
612
613 INIT_LIST_HEAD(&vsk->bound_table);
614 INIT_LIST_HEAD(&vsk->connected_table);
615 vsk->listener = NULL;
616 INIT_LIST_HEAD(&vsk->pending_links);
617 INIT_LIST_HEAD(&vsk->accept_queue);
618 vsk->rejected = false;
619 vsk->sent_request = false;
620 vsk->ignore_connecting_rst = false;
621 vsk->peer_shutdown = 0;
622
623 psk = parent ? vsock_sk(parent) : NULL;
624 if (parent) {
625 vsk->trusted = psk->trusted;
626 vsk->owner = get_cred(psk->owner);
627 vsk->connect_timeout = psk->connect_timeout;
628 } else {
629 vsk->trusted = capable(CAP_NET_ADMIN);
630 vsk->owner = get_current_cred();
631 vsk->connect_timeout = VSOCK_DEFAULT_CONNECT_TIMEOUT;
632 }
633
634 if (transport->init(vsk, psk) < 0) {
635 sk_free(sk);
636 return NULL;
637 }
638
639 if (sock)
640 vsock_insert_unbound(vsk);
641
642 return sk;
643}
644EXPORT_SYMBOL_GPL(__vsock_create);
645
646static void __vsock_release(struct sock *sk)
647{
648 if (sk) {
649 struct sk_buff *skb;
650 struct sock *pending;
651 struct vsock_sock *vsk;
652
653 vsk = vsock_sk(sk);
654 pending = NULL; /* Compiler warning. */
655
656 if (vsock_in_bound_table(vsk))
657 vsock_remove_bound(vsk);
658
659 if (vsock_in_connected_table(vsk))
660 vsock_remove_connected(vsk);
661
662 transport->release(vsk);
663
664 lock_sock(sk);
665 sock_orphan(sk);
666 sk->sk_shutdown = SHUTDOWN_MASK;
667
668 while ((skb = skb_dequeue(&sk->sk_receive_queue)))
669 kfree_skb(skb);
670
671 /* Clean up any sockets that never were accepted. */
672 while ((pending = vsock_dequeue_accept(sk)) != NULL) {
673 __vsock_release(pending);
674 sock_put(pending);
675 }
676
677 release_sock(sk);
678 sock_put(sk);
679 }
680}
681
682static void vsock_sk_destruct(struct sock *sk)
683{
684 struct vsock_sock *vsk = vsock_sk(sk);
685
686 transport->destruct(vsk);
687
688 /* When clearing these addresses, there's no need to set the family and
689 * possibly register the address family with the kernel.
690 */
691 vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
692 vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
693
694 put_cred(vsk->owner);
695}
696
697static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
698{
699 int err;
700
701 err = sock_queue_rcv_skb(sk, skb);
702 if (err)
703 kfree_skb(skb);
704
705 return err;
706}
707
708s64 vsock_stream_has_data(struct vsock_sock *vsk)
709{
710 return transport->stream_has_data(vsk);
711}
712EXPORT_SYMBOL_GPL(vsock_stream_has_data);
713
714s64 vsock_stream_has_space(struct vsock_sock *vsk)
715{
716 return transport->stream_has_space(vsk);
717}
718EXPORT_SYMBOL_GPL(vsock_stream_has_space);
719
720static int vsock_release(struct socket *sock)
721{
722 __vsock_release(sock->sk);
723 sock->sk = NULL;
724 sock->state = SS_FREE;
725
726 return 0;
727}
728
729static int
730vsock_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
731{
732 int err;
733 struct sock *sk;
734 struct sockaddr_vm *vm_addr;
735
736 sk = sock->sk;
737
738 if (vsock_addr_cast(addr, addr_len, &vm_addr) != 0)
739 return -EINVAL;
740
741 lock_sock(sk);
742 err = __vsock_bind(sk, vm_addr);
743 release_sock(sk);
744
745 return err;
746}
747
748static int vsock_getname(struct socket *sock,
749 struct sockaddr *addr, int *addr_len, int peer)
750{
751 int err;
752 struct sock *sk;
753 struct vsock_sock *vsk;
754 struct sockaddr_vm *vm_addr;
755
756 sk = sock->sk;
757 vsk = vsock_sk(sk);
758 err = 0;
759
760 lock_sock(sk);
761
762 if (peer) {
763 if (sock->state != SS_CONNECTED) {
764 err = -ENOTCONN;
765 goto out;
766 }
767 vm_addr = &vsk->remote_addr;
768 } else {
769 vm_addr = &vsk->local_addr;
770 }
771
772 if (!vm_addr) {
773 err = -EINVAL;
774 goto out;
775 }
776
777 /* sys_getsockname() and sys_getpeername() pass us a
778 * MAX_SOCK_ADDR-sized buffer and don't set addr_len. Unfortunately
779 * that macro is defined in socket.c instead of .h, so we hardcode its
780 * value here.
781 */
782 BUILD_BUG_ON(sizeof(*vm_addr) > 128);
783 memcpy(addr, vm_addr, sizeof(*vm_addr));
784 *addr_len = sizeof(*vm_addr);
785
786out:
787 release_sock(sk);
788 return err;
789}
790
791static int vsock_shutdown(struct socket *sock, int mode)
792{
793 int err;
794 struct sock *sk;
795
796 /* User level uses SHUT_RD (0) and SHUT_WR (1), but the kernel uses
797 * RCV_SHUTDOWN (1) and SEND_SHUTDOWN (2), so we must increment mode
798 * here like the other address families do. Note also that the
799 * increment makes SHUT_RDWR (2) into RCV_SHUTDOWN | SEND_SHUTDOWN (3),
800 * which is what we want.
801 */
802 mode++;
803
804 if ((mode & ~SHUTDOWN_MASK) || !mode)
805 return -EINVAL;
806
807 /* If this is a STREAM socket and it is not connected then bail out
808 * immediately. If it is a DGRAM socket then we must first kick the
809 * socket so that it wakes up from any sleeping calls, for example
810 * recv(), and then afterwards return the error.
811 */
812
813 sk = sock->sk;
814 if (sock->state == SS_UNCONNECTED) {
815 err = -ENOTCONN;
816 if (sk->sk_type == SOCK_STREAM)
817 return err;
818 } else {
819 sock->state = SS_DISCONNECTING;
820 err = 0;
821 }
822
823 /* Receive and send shutdowns are treated alike. */
824 mode = mode & (RCV_SHUTDOWN | SEND_SHUTDOWN);
825 if (mode) {
826 lock_sock(sk);
827 sk->sk_shutdown |= mode;
828 sk->sk_state_change(sk);
829 release_sock(sk);
830
831 if (sk->sk_type == SOCK_STREAM) {
832 sock_reset_flag(sk, SOCK_DONE);
833 vsock_send_shutdown(sk, mode);
834 }
835 }
836
837 return err;
838}
839
840static unsigned int vsock_poll(struct file *file, struct socket *sock,
841 poll_table *wait)
842{
843 struct sock *sk;
844 unsigned int mask;
845 struct vsock_sock *vsk;
846
847 sk = sock->sk;
848 vsk = vsock_sk(sk);
849
850 poll_wait(file, sk_sleep(sk), wait);
851 mask = 0;
852
853 if (sk->sk_err)
854 /* Signify that there has been an error on this socket. */
855 mask |= POLLERR;
856
857 /* INET sockets treat local write shutdown and peer write shutdown as a
858 * case of POLLHUP set.
859 */
860 if ((sk->sk_shutdown == SHUTDOWN_MASK) ||
861 ((sk->sk_shutdown & SEND_SHUTDOWN) &&
862 (vsk->peer_shutdown & SEND_SHUTDOWN))) {
863 mask |= POLLHUP;
864 }
865
866 if (sk->sk_shutdown & RCV_SHUTDOWN ||
867 vsk->peer_shutdown & SEND_SHUTDOWN) {
868 mask |= POLLRDHUP;
869 }
870
871 if (sock->type == SOCK_DGRAM) {
872 /* For datagram sockets we can read if there is something in
873 * the queue and write as long as the socket isn't shutdown for
874 * sending.
875 */
876 if (!skb_queue_empty(&sk->sk_receive_queue) ||
877 (sk->sk_shutdown & RCV_SHUTDOWN)) {
878 mask |= POLLIN | POLLRDNORM;
879 }
880
881 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
882 mask |= POLLOUT | POLLWRNORM | POLLWRBAND;
883
884 } else if (sock->type == SOCK_STREAM) {
885 lock_sock(sk);
886
887 /* Listening sockets that have connections in their accept
888 * queue can be read.
889 */
890 if (sk->sk_state == SS_LISTEN
891 && !vsock_is_accept_queue_empty(sk))
892 mask |= POLLIN | POLLRDNORM;
893
894 /* If there is something in the queue then we can read. */
895 if (transport->stream_is_active(vsk) &&
896 !(sk->sk_shutdown & RCV_SHUTDOWN)) {
897 bool data_ready_now = false;
898 int ret = transport->notify_poll_in(
899 vsk, 1, &data_ready_now);
900 if (ret < 0) {
901 mask |= POLLERR;
902 } else {
903 if (data_ready_now)
904 mask |= POLLIN | POLLRDNORM;
905
906 }
907 }
908
909 /* Sockets whose connections have been closed, reset, or
910 * terminated should also be considered read, and we check the
911 * shutdown flag for that.
912 */
913 if (sk->sk_shutdown & RCV_SHUTDOWN ||
914 vsk->peer_shutdown & SEND_SHUTDOWN) {
915 mask |= POLLIN | POLLRDNORM;
916 }
917
918 /* Connected sockets that can produce data can be written. */
919 if (sk->sk_state == SS_CONNECTED) {
920 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
921 bool space_avail_now = false;
922 int ret = transport->notify_poll_out(
923 vsk, 1, &space_avail_now);
924 if (ret < 0) {
925 mask |= POLLERR;
926 } else {
927 if (space_avail_now)
928 /* Remove POLLWRBAND since INET
929 * sockets are not setting it.
930 */
931 mask |= POLLOUT | POLLWRNORM;
932
933 }
934 }
935 }
936
937 /* Simulate INET socket poll behaviors, which sets
938 * POLLOUT|POLLWRNORM when peer is closed and nothing to read,
939 * but local send is not shutdown.
940 */
941 if (sk->sk_state == SS_UNCONNECTED) {
942 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
943 mask |= POLLOUT | POLLWRNORM;
944
945 }
946
947 release_sock(sk);
948 }
949
950 return mask;
951}
952
Ying Xue1b784142015-03-02 15:37:48 +0800953static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg,
954 size_t len)
Andy Kingd021c342013-02-06 14:23:56 +0000955{
956 int err;
957 struct sock *sk;
958 struct vsock_sock *vsk;
959 struct sockaddr_vm *remote_addr;
960
961 if (msg->msg_flags & MSG_OOB)
962 return -EOPNOTSUPP;
963
964 /* For now, MSG_DONTWAIT is always assumed... */
965 err = 0;
966 sk = sock->sk;
967 vsk = vsock_sk(sk);
968
969 lock_sock(sk);
970
Asias Heb3a6dfe2013-06-20 17:20:30 +0800971 err = vsock_auto_bind(vsk);
972 if (err)
973 goto out;
Andy Kingd021c342013-02-06 14:23:56 +0000974
Andy Kingd021c342013-02-06 14:23:56 +0000975
976 /* If the provided message contains an address, use that. Otherwise
977 * fall back on the socket's remote handle (if it has been connected).
978 */
979 if (msg->msg_name &&
980 vsock_addr_cast(msg->msg_name, msg->msg_namelen,
981 &remote_addr) == 0) {
982 /* Ensure this address is of the right type and is a valid
983 * destination.
984 */
985
986 if (remote_addr->svm_cid == VMADDR_CID_ANY)
987 remote_addr->svm_cid = transport->get_local_cid();
988
989 if (!vsock_addr_bound(remote_addr)) {
990 err = -EINVAL;
991 goto out;
992 }
993 } else if (sock->state == SS_CONNECTED) {
994 remote_addr = &vsk->remote_addr;
995
996 if (remote_addr->svm_cid == VMADDR_CID_ANY)
997 remote_addr->svm_cid = transport->get_local_cid();
998
999 /* XXX Should connect() or this function ensure remote_addr is
1000 * bound?
1001 */
1002 if (!vsock_addr_bound(&vsk->remote_addr)) {
1003 err = -EINVAL;
1004 goto out;
1005 }
1006 } else {
1007 err = -EINVAL;
1008 goto out;
1009 }
1010
1011 if (!transport->dgram_allow(remote_addr->svm_cid,
1012 remote_addr->svm_port)) {
1013 err = -EINVAL;
1014 goto out;
1015 }
1016
Al Viro0f7db232014-11-20 04:05:34 -05001017 err = transport->dgram_enqueue(vsk, remote_addr, msg, len);
Andy Kingd021c342013-02-06 14:23:56 +00001018
1019out:
1020 release_sock(sk);
1021 return err;
1022}
1023
1024static int vsock_dgram_connect(struct socket *sock,
1025 struct sockaddr *addr, int addr_len, int flags)
1026{
1027 int err;
1028 struct sock *sk;
1029 struct vsock_sock *vsk;
1030 struct sockaddr_vm *remote_addr;
1031
1032 sk = sock->sk;
1033 vsk = vsock_sk(sk);
1034
1035 err = vsock_addr_cast(addr, addr_len, &remote_addr);
1036 if (err == -EAFNOSUPPORT && remote_addr->svm_family == AF_UNSPEC) {
1037 lock_sock(sk);
1038 vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY,
1039 VMADDR_PORT_ANY);
1040 sock->state = SS_UNCONNECTED;
1041 release_sock(sk);
1042 return 0;
1043 } else if (err != 0)
1044 return -EINVAL;
1045
1046 lock_sock(sk);
1047
Asias Heb3a6dfe2013-06-20 17:20:30 +08001048 err = vsock_auto_bind(vsk);
1049 if (err)
1050 goto out;
Andy Kingd021c342013-02-06 14:23:56 +00001051
1052 if (!transport->dgram_allow(remote_addr->svm_cid,
1053 remote_addr->svm_port)) {
1054 err = -EINVAL;
1055 goto out;
1056 }
1057
1058 memcpy(&vsk->remote_addr, remote_addr, sizeof(vsk->remote_addr));
1059 sock->state = SS_CONNECTED;
1060
1061out:
1062 release_sock(sk);
1063 return err;
1064}
1065
Ying Xue1b784142015-03-02 15:37:48 +08001066static int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg,
1067 size_t len, int flags)
Andy Kingd021c342013-02-06 14:23:56 +00001068{
Ying Xue1b784142015-03-02 15:37:48 +08001069 return transport->dgram_dequeue(vsock_sk(sock->sk), msg, len, flags);
Andy Kingd021c342013-02-06 14:23:56 +00001070}
1071
1072static const struct proto_ops vsock_dgram_ops = {
1073 .family = PF_VSOCK,
1074 .owner = THIS_MODULE,
1075 .release = vsock_release,
1076 .bind = vsock_bind,
1077 .connect = vsock_dgram_connect,
1078 .socketpair = sock_no_socketpair,
1079 .accept = sock_no_accept,
1080 .getname = vsock_getname,
1081 .poll = vsock_poll,
1082 .ioctl = sock_no_ioctl,
1083 .listen = sock_no_listen,
1084 .shutdown = vsock_shutdown,
1085 .setsockopt = sock_no_setsockopt,
1086 .getsockopt = sock_no_getsockopt,
1087 .sendmsg = vsock_dgram_sendmsg,
1088 .recvmsg = vsock_dgram_recvmsg,
1089 .mmap = sock_no_mmap,
1090 .sendpage = sock_no_sendpage,
1091};
1092
1093static void vsock_connect_timeout(struct work_struct *work)
1094{
1095 struct sock *sk;
1096 struct vsock_sock *vsk;
1097
1098 vsk = container_of(work, struct vsock_sock, dwork.work);
1099 sk = sk_vsock(vsk);
1100
1101 lock_sock(sk);
1102 if (sk->sk_state == SS_CONNECTING &&
1103 (sk->sk_shutdown != SHUTDOWN_MASK)) {
1104 sk->sk_state = SS_UNCONNECTED;
1105 sk->sk_err = ETIMEDOUT;
1106 sk->sk_error_report(sk);
1107 }
1108 release_sock(sk);
1109
1110 sock_put(sk);
1111}
1112
1113static int vsock_stream_connect(struct socket *sock, struct sockaddr *addr,
1114 int addr_len, int flags)
1115{
1116 int err;
1117 struct sock *sk;
1118 struct vsock_sock *vsk;
1119 struct sockaddr_vm *remote_addr;
1120 long timeout;
1121 DEFINE_WAIT(wait);
1122
1123 err = 0;
1124 sk = sock->sk;
1125 vsk = vsock_sk(sk);
1126
1127 lock_sock(sk);
1128
1129 /* XXX AF_UNSPEC should make us disconnect like AF_INET. */
1130 switch (sock->state) {
1131 case SS_CONNECTED:
1132 err = -EISCONN;
1133 goto out;
1134 case SS_DISCONNECTING:
1135 err = -EINVAL;
1136 goto out;
1137 case SS_CONNECTING:
1138 /* This continues on so we can move sock into the SS_CONNECTED
1139 * state once the connection has completed (at which point err
1140 * will be set to zero also). Otherwise, we will either wait
1141 * for the connection or return -EALREADY should this be a
1142 * non-blocking call.
1143 */
1144 err = -EALREADY;
1145 break;
1146 default:
1147 if ((sk->sk_state == SS_LISTEN) ||
1148 vsock_addr_cast(addr, addr_len, &remote_addr) != 0) {
1149 err = -EINVAL;
1150 goto out;
1151 }
1152
1153 /* The hypervisor and well-known contexts do not have socket
1154 * endpoints.
1155 */
1156 if (!transport->stream_allow(remote_addr->svm_cid,
1157 remote_addr->svm_port)) {
1158 err = -ENETUNREACH;
1159 goto out;
1160 }
1161
1162 /* Set the remote address that we are connecting to. */
1163 memcpy(&vsk->remote_addr, remote_addr,
1164 sizeof(vsk->remote_addr));
1165
Asias Heb3a6dfe2013-06-20 17:20:30 +08001166 err = vsock_auto_bind(vsk);
1167 if (err)
1168 goto out;
Andy Kingd021c342013-02-06 14:23:56 +00001169
1170 sk->sk_state = SS_CONNECTING;
1171
1172 err = transport->connect(vsk);
1173 if (err < 0)
1174 goto out;
1175
1176 /* Mark sock as connecting and set the error code to in
1177 * progress in case this is a non-blocking connect.
1178 */
1179 sock->state = SS_CONNECTING;
1180 err = -EINPROGRESS;
1181 }
1182
1183 /* The receive path will handle all communication until we are able to
1184 * enter the connected state. Here we wait for the connection to be
1185 * completed or a notification of an error.
1186 */
1187 timeout = vsk->connect_timeout;
1188 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1189
1190 while (sk->sk_state != SS_CONNECTED && sk->sk_err == 0) {
1191 if (flags & O_NONBLOCK) {
1192 /* If we're not going to block, we schedule a timeout
1193 * function to generate a timeout on the connection
1194 * attempt, in case the peer doesn't respond in a
1195 * timely manner. We hold on to the socket until the
1196 * timeout fires.
1197 */
1198 sock_hold(sk);
1199 INIT_DELAYED_WORK(&vsk->dwork,
1200 vsock_connect_timeout);
1201 schedule_delayed_work(&vsk->dwork, timeout);
1202
1203 /* Skip ahead to preserve error code set above. */
1204 goto out_wait;
1205 }
1206
1207 release_sock(sk);
1208 timeout = schedule_timeout(timeout);
1209 lock_sock(sk);
1210
1211 if (signal_pending(current)) {
1212 err = sock_intr_errno(timeout);
1213 goto out_wait_error;
1214 } else if (timeout == 0) {
1215 err = -ETIMEDOUT;
1216 goto out_wait_error;
1217 }
1218
1219 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1220 }
1221
1222 if (sk->sk_err) {
1223 err = -sk->sk_err;
1224 goto out_wait_error;
1225 } else
1226 err = 0;
1227
1228out_wait:
1229 finish_wait(sk_sleep(sk), &wait);
1230out:
1231 release_sock(sk);
1232 return err;
1233
1234out_wait_error:
1235 sk->sk_state = SS_UNCONNECTED;
1236 sock->state = SS_UNCONNECTED;
1237 goto out_wait;
1238}
1239
1240static 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
1259 if (listener->sk_state != SS_LISTEN) {
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);
1274 lock_sock(listener);
1275
1276 if (signal_pending(current)) {
1277 err = sock_intr_errno(timeout);
1278 goto out_wait;
1279 } else if (timeout == 0) {
1280 err = -EAGAIN;
1281 goto out_wait;
1282 }
1283
1284 prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1285 }
1286
1287 if (listener->sk_err)
1288 err = -listener->sk_err;
1289
1290 if (connected) {
1291 listener->sk_ack_backlog--;
1292
1293 lock_sock(connected);
1294 vconnected = vsock_sk(connected);
1295
1296 /* If the listener socket has received an error, then we should
1297 * reject this socket and return. Note that we simply mark the
1298 * socket rejected, drop our reference, and let the cleanup
1299 * function handle the cleanup; the fact that we found it in
1300 * the listener's accept queue guarantees that the cleanup
1301 * function hasn't run yet.
1302 */
1303 if (err) {
1304 vconnected->rejected = true;
1305 release_sock(connected);
1306 sock_put(connected);
1307 goto out_wait;
1308 }
1309
1310 newsock->state = SS_CONNECTED;
1311 sock_graft(connected, newsock);
1312 release_sock(connected);
1313 sock_put(connected);
1314 }
1315
1316out_wait:
1317 finish_wait(sk_sleep(listener), &wait);
1318out:
1319 release_sock(listener);
1320 return err;
1321}
1322
1323static int vsock_listen(struct socket *sock, int backlog)
1324{
1325 int err;
1326 struct sock *sk;
1327 struct vsock_sock *vsk;
1328
1329 sk = sock->sk;
1330
1331 lock_sock(sk);
1332
1333 if (sock->type != SOCK_STREAM) {
1334 err = -EOPNOTSUPP;
1335 goto out;
1336 }
1337
1338 if (sock->state != SS_UNCONNECTED) {
1339 err = -EINVAL;
1340 goto out;
1341 }
1342
1343 vsk = vsock_sk(sk);
1344
1345 if (!vsock_addr_bound(&vsk->local_addr)) {
1346 err = -EINVAL;
1347 goto out;
1348 }
1349
1350 sk->sk_max_ack_backlog = backlog;
1351 sk->sk_state = SS_LISTEN;
1352
1353 err = 0;
1354
1355out:
1356 release_sock(sk);
1357 return err;
1358}
1359
1360static int vsock_stream_setsockopt(struct socket *sock,
1361 int level,
1362 int optname,
1363 char __user *optval,
1364 unsigned int optlen)
1365{
1366 int err;
1367 struct sock *sk;
1368 struct vsock_sock *vsk;
1369 u64 val;
1370
1371 if (level != AF_VSOCK)
1372 return -ENOPROTOOPT;
1373
1374#define COPY_IN(_v) \
1375 do { \
1376 if (optlen < sizeof(_v)) { \
1377 err = -EINVAL; \
1378 goto exit; \
1379 } \
1380 if (copy_from_user(&_v, optval, sizeof(_v)) != 0) { \
1381 err = -EFAULT; \
1382 goto exit; \
1383 } \
1384 } while (0)
1385
1386 err = 0;
1387 sk = sock->sk;
1388 vsk = vsock_sk(sk);
1389
1390 lock_sock(sk);
1391
1392 switch (optname) {
1393 case SO_VM_SOCKETS_BUFFER_SIZE:
1394 COPY_IN(val);
1395 transport->set_buffer_size(vsk, val);
1396 break;
1397
1398 case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1399 COPY_IN(val);
1400 transport->set_max_buffer_size(vsk, val);
1401 break;
1402
1403 case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1404 COPY_IN(val);
1405 transport->set_min_buffer_size(vsk, val);
1406 break;
1407
1408 case SO_VM_SOCKETS_CONNECT_TIMEOUT: {
1409 struct timeval tv;
1410 COPY_IN(tv);
1411 if (tv.tv_sec >= 0 && tv.tv_usec < USEC_PER_SEC &&
1412 tv.tv_sec < (MAX_SCHEDULE_TIMEOUT / HZ - 1)) {
1413 vsk->connect_timeout = tv.tv_sec * HZ +
1414 DIV_ROUND_UP(tv.tv_usec, (1000000 / HZ));
1415 if (vsk->connect_timeout == 0)
1416 vsk->connect_timeout =
1417 VSOCK_DEFAULT_CONNECT_TIMEOUT;
1418
1419 } else {
1420 err = -ERANGE;
1421 }
1422 break;
1423 }
1424
1425 default:
1426 err = -ENOPROTOOPT;
1427 break;
1428 }
1429
1430#undef COPY_IN
1431
1432exit:
1433 release_sock(sk);
1434 return err;
1435}
1436
1437static int vsock_stream_getsockopt(struct socket *sock,
1438 int level, int optname,
1439 char __user *optval,
1440 int __user *optlen)
1441{
1442 int err;
1443 int len;
1444 struct sock *sk;
1445 struct vsock_sock *vsk;
1446 u64 val;
1447
1448 if (level != AF_VSOCK)
1449 return -ENOPROTOOPT;
1450
1451 err = get_user(len, optlen);
1452 if (err != 0)
1453 return err;
1454
1455#define COPY_OUT(_v) \
1456 do { \
1457 if (len < sizeof(_v)) \
1458 return -EINVAL; \
1459 \
1460 len = sizeof(_v); \
1461 if (copy_to_user(optval, &_v, len) != 0) \
1462 return -EFAULT; \
1463 \
1464 } while (0)
1465
1466 err = 0;
1467 sk = sock->sk;
1468 vsk = vsock_sk(sk);
1469
1470 switch (optname) {
1471 case SO_VM_SOCKETS_BUFFER_SIZE:
1472 val = transport->get_buffer_size(vsk);
1473 COPY_OUT(val);
1474 break;
1475
1476 case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1477 val = transport->get_max_buffer_size(vsk);
1478 COPY_OUT(val);
1479 break;
1480
1481 case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1482 val = transport->get_min_buffer_size(vsk);
1483 COPY_OUT(val);
1484 break;
1485
1486 case SO_VM_SOCKETS_CONNECT_TIMEOUT: {
1487 struct timeval tv;
1488 tv.tv_sec = vsk->connect_timeout / HZ;
1489 tv.tv_usec =
1490 (vsk->connect_timeout -
1491 tv.tv_sec * HZ) * (1000000 / HZ);
1492 COPY_OUT(tv);
1493 break;
1494 }
1495 default:
1496 return -ENOPROTOOPT;
1497 }
1498
1499 err = put_user(len, optlen);
1500 if (err != 0)
1501 return -EFAULT;
1502
1503#undef COPY_OUT
1504
1505 return 0;
1506}
1507
Ying Xue1b784142015-03-02 15:37:48 +08001508static int vsock_stream_sendmsg(struct socket *sock, struct msghdr *msg,
1509 size_t len)
Andy Kingd021c342013-02-06 14:23:56 +00001510{
1511 struct sock *sk;
1512 struct vsock_sock *vsk;
1513 ssize_t total_written;
1514 long timeout;
1515 int err;
1516 struct vsock_transport_send_notify_data send_data;
1517
1518 DEFINE_WAIT(wait);
1519
1520 sk = sock->sk;
1521 vsk = vsock_sk(sk);
1522 total_written = 0;
1523 err = 0;
1524
1525 if (msg->msg_flags & MSG_OOB)
1526 return -EOPNOTSUPP;
1527
1528 lock_sock(sk);
1529
1530 /* Callers should not provide a destination with stream sockets. */
1531 if (msg->msg_namelen) {
1532 err = sk->sk_state == SS_CONNECTED ? -EISCONN : -EOPNOTSUPP;
1533 goto out;
1534 }
1535
1536 /* Send data only if both sides are not shutdown in the direction. */
1537 if (sk->sk_shutdown & SEND_SHUTDOWN ||
1538 vsk->peer_shutdown & RCV_SHUTDOWN) {
1539 err = -EPIPE;
1540 goto out;
1541 }
1542
1543 if (sk->sk_state != SS_CONNECTED ||
1544 !vsock_addr_bound(&vsk->local_addr)) {
1545 err = -ENOTCONN;
1546 goto out;
1547 }
1548
1549 if (!vsock_addr_bound(&vsk->remote_addr)) {
1550 err = -EDESTADDRREQ;
1551 goto out;
1552 }
1553
1554 /* Wait for room in the produce queue to enqueue our user's data. */
1555 timeout = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1556
1557 err = transport->notify_send_init(vsk, &send_data);
1558 if (err < 0)
1559 goto out;
1560
1561 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1562
1563 while (total_written < len) {
1564 ssize_t written;
1565
1566 while (vsock_stream_has_space(vsk) == 0 &&
1567 sk->sk_err == 0 &&
1568 !(sk->sk_shutdown & SEND_SHUTDOWN) &&
1569 !(vsk->peer_shutdown & RCV_SHUTDOWN)) {
1570
1571 /* Don't wait for non-blocking sockets. */
1572 if (timeout == 0) {
1573 err = -EAGAIN;
1574 goto out_wait;
1575 }
1576
1577 err = transport->notify_send_pre_block(vsk, &send_data);
1578 if (err < 0)
1579 goto out_wait;
1580
1581 release_sock(sk);
1582 timeout = schedule_timeout(timeout);
1583 lock_sock(sk);
1584 if (signal_pending(current)) {
1585 err = sock_intr_errno(timeout);
1586 goto out_wait;
1587 } else if (timeout == 0) {
1588 err = -EAGAIN;
1589 goto out_wait;
1590 }
1591
1592 prepare_to_wait(sk_sleep(sk), &wait,
1593 TASK_INTERRUPTIBLE);
1594 }
1595
1596 /* These checks occur both as part of and after the loop
1597 * conditional since we need to check before and after
1598 * sleeping.
1599 */
1600 if (sk->sk_err) {
1601 err = -sk->sk_err;
1602 goto out_wait;
1603 } else if ((sk->sk_shutdown & SEND_SHUTDOWN) ||
1604 (vsk->peer_shutdown & RCV_SHUTDOWN)) {
1605 err = -EPIPE;
1606 goto out_wait;
1607 }
1608
1609 err = transport->notify_send_pre_enqueue(vsk, &send_data);
1610 if (err < 0)
1611 goto out_wait;
1612
1613 /* Note that enqueue will only write as many bytes as are free
1614 * in the produce queue, so we don't need to ensure len is
1615 * smaller than the queue size. It is the caller's
1616 * responsibility to check how many bytes we were able to send.
1617 */
1618
1619 written = transport->stream_enqueue(
Al Viro0f7db232014-11-20 04:05:34 -05001620 vsk, msg,
Andy Kingd021c342013-02-06 14:23:56 +00001621 len - total_written);
1622 if (written < 0) {
1623 err = -ENOMEM;
1624 goto out_wait;
1625 }
1626
1627 total_written += written;
1628
1629 err = transport->notify_send_post_enqueue(
1630 vsk, written, &send_data);
1631 if (err < 0)
1632 goto out_wait;
1633
1634 }
1635
1636out_wait:
1637 if (total_written > 0)
1638 err = total_written;
1639 finish_wait(sk_sleep(sk), &wait);
1640out:
1641 release_sock(sk);
1642 return err;
1643}
1644
1645
1646static int
Ying Xue1b784142015-03-02 15:37:48 +08001647vsock_stream_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
1648 int flags)
Andy Kingd021c342013-02-06 14:23:56 +00001649{
1650 struct sock *sk;
1651 struct vsock_sock *vsk;
1652 int err;
1653 size_t target;
1654 ssize_t copied;
1655 long timeout;
1656 struct vsock_transport_recv_notify_data recv_data;
1657
1658 DEFINE_WAIT(wait);
1659
1660 sk = sock->sk;
1661 vsk = vsock_sk(sk);
1662 err = 0;
1663
1664 lock_sock(sk);
1665
1666 if (sk->sk_state != SS_CONNECTED) {
1667 /* Recvmsg is supposed to return 0 if a peer performs an
1668 * orderly shutdown. Differentiate between that case and when a
1669 * peer has not connected or a local shutdown occured with the
1670 * SOCK_DONE flag.
1671 */
1672 if (sock_flag(sk, SOCK_DONE))
1673 err = 0;
1674 else
1675 err = -ENOTCONN;
1676
1677 goto out;
1678 }
1679
1680 if (flags & MSG_OOB) {
1681 err = -EOPNOTSUPP;
1682 goto out;
1683 }
1684
1685 /* We don't check peer_shutdown flag here since peer may actually shut
1686 * down, but there can be data in the queue that a local socket can
1687 * receive.
1688 */
1689 if (sk->sk_shutdown & RCV_SHUTDOWN) {
1690 err = 0;
1691 goto out;
1692 }
1693
1694 /* It is valid on Linux to pass in a zero-length receive buffer. This
1695 * is not an error. We may as well bail out now.
1696 */
1697 if (!len) {
1698 err = 0;
1699 goto out;
1700 }
1701
1702 /* We must not copy less than target bytes into the user's buffer
1703 * before returning successfully, so we wait for the consume queue to
1704 * have that much data to consume before dequeueing. Note that this
1705 * makes it impossible to handle cases where target is greater than the
1706 * queue size.
1707 */
1708 target = sock_rcvlowat(sk, flags & MSG_WAITALL, len);
1709 if (target >= transport->stream_rcvhiwat(vsk)) {
1710 err = -ENOMEM;
1711 goto out;
1712 }
1713 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1714 copied = 0;
1715
1716 err = transport->notify_recv_init(vsk, target, &recv_data);
1717 if (err < 0)
1718 goto out;
1719
1720 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1721
1722 while (1) {
1723 s64 ready = vsock_stream_has_data(vsk);
1724
1725 if (ready < 0) {
1726 /* Invalid queue pair content. XXX This should be
1727 * changed to a connection reset in a later change.
1728 */
1729
1730 err = -ENOMEM;
1731 goto out_wait;
1732 } else if (ready > 0) {
1733 ssize_t read;
1734
1735 err = transport->notify_recv_pre_dequeue(
1736 vsk, target, &recv_data);
1737 if (err < 0)
1738 break;
1739
1740 read = transport->stream_dequeue(
Al Viro0f7db232014-11-20 04:05:34 -05001741 vsk, msg,
Andy Kingd021c342013-02-06 14:23:56 +00001742 len - copied, flags);
1743 if (read < 0) {
1744 err = -ENOMEM;
1745 break;
1746 }
1747
1748 copied += read;
1749
1750 err = transport->notify_recv_post_dequeue(
1751 vsk, target, read,
1752 !(flags & MSG_PEEK), &recv_data);
1753 if (err < 0)
1754 goto out_wait;
1755
1756 if (read >= target || flags & MSG_PEEK)
1757 break;
1758
1759 target -= read;
1760 } else {
1761 if (sk->sk_err != 0 || (sk->sk_shutdown & RCV_SHUTDOWN)
1762 || (vsk->peer_shutdown & SEND_SHUTDOWN)) {
1763 break;
1764 }
1765 /* Don't wait for non-blocking sockets. */
1766 if (timeout == 0) {
1767 err = -EAGAIN;
1768 break;
1769 }
1770
1771 err = transport->notify_recv_pre_block(
1772 vsk, target, &recv_data);
1773 if (err < 0)
1774 break;
1775
1776 release_sock(sk);
1777 timeout = schedule_timeout(timeout);
1778 lock_sock(sk);
1779
1780 if (signal_pending(current)) {
1781 err = sock_intr_errno(timeout);
1782 break;
1783 } else if (timeout == 0) {
1784 err = -EAGAIN;
1785 break;
1786 }
1787
1788 prepare_to_wait(sk_sleep(sk), &wait,
1789 TASK_INTERRUPTIBLE);
1790 }
1791 }
1792
1793 if (sk->sk_err)
1794 err = -sk->sk_err;
1795 else if (sk->sk_shutdown & RCV_SHUTDOWN)
1796 err = 0;
1797
1798 if (copied > 0) {
1799 /* We only do these additional bookkeeping/notification steps
1800 * if we actually copied something out of the queue pair
1801 * instead of just peeking ahead.
1802 */
1803
1804 if (!(flags & MSG_PEEK)) {
1805 /* If the other side has shutdown for sending and there
1806 * is nothing more to read, then modify the socket
1807 * state.
1808 */
1809 if (vsk->peer_shutdown & SEND_SHUTDOWN) {
1810 if (vsock_stream_has_data(vsk) <= 0) {
1811 sk->sk_state = SS_UNCONNECTED;
1812 sock_set_flag(sk, SOCK_DONE);
1813 sk->sk_state_change(sk);
1814 }
1815 }
1816 }
1817 err = copied;
1818 }
1819
1820out_wait:
1821 finish_wait(sk_sleep(sk), &wait);
1822out:
1823 release_sock(sk);
1824 return err;
1825}
1826
1827static const struct proto_ops vsock_stream_ops = {
1828 .family = PF_VSOCK,
1829 .owner = THIS_MODULE,
1830 .release = vsock_release,
1831 .bind = vsock_bind,
1832 .connect = vsock_stream_connect,
1833 .socketpair = sock_no_socketpair,
1834 .accept = vsock_accept,
1835 .getname = vsock_getname,
1836 .poll = vsock_poll,
1837 .ioctl = sock_no_ioctl,
1838 .listen = vsock_listen,
1839 .shutdown = vsock_shutdown,
1840 .setsockopt = vsock_stream_setsockopt,
1841 .getsockopt = vsock_stream_getsockopt,
1842 .sendmsg = vsock_stream_sendmsg,
1843 .recvmsg = vsock_stream_recvmsg,
1844 .mmap = sock_no_mmap,
1845 .sendpage = sock_no_sendpage,
1846};
1847
1848static int vsock_create(struct net *net, struct socket *sock,
1849 int protocol, int kern)
1850{
1851 if (!sock)
1852 return -EINVAL;
1853
Andy King6cf1c5f2013-02-18 06:04:13 +00001854 if (protocol && protocol != PF_VSOCK)
Andy Kingd021c342013-02-06 14:23:56 +00001855 return -EPROTONOSUPPORT;
1856
1857 switch (sock->type) {
1858 case SOCK_DGRAM:
1859 sock->ops = &vsock_dgram_ops;
1860 break;
1861 case SOCK_STREAM:
1862 sock->ops = &vsock_stream_ops;
1863 break;
1864 default:
1865 return -ESOCKTNOSUPPORT;
1866 }
1867
1868 sock->state = SS_UNCONNECTED;
1869
Eric W. Biederman11aa9c22015-05-08 21:09:13 -05001870 return __vsock_create(net, sock, NULL, GFP_KERNEL, 0, kern) ? 0 : -ENOMEM;
Andy Kingd021c342013-02-06 14:23:56 +00001871}
1872
1873static const struct net_proto_family vsock_family_ops = {
1874 .family = AF_VSOCK,
1875 .create = vsock_create,
1876 .owner = THIS_MODULE,
1877};
1878
1879static long vsock_dev_do_ioctl(struct file *filp,
1880 unsigned int cmd, void __user *ptr)
1881{
1882 u32 __user *p = ptr;
1883 int retval = 0;
1884
1885 switch (cmd) {
1886 case IOCTL_VM_SOCKETS_GET_LOCAL_CID:
1887 if (put_user(transport->get_local_cid(), p) != 0)
1888 retval = -EFAULT;
1889 break;
1890
1891 default:
1892 pr_err("Unknown ioctl %d\n", cmd);
1893 retval = -EINVAL;
1894 }
1895
1896 return retval;
1897}
1898
1899static long vsock_dev_ioctl(struct file *filp,
1900 unsigned int cmd, unsigned long arg)
1901{
1902 return vsock_dev_do_ioctl(filp, cmd, (void __user *)arg);
1903}
1904
1905#ifdef CONFIG_COMPAT
1906static long vsock_dev_compat_ioctl(struct file *filp,
1907 unsigned int cmd, unsigned long arg)
1908{
1909 return vsock_dev_do_ioctl(filp, cmd, compat_ptr(arg));
1910}
1911#endif
1912
1913static const struct file_operations vsock_device_ops = {
1914 .owner = THIS_MODULE,
1915 .unlocked_ioctl = vsock_dev_ioctl,
1916#ifdef CONFIG_COMPAT
1917 .compat_ioctl = vsock_dev_compat_ioctl,
1918#endif
1919 .open = nonseekable_open,
1920};
1921
1922static struct miscdevice vsock_device = {
1923 .name = "vsock",
Andy Kingd021c342013-02-06 14:23:56 +00001924 .fops = &vsock_device_ops,
1925};
1926
Andy King2c4a3362014-05-01 15:20:43 -07001927int __vsock_core_init(const struct vsock_transport *t, struct module *owner)
Andy Kingd021c342013-02-06 14:23:56 +00001928{
Andy King2c4a3362014-05-01 15:20:43 -07001929 int err = mutex_lock_interruptible(&vsock_register_mutex);
1930
1931 if (err)
1932 return err;
1933
1934 if (transport) {
1935 err = -EBUSY;
1936 goto err_busy;
1937 }
1938
1939 /* Transport must be the owner of the protocol so that it can't
1940 * unload while there are open sockets.
1941 */
1942 vsock_proto.owner = owner;
1943 transport = t;
Andy Kingd021c342013-02-06 14:23:56 +00001944
1945 vsock_init_tables();
1946
Asias He6ad0b2f2013-04-23 20:33:52 +00001947 vsock_device.minor = MISC_DYNAMIC_MINOR;
Andy Kingd021c342013-02-06 14:23:56 +00001948 err = misc_register(&vsock_device);
1949 if (err) {
1950 pr_err("Failed to register misc device\n");
1951 return -ENOENT;
1952 }
1953
1954 err = proto_register(&vsock_proto, 1); /* we want our slab */
1955 if (err) {
1956 pr_err("Cannot register vsock protocol\n");
1957 goto err_misc_deregister;
1958 }
1959
1960 err = sock_register(&vsock_family_ops);
1961 if (err) {
1962 pr_err("could not register af_vsock (%d) address family: %d\n",
1963 AF_VSOCK, err);
1964 goto err_unregister_proto;
1965 }
1966
Andy King2c4a3362014-05-01 15:20:43 -07001967 mutex_unlock(&vsock_register_mutex);
Andy Kingd021c342013-02-06 14:23:56 +00001968 return 0;
1969
1970err_unregister_proto:
1971 proto_unregister(&vsock_proto);
1972err_misc_deregister:
1973 misc_deregister(&vsock_device);
Andy King2c4a3362014-05-01 15:20:43 -07001974 transport = NULL;
1975err_busy:
1976 mutex_unlock(&vsock_register_mutex);
Andy Kingd021c342013-02-06 14:23:56 +00001977 return err;
1978}
Andy King2c4a3362014-05-01 15:20:43 -07001979EXPORT_SYMBOL_GPL(__vsock_core_init);
Andy Kingd021c342013-02-06 14:23:56 +00001980
1981void vsock_core_exit(void)
1982{
1983 mutex_lock(&vsock_register_mutex);
1984
1985 misc_deregister(&vsock_device);
1986 sock_unregister(AF_VSOCK);
1987 proto_unregister(&vsock_proto);
1988
1989 /* We do not want the assignment below re-ordered. */
1990 mb();
1991 transport = NULL;
1992
1993 mutex_unlock(&vsock_register_mutex);
1994}
1995EXPORT_SYMBOL_GPL(vsock_core_exit);
1996
1997MODULE_AUTHOR("VMware, Inc.");
1998MODULE_DESCRIPTION("VMware Virtual Socket Family");
Andy King2c4a3362014-05-01 15:20:43 -07001999MODULE_VERSION("1.0.1.0-k");
Andy Kingd021c342013-02-06 14:23:56 +00002000MODULE_LICENSE("GPL v2");