2 * VMware vSockets Driver
4 * Copyright (C) 2007-2013 VMware, Inc. All rights reserved.
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.
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
16 /* Implementation notes:
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.
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.
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).
38 * - "Server" sockets are referred to as listener sockets throughout this
39 * implementation because they are in the TCP_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.
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.
64 * - Lock ordering for pending or accept queue sockets is:
66 * lock_sock(listener);
67 * lock_sock_nested(pending, SINGLE_DEPTH_NESTING);
69 * Using explicit nested locking keeps lockdep happy since normally only one
70 * lock of a given class may be taken at a time.
72 * - Sockets created by user action will be cleaned up when the user process
73 * calls close(2), causing our release implementation to be called. Our release
74 * implementation will perform some cleanup then drop the last reference so our
75 * sk_destruct implementation is invoked. Our sk_destruct implementation will
76 * perform additional cleanup that's common for both types of sockets.
78 * - A socket's reference count is what ensures that the structure won't be
79 * freed. Each entry in a list (such as the "global" bound and connected tables
80 * and the listener socket's pending list and connected queue) ensures a
81 * reference. When we defer work until process context and pass a socket as our
82 * argument, we must ensure the reference count is increased to ensure the
83 * socket isn't freed before the function is run; the deferred function will
84 * then drop the reference.
86 * - sk->sk_state uses the TCP state constants because they are widely used by
87 * other address families and exposed to userspace tools like ss(8):
89 * TCP_CLOSE - unconnected
90 * TCP_SYN_SENT - connecting
91 * TCP_ESTABLISHED - connected
92 * TCP_CLOSING - disconnecting
93 * TCP_LISTEN - listening
96 #include <linux/types.h>
97 #include <linux/bitops.h>
98 #include <linux/cred.h>
99 #include <linux/init.h>
100 #include <linux/io.h>
101 #include <linux/kernel.h>
102 #include <linux/sched/signal.h>
103 #include <linux/kmod.h>
104 #include <linux/list.h>
105 #include <linux/miscdevice.h>
106 #include <linux/module.h>
107 #include <linux/mutex.h>
108 #include <linux/net.h>
109 #include <linux/poll.h>
110 #include <linux/random.h>
111 #include <linux/skbuff.h>
112 #include <linux/smp.h>
113 #include <linux/socket.h>
114 #include <linux/stddef.h>
115 #include <linux/unistd.h>
116 #include <linux/wait.h>
117 #include <linux/workqueue.h>
118 #include <net/sock.h>
119 #include <net/af_vsock.h>
121 static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr);
122 static void vsock_sk_destruct(struct sock *sk);
123 static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
125 /* Protocol family. */
126 static struct proto vsock_proto = {
128 .owner = THIS_MODULE,
129 .obj_size = sizeof(struct vsock_sock),
132 /* The default peer timeout indicates how long we will wait for a peer response
133 * to a control message.
135 #define VSOCK_DEFAULT_CONNECT_TIMEOUT (2 * HZ)
137 static const struct vsock_transport *transport;
138 static DEFINE_MUTEX(vsock_register_mutex);
142 /* Get the ID of the local context. This is transport dependent. */
144 int vm_sockets_get_local_cid(void)
146 return transport->get_local_cid();
148 EXPORT_SYMBOL_GPL(vm_sockets_get_local_cid);
152 /* Each bound VSocket is stored in the bind hash table and each connected
153 * VSocket is stored in the connected hash table.
155 * Unbound sockets are all put on the same list attached to the end of the hash
156 * table (vsock_unbound_sockets). Bound sockets are added to the hash table in
157 * the bucket that their local address hashes to (vsock_bound_sockets(addr)
158 * represents the list that addr hashes to).
160 * Specifically, we initialize the vsock_bind_table array to a size of
161 * VSOCK_HASH_SIZE + 1 so that vsock_bind_table[0] through
162 * vsock_bind_table[VSOCK_HASH_SIZE - 1] are for bound sockets and
163 * vsock_bind_table[VSOCK_HASH_SIZE] is for unbound sockets. The hash function
164 * mods with VSOCK_HASH_SIZE to ensure this.
166 #define VSOCK_HASH_SIZE 251
167 #define MAX_PORT_RETRIES 24
169 #define VSOCK_HASH(addr) ((addr)->svm_port % VSOCK_HASH_SIZE)
170 #define vsock_bound_sockets(addr) (&vsock_bind_table[VSOCK_HASH(addr)])
171 #define vsock_unbound_sockets (&vsock_bind_table[VSOCK_HASH_SIZE])
173 /* XXX This can probably be implemented in a better way. */
174 #define VSOCK_CONN_HASH(src, dst) \
175 (((src)->svm_cid ^ (dst)->svm_port) % VSOCK_HASH_SIZE)
176 #define vsock_connected_sockets(src, dst) \
177 (&vsock_connected_table[VSOCK_CONN_HASH(src, dst)])
178 #define vsock_connected_sockets_vsk(vsk) \
179 vsock_connected_sockets(&(vsk)->remote_addr, &(vsk)->local_addr)
181 static struct list_head vsock_bind_table[VSOCK_HASH_SIZE + 1];
182 static struct list_head vsock_connected_table[VSOCK_HASH_SIZE];
183 static DEFINE_SPINLOCK(vsock_table_lock);
185 /* Autobind this socket to the local address if necessary. */
186 static int vsock_auto_bind(struct vsock_sock *vsk)
188 struct sock *sk = sk_vsock(vsk);
189 struct sockaddr_vm local_addr;
191 if (vsock_addr_bound(&vsk->local_addr))
193 vsock_addr_init(&local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
194 return __vsock_bind(sk, &local_addr);
197 static void vsock_init_tables(void)
201 for (i = 0; i < ARRAY_SIZE(vsock_bind_table); i++)
202 INIT_LIST_HEAD(&vsock_bind_table[i]);
204 for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++)
205 INIT_LIST_HEAD(&vsock_connected_table[i]);
208 static void __vsock_insert_bound(struct list_head *list,
209 struct vsock_sock *vsk)
212 list_add(&vsk->bound_table, list);
215 static void __vsock_insert_connected(struct list_head *list,
216 struct vsock_sock *vsk)
219 list_add(&vsk->connected_table, list);
222 static void __vsock_remove_bound(struct vsock_sock *vsk)
224 list_del_init(&vsk->bound_table);
228 static void __vsock_remove_connected(struct vsock_sock *vsk)
230 list_del_init(&vsk->connected_table);
234 static struct sock *__vsock_find_bound_socket(struct sockaddr_vm *addr)
236 struct vsock_sock *vsk;
238 list_for_each_entry(vsk, vsock_bound_sockets(addr), bound_table)
239 if (addr->svm_port == vsk->local_addr.svm_port)
240 return sk_vsock(vsk);
245 static struct sock *__vsock_find_connected_socket(struct sockaddr_vm *src,
246 struct sockaddr_vm *dst)
248 struct vsock_sock *vsk;
250 list_for_each_entry(vsk, vsock_connected_sockets(src, dst),
252 if (vsock_addr_equals_addr(src, &vsk->remote_addr) &&
253 dst->svm_port == vsk->local_addr.svm_port) {
254 return sk_vsock(vsk);
261 static bool __vsock_in_bound_table(struct vsock_sock *vsk)
263 return !list_empty(&vsk->bound_table);
266 static bool __vsock_in_connected_table(struct vsock_sock *vsk)
268 return !list_empty(&vsk->connected_table);
271 static void vsock_insert_unbound(struct vsock_sock *vsk)
273 spin_lock_bh(&vsock_table_lock);
274 __vsock_insert_bound(vsock_unbound_sockets, vsk);
275 spin_unlock_bh(&vsock_table_lock);
278 void vsock_insert_connected(struct vsock_sock *vsk)
280 struct list_head *list = vsock_connected_sockets(
281 &vsk->remote_addr, &vsk->local_addr);
283 spin_lock_bh(&vsock_table_lock);
284 __vsock_insert_connected(list, vsk);
285 spin_unlock_bh(&vsock_table_lock);
287 EXPORT_SYMBOL_GPL(vsock_insert_connected);
289 void vsock_remove_bound(struct vsock_sock *vsk)
291 spin_lock_bh(&vsock_table_lock);
292 if (__vsock_in_bound_table(vsk))
293 __vsock_remove_bound(vsk);
294 spin_unlock_bh(&vsock_table_lock);
296 EXPORT_SYMBOL_GPL(vsock_remove_bound);
298 void vsock_remove_connected(struct vsock_sock *vsk)
300 spin_lock_bh(&vsock_table_lock);
301 if (__vsock_in_connected_table(vsk))
302 __vsock_remove_connected(vsk);
303 spin_unlock_bh(&vsock_table_lock);
305 EXPORT_SYMBOL_GPL(vsock_remove_connected);
307 struct sock *vsock_find_bound_socket(struct sockaddr_vm *addr)
311 spin_lock_bh(&vsock_table_lock);
312 sk = __vsock_find_bound_socket(addr);
316 spin_unlock_bh(&vsock_table_lock);
320 EXPORT_SYMBOL_GPL(vsock_find_bound_socket);
322 struct sock *vsock_find_connected_socket(struct sockaddr_vm *src,
323 struct sockaddr_vm *dst)
327 spin_lock_bh(&vsock_table_lock);
328 sk = __vsock_find_connected_socket(src, dst);
332 spin_unlock_bh(&vsock_table_lock);
336 EXPORT_SYMBOL_GPL(vsock_find_connected_socket);
338 void vsock_remove_sock(struct vsock_sock *vsk)
340 vsock_remove_bound(vsk);
341 vsock_remove_connected(vsk);
343 EXPORT_SYMBOL_GPL(vsock_remove_sock);
345 void vsock_for_each_connected_socket(void (*fn)(struct sock *sk))
349 spin_lock_bh(&vsock_table_lock);
351 for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) {
352 struct vsock_sock *vsk;
353 list_for_each_entry(vsk, &vsock_connected_table[i],
358 spin_unlock_bh(&vsock_table_lock);
360 EXPORT_SYMBOL_GPL(vsock_for_each_connected_socket);
362 void vsock_add_pending(struct sock *listener, struct sock *pending)
364 struct vsock_sock *vlistener;
365 struct vsock_sock *vpending;
367 vlistener = vsock_sk(listener);
368 vpending = vsock_sk(pending);
372 list_add_tail(&vpending->pending_links, &vlistener->pending_links);
374 EXPORT_SYMBOL_GPL(vsock_add_pending);
376 void vsock_remove_pending(struct sock *listener, struct sock *pending)
378 struct vsock_sock *vpending = vsock_sk(pending);
380 list_del_init(&vpending->pending_links);
384 EXPORT_SYMBOL_GPL(vsock_remove_pending);
386 void vsock_enqueue_accept(struct sock *listener, struct sock *connected)
388 struct vsock_sock *vlistener;
389 struct vsock_sock *vconnected;
391 vlistener = vsock_sk(listener);
392 vconnected = vsock_sk(connected);
394 sock_hold(connected);
396 list_add_tail(&vconnected->accept_queue, &vlistener->accept_queue);
398 EXPORT_SYMBOL_GPL(vsock_enqueue_accept);
400 static struct sock *vsock_dequeue_accept(struct sock *listener)
402 struct vsock_sock *vlistener;
403 struct vsock_sock *vconnected;
405 vlistener = vsock_sk(listener);
407 if (list_empty(&vlistener->accept_queue))
410 vconnected = list_entry(vlistener->accept_queue.next,
411 struct vsock_sock, accept_queue);
413 list_del_init(&vconnected->accept_queue);
415 /* The caller will need a reference on the connected socket so we let
416 * it call sock_put().
419 return sk_vsock(vconnected);
422 static bool vsock_is_accept_queue_empty(struct sock *sk)
424 struct vsock_sock *vsk = vsock_sk(sk);
425 return list_empty(&vsk->accept_queue);
428 static bool vsock_is_pending(struct sock *sk)
430 struct vsock_sock *vsk = vsock_sk(sk);
431 return !list_empty(&vsk->pending_links);
434 static int vsock_send_shutdown(struct sock *sk, int mode)
436 return transport->shutdown(vsock_sk(sk), mode);
439 static void vsock_pending_work(struct work_struct *work)
442 struct sock *listener;
443 struct vsock_sock *vsk;
446 vsk = container_of(work, struct vsock_sock, pending_work.work);
448 listener = vsk->listener;
452 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
454 if (vsock_is_pending(sk)) {
455 vsock_remove_pending(listener, sk);
457 listener->sk_ack_backlog--;
458 } else if (!vsk->rejected) {
459 /* We are not on the pending list and accept() did not reject
460 * us, so we must have been accepted by our user process. We
461 * just need to drop our references to the sockets and be on
468 /* We need to remove ourself from the global connected sockets list so
469 * incoming packets can't find this socket, and to reduce the reference
472 vsock_remove_connected(vsk);
474 sk->sk_state = TCP_CLOSE;
478 release_sock(listener);
486 /**** SOCKET OPERATIONS ****/
488 static int __vsock_bind_stream(struct vsock_sock *vsk,
489 struct sockaddr_vm *addr)
492 struct sockaddr_vm new_addr;
495 port = LAST_RESERVED_PORT + 1 +
496 prandom_u32_max(U32_MAX - LAST_RESERVED_PORT);
498 vsock_addr_init(&new_addr, addr->svm_cid, addr->svm_port);
500 if (addr->svm_port == VMADDR_PORT_ANY) {
504 for (i = 0; i < MAX_PORT_RETRIES; i++) {
505 if (port <= LAST_RESERVED_PORT)
506 port = LAST_RESERVED_PORT + 1;
508 new_addr.svm_port = port++;
510 if (!__vsock_find_bound_socket(&new_addr)) {
517 return -EADDRNOTAVAIL;
519 /* If port is in reserved range, ensure caller
520 * has necessary privileges.
522 if (addr->svm_port <= LAST_RESERVED_PORT &&
523 !capable(CAP_NET_BIND_SERVICE)) {
527 if (__vsock_find_bound_socket(&new_addr))
531 vsock_addr_init(&vsk->local_addr, new_addr.svm_cid, new_addr.svm_port);
533 /* Remove stream sockets from the unbound list and add them to the hash
534 * table for easy lookup by its address. The unbound list is simply an
535 * extra entry at the end of the hash table, a trick used by AF_UNIX.
537 __vsock_remove_bound(vsk);
538 __vsock_insert_bound(vsock_bound_sockets(&vsk->local_addr), vsk);
543 static int __vsock_bind_dgram(struct vsock_sock *vsk,
544 struct sockaddr_vm *addr)
546 return transport->dgram_bind(vsk, addr);
549 static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr)
551 struct vsock_sock *vsk = vsock_sk(sk);
555 /* First ensure this socket isn't already bound. */
556 if (vsock_addr_bound(&vsk->local_addr))
559 /* Now bind to the provided address or select appropriate values if
560 * none are provided (VMADDR_CID_ANY and VMADDR_PORT_ANY). Note that
561 * like AF_INET prevents binding to a non-local IP address (in most
562 * cases), we only allow binding to the local CID.
564 cid = transport->get_local_cid();
565 if (addr->svm_cid != cid && addr->svm_cid != VMADDR_CID_ANY)
566 return -EADDRNOTAVAIL;
568 switch (sk->sk_socket->type) {
570 spin_lock_bh(&vsock_table_lock);
571 retval = __vsock_bind_stream(vsk, addr);
572 spin_unlock_bh(&vsock_table_lock);
576 retval = __vsock_bind_dgram(vsk, addr);
587 static void vsock_connect_timeout(struct work_struct *work);
589 struct sock *__vsock_create(struct net *net,
597 struct vsock_sock *psk;
598 struct vsock_sock *vsk;
600 sk = sk_alloc(net, AF_VSOCK, priority, &vsock_proto, kern);
604 sock_init_data(sock, sk);
606 /* sk->sk_type is normally set in sock_init_data, but only if sock is
607 * non-NULL. We make sure that our sockets always have a type by
608 * setting it here if needed.
614 vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
615 vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
617 sk->sk_destruct = vsock_sk_destruct;
618 sk->sk_backlog_rcv = vsock_queue_rcv_skb;
619 sock_reset_flag(sk, SOCK_DONE);
621 INIT_LIST_HEAD(&vsk->bound_table);
622 INIT_LIST_HEAD(&vsk->connected_table);
623 vsk->listener = NULL;
624 INIT_LIST_HEAD(&vsk->pending_links);
625 INIT_LIST_HEAD(&vsk->accept_queue);
626 vsk->rejected = false;
627 vsk->sent_request = false;
628 vsk->ignore_connecting_rst = false;
629 vsk->peer_shutdown = 0;
630 INIT_DELAYED_WORK(&vsk->connect_work, vsock_connect_timeout);
631 INIT_DELAYED_WORK(&vsk->pending_work, vsock_pending_work);
633 psk = parent ? vsock_sk(parent) : NULL;
635 vsk->trusted = psk->trusted;
636 vsk->owner = get_cred(psk->owner);
637 vsk->connect_timeout = psk->connect_timeout;
638 security_sk_clone(parent, sk);
640 vsk->trusted = ns_capable_noaudit(&init_user_ns, CAP_NET_ADMIN);
641 vsk->owner = get_current_cred();
642 vsk->connect_timeout = VSOCK_DEFAULT_CONNECT_TIMEOUT;
645 if (transport->init(vsk, psk) < 0) {
651 vsock_insert_unbound(vsk);
655 EXPORT_SYMBOL_GPL(__vsock_create);
657 static void __vsock_release(struct sock *sk, int level)
661 struct sock *pending;
662 struct vsock_sock *vsk;
665 pending = NULL; /* Compiler warning. */
667 /* The release call is supposed to use lock_sock_nested()
668 * rather than lock_sock(), if a sock lock should be acquired.
670 transport->release(vsk);
672 /* When "level" is SINGLE_DEPTH_NESTING, use the nested
673 * version to avoid the warning "possible recursive locking
674 * detected". When "level" is 0, lock_sock_nested(sk, level)
675 * is the same as lock_sock(sk).
677 lock_sock_nested(sk, level);
679 sk->sk_shutdown = SHUTDOWN_MASK;
681 while ((skb = skb_dequeue(&sk->sk_receive_queue)))
684 /* Clean up any sockets that never were accepted. */
685 while ((pending = vsock_dequeue_accept(sk)) != NULL) {
686 __vsock_release(pending, SINGLE_DEPTH_NESTING);
695 static void vsock_sk_destruct(struct sock *sk)
697 struct vsock_sock *vsk = vsock_sk(sk);
699 transport->destruct(vsk);
701 /* When clearing these addresses, there's no need to set the family and
702 * possibly register the address family with the kernel.
704 vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
705 vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
707 put_cred(vsk->owner);
710 static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
714 err = sock_queue_rcv_skb(sk, skb);
721 s64 vsock_stream_has_data(struct vsock_sock *vsk)
723 return transport->stream_has_data(vsk);
725 EXPORT_SYMBOL_GPL(vsock_stream_has_data);
727 s64 vsock_stream_has_space(struct vsock_sock *vsk)
729 return transport->stream_has_space(vsk);
731 EXPORT_SYMBOL_GPL(vsock_stream_has_space);
733 static int vsock_release(struct socket *sock)
735 __vsock_release(sock->sk, 0);
737 sock->state = SS_FREE;
743 vsock_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
747 struct sockaddr_vm *vm_addr;
751 if (vsock_addr_cast(addr, addr_len, &vm_addr) != 0)
755 err = __vsock_bind(sk, vm_addr);
761 static int vsock_getname(struct socket *sock,
762 struct sockaddr *addr, int *addr_len, int peer)
766 struct vsock_sock *vsk;
767 struct sockaddr_vm *vm_addr;
776 if (sock->state != SS_CONNECTED) {
780 vm_addr = &vsk->remote_addr;
782 vm_addr = &vsk->local_addr;
790 /* sys_getsockname() and sys_getpeername() pass us a
791 * MAX_SOCK_ADDR-sized buffer and don't set addr_len. Unfortunately
792 * that macro is defined in socket.c instead of .h, so we hardcode its
795 BUILD_BUG_ON(sizeof(*vm_addr) > 128);
796 memcpy(addr, vm_addr, sizeof(*vm_addr));
797 *addr_len = sizeof(*vm_addr);
804 static int vsock_shutdown(struct socket *sock, int mode)
809 /* User level uses SHUT_RD (0) and SHUT_WR (1), but the kernel uses
810 * RCV_SHUTDOWN (1) and SEND_SHUTDOWN (2), so we must increment mode
811 * here like the other address families do. Note also that the
812 * increment makes SHUT_RDWR (2) into RCV_SHUTDOWN | SEND_SHUTDOWN (3),
813 * which is what we want.
817 if ((mode & ~SHUTDOWN_MASK) || !mode)
820 /* If this is a STREAM socket and it is not connected then bail out
821 * immediately. If it is a DGRAM socket then we must first kick the
822 * socket so that it wakes up from any sleeping calls, for example
823 * recv(), and then afterwards return the error.
829 if (sock->state == SS_UNCONNECTED) {
831 if (sk->sk_type == SOCK_STREAM)
834 sock->state = SS_DISCONNECTING;
838 /* Receive and send shutdowns are treated alike. */
839 mode = mode & (RCV_SHUTDOWN | SEND_SHUTDOWN);
841 sk->sk_shutdown |= mode;
842 sk->sk_state_change(sk);
844 if (sk->sk_type == SOCK_STREAM) {
845 sock_reset_flag(sk, SOCK_DONE);
846 vsock_send_shutdown(sk, mode);
855 static unsigned int vsock_poll(struct file *file, struct socket *sock,
860 struct vsock_sock *vsk;
865 poll_wait(file, sk_sleep(sk), wait);
869 /* Signify that there has been an error on this socket. */
872 /* INET sockets treat local write shutdown and peer write shutdown as a
873 * case of POLLHUP set.
875 if ((sk->sk_shutdown == SHUTDOWN_MASK) ||
876 ((sk->sk_shutdown & SEND_SHUTDOWN) &&
877 (vsk->peer_shutdown & SEND_SHUTDOWN))) {
881 if (sk->sk_shutdown & RCV_SHUTDOWN ||
882 vsk->peer_shutdown & SEND_SHUTDOWN) {
886 if (sock->type == SOCK_DGRAM) {
887 /* For datagram sockets we can read if there is something in
888 * the queue and write as long as the socket isn't shutdown for
891 if (!skb_queue_empty_lockless(&sk->sk_receive_queue) ||
892 (sk->sk_shutdown & RCV_SHUTDOWN)) {
893 mask |= POLLIN | POLLRDNORM;
896 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
897 mask |= POLLOUT | POLLWRNORM | POLLWRBAND;
899 } else if (sock->type == SOCK_STREAM) {
902 /* Listening sockets that have connections in their accept
905 if (sk->sk_state == TCP_LISTEN
906 && !vsock_is_accept_queue_empty(sk))
907 mask |= POLLIN | POLLRDNORM;
909 /* If there is something in the queue then we can read. */
910 if (transport->stream_is_active(vsk) &&
911 !(sk->sk_shutdown & RCV_SHUTDOWN)) {
912 bool data_ready_now = false;
913 int ret = transport->notify_poll_in(
914 vsk, 1, &data_ready_now);
919 mask |= POLLIN | POLLRDNORM;
924 /* Sockets whose connections have been closed, reset, or
925 * terminated should also be considered read, and we check the
926 * shutdown flag for that.
928 if (sk->sk_shutdown & RCV_SHUTDOWN ||
929 vsk->peer_shutdown & SEND_SHUTDOWN) {
930 mask |= POLLIN | POLLRDNORM;
933 /* Connected sockets that can produce data can be written. */
934 if (sk->sk_state == TCP_ESTABLISHED) {
935 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
936 bool space_avail_now = false;
937 int ret = transport->notify_poll_out(
938 vsk, 1, &space_avail_now);
943 /* Remove POLLWRBAND since INET
944 * sockets are not setting it.
946 mask |= POLLOUT | POLLWRNORM;
952 /* Simulate INET socket poll behaviors, which sets
953 * POLLOUT|POLLWRNORM when peer is closed and nothing to read,
954 * but local send is not shutdown.
956 if (sk->sk_state == TCP_CLOSE) {
957 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
958 mask |= POLLOUT | POLLWRNORM;
968 static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg,
973 struct vsock_sock *vsk;
974 struct sockaddr_vm *remote_addr;
976 if (msg->msg_flags & MSG_OOB)
979 /* For now, MSG_DONTWAIT is always assumed... */
986 err = vsock_auto_bind(vsk);
991 /* If the provided message contains an address, use that. Otherwise
992 * fall back on the socket's remote handle (if it has been connected).
995 vsock_addr_cast(msg->msg_name, msg->msg_namelen,
996 &remote_addr) == 0) {
997 /* Ensure this address is of the right type and is a valid
1001 if (remote_addr->svm_cid == VMADDR_CID_ANY)
1002 remote_addr->svm_cid = transport->get_local_cid();
1004 if (!vsock_addr_bound(remote_addr)) {
1008 } else if (sock->state == SS_CONNECTED) {
1009 remote_addr = &vsk->remote_addr;
1011 if (remote_addr->svm_cid == VMADDR_CID_ANY)
1012 remote_addr->svm_cid = transport->get_local_cid();
1014 /* XXX Should connect() or this function ensure remote_addr is
1017 if (!vsock_addr_bound(&vsk->remote_addr)) {
1026 if (!transport->dgram_allow(remote_addr->svm_cid,
1027 remote_addr->svm_port)) {
1032 err = transport->dgram_enqueue(vsk, remote_addr, msg, len);
1039 static int vsock_dgram_connect(struct socket *sock,
1040 struct sockaddr *addr, int addr_len, int flags)
1044 struct vsock_sock *vsk;
1045 struct sockaddr_vm *remote_addr;
1050 err = vsock_addr_cast(addr, addr_len, &remote_addr);
1051 if (err == -EAFNOSUPPORT && remote_addr->svm_family == AF_UNSPEC) {
1053 vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY,
1055 sock->state = SS_UNCONNECTED;
1058 } else if (err != 0)
1063 err = vsock_auto_bind(vsk);
1067 if (!transport->dgram_allow(remote_addr->svm_cid,
1068 remote_addr->svm_port)) {
1073 memcpy(&vsk->remote_addr, remote_addr, sizeof(vsk->remote_addr));
1074 sock->state = SS_CONNECTED;
1081 static int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg,
1082 size_t len, int flags)
1084 return transport->dgram_dequeue(vsock_sk(sock->sk), msg, len, flags);
1087 static const struct proto_ops vsock_dgram_ops = {
1089 .owner = THIS_MODULE,
1090 .release = vsock_release,
1092 .connect = vsock_dgram_connect,
1093 .socketpair = sock_no_socketpair,
1094 .accept = sock_no_accept,
1095 .getname = vsock_getname,
1097 .ioctl = sock_no_ioctl,
1098 .listen = sock_no_listen,
1099 .shutdown = vsock_shutdown,
1100 .setsockopt = sock_no_setsockopt,
1101 .getsockopt = sock_no_getsockopt,
1102 .sendmsg = vsock_dgram_sendmsg,
1103 .recvmsg = vsock_dgram_recvmsg,
1104 .mmap = sock_no_mmap,
1105 .sendpage = sock_no_sendpage,
1108 static int vsock_transport_cancel_pkt(struct vsock_sock *vsk)
1110 if (!transport->cancel_pkt)
1113 return transport->cancel_pkt(vsk);
1116 static void vsock_connect_timeout(struct work_struct *work)
1119 struct vsock_sock *vsk;
1121 vsk = container_of(work, struct vsock_sock, connect_work.work);
1125 if (sk->sk_state == TCP_SYN_SENT &&
1126 (sk->sk_shutdown != SHUTDOWN_MASK)) {
1127 sk->sk_state = TCP_CLOSE;
1128 sk->sk_err = ETIMEDOUT;
1129 sk->sk_error_report(sk);
1130 vsock_transport_cancel_pkt(vsk);
1137 static int vsock_stream_connect(struct socket *sock, struct sockaddr *addr,
1138 int addr_len, int flags)
1142 struct vsock_sock *vsk;
1143 struct sockaddr_vm *remote_addr;
1153 /* XXX AF_UNSPEC should make us disconnect like AF_INET. */
1154 switch (sock->state) {
1158 case SS_DISCONNECTING:
1162 /* This continues on so we can move sock into the SS_CONNECTED
1163 * state once the connection has completed (at which point err
1164 * will be set to zero also). Otherwise, we will either wait
1165 * for the connection or return -EALREADY should this be a
1166 * non-blocking call.
1171 if ((sk->sk_state == TCP_LISTEN) ||
1172 vsock_addr_cast(addr, addr_len, &remote_addr) != 0) {
1177 /* The hypervisor and well-known contexts do not have socket
1180 if (!transport->stream_allow(remote_addr->svm_cid,
1181 remote_addr->svm_port)) {
1186 /* Set the remote address that we are connecting to. */
1187 memcpy(&vsk->remote_addr, remote_addr,
1188 sizeof(vsk->remote_addr));
1190 err = vsock_auto_bind(vsk);
1194 sk->sk_state = TCP_SYN_SENT;
1196 err = transport->connect(vsk);
1200 /* Mark sock as connecting and set the error code to in
1201 * progress in case this is a non-blocking connect.
1203 sock->state = SS_CONNECTING;
1207 /* The receive path will handle all communication until we are able to
1208 * enter the connected state. Here we wait for the connection to be
1209 * completed or a notification of an error.
1211 timeout = vsk->connect_timeout;
1212 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1214 while (sk->sk_state != TCP_ESTABLISHED && sk->sk_err == 0) {
1215 if (flags & O_NONBLOCK) {
1216 /* If we're not going to block, we schedule a timeout
1217 * function to generate a timeout on the connection
1218 * attempt, in case the peer doesn't respond in a
1219 * timely manner. We hold on to the socket until the
1223 schedule_delayed_work(&vsk->connect_work, timeout);
1225 /* Skip ahead to preserve error code set above. */
1230 timeout = schedule_timeout(timeout);
1233 if (signal_pending(current)) {
1234 err = sock_intr_errno(timeout);
1235 sk->sk_state = sk->sk_state == TCP_ESTABLISHED ? TCP_CLOSING : TCP_CLOSE;
1236 sock->state = SS_UNCONNECTED;
1237 vsock_transport_cancel_pkt(vsk);
1239 } else if (timeout == 0) {
1241 sk->sk_state = TCP_CLOSE;
1242 sock->state = SS_UNCONNECTED;
1243 vsock_transport_cancel_pkt(vsk);
1247 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1252 sk->sk_state = TCP_CLOSE;
1253 sock->state = SS_UNCONNECTED;
1259 finish_wait(sk_sleep(sk), &wait);
1265 static int vsock_accept(struct socket *sock, struct socket *newsock, int flags,
1268 struct sock *listener;
1270 struct sock *connected;
1271 struct vsock_sock *vconnected;
1276 listener = sock->sk;
1278 lock_sock(listener);
1280 if (sock->type != SOCK_STREAM) {
1285 if (listener->sk_state != TCP_LISTEN) {
1290 /* Wait for children sockets to appear; these are the new sockets
1291 * created upon connection establishment.
1293 timeout = sock_rcvtimeo(listener, flags & O_NONBLOCK);
1294 prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1296 while ((connected = vsock_dequeue_accept(listener)) == NULL &&
1297 listener->sk_err == 0) {
1298 release_sock(listener);
1299 timeout = schedule_timeout(timeout);
1300 finish_wait(sk_sleep(listener), &wait);
1301 lock_sock(listener);
1303 if (signal_pending(current)) {
1304 err = sock_intr_errno(timeout);
1306 } else if (timeout == 0) {
1311 prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1313 finish_wait(sk_sleep(listener), &wait);
1315 if (listener->sk_err)
1316 err = -listener->sk_err;
1319 listener->sk_ack_backlog--;
1321 lock_sock_nested(connected, SINGLE_DEPTH_NESTING);
1322 vconnected = vsock_sk(connected);
1324 /* If the listener socket has received an error, then we should
1325 * reject this socket and return. Note that we simply mark the
1326 * socket rejected, drop our reference, and let the cleanup
1327 * function handle the cleanup; the fact that we found it in
1328 * the listener's accept queue guarantees that the cleanup
1329 * function hasn't run yet.
1332 vconnected->rejected = true;
1334 newsock->state = SS_CONNECTED;
1335 sock_graft(connected, newsock);
1338 release_sock(connected);
1339 sock_put(connected);
1343 release_sock(listener);
1347 static int vsock_listen(struct socket *sock, int backlog)
1351 struct vsock_sock *vsk;
1357 if (sock->type != SOCK_STREAM) {
1362 if (sock->state != SS_UNCONNECTED) {
1369 if (!vsock_addr_bound(&vsk->local_addr)) {
1374 sk->sk_max_ack_backlog = backlog;
1375 sk->sk_state = TCP_LISTEN;
1384 static int vsock_stream_setsockopt(struct socket *sock,
1387 char __user *optval,
1388 unsigned int optlen)
1392 struct vsock_sock *vsk;
1395 if (level != AF_VSOCK)
1396 return -ENOPROTOOPT;
1398 #define COPY_IN(_v) \
1400 if (optlen < sizeof(_v)) { \
1404 if (copy_from_user(&_v, optval, sizeof(_v)) != 0) { \
1417 case SO_VM_SOCKETS_BUFFER_SIZE:
1419 transport->set_buffer_size(vsk, val);
1422 case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1424 transport->set_max_buffer_size(vsk, val);
1427 case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1429 transport->set_min_buffer_size(vsk, val);
1432 case SO_VM_SOCKETS_CONNECT_TIMEOUT: {
1435 if (tv.tv_sec >= 0 && tv.tv_usec < USEC_PER_SEC &&
1436 tv.tv_sec < (MAX_SCHEDULE_TIMEOUT / HZ - 1)) {
1437 vsk->connect_timeout = tv.tv_sec * HZ +
1438 DIV_ROUND_UP(tv.tv_usec, (1000000 / HZ));
1439 if (vsk->connect_timeout == 0)
1440 vsk->connect_timeout =
1441 VSOCK_DEFAULT_CONNECT_TIMEOUT;
1461 static int vsock_stream_getsockopt(struct socket *sock,
1462 int level, int optname,
1463 char __user *optval,
1469 struct vsock_sock *vsk;
1472 if (level != AF_VSOCK)
1473 return -ENOPROTOOPT;
1475 err = get_user(len, optlen);
1479 #define COPY_OUT(_v) \
1481 if (len < sizeof(_v)) \
1485 if (copy_to_user(optval, &_v, len) != 0) \
1495 case SO_VM_SOCKETS_BUFFER_SIZE:
1496 val = transport->get_buffer_size(vsk);
1500 case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1501 val = transport->get_max_buffer_size(vsk);
1505 case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1506 val = transport->get_min_buffer_size(vsk);
1510 case SO_VM_SOCKETS_CONNECT_TIMEOUT: {
1512 tv.tv_sec = vsk->connect_timeout / HZ;
1514 (vsk->connect_timeout -
1515 tv.tv_sec * HZ) * (1000000 / HZ);
1520 return -ENOPROTOOPT;
1523 err = put_user(len, optlen);
1532 static int vsock_stream_sendmsg(struct socket *sock, struct msghdr *msg,
1536 struct vsock_sock *vsk;
1537 ssize_t total_written;
1540 struct vsock_transport_send_notify_data send_data;
1541 DEFINE_WAIT_FUNC(wait, woken_wake_function);
1548 if (msg->msg_flags & MSG_OOB)
1553 /* Callers should not provide a destination with stream sockets. */
1554 if (msg->msg_namelen) {
1555 err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP;
1559 /* Send data only if both sides are not shutdown in the direction. */
1560 if (sk->sk_shutdown & SEND_SHUTDOWN ||
1561 vsk->peer_shutdown & RCV_SHUTDOWN) {
1566 if (sk->sk_state != TCP_ESTABLISHED ||
1567 !vsock_addr_bound(&vsk->local_addr)) {
1572 if (!vsock_addr_bound(&vsk->remote_addr)) {
1573 err = -EDESTADDRREQ;
1577 /* Wait for room in the produce queue to enqueue our user's data. */
1578 timeout = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1580 err = transport->notify_send_init(vsk, &send_data);
1584 while (total_written < len) {
1587 add_wait_queue(sk_sleep(sk), &wait);
1588 while (vsock_stream_has_space(vsk) == 0 &&
1590 !(sk->sk_shutdown & SEND_SHUTDOWN) &&
1591 !(vsk->peer_shutdown & RCV_SHUTDOWN)) {
1593 /* Don't wait for non-blocking sockets. */
1596 remove_wait_queue(sk_sleep(sk), &wait);
1600 err = transport->notify_send_pre_block(vsk, &send_data);
1602 remove_wait_queue(sk_sleep(sk), &wait);
1607 timeout = wait_woken(&wait, TASK_INTERRUPTIBLE, timeout);
1609 if (signal_pending(current)) {
1610 err = sock_intr_errno(timeout);
1611 remove_wait_queue(sk_sleep(sk), &wait);
1613 } else if (timeout == 0) {
1615 remove_wait_queue(sk_sleep(sk), &wait);
1619 remove_wait_queue(sk_sleep(sk), &wait);
1621 /* These checks occur both as part of and after the loop
1622 * conditional since we need to check before and after
1628 } else if ((sk->sk_shutdown & SEND_SHUTDOWN) ||
1629 (vsk->peer_shutdown & RCV_SHUTDOWN)) {
1634 err = transport->notify_send_pre_enqueue(vsk, &send_data);
1638 /* Note that enqueue will only write as many bytes as are free
1639 * in the produce queue, so we don't need to ensure len is
1640 * smaller than the queue size. It is the caller's
1641 * responsibility to check how many bytes we were able to send.
1644 written = transport->stream_enqueue(
1646 len - total_written);
1652 total_written += written;
1654 err = transport->notify_send_post_enqueue(
1655 vsk, written, &send_data);
1662 if (total_written > 0)
1663 err = total_written;
1671 vsock_stream_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
1675 struct vsock_sock *vsk;
1680 struct vsock_transport_recv_notify_data recv_data;
1690 if (sk->sk_state != TCP_ESTABLISHED) {
1691 /* Recvmsg is supposed to return 0 if a peer performs an
1692 * orderly shutdown. Differentiate between that case and when a
1693 * peer has not connected or a local shutdown occured with the
1696 if (sock_flag(sk, SOCK_DONE))
1704 if (flags & MSG_OOB) {
1709 /* We don't check peer_shutdown flag here since peer may actually shut
1710 * down, but there can be data in the queue that a local socket can
1713 if (sk->sk_shutdown & RCV_SHUTDOWN) {
1718 /* It is valid on Linux to pass in a zero-length receive buffer. This
1719 * is not an error. We may as well bail out now.
1726 /* We must not copy less than target bytes into the user's buffer
1727 * before returning successfully, so we wait for the consume queue to
1728 * have that much data to consume before dequeueing. Note that this
1729 * makes it impossible to handle cases where target is greater than the
1732 target = sock_rcvlowat(sk, flags & MSG_WAITALL, len);
1733 if (target >= transport->stream_rcvhiwat(vsk)) {
1737 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1740 err = transport->notify_recv_init(vsk, target, &recv_data);
1748 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1749 ready = vsock_stream_has_data(vsk);
1752 if (sk->sk_err != 0 ||
1753 (sk->sk_shutdown & RCV_SHUTDOWN) ||
1754 (vsk->peer_shutdown & SEND_SHUTDOWN)) {
1755 finish_wait(sk_sleep(sk), &wait);
1758 /* Don't wait for non-blocking sockets. */
1761 finish_wait(sk_sleep(sk), &wait);
1765 err = transport->notify_recv_pre_block(
1766 vsk, target, &recv_data);
1768 finish_wait(sk_sleep(sk), &wait);
1772 timeout = schedule_timeout(timeout);
1775 if (signal_pending(current)) {
1776 err = sock_intr_errno(timeout);
1777 finish_wait(sk_sleep(sk), &wait);
1779 } else if (timeout == 0) {
1781 finish_wait(sk_sleep(sk), &wait);
1787 finish_wait(sk_sleep(sk), &wait);
1790 /* Invalid queue pair content. XXX This should
1791 * be changed to a connection reset in a later
1799 err = transport->notify_recv_pre_dequeue(
1800 vsk, target, &recv_data);
1804 read = transport->stream_dequeue(
1806 len - copied, flags);
1814 err = transport->notify_recv_post_dequeue(
1816 !(flags & MSG_PEEK), &recv_data);
1820 if (read >= target || flags & MSG_PEEK)
1829 else if (sk->sk_shutdown & RCV_SHUTDOWN)
1840 static const struct proto_ops vsock_stream_ops = {
1842 .owner = THIS_MODULE,
1843 .release = vsock_release,
1845 .connect = vsock_stream_connect,
1846 .socketpair = sock_no_socketpair,
1847 .accept = vsock_accept,
1848 .getname = vsock_getname,
1850 .ioctl = sock_no_ioctl,
1851 .listen = vsock_listen,
1852 .shutdown = vsock_shutdown,
1853 .setsockopt = vsock_stream_setsockopt,
1854 .getsockopt = vsock_stream_getsockopt,
1855 .sendmsg = vsock_stream_sendmsg,
1856 .recvmsg = vsock_stream_recvmsg,
1857 .mmap = sock_no_mmap,
1858 .sendpage = sock_no_sendpage,
1861 static int vsock_create(struct net *net, struct socket *sock,
1862 int protocol, int kern)
1867 if (protocol && protocol != PF_VSOCK)
1868 return -EPROTONOSUPPORT;
1870 switch (sock->type) {
1872 sock->ops = &vsock_dgram_ops;
1875 sock->ops = &vsock_stream_ops;
1878 return -ESOCKTNOSUPPORT;
1881 sock->state = SS_UNCONNECTED;
1883 return __vsock_create(net, sock, NULL, GFP_KERNEL, 0, kern) ? 0 : -ENOMEM;
1886 static const struct net_proto_family vsock_family_ops = {
1888 .create = vsock_create,
1889 .owner = THIS_MODULE,
1892 static long vsock_dev_do_ioctl(struct file *filp,
1893 unsigned int cmd, void __user *ptr)
1895 u32 __user *p = ptr;
1899 case IOCTL_VM_SOCKETS_GET_LOCAL_CID:
1900 if (put_user(transport->get_local_cid(), p) != 0)
1905 pr_err("Unknown ioctl %d\n", cmd);
1912 static long vsock_dev_ioctl(struct file *filp,
1913 unsigned int cmd, unsigned long arg)
1915 return vsock_dev_do_ioctl(filp, cmd, (void __user *)arg);
1918 #ifdef CONFIG_COMPAT
1919 static long vsock_dev_compat_ioctl(struct file *filp,
1920 unsigned int cmd, unsigned long arg)
1922 return vsock_dev_do_ioctl(filp, cmd, compat_ptr(arg));
1926 static const struct file_operations vsock_device_ops = {
1927 .owner = THIS_MODULE,
1928 .unlocked_ioctl = vsock_dev_ioctl,
1929 #ifdef CONFIG_COMPAT
1930 .compat_ioctl = vsock_dev_compat_ioctl,
1932 .open = nonseekable_open,
1935 static struct miscdevice vsock_device = {
1937 .fops = &vsock_device_ops,
1940 int __vsock_core_init(const struct vsock_transport *t, struct module *owner)
1942 int err = mutex_lock_interruptible(&vsock_register_mutex);
1952 /* Transport must be the owner of the protocol so that it can't
1953 * unload while there are open sockets.
1955 vsock_proto.owner = owner;
1958 vsock_init_tables();
1960 vsock_device.minor = MISC_DYNAMIC_MINOR;
1961 err = misc_register(&vsock_device);
1963 pr_err("Failed to register misc device\n");
1964 goto err_reset_transport;
1967 err = proto_register(&vsock_proto, 1); /* we want our slab */
1969 pr_err("Cannot register vsock protocol\n");
1970 goto err_deregister_misc;
1973 err = sock_register(&vsock_family_ops);
1975 pr_err("could not register af_vsock (%d) address family: %d\n",
1977 goto err_unregister_proto;
1980 mutex_unlock(&vsock_register_mutex);
1983 err_unregister_proto:
1984 proto_unregister(&vsock_proto);
1985 err_deregister_misc:
1986 misc_deregister(&vsock_device);
1987 err_reset_transport:
1990 mutex_unlock(&vsock_register_mutex);
1993 EXPORT_SYMBOL_GPL(__vsock_core_init);
1995 void vsock_core_exit(void)
1997 mutex_lock(&vsock_register_mutex);
1999 misc_deregister(&vsock_device);
2000 sock_unregister(AF_VSOCK);
2001 proto_unregister(&vsock_proto);
2003 /* We do not want the assignment below re-ordered. */
2007 mutex_unlock(&vsock_register_mutex);
2009 EXPORT_SYMBOL_GPL(vsock_core_exit);
2011 const struct vsock_transport *vsock_core_get_transport(void)
2013 /* vsock_register_mutex not taken since only the transport uses this
2014 * function and only while registered.
2018 EXPORT_SYMBOL_GPL(vsock_core_get_transport);
2020 MODULE_AUTHOR("VMware, Inc.");
2021 MODULE_DESCRIPTION("VMware Virtual Socket Family");
2022 MODULE_VERSION("1.0.2.0-k");
2023 MODULE_LICENSE("GPL v2");