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 MAX_PORT_RETRIES 24
168 #define VSOCK_HASH(addr) ((addr)->svm_port % VSOCK_HASH_SIZE)
169 #define vsock_bound_sockets(addr) (&vsock_bind_table[VSOCK_HASH(addr)])
170 #define vsock_unbound_sockets (&vsock_bind_table[VSOCK_HASH_SIZE])
172 /* XXX This can probably be implemented in a better way. */
173 #define VSOCK_CONN_HASH(src, dst) \
174 (((src)->svm_cid ^ (dst)->svm_port) % VSOCK_HASH_SIZE)
175 #define vsock_connected_sockets(src, dst) \
176 (&vsock_connected_table[VSOCK_CONN_HASH(src, dst)])
177 #define vsock_connected_sockets_vsk(vsk) \
178 vsock_connected_sockets(&(vsk)->remote_addr, &(vsk)->local_addr)
180 struct list_head vsock_bind_table[VSOCK_HASH_SIZE + 1];
181 EXPORT_SYMBOL_GPL(vsock_bind_table);
182 struct list_head vsock_connected_table[VSOCK_HASH_SIZE];
183 EXPORT_SYMBOL_GPL(vsock_connected_table);
184 DEFINE_SPINLOCK(vsock_table_lock);
185 EXPORT_SYMBOL_GPL(vsock_table_lock);
187 /* Autobind this socket to the local address if necessary. */
188 static int vsock_auto_bind(struct vsock_sock *vsk)
190 struct sock *sk = sk_vsock(vsk);
191 struct sockaddr_vm local_addr;
193 if (vsock_addr_bound(&vsk->local_addr))
195 vsock_addr_init(&local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
196 return __vsock_bind(sk, &local_addr);
199 static int __init vsock_init_tables(void)
203 for (i = 0; i < ARRAY_SIZE(vsock_bind_table); i++)
204 INIT_LIST_HEAD(&vsock_bind_table[i]);
206 for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++)
207 INIT_LIST_HEAD(&vsock_connected_table[i]);
211 static void __vsock_insert_bound(struct list_head *list,
212 struct vsock_sock *vsk)
215 list_add(&vsk->bound_table, list);
218 static void __vsock_insert_connected(struct list_head *list,
219 struct vsock_sock *vsk)
222 list_add(&vsk->connected_table, list);
225 static void __vsock_remove_bound(struct vsock_sock *vsk)
227 list_del_init(&vsk->bound_table);
231 static void __vsock_remove_connected(struct vsock_sock *vsk)
233 list_del_init(&vsk->connected_table);
237 static struct sock *__vsock_find_bound_socket(struct sockaddr_vm *addr)
239 struct vsock_sock *vsk;
241 list_for_each_entry(vsk, vsock_bound_sockets(addr), bound_table)
242 if (addr->svm_port == vsk->local_addr.svm_port)
243 return sk_vsock(vsk);
248 static struct sock *__vsock_find_connected_socket(struct sockaddr_vm *src,
249 struct sockaddr_vm *dst)
251 struct vsock_sock *vsk;
253 list_for_each_entry(vsk, vsock_connected_sockets(src, dst),
255 if (vsock_addr_equals_addr(src, &vsk->remote_addr) &&
256 dst->svm_port == vsk->local_addr.svm_port) {
257 return sk_vsock(vsk);
264 static void vsock_insert_unbound(struct vsock_sock *vsk)
266 spin_lock_bh(&vsock_table_lock);
267 __vsock_insert_bound(vsock_unbound_sockets, vsk);
268 spin_unlock_bh(&vsock_table_lock);
271 void vsock_insert_connected(struct vsock_sock *vsk)
273 struct list_head *list = vsock_connected_sockets(
274 &vsk->remote_addr, &vsk->local_addr);
276 spin_lock_bh(&vsock_table_lock);
277 __vsock_insert_connected(list, vsk);
278 spin_unlock_bh(&vsock_table_lock);
280 EXPORT_SYMBOL_GPL(vsock_insert_connected);
282 void vsock_remove_bound(struct vsock_sock *vsk)
284 spin_lock_bh(&vsock_table_lock);
285 if (__vsock_in_bound_table(vsk))
286 __vsock_remove_bound(vsk);
287 spin_unlock_bh(&vsock_table_lock);
289 EXPORT_SYMBOL_GPL(vsock_remove_bound);
291 void vsock_remove_connected(struct vsock_sock *vsk)
293 spin_lock_bh(&vsock_table_lock);
294 if (__vsock_in_connected_table(vsk))
295 __vsock_remove_connected(vsk);
296 spin_unlock_bh(&vsock_table_lock);
298 EXPORT_SYMBOL_GPL(vsock_remove_connected);
300 struct sock *vsock_find_bound_socket(struct sockaddr_vm *addr)
304 spin_lock_bh(&vsock_table_lock);
305 sk = __vsock_find_bound_socket(addr);
309 spin_unlock_bh(&vsock_table_lock);
313 EXPORT_SYMBOL_GPL(vsock_find_bound_socket);
315 struct sock *vsock_find_connected_socket(struct sockaddr_vm *src,
316 struct sockaddr_vm *dst)
320 spin_lock_bh(&vsock_table_lock);
321 sk = __vsock_find_connected_socket(src, dst);
325 spin_unlock_bh(&vsock_table_lock);
329 EXPORT_SYMBOL_GPL(vsock_find_connected_socket);
331 void vsock_remove_sock(struct vsock_sock *vsk)
333 vsock_remove_bound(vsk);
334 vsock_remove_connected(vsk);
336 EXPORT_SYMBOL_GPL(vsock_remove_sock);
338 void vsock_for_each_connected_socket(void (*fn)(struct sock *sk))
342 spin_lock_bh(&vsock_table_lock);
344 for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) {
345 struct vsock_sock *vsk;
346 list_for_each_entry(vsk, &vsock_connected_table[i],
351 spin_unlock_bh(&vsock_table_lock);
353 EXPORT_SYMBOL_GPL(vsock_for_each_connected_socket);
355 void vsock_add_pending(struct sock *listener, struct sock *pending)
357 struct vsock_sock *vlistener;
358 struct vsock_sock *vpending;
360 vlistener = vsock_sk(listener);
361 vpending = vsock_sk(pending);
365 list_add_tail(&vpending->pending_links, &vlistener->pending_links);
367 EXPORT_SYMBOL_GPL(vsock_add_pending);
369 void vsock_remove_pending(struct sock *listener, struct sock *pending)
371 struct vsock_sock *vpending = vsock_sk(pending);
373 list_del_init(&vpending->pending_links);
377 EXPORT_SYMBOL_GPL(vsock_remove_pending);
379 void vsock_enqueue_accept(struct sock *listener, struct sock *connected)
381 struct vsock_sock *vlistener;
382 struct vsock_sock *vconnected;
384 vlistener = vsock_sk(listener);
385 vconnected = vsock_sk(connected);
387 sock_hold(connected);
389 list_add_tail(&vconnected->accept_queue, &vlistener->accept_queue);
391 EXPORT_SYMBOL_GPL(vsock_enqueue_accept);
393 static struct sock *vsock_dequeue_accept(struct sock *listener)
395 struct vsock_sock *vlistener;
396 struct vsock_sock *vconnected;
398 vlistener = vsock_sk(listener);
400 if (list_empty(&vlistener->accept_queue))
403 vconnected = list_entry(vlistener->accept_queue.next,
404 struct vsock_sock, accept_queue);
406 list_del_init(&vconnected->accept_queue);
408 /* The caller will need a reference on the connected socket so we let
409 * it call sock_put().
412 return sk_vsock(vconnected);
415 static bool vsock_is_accept_queue_empty(struct sock *sk)
417 struct vsock_sock *vsk = vsock_sk(sk);
418 return list_empty(&vsk->accept_queue);
421 static bool vsock_is_pending(struct sock *sk)
423 struct vsock_sock *vsk = vsock_sk(sk);
424 return !list_empty(&vsk->pending_links);
427 static int vsock_send_shutdown(struct sock *sk, int mode)
429 return transport->shutdown(vsock_sk(sk), mode);
432 static void vsock_pending_work(struct work_struct *work)
435 struct sock *listener;
436 struct vsock_sock *vsk;
439 vsk = container_of(work, struct vsock_sock, pending_work.work);
441 listener = vsk->listener;
445 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
447 if (vsock_is_pending(sk)) {
448 vsock_remove_pending(listener, sk);
450 listener->sk_ack_backlog--;
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
461 /* We need to remove ourself from the global connected sockets list so
462 * incoming packets can't find this socket, and to reduce the reference
465 vsock_remove_connected(vsk);
467 sk->sk_state = TCP_CLOSE;
471 release_sock(listener);
479 /**** SOCKET OPERATIONS ****/
481 static int __vsock_bind_stream(struct vsock_sock *vsk,
482 struct sockaddr_vm *addr)
485 struct sockaddr_vm new_addr;
488 port = LAST_RESERVED_PORT + 1 +
489 prandom_u32_max(U32_MAX - LAST_RESERVED_PORT);
491 vsock_addr_init(&new_addr, addr->svm_cid, addr->svm_port);
493 if (addr->svm_port == VMADDR_PORT_ANY) {
497 for (i = 0; i < MAX_PORT_RETRIES; i++) {
498 if (port <= LAST_RESERVED_PORT)
499 port = LAST_RESERVED_PORT + 1;
501 new_addr.svm_port = port++;
503 if (!__vsock_find_bound_socket(&new_addr)) {
510 return -EADDRNOTAVAIL;
512 /* If port is in reserved range, ensure caller
513 * has necessary privileges.
515 if (addr->svm_port <= LAST_RESERVED_PORT &&
516 !capable(CAP_NET_BIND_SERVICE)) {
520 if (__vsock_find_bound_socket(&new_addr))
524 vsock_addr_init(&vsk->local_addr, new_addr.svm_cid, new_addr.svm_port);
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.
530 __vsock_remove_bound(vsk);
531 __vsock_insert_bound(vsock_bound_sockets(&vsk->local_addr), vsk);
536 static int __vsock_bind_dgram(struct vsock_sock *vsk,
537 struct sockaddr_vm *addr)
539 return transport->dgram_bind(vsk, addr);
542 static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr)
544 struct vsock_sock *vsk = vsock_sk(sk);
548 /* First ensure this socket isn't already bound. */
549 if (vsock_addr_bound(&vsk->local_addr))
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.
557 cid = transport->get_local_cid();
558 if (addr->svm_cid != cid && addr->svm_cid != VMADDR_CID_ANY)
559 return -EADDRNOTAVAIL;
561 switch (sk->sk_socket->type) {
563 spin_lock_bh(&vsock_table_lock);
564 retval = __vsock_bind_stream(vsk, addr);
565 spin_unlock_bh(&vsock_table_lock);
569 retval = __vsock_bind_dgram(vsk, addr);
580 static void vsock_connect_timeout(struct work_struct *work);
582 struct sock *__vsock_create(struct net *net,
590 struct vsock_sock *psk;
591 struct vsock_sock *vsk;
593 sk = sk_alloc(net, AF_VSOCK, priority, &vsock_proto, kern);
597 sock_init_data(sock, sk);
599 /* sk->sk_type is normally set in sock_init_data, but only if sock is
600 * non-NULL. We make sure that our sockets always have a type by
601 * setting it here if needed.
607 vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
608 vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
610 sk->sk_destruct = vsock_sk_destruct;
611 sk->sk_backlog_rcv = vsock_queue_rcv_skb;
612 sock_reset_flag(sk, SOCK_DONE);
614 INIT_LIST_HEAD(&vsk->bound_table);
615 INIT_LIST_HEAD(&vsk->connected_table);
616 vsk->listener = NULL;
617 INIT_LIST_HEAD(&vsk->pending_links);
618 INIT_LIST_HEAD(&vsk->accept_queue);
619 vsk->rejected = false;
620 vsk->sent_request = false;
621 vsk->ignore_connecting_rst = false;
622 vsk->peer_shutdown = 0;
623 INIT_DELAYED_WORK(&vsk->connect_work, vsock_connect_timeout);
624 INIT_DELAYED_WORK(&vsk->pending_work, vsock_pending_work);
626 psk = parent ? vsock_sk(parent) : NULL;
628 vsk->trusted = psk->trusted;
629 vsk->owner = get_cred(psk->owner);
630 vsk->connect_timeout = psk->connect_timeout;
631 security_sk_clone(parent, sk);
633 vsk->trusted = ns_capable_noaudit(&init_user_ns, CAP_NET_ADMIN);
634 vsk->owner = get_current_cred();
635 vsk->connect_timeout = VSOCK_DEFAULT_CONNECT_TIMEOUT;
638 if (transport->init(vsk, psk) < 0) {
644 vsock_insert_unbound(vsk);
648 EXPORT_SYMBOL_GPL(__vsock_create);
650 static void __vsock_release(struct sock *sk, int level)
654 struct sock *pending;
655 struct vsock_sock *vsk;
658 pending = NULL; /* Compiler warning. */
660 /* The release call is supposed to use lock_sock_nested()
661 * rather than lock_sock(), if a sock lock should be acquired.
663 transport->release(vsk);
665 /* When "level" is SINGLE_DEPTH_NESTING, use the nested
666 * version to avoid the warning "possible recursive locking
667 * detected". When "level" is 0, lock_sock_nested(sk, level)
668 * is the same as lock_sock(sk).
670 lock_sock_nested(sk, level);
672 sk->sk_shutdown = SHUTDOWN_MASK;
674 while ((skb = skb_dequeue(&sk->sk_receive_queue)))
677 /* Clean up any sockets that never were accepted. */
678 while ((pending = vsock_dequeue_accept(sk)) != NULL) {
679 __vsock_release(pending, SINGLE_DEPTH_NESTING);
688 static void vsock_sk_destruct(struct sock *sk)
690 struct vsock_sock *vsk = vsock_sk(sk);
692 transport->destruct(vsk);
694 /* When clearing these addresses, there's no need to set the family and
695 * possibly register the address family with the kernel.
697 vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
698 vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
700 put_cred(vsk->owner);
703 static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
707 err = sock_queue_rcv_skb(sk, skb);
714 s64 vsock_stream_has_data(struct vsock_sock *vsk)
716 return transport->stream_has_data(vsk);
718 EXPORT_SYMBOL_GPL(vsock_stream_has_data);
720 s64 vsock_stream_has_space(struct vsock_sock *vsk)
722 return transport->stream_has_space(vsk);
724 EXPORT_SYMBOL_GPL(vsock_stream_has_space);
726 static int vsock_release(struct socket *sock)
728 __vsock_release(sock->sk, 0);
730 sock->state = SS_FREE;
736 vsock_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
740 struct sockaddr_vm *vm_addr;
744 if (vsock_addr_cast(addr, addr_len, &vm_addr) != 0)
748 err = __vsock_bind(sk, vm_addr);
754 static int vsock_getname(struct socket *sock,
755 struct sockaddr *addr, int peer)
759 struct vsock_sock *vsk;
760 struct sockaddr_vm *vm_addr;
769 if (sock->state != SS_CONNECTED) {
773 vm_addr = &vsk->remote_addr;
775 vm_addr = &vsk->local_addr;
783 /* sys_getsockname() and sys_getpeername() pass us a
784 * MAX_SOCK_ADDR-sized buffer and don't set addr_len. Unfortunately
785 * that macro is defined in socket.c instead of .h, so we hardcode its
788 BUILD_BUG_ON(sizeof(*vm_addr) > 128);
789 memcpy(addr, vm_addr, sizeof(*vm_addr));
790 err = sizeof(*vm_addr);
797 static int vsock_shutdown(struct socket *sock, int mode)
802 /* User level uses SHUT_RD (0) and SHUT_WR (1), but the kernel uses
803 * RCV_SHUTDOWN (1) and SEND_SHUTDOWN (2), so we must increment mode
804 * here like the other address families do. Note also that the
805 * increment makes SHUT_RDWR (2) into RCV_SHUTDOWN | SEND_SHUTDOWN (3),
806 * which is what we want.
810 if ((mode & ~SHUTDOWN_MASK) || !mode)
813 /* If this is a STREAM socket and it is not connected then bail out
814 * immediately. If it is a DGRAM socket then we must first kick the
815 * socket so that it wakes up from any sleeping calls, for example
816 * recv(), and then afterwards return the error.
822 if (sock->state == SS_UNCONNECTED) {
824 if (sk->sk_type == SOCK_STREAM)
827 sock->state = SS_DISCONNECTING;
831 /* Receive and send shutdowns are treated alike. */
832 mode = mode & (RCV_SHUTDOWN | SEND_SHUTDOWN);
834 sk->sk_shutdown |= mode;
835 sk->sk_state_change(sk);
837 if (sk->sk_type == SOCK_STREAM) {
838 sock_reset_flag(sk, SOCK_DONE);
839 vsock_send_shutdown(sk, mode);
848 static __poll_t vsock_poll(struct file *file, struct socket *sock,
853 struct vsock_sock *vsk;
858 poll_wait(file, sk_sleep(sk), wait);
862 /* Signify that there has been an error on this socket. */
865 /* INET sockets treat local write shutdown and peer write shutdown as a
866 * case of EPOLLHUP set.
868 if ((sk->sk_shutdown == SHUTDOWN_MASK) ||
869 ((sk->sk_shutdown & SEND_SHUTDOWN) &&
870 (vsk->peer_shutdown & SEND_SHUTDOWN))) {
874 if (sk->sk_shutdown & RCV_SHUTDOWN ||
875 vsk->peer_shutdown & SEND_SHUTDOWN) {
879 if (sock->type == SOCK_DGRAM) {
880 /* For datagram sockets we can read if there is something in
881 * the queue and write as long as the socket isn't shutdown for
884 if (!skb_queue_empty_lockless(&sk->sk_receive_queue) ||
885 (sk->sk_shutdown & RCV_SHUTDOWN)) {
886 mask |= EPOLLIN | EPOLLRDNORM;
889 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
890 mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND;
892 } else if (sock->type == SOCK_STREAM) {
895 /* Listening sockets that have connections in their accept
898 if (sk->sk_state == TCP_LISTEN
899 && !vsock_is_accept_queue_empty(sk))
900 mask |= EPOLLIN | EPOLLRDNORM;
902 /* If there is something in the queue then we can read. */
903 if (transport->stream_is_active(vsk) &&
904 !(sk->sk_shutdown & RCV_SHUTDOWN)) {
905 bool data_ready_now = false;
906 int ret = transport->notify_poll_in(
907 vsk, 1, &data_ready_now);
912 mask |= EPOLLIN | EPOLLRDNORM;
917 /* Sockets whose connections have been closed, reset, or
918 * terminated should also be considered read, and we check the
919 * shutdown flag for that.
921 if (sk->sk_shutdown & RCV_SHUTDOWN ||
922 vsk->peer_shutdown & SEND_SHUTDOWN) {
923 mask |= EPOLLIN | EPOLLRDNORM;
926 /* Connected sockets that can produce data can be written. */
927 if (sk->sk_state == TCP_ESTABLISHED) {
928 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
929 bool space_avail_now = false;
930 int ret = transport->notify_poll_out(
931 vsk, 1, &space_avail_now);
936 /* Remove EPOLLWRBAND since INET
937 * sockets are not setting it.
939 mask |= EPOLLOUT | EPOLLWRNORM;
945 /* Simulate INET socket poll behaviors, which sets
946 * EPOLLOUT|EPOLLWRNORM when peer is closed and nothing to read,
947 * but local send is not shutdown.
949 if (sk->sk_state == TCP_CLOSE || sk->sk_state == TCP_CLOSING) {
950 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
951 mask |= EPOLLOUT | EPOLLWRNORM;
961 static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg,
966 struct vsock_sock *vsk;
967 struct sockaddr_vm *remote_addr;
969 if (msg->msg_flags & MSG_OOB)
972 /* For now, MSG_DONTWAIT is always assumed... */
979 err = vsock_auto_bind(vsk);
984 /* If the provided message contains an address, use that. Otherwise
985 * fall back on the socket's remote handle (if it has been connected).
988 vsock_addr_cast(msg->msg_name, msg->msg_namelen,
989 &remote_addr) == 0) {
990 /* Ensure this address is of the right type and is a valid
994 if (remote_addr->svm_cid == VMADDR_CID_ANY)
995 remote_addr->svm_cid = transport->get_local_cid();
997 if (!vsock_addr_bound(remote_addr)) {
1001 } else if (sock->state == SS_CONNECTED) {
1002 remote_addr = &vsk->remote_addr;
1004 if (remote_addr->svm_cid == VMADDR_CID_ANY)
1005 remote_addr->svm_cid = transport->get_local_cid();
1007 /* XXX Should connect() or this function ensure remote_addr is
1010 if (!vsock_addr_bound(&vsk->remote_addr)) {
1019 if (!transport->dgram_allow(remote_addr->svm_cid,
1020 remote_addr->svm_port)) {
1025 err = transport->dgram_enqueue(vsk, remote_addr, msg, len);
1032 static int vsock_dgram_connect(struct socket *sock,
1033 struct sockaddr *addr, int addr_len, int flags)
1037 struct vsock_sock *vsk;
1038 struct sockaddr_vm *remote_addr;
1043 err = vsock_addr_cast(addr, addr_len, &remote_addr);
1044 if (err == -EAFNOSUPPORT && remote_addr->svm_family == AF_UNSPEC) {
1046 vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY,
1048 sock->state = SS_UNCONNECTED;
1051 } else if (err != 0)
1056 err = vsock_auto_bind(vsk);
1060 if (!transport->dgram_allow(remote_addr->svm_cid,
1061 remote_addr->svm_port)) {
1066 memcpy(&vsk->remote_addr, remote_addr, sizeof(vsk->remote_addr));
1067 sock->state = SS_CONNECTED;
1074 static int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg,
1075 size_t len, int flags)
1077 return transport->dgram_dequeue(vsock_sk(sock->sk), msg, len, flags);
1080 static const struct proto_ops vsock_dgram_ops = {
1082 .owner = THIS_MODULE,
1083 .release = vsock_release,
1085 .connect = vsock_dgram_connect,
1086 .socketpair = sock_no_socketpair,
1087 .accept = sock_no_accept,
1088 .getname = vsock_getname,
1090 .ioctl = sock_no_ioctl,
1091 .listen = sock_no_listen,
1092 .shutdown = vsock_shutdown,
1093 .setsockopt = sock_no_setsockopt,
1094 .getsockopt = sock_no_getsockopt,
1095 .sendmsg = vsock_dgram_sendmsg,
1096 .recvmsg = vsock_dgram_recvmsg,
1097 .mmap = sock_no_mmap,
1098 .sendpage = sock_no_sendpage,
1101 static int vsock_transport_cancel_pkt(struct vsock_sock *vsk)
1103 if (!transport->cancel_pkt)
1106 return transport->cancel_pkt(vsk);
1109 static void vsock_connect_timeout(struct work_struct *work)
1112 struct vsock_sock *vsk;
1114 vsk = container_of(work, struct vsock_sock, connect_work.work);
1118 if (sk->sk_state == TCP_SYN_SENT &&
1119 (sk->sk_shutdown != SHUTDOWN_MASK)) {
1120 sk->sk_state = TCP_CLOSE;
1121 sk->sk_err = ETIMEDOUT;
1122 sk->sk_error_report(sk);
1123 vsock_transport_cancel_pkt(vsk);
1130 static int vsock_stream_connect(struct socket *sock, struct sockaddr *addr,
1131 int addr_len, int flags)
1135 struct vsock_sock *vsk;
1136 struct sockaddr_vm *remote_addr;
1146 /* XXX AF_UNSPEC should make us disconnect like AF_INET. */
1147 switch (sock->state) {
1151 case SS_DISCONNECTING:
1155 /* This continues on so we can move sock into the SS_CONNECTED
1156 * state once the connection has completed (at which point err
1157 * will be set to zero also). Otherwise, we will either wait
1158 * for the connection or return -EALREADY should this be a
1159 * non-blocking call.
1162 if (flags & O_NONBLOCK)
1166 if ((sk->sk_state == TCP_LISTEN) ||
1167 vsock_addr_cast(addr, addr_len, &remote_addr) != 0) {
1172 /* The hypervisor and well-known contexts do not have socket
1175 if (!transport->stream_allow(remote_addr->svm_cid,
1176 remote_addr->svm_port)) {
1181 /* Set the remote address that we are connecting to. */
1182 memcpy(&vsk->remote_addr, remote_addr,
1183 sizeof(vsk->remote_addr));
1185 err = vsock_auto_bind(vsk);
1189 sk->sk_state = TCP_SYN_SENT;
1191 err = transport->connect(vsk);
1195 /* Mark sock as connecting and set the error code to in
1196 * progress in case this is a non-blocking connect.
1198 sock->state = SS_CONNECTING;
1202 /* The receive path will handle all communication until we are able to
1203 * enter the connected state. Here we wait for the connection to be
1204 * completed or a notification of an error.
1206 timeout = vsk->connect_timeout;
1207 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1209 while (sk->sk_state != TCP_ESTABLISHED && sk->sk_err == 0) {
1210 if (flags & O_NONBLOCK) {
1211 /* If we're not going to block, we schedule a timeout
1212 * function to generate a timeout on the connection
1213 * attempt, in case the peer doesn't respond in a
1214 * timely manner. We hold on to the socket until the
1218 schedule_delayed_work(&vsk->connect_work, timeout);
1220 /* Skip ahead to preserve error code set above. */
1225 timeout = schedule_timeout(timeout);
1228 if (signal_pending(current)) {
1229 err = sock_intr_errno(timeout);
1230 sk->sk_state = sk->sk_state == TCP_ESTABLISHED ? TCP_CLOSING : TCP_CLOSE;
1231 sock->state = SS_UNCONNECTED;
1232 vsock_transport_cancel_pkt(vsk);
1233 vsock_remove_connected(vsk);
1235 } else if (timeout == 0) {
1237 sk->sk_state = TCP_CLOSE;
1238 sock->state = SS_UNCONNECTED;
1239 vsock_transport_cancel_pkt(vsk);
1243 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1248 sk->sk_state = TCP_CLOSE;
1249 sock->state = SS_UNCONNECTED;
1255 finish_wait(sk_sleep(sk), &wait);
1261 static int vsock_accept(struct socket *sock, struct socket *newsock, int flags,
1264 struct sock *listener;
1266 struct sock *connected;
1267 struct vsock_sock *vconnected;
1272 listener = sock->sk;
1274 lock_sock(listener);
1276 if (sock->type != SOCK_STREAM) {
1281 if (listener->sk_state != TCP_LISTEN) {
1286 /* Wait for children sockets to appear; these are the new sockets
1287 * created upon connection establishment.
1289 timeout = sock_rcvtimeo(listener, flags & O_NONBLOCK);
1290 prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1292 while ((connected = vsock_dequeue_accept(listener)) == NULL &&
1293 listener->sk_err == 0) {
1294 release_sock(listener);
1295 timeout = schedule_timeout(timeout);
1296 finish_wait(sk_sleep(listener), &wait);
1297 lock_sock(listener);
1299 if (signal_pending(current)) {
1300 err = sock_intr_errno(timeout);
1302 } else if (timeout == 0) {
1307 prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1309 finish_wait(sk_sleep(listener), &wait);
1311 if (listener->sk_err)
1312 err = -listener->sk_err;
1315 listener->sk_ack_backlog--;
1317 lock_sock_nested(connected, SINGLE_DEPTH_NESTING);
1318 vconnected = vsock_sk(connected);
1320 /* If the listener socket has received an error, then we should
1321 * reject this socket and return. Note that we simply mark the
1322 * socket rejected, drop our reference, and let the cleanup
1323 * function handle the cleanup; the fact that we found it in
1324 * the listener's accept queue guarantees that the cleanup
1325 * function hasn't run yet.
1328 vconnected->rejected = true;
1330 newsock->state = SS_CONNECTED;
1331 sock_graft(connected, newsock);
1334 release_sock(connected);
1335 sock_put(connected);
1339 release_sock(listener);
1343 static int vsock_listen(struct socket *sock, int backlog)
1347 struct vsock_sock *vsk;
1353 if (sock->type != SOCK_STREAM) {
1358 if (sock->state != SS_UNCONNECTED) {
1365 if (!vsock_addr_bound(&vsk->local_addr)) {
1370 sk->sk_max_ack_backlog = backlog;
1371 sk->sk_state = TCP_LISTEN;
1380 static int vsock_stream_setsockopt(struct socket *sock,
1383 char __user *optval,
1384 unsigned int optlen)
1388 struct vsock_sock *vsk;
1391 if (level != AF_VSOCK)
1392 return -ENOPROTOOPT;
1394 #define COPY_IN(_v) \
1396 if (optlen < sizeof(_v)) { \
1400 if (copy_from_user(&_v, optval, sizeof(_v)) != 0) { \
1413 case SO_VM_SOCKETS_BUFFER_SIZE:
1415 transport->set_buffer_size(vsk, val);
1418 case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1420 transport->set_max_buffer_size(vsk, val);
1423 case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1425 transport->set_min_buffer_size(vsk, val);
1428 case SO_VM_SOCKETS_CONNECT_TIMEOUT: {
1431 if (tv.tv_sec >= 0 && tv.tv_usec < USEC_PER_SEC &&
1432 tv.tv_sec < (MAX_SCHEDULE_TIMEOUT / HZ - 1)) {
1433 vsk->connect_timeout = tv.tv_sec * HZ +
1434 DIV_ROUND_UP(tv.tv_usec, (1000000 / HZ));
1435 if (vsk->connect_timeout == 0)
1436 vsk->connect_timeout =
1437 VSOCK_DEFAULT_CONNECT_TIMEOUT;
1457 static int vsock_stream_getsockopt(struct socket *sock,
1458 int level, int optname,
1459 char __user *optval,
1465 struct vsock_sock *vsk;
1468 if (level != AF_VSOCK)
1469 return -ENOPROTOOPT;
1471 err = get_user(len, optlen);
1475 #define COPY_OUT(_v) \
1477 if (len < sizeof(_v)) \
1481 if (copy_to_user(optval, &_v, len) != 0) \
1491 case SO_VM_SOCKETS_BUFFER_SIZE:
1492 val = transport->get_buffer_size(vsk);
1496 case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1497 val = transport->get_max_buffer_size(vsk);
1501 case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1502 val = transport->get_min_buffer_size(vsk);
1506 case SO_VM_SOCKETS_CONNECT_TIMEOUT: {
1508 tv.tv_sec = vsk->connect_timeout / HZ;
1510 (vsk->connect_timeout -
1511 tv.tv_sec * HZ) * (1000000 / HZ);
1516 return -ENOPROTOOPT;
1519 err = put_user(len, optlen);
1528 static int vsock_stream_sendmsg(struct socket *sock, struct msghdr *msg,
1532 struct vsock_sock *vsk;
1533 ssize_t total_written;
1536 struct vsock_transport_send_notify_data send_data;
1537 DEFINE_WAIT_FUNC(wait, woken_wake_function);
1544 if (msg->msg_flags & MSG_OOB)
1549 /* Callers should not provide a destination with stream sockets. */
1550 if (msg->msg_namelen) {
1551 err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP;
1555 /* Send data only if both sides are not shutdown in the direction. */
1556 if (sk->sk_shutdown & SEND_SHUTDOWN ||
1557 vsk->peer_shutdown & RCV_SHUTDOWN) {
1562 if (sk->sk_state != TCP_ESTABLISHED ||
1563 !vsock_addr_bound(&vsk->local_addr)) {
1568 if (!vsock_addr_bound(&vsk->remote_addr)) {
1569 err = -EDESTADDRREQ;
1573 /* Wait for room in the produce queue to enqueue our user's data. */
1574 timeout = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1576 err = transport->notify_send_init(vsk, &send_data);
1580 while (total_written < len) {
1583 add_wait_queue(sk_sleep(sk), &wait);
1584 while (vsock_stream_has_space(vsk) == 0 &&
1586 !(sk->sk_shutdown & SEND_SHUTDOWN) &&
1587 !(vsk->peer_shutdown & RCV_SHUTDOWN)) {
1589 /* Don't wait for non-blocking sockets. */
1592 remove_wait_queue(sk_sleep(sk), &wait);
1596 err = transport->notify_send_pre_block(vsk, &send_data);
1598 remove_wait_queue(sk_sleep(sk), &wait);
1603 timeout = wait_woken(&wait, TASK_INTERRUPTIBLE, timeout);
1605 if (signal_pending(current)) {
1606 err = sock_intr_errno(timeout);
1607 remove_wait_queue(sk_sleep(sk), &wait);
1609 } else if (timeout == 0) {
1611 remove_wait_queue(sk_sleep(sk), &wait);
1615 remove_wait_queue(sk_sleep(sk), &wait);
1617 /* These checks occur both as part of and after the loop
1618 * conditional since we need to check before and after
1624 } else if ((sk->sk_shutdown & SEND_SHUTDOWN) ||
1625 (vsk->peer_shutdown & RCV_SHUTDOWN)) {
1630 err = transport->notify_send_pre_enqueue(vsk, &send_data);
1634 /* Note that enqueue will only write as many bytes as are free
1635 * in the produce queue, so we don't need to ensure len is
1636 * smaller than the queue size. It is the caller's
1637 * responsibility to check how many bytes we were able to send.
1640 written = transport->stream_enqueue(
1642 len - total_written);
1648 total_written += written;
1650 err = transport->notify_send_post_enqueue(
1651 vsk, written, &send_data);
1658 if (total_written > 0)
1659 err = total_written;
1667 vsock_stream_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
1671 struct vsock_sock *vsk;
1676 struct vsock_transport_recv_notify_data recv_data;
1686 if (sk->sk_state != TCP_ESTABLISHED) {
1687 /* Recvmsg is supposed to return 0 if a peer performs an
1688 * orderly shutdown. Differentiate between that case and when a
1689 * peer has not connected or a local shutdown occured with the
1692 if (sock_flag(sk, SOCK_DONE))
1700 if (flags & MSG_OOB) {
1705 /* We don't check peer_shutdown flag here since peer may actually shut
1706 * down, but there can be data in the queue that a local socket can
1709 if (sk->sk_shutdown & RCV_SHUTDOWN) {
1714 /* It is valid on Linux to pass in a zero-length receive buffer. This
1715 * is not an error. We may as well bail out now.
1722 /* We must not copy less than target bytes into the user's buffer
1723 * before returning successfully, so we wait for the consume queue to
1724 * have that much data to consume before dequeueing. Note that this
1725 * makes it impossible to handle cases where target is greater than the
1728 target = sock_rcvlowat(sk, flags & MSG_WAITALL, len);
1729 if (target >= transport->stream_rcvhiwat(vsk)) {
1733 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1736 err = transport->notify_recv_init(vsk, target, &recv_data);
1744 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1745 ready = vsock_stream_has_data(vsk);
1748 if (sk->sk_err != 0 ||
1749 (sk->sk_shutdown & RCV_SHUTDOWN) ||
1750 (vsk->peer_shutdown & SEND_SHUTDOWN)) {
1751 finish_wait(sk_sleep(sk), &wait);
1754 /* Don't wait for non-blocking sockets. */
1757 finish_wait(sk_sleep(sk), &wait);
1761 err = transport->notify_recv_pre_block(
1762 vsk, target, &recv_data);
1764 finish_wait(sk_sleep(sk), &wait);
1768 timeout = schedule_timeout(timeout);
1771 if (signal_pending(current)) {
1772 err = sock_intr_errno(timeout);
1773 finish_wait(sk_sleep(sk), &wait);
1775 } else if (timeout == 0) {
1777 finish_wait(sk_sleep(sk), &wait);
1783 finish_wait(sk_sleep(sk), &wait);
1786 /* Invalid queue pair content. XXX This should
1787 * be changed to a connection reset in a later
1795 err = transport->notify_recv_pre_dequeue(
1796 vsk, target, &recv_data);
1800 read = transport->stream_dequeue(
1802 len - copied, flags);
1810 err = transport->notify_recv_post_dequeue(
1812 !(flags & MSG_PEEK), &recv_data);
1816 if (read >= target || flags & MSG_PEEK)
1825 else if (sk->sk_shutdown & RCV_SHUTDOWN)
1836 static const struct proto_ops vsock_stream_ops = {
1838 .owner = THIS_MODULE,
1839 .release = vsock_release,
1841 .connect = vsock_stream_connect,
1842 .socketpair = sock_no_socketpair,
1843 .accept = vsock_accept,
1844 .getname = vsock_getname,
1846 .ioctl = sock_no_ioctl,
1847 .listen = vsock_listen,
1848 .shutdown = vsock_shutdown,
1849 .setsockopt = vsock_stream_setsockopt,
1850 .getsockopt = vsock_stream_getsockopt,
1851 .sendmsg = vsock_stream_sendmsg,
1852 .recvmsg = vsock_stream_recvmsg,
1853 .mmap = sock_no_mmap,
1854 .sendpage = sock_no_sendpage,
1857 static int vsock_create(struct net *net, struct socket *sock,
1858 int protocol, int kern)
1863 if (protocol && protocol != PF_VSOCK)
1864 return -EPROTONOSUPPORT;
1866 switch (sock->type) {
1868 sock->ops = &vsock_dgram_ops;
1871 sock->ops = &vsock_stream_ops;
1874 return -ESOCKTNOSUPPORT;
1877 sock->state = SS_UNCONNECTED;
1879 return __vsock_create(net, sock, NULL, GFP_KERNEL, 0, kern) ? 0 : -ENOMEM;
1882 static const struct net_proto_family vsock_family_ops = {
1884 .create = vsock_create,
1885 .owner = THIS_MODULE,
1888 static long vsock_dev_do_ioctl(struct file *filp,
1889 unsigned int cmd, void __user *ptr)
1891 u32 __user *p = ptr;
1895 case IOCTL_VM_SOCKETS_GET_LOCAL_CID:
1896 if (put_user(transport->get_local_cid(), p) != 0)
1901 pr_err("Unknown ioctl %d\n", cmd);
1908 static long vsock_dev_ioctl(struct file *filp,
1909 unsigned int cmd, unsigned long arg)
1911 return vsock_dev_do_ioctl(filp, cmd, (void __user *)arg);
1914 #ifdef CONFIG_COMPAT
1915 static long vsock_dev_compat_ioctl(struct file *filp,
1916 unsigned int cmd, unsigned long arg)
1918 return vsock_dev_do_ioctl(filp, cmd, compat_ptr(arg));
1922 static const struct file_operations vsock_device_ops = {
1923 .owner = THIS_MODULE,
1924 .unlocked_ioctl = vsock_dev_ioctl,
1925 #ifdef CONFIG_COMPAT
1926 .compat_ioctl = vsock_dev_compat_ioctl,
1928 .open = nonseekable_open,
1931 static struct miscdevice vsock_device = {
1933 .fops = &vsock_device_ops,
1936 int __vsock_core_init(const struct vsock_transport *t, struct module *owner)
1938 int err = mutex_lock_interruptible(&vsock_register_mutex);
1948 /* Transport must be the owner of the protocol so that it can't
1949 * unload while there are open sockets.
1951 vsock_proto.owner = owner;
1954 vsock_device.minor = MISC_DYNAMIC_MINOR;
1955 err = misc_register(&vsock_device);
1957 pr_err("Failed to register misc device\n");
1958 goto err_reset_transport;
1961 err = proto_register(&vsock_proto, 1); /* we want our slab */
1963 pr_err("Cannot register vsock protocol\n");
1964 goto err_deregister_misc;
1967 err = sock_register(&vsock_family_ops);
1969 pr_err("could not register af_vsock (%d) address family: %d\n",
1971 goto err_unregister_proto;
1974 mutex_unlock(&vsock_register_mutex);
1977 err_unregister_proto:
1978 proto_unregister(&vsock_proto);
1979 err_deregister_misc:
1980 misc_deregister(&vsock_device);
1981 err_reset_transport:
1984 mutex_unlock(&vsock_register_mutex);
1987 EXPORT_SYMBOL_GPL(__vsock_core_init);
1989 void vsock_core_exit(void)
1991 mutex_lock(&vsock_register_mutex);
1993 misc_deregister(&vsock_device);
1994 sock_unregister(AF_VSOCK);
1995 proto_unregister(&vsock_proto);
1997 /* We do not want the assignment below re-ordered. */
2001 mutex_unlock(&vsock_register_mutex);
2003 EXPORT_SYMBOL_GPL(vsock_core_exit);
2005 const struct vsock_transport *vsock_core_get_transport(void)
2007 /* vsock_register_mutex not taken since only the transport uses this
2008 * function and only while registered.
2012 EXPORT_SYMBOL_GPL(vsock_core_get_transport);
2014 static void __exit vsock_exit(void)
2016 /* Do nothing. This function makes this module removable. */
2019 module_init(vsock_init_tables);
2020 module_exit(vsock_exit);
2022 MODULE_AUTHOR("VMware, Inc.");
2023 MODULE_DESCRIPTION("VMware Virtual Socket Family");
2024 MODULE_VERSION("1.0.2.0-k");
2025 MODULE_LICENSE("GPL v2");