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 VSOCK_SS_LISTEN state. When a
40 * connection request is received (the second kind of socket mentioned above),
41 * we create a new socket and refer to it as a pending socket. These pending
42 * sockets are placed on the pending connection list of the listener socket.
43 * When future packets are received for the address the listener socket is
44 * bound to, we check if the source of the packet is from one that has an
45 * existing pending connection. If it does, we process the packet for the
46 * pending socket. When that socket reaches the connected state, it is removed
47 * from the listener socket's pending list and enqueued in the listener
48 * socket's accept queue. Callers of accept(2) will accept connected sockets
49 * from the listener socket's accept queue. If the socket cannot be accepted
50 * for some reason then it is marked rejected. Once the connection is
51 * accepted, it is owned by the user process and the responsibility for cleanup
52 * falls with that user process.
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.
87 #include <linux/types.h>
88 #include <linux/bitops.h>
89 #include <linux/cred.h>
90 #include <linux/init.h>
92 #include <linux/kernel.h>
93 #include <linux/sched/signal.h>
94 #include <linux/kmod.h>
95 #include <linux/list.h>
96 #include <linux/miscdevice.h>
97 #include <linux/module.h>
98 #include <linux/mutex.h>
99 #include <linux/net.h>
100 #include <linux/poll.h>
101 #include <linux/skbuff.h>
102 #include <linux/smp.h>
103 #include <linux/socket.h>
104 #include <linux/stddef.h>
105 #include <linux/unistd.h>
106 #include <linux/wait.h>
107 #include <linux/workqueue.h>
108 #include <net/sock.h>
109 #include <net/af_vsock.h>
111 static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr);
112 static void vsock_sk_destruct(struct sock *sk);
113 static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
115 /* Protocol family. */
116 static struct proto vsock_proto = {
118 .owner = THIS_MODULE,
119 .obj_size = sizeof(struct vsock_sock),
122 /* The default peer timeout indicates how long we will wait for a peer response
123 * to a control message.
125 #define VSOCK_DEFAULT_CONNECT_TIMEOUT (2 * HZ)
127 static const struct vsock_transport *transport;
128 static DEFINE_MUTEX(vsock_register_mutex);
132 /* Get the ID of the local context. This is transport dependent. */
134 int vm_sockets_get_local_cid(void)
136 return transport->get_local_cid();
138 EXPORT_SYMBOL_GPL(vm_sockets_get_local_cid);
142 /* Each bound VSocket is stored in the bind hash table and each connected
143 * VSocket is stored in the connected hash table.
145 * Unbound sockets are all put on the same list attached to the end of the hash
146 * table (vsock_unbound_sockets). Bound sockets are added to the hash table in
147 * the bucket that their local address hashes to (vsock_bound_sockets(addr)
148 * represents the list that addr hashes to).
150 * Specifically, we initialize the vsock_bind_table array to a size of
151 * VSOCK_HASH_SIZE + 1 so that vsock_bind_table[0] through
152 * vsock_bind_table[VSOCK_HASH_SIZE - 1] are for bound sockets and
153 * vsock_bind_table[VSOCK_HASH_SIZE] is for unbound sockets. The hash function
154 * mods with VSOCK_HASH_SIZE to ensure this.
156 #define VSOCK_HASH_SIZE 251
157 #define MAX_PORT_RETRIES 24
159 #define VSOCK_HASH(addr) ((addr)->svm_port % VSOCK_HASH_SIZE)
160 #define vsock_bound_sockets(addr) (&vsock_bind_table[VSOCK_HASH(addr)])
161 #define vsock_unbound_sockets (&vsock_bind_table[VSOCK_HASH_SIZE])
163 /* XXX This can probably be implemented in a better way. */
164 #define VSOCK_CONN_HASH(src, dst) \
165 (((src)->svm_cid ^ (dst)->svm_port) % VSOCK_HASH_SIZE)
166 #define vsock_connected_sockets(src, dst) \
167 (&vsock_connected_table[VSOCK_CONN_HASH(src, dst)])
168 #define vsock_connected_sockets_vsk(vsk) \
169 vsock_connected_sockets(&(vsk)->remote_addr, &(vsk)->local_addr)
171 static struct list_head vsock_bind_table[VSOCK_HASH_SIZE + 1];
172 static struct list_head vsock_connected_table[VSOCK_HASH_SIZE];
173 static DEFINE_SPINLOCK(vsock_table_lock);
175 /* Autobind this socket to the local address if necessary. */
176 static int vsock_auto_bind(struct vsock_sock *vsk)
178 struct sock *sk = sk_vsock(vsk);
179 struct sockaddr_vm local_addr;
181 if (vsock_addr_bound(&vsk->local_addr))
183 vsock_addr_init(&local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
184 return __vsock_bind(sk, &local_addr);
187 static void vsock_init_tables(void)
191 for (i = 0; i < ARRAY_SIZE(vsock_bind_table); i++)
192 INIT_LIST_HEAD(&vsock_bind_table[i]);
194 for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++)
195 INIT_LIST_HEAD(&vsock_connected_table[i]);
198 static void __vsock_insert_bound(struct list_head *list,
199 struct vsock_sock *vsk)
202 list_add(&vsk->bound_table, list);
205 static void __vsock_insert_connected(struct list_head *list,
206 struct vsock_sock *vsk)
209 list_add(&vsk->connected_table, list);
212 static void __vsock_remove_bound(struct vsock_sock *vsk)
214 list_del_init(&vsk->bound_table);
218 static void __vsock_remove_connected(struct vsock_sock *vsk)
220 list_del_init(&vsk->connected_table);
224 static struct sock *__vsock_find_bound_socket(struct sockaddr_vm *addr)
226 struct vsock_sock *vsk;
228 list_for_each_entry(vsk, vsock_bound_sockets(addr), bound_table)
229 if (addr->svm_port == vsk->local_addr.svm_port)
230 return sk_vsock(vsk);
235 static struct sock *__vsock_find_connected_socket(struct sockaddr_vm *src,
236 struct sockaddr_vm *dst)
238 struct vsock_sock *vsk;
240 list_for_each_entry(vsk, vsock_connected_sockets(src, dst),
242 if (vsock_addr_equals_addr(src, &vsk->remote_addr) &&
243 dst->svm_port == vsk->local_addr.svm_port) {
244 return sk_vsock(vsk);
251 static bool __vsock_in_bound_table(struct vsock_sock *vsk)
253 return !list_empty(&vsk->bound_table);
256 static bool __vsock_in_connected_table(struct vsock_sock *vsk)
258 return !list_empty(&vsk->connected_table);
261 static void vsock_insert_unbound(struct vsock_sock *vsk)
263 spin_lock_bh(&vsock_table_lock);
264 __vsock_insert_bound(vsock_unbound_sockets, vsk);
265 spin_unlock_bh(&vsock_table_lock);
268 void vsock_insert_connected(struct vsock_sock *vsk)
270 struct list_head *list = vsock_connected_sockets(
271 &vsk->remote_addr, &vsk->local_addr);
273 spin_lock_bh(&vsock_table_lock);
274 __vsock_insert_connected(list, vsk);
275 spin_unlock_bh(&vsock_table_lock);
277 EXPORT_SYMBOL_GPL(vsock_insert_connected);
279 void vsock_remove_bound(struct vsock_sock *vsk)
281 spin_lock_bh(&vsock_table_lock);
282 __vsock_remove_bound(vsk);
283 spin_unlock_bh(&vsock_table_lock);
285 EXPORT_SYMBOL_GPL(vsock_remove_bound);
287 void vsock_remove_connected(struct vsock_sock *vsk)
289 spin_lock_bh(&vsock_table_lock);
290 __vsock_remove_connected(vsk);
291 spin_unlock_bh(&vsock_table_lock);
293 EXPORT_SYMBOL_GPL(vsock_remove_connected);
295 struct sock *vsock_find_bound_socket(struct sockaddr_vm *addr)
299 spin_lock_bh(&vsock_table_lock);
300 sk = __vsock_find_bound_socket(addr);
304 spin_unlock_bh(&vsock_table_lock);
308 EXPORT_SYMBOL_GPL(vsock_find_bound_socket);
310 struct sock *vsock_find_connected_socket(struct sockaddr_vm *src,
311 struct sockaddr_vm *dst)
315 spin_lock_bh(&vsock_table_lock);
316 sk = __vsock_find_connected_socket(src, dst);
320 spin_unlock_bh(&vsock_table_lock);
324 EXPORT_SYMBOL_GPL(vsock_find_connected_socket);
326 static bool vsock_in_bound_table(struct vsock_sock *vsk)
330 spin_lock_bh(&vsock_table_lock);
331 ret = __vsock_in_bound_table(vsk);
332 spin_unlock_bh(&vsock_table_lock);
337 static bool vsock_in_connected_table(struct vsock_sock *vsk)
341 spin_lock_bh(&vsock_table_lock);
342 ret = __vsock_in_connected_table(vsk);
343 spin_unlock_bh(&vsock_table_lock);
348 void vsock_remove_sock(struct vsock_sock *vsk)
350 if (vsock_in_bound_table(vsk))
351 vsock_remove_bound(vsk);
353 if (vsock_in_connected_table(vsk))
354 vsock_remove_connected(vsk);
356 EXPORT_SYMBOL_GPL(vsock_remove_sock);
358 void vsock_for_each_connected_socket(void (*fn)(struct sock *sk))
362 spin_lock_bh(&vsock_table_lock);
364 for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) {
365 struct vsock_sock *vsk;
366 list_for_each_entry(vsk, &vsock_connected_table[i],
371 spin_unlock_bh(&vsock_table_lock);
373 EXPORT_SYMBOL_GPL(vsock_for_each_connected_socket);
375 void vsock_add_pending(struct sock *listener, struct sock *pending)
377 struct vsock_sock *vlistener;
378 struct vsock_sock *vpending;
380 vlistener = vsock_sk(listener);
381 vpending = vsock_sk(pending);
385 list_add_tail(&vpending->pending_links, &vlistener->pending_links);
387 EXPORT_SYMBOL_GPL(vsock_add_pending);
389 void vsock_remove_pending(struct sock *listener, struct sock *pending)
391 struct vsock_sock *vpending = vsock_sk(pending);
393 list_del_init(&vpending->pending_links);
397 EXPORT_SYMBOL_GPL(vsock_remove_pending);
399 void vsock_enqueue_accept(struct sock *listener, struct sock *connected)
401 struct vsock_sock *vlistener;
402 struct vsock_sock *vconnected;
404 vlistener = vsock_sk(listener);
405 vconnected = vsock_sk(connected);
407 sock_hold(connected);
409 list_add_tail(&vconnected->accept_queue, &vlistener->accept_queue);
411 EXPORT_SYMBOL_GPL(vsock_enqueue_accept);
413 static struct sock *vsock_dequeue_accept(struct sock *listener)
415 struct vsock_sock *vlistener;
416 struct vsock_sock *vconnected;
418 vlistener = vsock_sk(listener);
420 if (list_empty(&vlistener->accept_queue))
423 vconnected = list_entry(vlistener->accept_queue.next,
424 struct vsock_sock, accept_queue);
426 list_del_init(&vconnected->accept_queue);
428 /* The caller will need a reference on the connected socket so we let
429 * it call sock_put().
432 return sk_vsock(vconnected);
435 static bool vsock_is_accept_queue_empty(struct sock *sk)
437 struct vsock_sock *vsk = vsock_sk(sk);
438 return list_empty(&vsk->accept_queue);
441 static bool vsock_is_pending(struct sock *sk)
443 struct vsock_sock *vsk = vsock_sk(sk);
444 return !list_empty(&vsk->pending_links);
447 static int vsock_send_shutdown(struct sock *sk, int mode)
449 return transport->shutdown(vsock_sk(sk), mode);
452 void vsock_pending_work(struct work_struct *work)
455 struct sock *listener;
456 struct vsock_sock *vsk;
459 vsk = container_of(work, struct vsock_sock, dwork.work);
461 listener = vsk->listener;
465 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
467 if (vsock_is_pending(sk)) {
468 vsock_remove_pending(listener, sk);
470 listener->sk_ack_backlog--;
471 } else if (!vsk->rejected) {
472 /* We are not on the pending list and accept() did not reject
473 * us, so we must have been accepted by our user process. We
474 * just need to drop our references to the sockets and be on
481 /* We need to remove ourself from the global connected sockets list so
482 * incoming packets can't find this socket, and to reduce the reference
485 if (vsock_in_connected_table(vsk))
486 vsock_remove_connected(vsk);
488 sk->sk_state = SS_FREE;
492 release_sock(listener);
499 EXPORT_SYMBOL_GPL(vsock_pending_work);
501 /**** SOCKET OPERATIONS ****/
503 static int __vsock_bind_stream(struct vsock_sock *vsk,
504 struct sockaddr_vm *addr)
506 static u32 port = LAST_RESERVED_PORT + 1;
507 struct sockaddr_vm new_addr;
509 vsock_addr_init(&new_addr, addr->svm_cid, addr->svm_port);
511 if (addr->svm_port == VMADDR_PORT_ANY) {
515 for (i = 0; i < MAX_PORT_RETRIES; i++) {
516 if (port <= LAST_RESERVED_PORT)
517 port = LAST_RESERVED_PORT + 1;
519 new_addr.svm_port = port++;
521 if (!__vsock_find_bound_socket(&new_addr)) {
528 return -EADDRNOTAVAIL;
530 /* If port is in reserved range, ensure caller
531 * has necessary privileges.
533 if (addr->svm_port <= LAST_RESERVED_PORT &&
534 !capable(CAP_NET_BIND_SERVICE)) {
538 if (__vsock_find_bound_socket(&new_addr))
542 vsock_addr_init(&vsk->local_addr, new_addr.svm_cid, new_addr.svm_port);
544 /* Remove stream sockets from the unbound list and add them to the hash
545 * table for easy lookup by its address. The unbound list is simply an
546 * extra entry at the end of the hash table, a trick used by AF_UNIX.
548 __vsock_remove_bound(vsk);
549 __vsock_insert_bound(vsock_bound_sockets(&vsk->local_addr), vsk);
554 static int __vsock_bind_dgram(struct vsock_sock *vsk,
555 struct sockaddr_vm *addr)
557 return transport->dgram_bind(vsk, addr);
560 static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr)
562 struct vsock_sock *vsk = vsock_sk(sk);
566 /* First ensure this socket isn't already bound. */
567 if (vsock_addr_bound(&vsk->local_addr))
570 /* Now bind to the provided address or select appropriate values if
571 * none are provided (VMADDR_CID_ANY and VMADDR_PORT_ANY). Note that
572 * like AF_INET prevents binding to a non-local IP address (in most
573 * cases), we only allow binding to the local CID.
575 cid = transport->get_local_cid();
576 if (addr->svm_cid != cid && addr->svm_cid != VMADDR_CID_ANY)
577 return -EADDRNOTAVAIL;
579 switch (sk->sk_socket->type) {
581 spin_lock_bh(&vsock_table_lock);
582 retval = __vsock_bind_stream(vsk, addr);
583 spin_unlock_bh(&vsock_table_lock);
587 retval = __vsock_bind_dgram(vsk, addr);
598 struct sock *__vsock_create(struct net *net,
606 struct vsock_sock *psk;
607 struct vsock_sock *vsk;
609 sk = sk_alloc(net, AF_VSOCK, priority, &vsock_proto, kern);
613 sock_init_data(sock, sk);
615 /* sk->sk_type is normally set in sock_init_data, but only if sock is
616 * non-NULL. We make sure that our sockets always have a type by
617 * setting it here if needed.
623 vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
624 vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
626 sk->sk_destruct = vsock_sk_destruct;
627 sk->sk_backlog_rcv = vsock_queue_rcv_skb;
629 sock_reset_flag(sk, SOCK_DONE);
631 INIT_LIST_HEAD(&vsk->bound_table);
632 INIT_LIST_HEAD(&vsk->connected_table);
633 vsk->listener = NULL;
634 INIT_LIST_HEAD(&vsk->pending_links);
635 INIT_LIST_HEAD(&vsk->accept_queue);
636 vsk->rejected = false;
637 vsk->sent_request = false;
638 vsk->ignore_connecting_rst = false;
639 vsk->peer_shutdown = 0;
641 psk = parent ? vsock_sk(parent) : NULL;
643 vsk->trusted = psk->trusted;
644 vsk->owner = get_cred(psk->owner);
645 vsk->connect_timeout = psk->connect_timeout;
647 vsk->trusted = capable(CAP_NET_ADMIN);
648 vsk->owner = get_current_cred();
649 vsk->connect_timeout = VSOCK_DEFAULT_CONNECT_TIMEOUT;
652 if (transport->init(vsk, psk) < 0) {
658 vsock_insert_unbound(vsk);
662 EXPORT_SYMBOL_GPL(__vsock_create);
664 static void __vsock_release(struct sock *sk)
668 struct sock *pending;
669 struct vsock_sock *vsk;
672 pending = NULL; /* Compiler warning. */
674 transport->release(vsk);
678 sk->sk_shutdown = SHUTDOWN_MASK;
680 while ((skb = skb_dequeue(&sk->sk_receive_queue)))
683 /* Clean up any sockets that never were accepted. */
684 while ((pending = vsock_dequeue_accept(sk)) != NULL) {
685 __vsock_release(pending);
694 static void vsock_sk_destruct(struct sock *sk)
696 struct vsock_sock *vsk = vsock_sk(sk);
698 transport->destruct(vsk);
700 /* When clearing these addresses, there's no need to set the family and
701 * possibly register the address family with the kernel.
703 vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
704 vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
706 put_cred(vsk->owner);
709 static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
713 err = sock_queue_rcv_skb(sk, skb);
720 s64 vsock_stream_has_data(struct vsock_sock *vsk)
722 return transport->stream_has_data(vsk);
724 EXPORT_SYMBOL_GPL(vsock_stream_has_data);
726 s64 vsock_stream_has_space(struct vsock_sock *vsk)
728 return transport->stream_has_space(vsk);
730 EXPORT_SYMBOL_GPL(vsock_stream_has_space);
732 static int vsock_release(struct socket *sock)
734 __vsock_release(sock->sk);
736 sock->state = SS_FREE;
742 vsock_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
746 struct sockaddr_vm *vm_addr;
750 if (vsock_addr_cast(addr, addr_len, &vm_addr) != 0)
754 err = __vsock_bind(sk, vm_addr);
760 static int vsock_getname(struct socket *sock,
761 struct sockaddr *addr, int *addr_len, int peer)
765 struct vsock_sock *vsk;
766 struct sockaddr_vm *vm_addr;
775 if (sock->state != SS_CONNECTED) {
779 vm_addr = &vsk->remote_addr;
781 vm_addr = &vsk->local_addr;
789 /* sys_getsockname() and sys_getpeername() pass us a
790 * MAX_SOCK_ADDR-sized buffer and don't set addr_len. Unfortunately
791 * that macro is defined in socket.c instead of .h, so we hardcode its
794 BUILD_BUG_ON(sizeof(*vm_addr) > 128);
795 memcpy(addr, vm_addr, sizeof(*vm_addr));
796 *addr_len = sizeof(*vm_addr);
803 static int vsock_shutdown(struct socket *sock, int mode)
808 /* User level uses SHUT_RD (0) and SHUT_WR (1), but the kernel uses
809 * RCV_SHUTDOWN (1) and SEND_SHUTDOWN (2), so we must increment mode
810 * here like the other address families do. Note also that the
811 * increment makes SHUT_RDWR (2) into RCV_SHUTDOWN | SEND_SHUTDOWN (3),
812 * which is what we want.
816 if ((mode & ~SHUTDOWN_MASK) || !mode)
819 /* If this is a STREAM socket and it is not connected then bail out
820 * immediately. If it is a DGRAM socket then we must first kick the
821 * socket so that it wakes up from any sleeping calls, for example
822 * recv(), and then afterwards return the error.
826 if (sock->state == SS_UNCONNECTED) {
828 if (sk->sk_type == SOCK_STREAM)
831 sock->state = SS_DISCONNECTING;
835 /* Receive and send shutdowns are treated alike. */
836 mode = mode & (RCV_SHUTDOWN | SEND_SHUTDOWN);
839 sk->sk_shutdown |= mode;
840 sk->sk_state_change(sk);
843 if (sk->sk_type == SOCK_STREAM) {
844 sock_reset_flag(sk, SOCK_DONE);
845 vsock_send_shutdown(sk, mode);
852 static unsigned int vsock_poll(struct file *file, struct socket *sock,
857 struct vsock_sock *vsk;
862 poll_wait(file, sk_sleep(sk), wait);
866 /* Signify that there has been an error on this socket. */
869 /* INET sockets treat local write shutdown and peer write shutdown as a
870 * case of POLLHUP set.
872 if ((sk->sk_shutdown == SHUTDOWN_MASK) ||
873 ((sk->sk_shutdown & SEND_SHUTDOWN) &&
874 (vsk->peer_shutdown & SEND_SHUTDOWN))) {
878 if (sk->sk_shutdown & RCV_SHUTDOWN ||
879 vsk->peer_shutdown & SEND_SHUTDOWN) {
883 if (sock->type == SOCK_DGRAM) {
884 /* For datagram sockets we can read if there is something in
885 * the queue and write as long as the socket isn't shutdown for
888 if (!skb_queue_empty(&sk->sk_receive_queue) ||
889 (sk->sk_shutdown & RCV_SHUTDOWN)) {
890 mask |= POLLIN | POLLRDNORM;
893 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
894 mask |= POLLOUT | POLLWRNORM | POLLWRBAND;
896 } else if (sock->type == SOCK_STREAM) {
899 /* Listening sockets that have connections in their accept
902 if (sk->sk_state == VSOCK_SS_LISTEN
903 && !vsock_is_accept_queue_empty(sk))
904 mask |= POLLIN | POLLRDNORM;
906 /* If there is something in the queue then we can read. */
907 if (transport->stream_is_active(vsk) &&
908 !(sk->sk_shutdown & RCV_SHUTDOWN)) {
909 bool data_ready_now = false;
910 int ret = transport->notify_poll_in(
911 vsk, 1, &data_ready_now);
916 mask |= POLLIN | POLLRDNORM;
921 /* Sockets whose connections have been closed, reset, or
922 * terminated should also be considered read, and we check the
923 * shutdown flag for that.
925 if (sk->sk_shutdown & RCV_SHUTDOWN ||
926 vsk->peer_shutdown & SEND_SHUTDOWN) {
927 mask |= POLLIN | POLLRDNORM;
930 /* Connected sockets that can produce data can be written. */
931 if (sk->sk_state == SS_CONNECTED) {
932 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
933 bool space_avail_now = false;
934 int ret = transport->notify_poll_out(
935 vsk, 1, &space_avail_now);
940 /* Remove POLLWRBAND since INET
941 * sockets are not setting it.
943 mask |= POLLOUT | POLLWRNORM;
949 /* Simulate INET socket poll behaviors, which sets
950 * POLLOUT|POLLWRNORM when peer is closed and nothing to read,
951 * but local send is not shutdown.
953 if (sk->sk_state == SS_UNCONNECTED) {
954 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
955 mask |= POLLOUT | POLLWRNORM;
965 static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg,
970 struct vsock_sock *vsk;
971 struct sockaddr_vm *remote_addr;
973 if (msg->msg_flags & MSG_OOB)
976 /* For now, MSG_DONTWAIT is always assumed... */
983 err = vsock_auto_bind(vsk);
988 /* If the provided message contains an address, use that. Otherwise
989 * fall back on the socket's remote handle (if it has been connected).
992 vsock_addr_cast(msg->msg_name, msg->msg_namelen,
993 &remote_addr) == 0) {
994 /* Ensure this address is of the right type and is a valid
998 if (remote_addr->svm_cid == VMADDR_CID_ANY)
999 remote_addr->svm_cid = transport->get_local_cid();
1001 if (!vsock_addr_bound(remote_addr)) {
1005 } else if (sock->state == SS_CONNECTED) {
1006 remote_addr = &vsk->remote_addr;
1008 if (remote_addr->svm_cid == VMADDR_CID_ANY)
1009 remote_addr->svm_cid = transport->get_local_cid();
1011 /* XXX Should connect() or this function ensure remote_addr is
1014 if (!vsock_addr_bound(&vsk->remote_addr)) {
1023 if (!transport->dgram_allow(remote_addr->svm_cid,
1024 remote_addr->svm_port)) {
1029 err = transport->dgram_enqueue(vsk, remote_addr, msg, len);
1036 static int vsock_dgram_connect(struct socket *sock,
1037 struct sockaddr *addr, int addr_len, int flags)
1041 struct vsock_sock *vsk;
1042 struct sockaddr_vm *remote_addr;
1047 err = vsock_addr_cast(addr, addr_len, &remote_addr);
1048 if (err == -EAFNOSUPPORT && remote_addr->svm_family == AF_UNSPEC) {
1050 vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY,
1052 sock->state = SS_UNCONNECTED;
1055 } else if (err != 0)
1060 err = vsock_auto_bind(vsk);
1064 if (!transport->dgram_allow(remote_addr->svm_cid,
1065 remote_addr->svm_port)) {
1070 memcpy(&vsk->remote_addr, remote_addr, sizeof(vsk->remote_addr));
1071 sock->state = SS_CONNECTED;
1078 static int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg,
1079 size_t len, int flags)
1081 return transport->dgram_dequeue(vsock_sk(sock->sk), msg, len, flags);
1084 static const struct proto_ops vsock_dgram_ops = {
1086 .owner = THIS_MODULE,
1087 .release = vsock_release,
1089 .connect = vsock_dgram_connect,
1090 .socketpair = sock_no_socketpair,
1091 .accept = sock_no_accept,
1092 .getname = vsock_getname,
1094 .ioctl = sock_no_ioctl,
1095 .listen = sock_no_listen,
1096 .shutdown = vsock_shutdown,
1097 .setsockopt = sock_no_setsockopt,
1098 .getsockopt = sock_no_getsockopt,
1099 .sendmsg = vsock_dgram_sendmsg,
1100 .recvmsg = vsock_dgram_recvmsg,
1101 .mmap = sock_no_mmap,
1102 .sendpage = sock_no_sendpage,
1105 static void vsock_connect_timeout(struct work_struct *work)
1108 struct vsock_sock *vsk;
1110 vsk = container_of(work, struct vsock_sock, dwork.work);
1114 if (sk->sk_state == SS_CONNECTING &&
1115 (sk->sk_shutdown != SHUTDOWN_MASK)) {
1116 sk->sk_state = SS_UNCONNECTED;
1117 sk->sk_err = ETIMEDOUT;
1118 sk->sk_error_report(sk);
1125 static int vsock_stream_connect(struct socket *sock, struct sockaddr *addr,
1126 int addr_len, int flags)
1130 struct vsock_sock *vsk;
1131 struct sockaddr_vm *remote_addr;
1141 /* XXX AF_UNSPEC should make us disconnect like AF_INET. */
1142 switch (sock->state) {
1146 case SS_DISCONNECTING:
1150 /* This continues on so we can move sock into the SS_CONNECTED
1151 * state once the connection has completed (at which point err
1152 * will be set to zero also). Otherwise, we will either wait
1153 * for the connection or return -EALREADY should this be a
1154 * non-blocking call.
1159 if ((sk->sk_state == VSOCK_SS_LISTEN) ||
1160 vsock_addr_cast(addr, addr_len, &remote_addr) != 0) {
1165 /* The hypervisor and well-known contexts do not have socket
1168 if (!transport->stream_allow(remote_addr->svm_cid,
1169 remote_addr->svm_port)) {
1174 /* Set the remote address that we are connecting to. */
1175 memcpy(&vsk->remote_addr, remote_addr,
1176 sizeof(vsk->remote_addr));
1178 err = vsock_auto_bind(vsk);
1182 sk->sk_state = SS_CONNECTING;
1184 err = transport->connect(vsk);
1188 /* Mark sock as connecting and set the error code to in
1189 * progress in case this is a non-blocking connect.
1191 sock->state = SS_CONNECTING;
1195 /* The receive path will handle all communication until we are able to
1196 * enter the connected state. Here we wait for the connection to be
1197 * completed or a notification of an error.
1199 timeout = vsk->connect_timeout;
1200 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1202 while (sk->sk_state != SS_CONNECTED && sk->sk_err == 0) {
1203 if (flags & O_NONBLOCK) {
1204 /* If we're not going to block, we schedule a timeout
1205 * function to generate a timeout on the connection
1206 * attempt, in case the peer doesn't respond in a
1207 * timely manner. We hold on to the socket until the
1211 INIT_DELAYED_WORK(&vsk->dwork,
1212 vsock_connect_timeout);
1213 schedule_delayed_work(&vsk->dwork, timeout);
1215 /* Skip ahead to preserve error code set above. */
1220 timeout = schedule_timeout(timeout);
1223 if (signal_pending(current)) {
1224 err = sock_intr_errno(timeout);
1225 sk->sk_state = SS_UNCONNECTED;
1226 sock->state = SS_UNCONNECTED;
1228 } else if (timeout == 0) {
1230 sk->sk_state = SS_UNCONNECTED;
1231 sock->state = SS_UNCONNECTED;
1235 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1240 sk->sk_state = SS_UNCONNECTED;
1241 sock->state = SS_UNCONNECTED;
1247 finish_wait(sk_sleep(sk), &wait);
1253 static int vsock_accept(struct socket *sock, struct socket *newsock, int flags,
1256 struct sock *listener;
1258 struct sock *connected;
1259 struct vsock_sock *vconnected;
1264 listener = sock->sk;
1266 lock_sock(listener);
1268 if (sock->type != SOCK_STREAM) {
1273 if (listener->sk_state != VSOCK_SS_LISTEN) {
1278 /* Wait for children sockets to appear; these are the new sockets
1279 * created upon connection establishment.
1281 timeout = sock_sndtimeo(listener, flags & O_NONBLOCK);
1282 prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1284 while ((connected = vsock_dequeue_accept(listener)) == NULL &&
1285 listener->sk_err == 0) {
1286 release_sock(listener);
1287 timeout = schedule_timeout(timeout);
1288 finish_wait(sk_sleep(listener), &wait);
1289 lock_sock(listener);
1291 if (signal_pending(current)) {
1292 err = sock_intr_errno(timeout);
1294 } else if (timeout == 0) {
1299 prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1301 finish_wait(sk_sleep(listener), &wait);
1303 if (listener->sk_err)
1304 err = -listener->sk_err;
1307 listener->sk_ack_backlog--;
1309 lock_sock_nested(connected, SINGLE_DEPTH_NESTING);
1310 vconnected = vsock_sk(connected);
1312 /* If the listener socket has received an error, then we should
1313 * reject this socket and return. Note that we simply mark the
1314 * socket rejected, drop our reference, and let the cleanup
1315 * function handle the cleanup; the fact that we found it in
1316 * the listener's accept queue guarantees that the cleanup
1317 * function hasn't run yet.
1320 vconnected->rejected = true;
1322 newsock->state = SS_CONNECTED;
1323 sock_graft(connected, newsock);
1326 release_sock(connected);
1327 sock_put(connected);
1331 release_sock(listener);
1335 static int vsock_listen(struct socket *sock, int backlog)
1339 struct vsock_sock *vsk;
1345 if (sock->type != SOCK_STREAM) {
1350 if (sock->state != SS_UNCONNECTED) {
1357 if (!vsock_addr_bound(&vsk->local_addr)) {
1362 sk->sk_max_ack_backlog = backlog;
1363 sk->sk_state = VSOCK_SS_LISTEN;
1372 static int vsock_stream_setsockopt(struct socket *sock,
1375 char __user *optval,
1376 unsigned int optlen)
1380 struct vsock_sock *vsk;
1383 if (level != AF_VSOCK)
1384 return -ENOPROTOOPT;
1386 #define COPY_IN(_v) \
1388 if (optlen < sizeof(_v)) { \
1392 if (copy_from_user(&_v, optval, sizeof(_v)) != 0) { \
1405 case SO_VM_SOCKETS_BUFFER_SIZE:
1407 transport->set_buffer_size(vsk, val);
1410 case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1412 transport->set_max_buffer_size(vsk, val);
1415 case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1417 transport->set_min_buffer_size(vsk, val);
1420 case SO_VM_SOCKETS_CONNECT_TIMEOUT: {
1423 if (tv.tv_sec >= 0 && tv.tv_usec < USEC_PER_SEC &&
1424 tv.tv_sec < (MAX_SCHEDULE_TIMEOUT / HZ - 1)) {
1425 vsk->connect_timeout = tv.tv_sec * HZ +
1426 DIV_ROUND_UP(tv.tv_usec, (1000000 / HZ));
1427 if (vsk->connect_timeout == 0)
1428 vsk->connect_timeout =
1429 VSOCK_DEFAULT_CONNECT_TIMEOUT;
1449 static int vsock_stream_getsockopt(struct socket *sock,
1450 int level, int optname,
1451 char __user *optval,
1457 struct vsock_sock *vsk;
1460 if (level != AF_VSOCK)
1461 return -ENOPROTOOPT;
1463 err = get_user(len, optlen);
1467 #define COPY_OUT(_v) \
1469 if (len < sizeof(_v)) \
1473 if (copy_to_user(optval, &_v, len) != 0) \
1483 case SO_VM_SOCKETS_BUFFER_SIZE:
1484 val = transport->get_buffer_size(vsk);
1488 case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1489 val = transport->get_max_buffer_size(vsk);
1493 case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1494 val = transport->get_min_buffer_size(vsk);
1498 case SO_VM_SOCKETS_CONNECT_TIMEOUT: {
1500 tv.tv_sec = vsk->connect_timeout / HZ;
1502 (vsk->connect_timeout -
1503 tv.tv_sec * HZ) * (1000000 / HZ);
1508 return -ENOPROTOOPT;
1511 err = put_user(len, optlen);
1520 static int vsock_stream_sendmsg(struct socket *sock, struct msghdr *msg,
1524 struct vsock_sock *vsk;
1525 ssize_t total_written;
1528 struct vsock_transport_send_notify_data send_data;
1537 if (msg->msg_flags & MSG_OOB)
1542 /* Callers should not provide a destination with stream sockets. */
1543 if (msg->msg_namelen) {
1544 err = sk->sk_state == SS_CONNECTED ? -EISCONN : -EOPNOTSUPP;
1548 /* Send data only if both sides are not shutdown in the direction. */
1549 if (sk->sk_shutdown & SEND_SHUTDOWN ||
1550 vsk->peer_shutdown & RCV_SHUTDOWN) {
1555 if (sk->sk_state != SS_CONNECTED ||
1556 !vsock_addr_bound(&vsk->local_addr)) {
1561 if (!vsock_addr_bound(&vsk->remote_addr)) {
1562 err = -EDESTADDRREQ;
1566 /* Wait for room in the produce queue to enqueue our user's data. */
1567 timeout = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1569 err = transport->notify_send_init(vsk, &send_data);
1574 while (total_written < len) {
1577 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1578 while (vsock_stream_has_space(vsk) == 0 &&
1580 !(sk->sk_shutdown & SEND_SHUTDOWN) &&
1581 !(vsk->peer_shutdown & RCV_SHUTDOWN)) {
1583 /* Don't wait for non-blocking sockets. */
1586 finish_wait(sk_sleep(sk), &wait);
1590 err = transport->notify_send_pre_block(vsk, &send_data);
1592 finish_wait(sk_sleep(sk), &wait);
1597 timeout = schedule_timeout(timeout);
1599 if (signal_pending(current)) {
1600 err = sock_intr_errno(timeout);
1601 finish_wait(sk_sleep(sk), &wait);
1603 } else if (timeout == 0) {
1605 finish_wait(sk_sleep(sk), &wait);
1609 prepare_to_wait(sk_sleep(sk), &wait,
1610 TASK_INTERRUPTIBLE);
1612 finish_wait(sk_sleep(sk), &wait);
1614 /* These checks occur both as part of and after the loop
1615 * conditional since we need to check before and after
1621 } else if ((sk->sk_shutdown & SEND_SHUTDOWN) ||
1622 (vsk->peer_shutdown & RCV_SHUTDOWN)) {
1627 err = transport->notify_send_pre_enqueue(vsk, &send_data);
1631 /* Note that enqueue will only write as many bytes as are free
1632 * in the produce queue, so we don't need to ensure len is
1633 * smaller than the queue size. It is the caller's
1634 * responsibility to check how many bytes we were able to send.
1637 written = transport->stream_enqueue(
1639 len - total_written);
1645 total_written += written;
1647 err = transport->notify_send_post_enqueue(
1648 vsk, written, &send_data);
1655 if (total_written > 0)
1656 err = total_written;
1664 vsock_stream_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
1668 struct vsock_sock *vsk;
1673 struct vsock_transport_recv_notify_data recv_data;
1683 if (sk->sk_state != SS_CONNECTED) {
1684 /* Recvmsg is supposed to return 0 if a peer performs an
1685 * orderly shutdown. Differentiate between that case and when a
1686 * peer has not connected or a local shutdown occured with the
1689 if (sock_flag(sk, SOCK_DONE))
1697 if (flags & MSG_OOB) {
1702 /* We don't check peer_shutdown flag here since peer may actually shut
1703 * down, but there can be data in the queue that a local socket can
1706 if (sk->sk_shutdown & RCV_SHUTDOWN) {
1711 /* It is valid on Linux to pass in a zero-length receive buffer. This
1712 * is not an error. We may as well bail out now.
1719 /* We must not copy less than target bytes into the user's buffer
1720 * before returning successfully, so we wait for the consume queue to
1721 * have that much data to consume before dequeueing. Note that this
1722 * makes it impossible to handle cases where target is greater than the
1725 target = sock_rcvlowat(sk, flags & MSG_WAITALL, len);
1726 if (target >= transport->stream_rcvhiwat(vsk)) {
1730 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1733 err = transport->notify_recv_init(vsk, target, &recv_data);
1741 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1742 ready = vsock_stream_has_data(vsk);
1745 if (sk->sk_err != 0 ||
1746 (sk->sk_shutdown & RCV_SHUTDOWN) ||
1747 (vsk->peer_shutdown & SEND_SHUTDOWN)) {
1748 finish_wait(sk_sleep(sk), &wait);
1751 /* Don't wait for non-blocking sockets. */
1754 finish_wait(sk_sleep(sk), &wait);
1758 err = transport->notify_recv_pre_block(
1759 vsk, target, &recv_data);
1761 finish_wait(sk_sleep(sk), &wait);
1765 timeout = schedule_timeout(timeout);
1768 if (signal_pending(current)) {
1769 err = sock_intr_errno(timeout);
1770 finish_wait(sk_sleep(sk), &wait);
1772 } else if (timeout == 0) {
1774 finish_wait(sk_sleep(sk), &wait);
1780 finish_wait(sk_sleep(sk), &wait);
1783 /* Invalid queue pair content. XXX This should
1784 * be changed to a connection reset in a later
1792 err = transport->notify_recv_pre_dequeue(
1793 vsk, target, &recv_data);
1797 read = transport->stream_dequeue(
1799 len - copied, flags);
1807 err = transport->notify_recv_post_dequeue(
1809 !(flags & MSG_PEEK), &recv_data);
1813 if (read >= target || flags & MSG_PEEK)
1822 else if (sk->sk_shutdown & RCV_SHUTDOWN)
1833 static const struct proto_ops vsock_stream_ops = {
1835 .owner = THIS_MODULE,
1836 .release = vsock_release,
1838 .connect = vsock_stream_connect,
1839 .socketpair = sock_no_socketpair,
1840 .accept = vsock_accept,
1841 .getname = vsock_getname,
1843 .ioctl = sock_no_ioctl,
1844 .listen = vsock_listen,
1845 .shutdown = vsock_shutdown,
1846 .setsockopt = vsock_stream_setsockopt,
1847 .getsockopt = vsock_stream_getsockopt,
1848 .sendmsg = vsock_stream_sendmsg,
1849 .recvmsg = vsock_stream_recvmsg,
1850 .mmap = sock_no_mmap,
1851 .sendpage = sock_no_sendpage,
1854 static int vsock_create(struct net *net, struct socket *sock,
1855 int protocol, int kern)
1860 if (protocol && protocol != PF_VSOCK)
1861 return -EPROTONOSUPPORT;
1863 switch (sock->type) {
1865 sock->ops = &vsock_dgram_ops;
1868 sock->ops = &vsock_stream_ops;
1871 return -ESOCKTNOSUPPORT;
1874 sock->state = SS_UNCONNECTED;
1876 return __vsock_create(net, sock, NULL, GFP_KERNEL, 0, kern) ? 0 : -ENOMEM;
1879 static const struct net_proto_family vsock_family_ops = {
1881 .create = vsock_create,
1882 .owner = THIS_MODULE,
1885 static long vsock_dev_do_ioctl(struct file *filp,
1886 unsigned int cmd, void __user *ptr)
1888 u32 __user *p = ptr;
1892 case IOCTL_VM_SOCKETS_GET_LOCAL_CID:
1893 if (put_user(transport->get_local_cid(), p) != 0)
1898 pr_err("Unknown ioctl %d\n", cmd);
1905 static long vsock_dev_ioctl(struct file *filp,
1906 unsigned int cmd, unsigned long arg)
1908 return vsock_dev_do_ioctl(filp, cmd, (void __user *)arg);
1911 #ifdef CONFIG_COMPAT
1912 static long vsock_dev_compat_ioctl(struct file *filp,
1913 unsigned int cmd, unsigned long arg)
1915 return vsock_dev_do_ioctl(filp, cmd, compat_ptr(arg));
1919 static const struct file_operations vsock_device_ops = {
1920 .owner = THIS_MODULE,
1921 .unlocked_ioctl = vsock_dev_ioctl,
1922 #ifdef CONFIG_COMPAT
1923 .compat_ioctl = vsock_dev_compat_ioctl,
1925 .open = nonseekable_open,
1928 static struct miscdevice vsock_device = {
1930 .fops = &vsock_device_ops,
1933 int __vsock_core_init(const struct vsock_transport *t, struct module *owner)
1935 int err = mutex_lock_interruptible(&vsock_register_mutex);
1945 /* Transport must be the owner of the protocol so that it can't
1946 * unload while there are open sockets.
1948 vsock_proto.owner = owner;
1951 vsock_init_tables();
1953 vsock_device.minor = MISC_DYNAMIC_MINOR;
1954 err = misc_register(&vsock_device);
1956 pr_err("Failed to register misc device\n");
1957 goto err_reset_transport;
1960 err = proto_register(&vsock_proto, 1); /* we want our slab */
1962 pr_err("Cannot register vsock protocol\n");
1963 goto err_deregister_misc;
1966 err = sock_register(&vsock_family_ops);
1968 pr_err("could not register af_vsock (%d) address family: %d\n",
1970 goto err_unregister_proto;
1973 mutex_unlock(&vsock_register_mutex);
1976 err_unregister_proto:
1977 proto_unregister(&vsock_proto);
1978 err_deregister_misc:
1979 misc_deregister(&vsock_device);
1980 err_reset_transport:
1983 mutex_unlock(&vsock_register_mutex);
1986 EXPORT_SYMBOL_GPL(__vsock_core_init);
1988 void vsock_core_exit(void)
1990 mutex_lock(&vsock_register_mutex);
1992 misc_deregister(&vsock_device);
1993 sock_unregister(AF_VSOCK);
1994 proto_unregister(&vsock_proto);
1996 /* We do not want the assignment below re-ordered. */
2000 mutex_unlock(&vsock_register_mutex);
2002 EXPORT_SYMBOL_GPL(vsock_core_exit);
2004 const struct vsock_transport *vsock_core_get_transport(void)
2006 /* vsock_register_mutex not taken since only the transport uses this
2007 * function and only while registered.
2011 EXPORT_SYMBOL_GPL(vsock_core_get_transport);
2013 MODULE_AUTHOR("VMware, Inc.");
2014 MODULE_DESCRIPTION("VMware Virtual Socket Family");
2015 MODULE_VERSION("1.0.2.0-k");
2016 MODULE_LICENSE("GPL v2");