1 // SPDX-License-Identifier: GPL-2.0-only
3 * VMware vSockets Driver
5 * Copyright (C) 2007-2013 VMware, Inc. All rights reserved.
8 /* Implementation notes:
10 * - There are two kinds of sockets: those created by user action (such as
11 * calling socket(2)) and those created by incoming connection request packets.
13 * - There are two "global" tables, one for bound sockets (sockets that have
14 * specified an address that they are responsible for) and one for connected
15 * sockets (sockets that have established a connection with another socket).
16 * These tables are "global" in that all sockets on the system are placed
17 * within them. - Note, though, that the bound table contains an extra entry
18 * for a list of unbound sockets and SOCK_DGRAM sockets will always remain in
19 * that list. The bound table is used solely for lookup of sockets when packets
20 * are received and that's not necessary for SOCK_DGRAM sockets since we create
21 * a datagram handle for each and need not perform a lookup. Keeping SOCK_DGRAM
22 * sockets out of the bound hash buckets will reduce the chance of collisions
23 * when looking for SOCK_STREAM sockets and prevents us from having to check the
24 * socket type in the hash table lookups.
26 * - Sockets created by user action will either be "client" sockets that
27 * initiate a connection or "server" sockets that listen for connections; we do
28 * not support simultaneous connects (two "client" sockets connecting).
30 * - "Server" sockets are referred to as listener sockets throughout this
31 * implementation because they are in the TCP_LISTEN state. When a
32 * connection request is received (the second kind of socket mentioned above),
33 * we create a new socket and refer to it as a pending socket. These pending
34 * sockets are placed on the pending connection list of the listener socket.
35 * When future packets are received for the address the listener socket is
36 * bound to, we check if the source of the packet is from one that has an
37 * existing pending connection. If it does, we process the packet for the
38 * pending socket. When that socket reaches the connected state, it is removed
39 * from the listener socket's pending list and enqueued in the listener
40 * socket's accept queue. Callers of accept(2) will accept connected sockets
41 * from the listener socket's accept queue. If the socket cannot be accepted
42 * for some reason then it is marked rejected. Once the connection is
43 * accepted, it is owned by the user process and the responsibility for cleanup
44 * falls with that user process.
46 * - It is possible that these pending sockets will never reach the connected
47 * state; in fact, we may never receive another packet after the connection
48 * request. Because of this, we must schedule a cleanup function to run in the
49 * future, after some amount of time passes where a connection should have been
50 * established. This function ensures that the socket is off all lists so it
51 * cannot be retrieved, then drops all references to the socket so it is cleaned
52 * up (sock_put() -> sk_free() -> our sk_destruct implementation). Note this
53 * function will also cleanup rejected sockets, those that reach the connected
54 * state but leave it before they have been accepted.
56 * - Lock ordering for pending or accept queue sockets is:
58 * lock_sock(listener);
59 * lock_sock_nested(pending, SINGLE_DEPTH_NESTING);
61 * Using explicit nested locking keeps lockdep happy since normally only one
62 * lock of a given class may be taken at a time.
64 * - Sockets created by user action will be cleaned up when the user process
65 * calls close(2), causing our release implementation to be called. Our release
66 * implementation will perform some cleanup then drop the last reference so our
67 * sk_destruct implementation is invoked. Our sk_destruct implementation will
68 * perform additional cleanup that's common for both types of sockets.
70 * - A socket's reference count is what ensures that the structure won't be
71 * freed. Each entry in a list (such as the "global" bound and connected tables
72 * and the listener socket's pending list and connected queue) ensures a
73 * reference. When we defer work until process context and pass a socket as our
74 * argument, we must ensure the reference count is increased to ensure the
75 * socket isn't freed before the function is run; the deferred function will
76 * then drop the reference.
78 * - sk->sk_state uses the TCP state constants because they are widely used by
79 * other address families and exposed to userspace tools like ss(8):
81 * TCP_CLOSE - unconnected
82 * TCP_SYN_SENT - connecting
83 * TCP_ESTABLISHED - connected
84 * TCP_CLOSING - disconnecting
85 * TCP_LISTEN - listening
88 #include <linux/types.h>
89 #include <linux/bitops.h>
90 #include <linux/cred.h>
91 #include <linux/init.h>
93 #include <linux/kernel.h>
94 #include <linux/sched/signal.h>
95 #include <linux/kmod.h>
96 #include <linux/list.h>
97 #include <linux/miscdevice.h>
98 #include <linux/module.h>
99 #include <linux/mutex.h>
100 #include <linux/net.h>
101 #include <linux/poll.h>
102 #include <linux/random.h>
103 #include <linux/skbuff.h>
104 #include <linux/smp.h>
105 #include <linux/socket.h>
106 #include <linux/stddef.h>
107 #include <linux/unistd.h>
108 #include <linux/wait.h>
109 #include <linux/workqueue.h>
110 #include <net/sock.h>
111 #include <net/af_vsock.h>
113 static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr);
114 static void vsock_sk_destruct(struct sock *sk);
115 static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
117 /* Protocol family. */
118 static struct proto vsock_proto = {
120 .owner = THIS_MODULE,
121 .obj_size = sizeof(struct vsock_sock),
124 /* The default peer timeout indicates how long we will wait for a peer response
125 * to a control message.
127 #define VSOCK_DEFAULT_CONNECT_TIMEOUT (2 * HZ)
129 #define VSOCK_DEFAULT_BUFFER_SIZE (1024 * 256)
130 #define VSOCK_DEFAULT_BUFFER_MAX_SIZE (1024 * 256)
131 #define VSOCK_DEFAULT_BUFFER_MIN_SIZE 128
133 /* Transport used for host->guest communication */
134 static const struct vsock_transport *transport_h2g;
135 /* Transport used for guest->host communication */
136 static const struct vsock_transport *transport_g2h;
137 /* Transport used for DGRAM communication */
138 static const struct vsock_transport *transport_dgram;
139 /* Transport used for local communication */
140 static const struct vsock_transport *transport_local;
141 static DEFINE_MUTEX(vsock_register_mutex);
145 /* Each bound VSocket is stored in the bind hash table and each connected
146 * VSocket is stored in the connected hash table.
148 * Unbound sockets are all put on the same list attached to the end of the hash
149 * table (vsock_unbound_sockets). Bound sockets are added to the hash table in
150 * the bucket that their local address hashes to (vsock_bound_sockets(addr)
151 * represents the list that addr hashes to).
153 * Specifically, we initialize the vsock_bind_table array to a size of
154 * VSOCK_HASH_SIZE + 1 so that vsock_bind_table[0] through
155 * vsock_bind_table[VSOCK_HASH_SIZE - 1] are for bound sockets and
156 * vsock_bind_table[VSOCK_HASH_SIZE] is for unbound sockets. The hash function
157 * mods with VSOCK_HASH_SIZE to ensure this.
159 #define MAX_PORT_RETRIES 24
161 #define VSOCK_HASH(addr) ((addr)->svm_port % VSOCK_HASH_SIZE)
162 #define vsock_bound_sockets(addr) (&vsock_bind_table[VSOCK_HASH(addr)])
163 #define vsock_unbound_sockets (&vsock_bind_table[VSOCK_HASH_SIZE])
165 /* XXX This can probably be implemented in a better way. */
166 #define VSOCK_CONN_HASH(src, dst) \
167 (((src)->svm_cid ^ (dst)->svm_port) % VSOCK_HASH_SIZE)
168 #define vsock_connected_sockets(src, dst) \
169 (&vsock_connected_table[VSOCK_CONN_HASH(src, dst)])
170 #define vsock_connected_sockets_vsk(vsk) \
171 vsock_connected_sockets(&(vsk)->remote_addr, &(vsk)->local_addr)
173 struct list_head vsock_bind_table[VSOCK_HASH_SIZE + 1];
174 EXPORT_SYMBOL_GPL(vsock_bind_table);
175 struct list_head vsock_connected_table[VSOCK_HASH_SIZE];
176 EXPORT_SYMBOL_GPL(vsock_connected_table);
177 DEFINE_SPINLOCK(vsock_table_lock);
178 EXPORT_SYMBOL_GPL(vsock_table_lock);
180 /* Autobind this socket to the local address if necessary. */
181 static int vsock_auto_bind(struct vsock_sock *vsk)
183 struct sock *sk = sk_vsock(vsk);
184 struct sockaddr_vm local_addr;
186 if (vsock_addr_bound(&vsk->local_addr))
188 vsock_addr_init(&local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
189 return __vsock_bind(sk, &local_addr);
192 static void vsock_init_tables(void)
196 for (i = 0; i < ARRAY_SIZE(vsock_bind_table); i++)
197 INIT_LIST_HEAD(&vsock_bind_table[i]);
199 for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++)
200 INIT_LIST_HEAD(&vsock_connected_table[i]);
203 static void __vsock_insert_bound(struct list_head *list,
204 struct vsock_sock *vsk)
207 list_add(&vsk->bound_table, list);
210 static void __vsock_insert_connected(struct list_head *list,
211 struct vsock_sock *vsk)
214 list_add(&vsk->connected_table, list);
217 static void __vsock_remove_bound(struct vsock_sock *vsk)
219 list_del_init(&vsk->bound_table);
223 static void __vsock_remove_connected(struct vsock_sock *vsk)
225 list_del_init(&vsk->connected_table);
229 static struct sock *__vsock_find_bound_socket(struct sockaddr_vm *addr)
231 struct vsock_sock *vsk;
233 list_for_each_entry(vsk, vsock_bound_sockets(addr), bound_table) {
234 if (vsock_addr_equals_addr(addr, &vsk->local_addr))
235 return sk_vsock(vsk);
237 if (addr->svm_port == vsk->local_addr.svm_port &&
238 (vsk->local_addr.svm_cid == VMADDR_CID_ANY ||
239 addr->svm_cid == VMADDR_CID_ANY))
240 return sk_vsock(vsk);
246 static struct sock *__vsock_find_connected_socket(struct sockaddr_vm *src,
247 struct sockaddr_vm *dst)
249 struct vsock_sock *vsk;
251 list_for_each_entry(vsk, vsock_connected_sockets(src, dst),
253 if (vsock_addr_equals_addr(src, &vsk->remote_addr) &&
254 dst->svm_port == vsk->local_addr.svm_port) {
255 return sk_vsock(vsk);
262 static void vsock_insert_unbound(struct vsock_sock *vsk)
264 spin_lock_bh(&vsock_table_lock);
265 __vsock_insert_bound(vsock_unbound_sockets, vsk);
266 spin_unlock_bh(&vsock_table_lock);
269 void vsock_insert_connected(struct vsock_sock *vsk)
271 struct list_head *list = vsock_connected_sockets(
272 &vsk->remote_addr, &vsk->local_addr);
274 spin_lock_bh(&vsock_table_lock);
275 __vsock_insert_connected(list, vsk);
276 spin_unlock_bh(&vsock_table_lock);
278 EXPORT_SYMBOL_GPL(vsock_insert_connected);
280 void vsock_remove_bound(struct vsock_sock *vsk)
282 spin_lock_bh(&vsock_table_lock);
283 if (__vsock_in_bound_table(vsk))
284 __vsock_remove_bound(vsk);
285 spin_unlock_bh(&vsock_table_lock);
287 EXPORT_SYMBOL_GPL(vsock_remove_bound);
289 void vsock_remove_connected(struct vsock_sock *vsk)
291 spin_lock_bh(&vsock_table_lock);
292 if (__vsock_in_connected_table(vsk))
293 __vsock_remove_connected(vsk);
294 spin_unlock_bh(&vsock_table_lock);
296 EXPORT_SYMBOL_GPL(vsock_remove_connected);
298 struct sock *vsock_find_bound_socket(struct sockaddr_vm *addr)
302 spin_lock_bh(&vsock_table_lock);
303 sk = __vsock_find_bound_socket(addr);
307 spin_unlock_bh(&vsock_table_lock);
311 EXPORT_SYMBOL_GPL(vsock_find_bound_socket);
313 struct sock *vsock_find_connected_socket(struct sockaddr_vm *src,
314 struct sockaddr_vm *dst)
318 spin_lock_bh(&vsock_table_lock);
319 sk = __vsock_find_connected_socket(src, dst);
323 spin_unlock_bh(&vsock_table_lock);
327 EXPORT_SYMBOL_GPL(vsock_find_connected_socket);
329 void vsock_remove_sock(struct vsock_sock *vsk)
331 vsock_remove_bound(vsk);
332 vsock_remove_connected(vsk);
334 EXPORT_SYMBOL_GPL(vsock_remove_sock);
336 void vsock_for_each_connected_socket(struct vsock_transport *transport,
337 void (*fn)(struct sock *sk))
341 spin_lock_bh(&vsock_table_lock);
343 for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) {
344 struct vsock_sock *vsk;
345 list_for_each_entry(vsk, &vsock_connected_table[i],
347 if (vsk->transport != transport)
354 spin_unlock_bh(&vsock_table_lock);
356 EXPORT_SYMBOL_GPL(vsock_for_each_connected_socket);
358 void vsock_add_pending(struct sock *listener, struct sock *pending)
360 struct vsock_sock *vlistener;
361 struct vsock_sock *vpending;
363 vlistener = vsock_sk(listener);
364 vpending = vsock_sk(pending);
368 list_add_tail(&vpending->pending_links, &vlistener->pending_links);
370 EXPORT_SYMBOL_GPL(vsock_add_pending);
372 void vsock_remove_pending(struct sock *listener, struct sock *pending)
374 struct vsock_sock *vpending = vsock_sk(pending);
376 list_del_init(&vpending->pending_links);
380 EXPORT_SYMBOL_GPL(vsock_remove_pending);
382 void vsock_enqueue_accept(struct sock *listener, struct sock *connected)
384 struct vsock_sock *vlistener;
385 struct vsock_sock *vconnected;
387 vlistener = vsock_sk(listener);
388 vconnected = vsock_sk(connected);
390 sock_hold(connected);
392 list_add_tail(&vconnected->accept_queue, &vlistener->accept_queue);
394 EXPORT_SYMBOL_GPL(vsock_enqueue_accept);
396 static bool vsock_use_local_transport(unsigned int remote_cid)
398 if (!transport_local)
401 if (remote_cid == VMADDR_CID_LOCAL)
405 return remote_cid == transport_g2h->get_local_cid();
407 return remote_cid == VMADDR_CID_HOST;
411 static void vsock_deassign_transport(struct vsock_sock *vsk)
416 vsk->transport->destruct(vsk);
417 module_put(vsk->transport->module);
418 vsk->transport = NULL;
421 /* Assign a transport to a socket and call the .init transport callback.
423 * Note: for stream socket this must be called when vsk->remote_addr is set
424 * (e.g. during the connect() or when a connection request on a listener
425 * socket is received).
426 * The vsk->remote_addr is used to decide which transport to use:
427 * - remote CID == VMADDR_CID_LOCAL or g2h->local_cid or VMADDR_CID_HOST if
428 * g2h is not loaded, will use local transport;
429 * - remote CID <= VMADDR_CID_HOST will use guest->host transport;
430 * - remote CID > VMADDR_CID_HOST will use host->guest transport;
432 int vsock_assign_transport(struct vsock_sock *vsk, struct vsock_sock *psk)
434 const struct vsock_transport *new_transport;
435 struct sock *sk = sk_vsock(vsk);
436 unsigned int remote_cid = vsk->remote_addr.svm_cid;
439 switch (sk->sk_type) {
441 new_transport = transport_dgram;
444 if (vsock_use_local_transport(remote_cid))
445 new_transport = transport_local;
446 else if (remote_cid <= VMADDR_CID_HOST || !transport_h2g)
447 new_transport = transport_g2h;
449 new_transport = transport_h2g;
452 return -ESOCKTNOSUPPORT;
455 if (vsk->transport) {
456 if (vsk->transport == new_transport)
459 /* transport->release() must be called with sock lock acquired.
460 * This path can only be taken during vsock_stream_connect(),
461 * where we have already held the sock lock.
462 * In the other cases, this function is called on a new socket
463 * which is not assigned to any transport.
465 vsk->transport->release(vsk);
466 vsock_deassign_transport(vsk);
469 /* We increase the module refcnt to prevent the transport unloading
470 * while there are open sockets assigned to it.
472 if (!new_transport || !try_module_get(new_transport->module))
475 ret = new_transport->init(vsk, psk);
477 module_put(new_transport->module);
481 vsk->transport = new_transport;
485 EXPORT_SYMBOL_GPL(vsock_assign_transport);
487 bool vsock_find_cid(unsigned int cid)
489 if (transport_g2h && cid == transport_g2h->get_local_cid())
492 if (transport_h2g && cid == VMADDR_CID_HOST)
495 if (transport_local && cid == VMADDR_CID_LOCAL)
500 EXPORT_SYMBOL_GPL(vsock_find_cid);
502 static struct sock *vsock_dequeue_accept(struct sock *listener)
504 struct vsock_sock *vlistener;
505 struct vsock_sock *vconnected;
507 vlistener = vsock_sk(listener);
509 if (list_empty(&vlistener->accept_queue))
512 vconnected = list_entry(vlistener->accept_queue.next,
513 struct vsock_sock, accept_queue);
515 list_del_init(&vconnected->accept_queue);
517 /* The caller will need a reference on the connected socket so we let
518 * it call sock_put().
521 return sk_vsock(vconnected);
524 static bool vsock_is_accept_queue_empty(struct sock *sk)
526 struct vsock_sock *vsk = vsock_sk(sk);
527 return list_empty(&vsk->accept_queue);
530 static bool vsock_is_pending(struct sock *sk)
532 struct vsock_sock *vsk = vsock_sk(sk);
533 return !list_empty(&vsk->pending_links);
536 static int vsock_send_shutdown(struct sock *sk, int mode)
538 struct vsock_sock *vsk = vsock_sk(sk);
543 return vsk->transport->shutdown(vsk, mode);
546 static void vsock_pending_work(struct work_struct *work)
549 struct sock *listener;
550 struct vsock_sock *vsk;
553 vsk = container_of(work, struct vsock_sock, pending_work.work);
555 listener = vsk->listener;
559 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
561 if (vsock_is_pending(sk)) {
562 vsock_remove_pending(listener, sk);
564 sk_acceptq_removed(listener);
565 } else if (!vsk->rejected) {
566 /* We are not on the pending list and accept() did not reject
567 * us, so we must have been accepted by our user process. We
568 * just need to drop our references to the sockets and be on
575 /* We need to remove ourself from the global connected sockets list so
576 * incoming packets can't find this socket, and to reduce the reference
579 vsock_remove_connected(vsk);
581 sk->sk_state = TCP_CLOSE;
585 release_sock(listener);
593 /**** SOCKET OPERATIONS ****/
595 static int __vsock_bind_stream(struct vsock_sock *vsk,
596 struct sockaddr_vm *addr)
599 struct sockaddr_vm new_addr;
602 port = LAST_RESERVED_PORT + 1 +
603 prandom_u32_max(U32_MAX - LAST_RESERVED_PORT);
605 vsock_addr_init(&new_addr, addr->svm_cid, addr->svm_port);
607 if (addr->svm_port == VMADDR_PORT_ANY) {
611 for (i = 0; i < MAX_PORT_RETRIES; i++) {
612 if (port <= LAST_RESERVED_PORT)
613 port = LAST_RESERVED_PORT + 1;
615 new_addr.svm_port = port++;
617 if (!__vsock_find_bound_socket(&new_addr)) {
624 return -EADDRNOTAVAIL;
626 /* If port is in reserved range, ensure caller
627 * has necessary privileges.
629 if (addr->svm_port <= LAST_RESERVED_PORT &&
630 !capable(CAP_NET_BIND_SERVICE)) {
634 if (__vsock_find_bound_socket(&new_addr))
638 vsock_addr_init(&vsk->local_addr, new_addr.svm_cid, new_addr.svm_port);
640 /* Remove stream sockets from the unbound list and add them to the hash
641 * table for easy lookup by its address. The unbound list is simply an
642 * extra entry at the end of the hash table, a trick used by AF_UNIX.
644 __vsock_remove_bound(vsk);
645 __vsock_insert_bound(vsock_bound_sockets(&vsk->local_addr), vsk);
650 static int __vsock_bind_dgram(struct vsock_sock *vsk,
651 struct sockaddr_vm *addr)
653 return vsk->transport->dgram_bind(vsk, addr);
656 static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr)
658 struct vsock_sock *vsk = vsock_sk(sk);
661 /* First ensure this socket isn't already bound. */
662 if (vsock_addr_bound(&vsk->local_addr))
665 /* Now bind to the provided address or select appropriate values if
666 * none are provided (VMADDR_CID_ANY and VMADDR_PORT_ANY). Note that
667 * like AF_INET prevents binding to a non-local IP address (in most
668 * cases), we only allow binding to a local CID.
670 if (addr->svm_cid != VMADDR_CID_ANY && !vsock_find_cid(addr->svm_cid))
671 return -EADDRNOTAVAIL;
673 switch (sk->sk_socket->type) {
675 spin_lock_bh(&vsock_table_lock);
676 retval = __vsock_bind_stream(vsk, addr);
677 spin_unlock_bh(&vsock_table_lock);
681 retval = __vsock_bind_dgram(vsk, addr);
692 static void vsock_connect_timeout(struct work_struct *work);
694 static struct sock *__vsock_create(struct net *net,
702 struct vsock_sock *psk;
703 struct vsock_sock *vsk;
705 sk = sk_alloc(net, AF_VSOCK, priority, &vsock_proto, kern);
709 sock_init_data(sock, sk);
711 /* sk->sk_type is normally set in sock_init_data, but only if sock is
712 * non-NULL. We make sure that our sockets always have a type by
713 * setting it here if needed.
719 vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
720 vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
722 sk->sk_destruct = vsock_sk_destruct;
723 sk->sk_backlog_rcv = vsock_queue_rcv_skb;
724 sock_reset_flag(sk, SOCK_DONE);
726 INIT_LIST_HEAD(&vsk->bound_table);
727 INIT_LIST_HEAD(&vsk->connected_table);
728 vsk->listener = NULL;
729 INIT_LIST_HEAD(&vsk->pending_links);
730 INIT_LIST_HEAD(&vsk->accept_queue);
731 vsk->rejected = false;
732 vsk->sent_request = false;
733 vsk->ignore_connecting_rst = false;
734 vsk->peer_shutdown = 0;
735 INIT_DELAYED_WORK(&vsk->connect_work, vsock_connect_timeout);
736 INIT_DELAYED_WORK(&vsk->pending_work, vsock_pending_work);
738 psk = parent ? vsock_sk(parent) : NULL;
740 vsk->trusted = psk->trusted;
741 vsk->owner = get_cred(psk->owner);
742 vsk->connect_timeout = psk->connect_timeout;
743 vsk->buffer_size = psk->buffer_size;
744 vsk->buffer_min_size = psk->buffer_min_size;
745 vsk->buffer_max_size = psk->buffer_max_size;
746 security_sk_clone(parent, sk);
748 vsk->trusted = ns_capable_noaudit(&init_user_ns, CAP_NET_ADMIN);
749 vsk->owner = get_current_cred();
750 vsk->connect_timeout = VSOCK_DEFAULT_CONNECT_TIMEOUT;
751 vsk->buffer_size = VSOCK_DEFAULT_BUFFER_SIZE;
752 vsk->buffer_min_size = VSOCK_DEFAULT_BUFFER_MIN_SIZE;
753 vsk->buffer_max_size = VSOCK_DEFAULT_BUFFER_MAX_SIZE;
759 static void __vsock_release(struct sock *sk, int level)
762 struct sock *pending;
763 struct vsock_sock *vsk;
766 pending = NULL; /* Compiler warning. */
768 /* When "level" is SINGLE_DEPTH_NESTING, use the nested
769 * version to avoid the warning "possible recursive locking
770 * detected". When "level" is 0, lock_sock_nested(sk, level)
771 * is the same as lock_sock(sk).
773 lock_sock_nested(sk, level);
776 vsk->transport->release(vsk);
777 else if (sk->sk_type == SOCK_STREAM)
778 vsock_remove_sock(vsk);
781 sk->sk_shutdown = SHUTDOWN_MASK;
783 skb_queue_purge(&sk->sk_receive_queue);
785 /* Clean up any sockets that never were accepted. */
786 while ((pending = vsock_dequeue_accept(sk)) != NULL) {
787 __vsock_release(pending, SINGLE_DEPTH_NESTING);
796 static void vsock_sk_destruct(struct sock *sk)
798 struct vsock_sock *vsk = vsock_sk(sk);
800 vsock_deassign_transport(vsk);
802 /* When clearing these addresses, there's no need to set the family and
803 * possibly register the address family with the kernel.
805 vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
806 vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
808 put_cred(vsk->owner);
811 static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
815 err = sock_queue_rcv_skb(sk, skb);
822 struct sock *vsock_create_connected(struct sock *parent)
824 return __vsock_create(sock_net(parent), NULL, parent, GFP_KERNEL,
827 EXPORT_SYMBOL_GPL(vsock_create_connected);
829 s64 vsock_stream_has_data(struct vsock_sock *vsk)
831 return vsk->transport->stream_has_data(vsk);
833 EXPORT_SYMBOL_GPL(vsock_stream_has_data);
835 s64 vsock_stream_has_space(struct vsock_sock *vsk)
837 return vsk->transport->stream_has_space(vsk);
839 EXPORT_SYMBOL_GPL(vsock_stream_has_space);
841 static int vsock_release(struct socket *sock)
843 __vsock_release(sock->sk, 0);
845 sock->state = SS_FREE;
851 vsock_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
855 struct sockaddr_vm *vm_addr;
859 if (vsock_addr_cast(addr, addr_len, &vm_addr) != 0)
863 err = __vsock_bind(sk, vm_addr);
869 static int vsock_getname(struct socket *sock,
870 struct sockaddr *addr, int peer)
874 struct vsock_sock *vsk;
875 struct sockaddr_vm *vm_addr;
884 if (sock->state != SS_CONNECTED) {
888 vm_addr = &vsk->remote_addr;
890 vm_addr = &vsk->local_addr;
898 /* sys_getsockname() and sys_getpeername() pass us a
899 * MAX_SOCK_ADDR-sized buffer and don't set addr_len. Unfortunately
900 * that macro is defined in socket.c instead of .h, so we hardcode its
903 BUILD_BUG_ON(sizeof(*vm_addr) > 128);
904 memcpy(addr, vm_addr, sizeof(*vm_addr));
905 err = sizeof(*vm_addr);
912 static int vsock_shutdown(struct socket *sock, int mode)
917 /* User level uses SHUT_RD (0) and SHUT_WR (1), but the kernel uses
918 * RCV_SHUTDOWN (1) and SEND_SHUTDOWN (2), so we must increment mode
919 * here like the other address families do. Note also that the
920 * increment makes SHUT_RDWR (2) into RCV_SHUTDOWN | SEND_SHUTDOWN (3),
921 * which is what we want.
925 if ((mode & ~SHUTDOWN_MASK) || !mode)
928 /* If this is a STREAM socket and it is not connected then bail out
929 * immediately. If it is a DGRAM socket then we must first kick the
930 * socket so that it wakes up from any sleeping calls, for example
931 * recv(), and then afterwards return the error.
937 if (sock->state == SS_UNCONNECTED) {
939 if (sk->sk_type == SOCK_STREAM)
942 sock->state = SS_DISCONNECTING;
946 /* Receive and send shutdowns are treated alike. */
947 mode = mode & (RCV_SHUTDOWN | SEND_SHUTDOWN);
949 sk->sk_shutdown |= mode;
950 sk->sk_state_change(sk);
952 if (sk->sk_type == SOCK_STREAM) {
953 sock_reset_flag(sk, SOCK_DONE);
954 vsock_send_shutdown(sk, mode);
963 static __poll_t vsock_poll(struct file *file, struct socket *sock,
968 struct vsock_sock *vsk;
973 poll_wait(file, sk_sleep(sk), wait);
977 /* Signify that there has been an error on this socket. */
980 /* INET sockets treat local write shutdown and peer write shutdown as a
981 * case of EPOLLHUP set.
983 if ((sk->sk_shutdown == SHUTDOWN_MASK) ||
984 ((sk->sk_shutdown & SEND_SHUTDOWN) &&
985 (vsk->peer_shutdown & SEND_SHUTDOWN))) {
989 if (sk->sk_shutdown & RCV_SHUTDOWN ||
990 vsk->peer_shutdown & SEND_SHUTDOWN) {
994 if (sock->type == SOCK_DGRAM) {
995 /* For datagram sockets we can read if there is something in
996 * the queue and write as long as the socket isn't shutdown for
999 if (!skb_queue_empty_lockless(&sk->sk_receive_queue) ||
1000 (sk->sk_shutdown & RCV_SHUTDOWN)) {
1001 mask |= EPOLLIN | EPOLLRDNORM;
1004 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
1005 mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND;
1007 } else if (sock->type == SOCK_STREAM) {
1008 const struct vsock_transport *transport;
1012 transport = vsk->transport;
1014 /* Listening sockets that have connections in their accept
1015 * queue can be read.
1017 if (sk->sk_state == TCP_LISTEN
1018 && !vsock_is_accept_queue_empty(sk))
1019 mask |= EPOLLIN | EPOLLRDNORM;
1021 /* If there is something in the queue then we can read. */
1022 if (transport && transport->stream_is_active(vsk) &&
1023 !(sk->sk_shutdown & RCV_SHUTDOWN)) {
1024 bool data_ready_now = false;
1025 int ret = transport->notify_poll_in(
1026 vsk, 1, &data_ready_now);
1031 mask |= EPOLLIN | EPOLLRDNORM;
1036 /* Sockets whose connections have been closed, reset, or
1037 * terminated should also be considered read, and we check the
1038 * shutdown flag for that.
1040 if (sk->sk_shutdown & RCV_SHUTDOWN ||
1041 vsk->peer_shutdown & SEND_SHUTDOWN) {
1042 mask |= EPOLLIN | EPOLLRDNORM;
1045 /* Connected sockets that can produce data can be written. */
1046 if (transport && sk->sk_state == TCP_ESTABLISHED) {
1047 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1048 bool space_avail_now = false;
1049 int ret = transport->notify_poll_out(
1050 vsk, 1, &space_avail_now);
1054 if (space_avail_now)
1055 /* Remove EPOLLWRBAND since INET
1056 * sockets are not setting it.
1058 mask |= EPOLLOUT | EPOLLWRNORM;
1064 /* Simulate INET socket poll behaviors, which sets
1065 * EPOLLOUT|EPOLLWRNORM when peer is closed and nothing to read,
1066 * but local send is not shutdown.
1068 if (sk->sk_state == TCP_CLOSE || sk->sk_state == TCP_CLOSING) {
1069 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
1070 mask |= EPOLLOUT | EPOLLWRNORM;
1080 static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg,
1085 struct vsock_sock *vsk;
1086 struct sockaddr_vm *remote_addr;
1087 const struct vsock_transport *transport;
1089 if (msg->msg_flags & MSG_OOB)
1092 /* For now, MSG_DONTWAIT is always assumed... */
1099 transport = vsk->transport;
1101 err = vsock_auto_bind(vsk);
1106 /* If the provided message contains an address, use that. Otherwise
1107 * fall back on the socket's remote handle (if it has been connected).
1109 if (msg->msg_name &&
1110 vsock_addr_cast(msg->msg_name, msg->msg_namelen,
1111 &remote_addr) == 0) {
1112 /* Ensure this address is of the right type and is a valid
1116 if (remote_addr->svm_cid == VMADDR_CID_ANY)
1117 remote_addr->svm_cid = transport->get_local_cid();
1119 if (!vsock_addr_bound(remote_addr)) {
1123 } else if (sock->state == SS_CONNECTED) {
1124 remote_addr = &vsk->remote_addr;
1126 if (remote_addr->svm_cid == VMADDR_CID_ANY)
1127 remote_addr->svm_cid = transport->get_local_cid();
1129 /* XXX Should connect() or this function ensure remote_addr is
1132 if (!vsock_addr_bound(&vsk->remote_addr)) {
1141 if (!transport->dgram_allow(remote_addr->svm_cid,
1142 remote_addr->svm_port)) {
1147 err = transport->dgram_enqueue(vsk, remote_addr, msg, len);
1154 static int vsock_dgram_connect(struct socket *sock,
1155 struct sockaddr *addr, int addr_len, int flags)
1159 struct vsock_sock *vsk;
1160 struct sockaddr_vm *remote_addr;
1165 err = vsock_addr_cast(addr, addr_len, &remote_addr);
1166 if (err == -EAFNOSUPPORT && remote_addr->svm_family == AF_UNSPEC) {
1168 vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY,
1170 sock->state = SS_UNCONNECTED;
1173 } else if (err != 0)
1178 err = vsock_auto_bind(vsk);
1182 if (!vsk->transport->dgram_allow(remote_addr->svm_cid,
1183 remote_addr->svm_port)) {
1188 memcpy(&vsk->remote_addr, remote_addr, sizeof(vsk->remote_addr));
1189 sock->state = SS_CONNECTED;
1196 static int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg,
1197 size_t len, int flags)
1199 struct vsock_sock *vsk = vsock_sk(sock->sk);
1201 return vsk->transport->dgram_dequeue(vsk, msg, len, flags);
1204 static const struct proto_ops vsock_dgram_ops = {
1206 .owner = THIS_MODULE,
1207 .release = vsock_release,
1209 .connect = vsock_dgram_connect,
1210 .socketpair = sock_no_socketpair,
1211 .accept = sock_no_accept,
1212 .getname = vsock_getname,
1214 .ioctl = sock_no_ioctl,
1215 .listen = sock_no_listen,
1216 .shutdown = vsock_shutdown,
1217 .sendmsg = vsock_dgram_sendmsg,
1218 .recvmsg = vsock_dgram_recvmsg,
1219 .mmap = sock_no_mmap,
1220 .sendpage = sock_no_sendpage,
1223 static int vsock_transport_cancel_pkt(struct vsock_sock *vsk)
1225 const struct vsock_transport *transport = vsk->transport;
1227 if (!transport || !transport->cancel_pkt)
1230 return transport->cancel_pkt(vsk);
1233 static void vsock_connect_timeout(struct work_struct *work)
1236 struct vsock_sock *vsk;
1238 vsk = container_of(work, struct vsock_sock, connect_work.work);
1242 if (sk->sk_state == TCP_SYN_SENT &&
1243 (sk->sk_shutdown != SHUTDOWN_MASK)) {
1244 sk->sk_state = TCP_CLOSE;
1245 sk->sk_socket->state = SS_UNCONNECTED;
1246 sk->sk_err = ETIMEDOUT;
1247 sk->sk_error_report(sk);
1248 vsock_transport_cancel_pkt(vsk);
1255 static int vsock_stream_connect(struct socket *sock, struct sockaddr *addr,
1256 int addr_len, int flags)
1260 struct vsock_sock *vsk;
1261 const struct vsock_transport *transport;
1262 struct sockaddr_vm *remote_addr;
1272 /* XXX AF_UNSPEC should make us disconnect like AF_INET. */
1273 switch (sock->state) {
1277 case SS_DISCONNECTING:
1281 /* This continues on so we can move sock into the SS_CONNECTED
1282 * state once the connection has completed (at which point err
1283 * will be set to zero also). Otherwise, we will either wait
1284 * for the connection or return -EALREADY should this be a
1285 * non-blocking call.
1288 if (flags & O_NONBLOCK)
1292 if ((sk->sk_state == TCP_LISTEN) ||
1293 vsock_addr_cast(addr, addr_len, &remote_addr) != 0) {
1298 /* Set the remote address that we are connecting to. */
1299 memcpy(&vsk->remote_addr, remote_addr,
1300 sizeof(vsk->remote_addr));
1302 err = vsock_assign_transport(vsk, NULL);
1306 transport = vsk->transport;
1308 /* The hypervisor and well-known contexts do not have socket
1312 !transport->stream_allow(remote_addr->svm_cid,
1313 remote_addr->svm_port)) {
1318 err = vsock_auto_bind(vsk);
1322 sk->sk_state = TCP_SYN_SENT;
1324 err = transport->connect(vsk);
1328 /* Mark sock as connecting and set the error code to in
1329 * progress in case this is a non-blocking connect.
1331 sock->state = SS_CONNECTING;
1335 /* The receive path will handle all communication until we are able to
1336 * enter the connected state. Here we wait for the connection to be
1337 * completed or a notification of an error.
1339 timeout = vsk->connect_timeout;
1340 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1342 while (sk->sk_state != TCP_ESTABLISHED && sk->sk_err == 0) {
1343 if (flags & O_NONBLOCK) {
1344 /* If we're not going to block, we schedule a timeout
1345 * function to generate a timeout on the connection
1346 * attempt, in case the peer doesn't respond in a
1347 * timely manner. We hold on to the socket until the
1352 /* If the timeout function is already scheduled,
1353 * reschedule it, then ungrab the socket refcount to
1356 if (mod_delayed_work(system_wq, &vsk->connect_work,
1360 /* Skip ahead to preserve error code set above. */
1365 timeout = schedule_timeout(timeout);
1368 if (signal_pending(current)) {
1369 err = sock_intr_errno(timeout);
1370 sk->sk_state = sk->sk_state == TCP_ESTABLISHED ? TCP_CLOSING : TCP_CLOSE;
1371 sock->state = SS_UNCONNECTED;
1372 vsock_transport_cancel_pkt(vsk);
1373 vsock_remove_connected(vsk);
1375 } else if ((sk->sk_state != TCP_ESTABLISHED) && (timeout == 0)) {
1377 sk->sk_state = TCP_CLOSE;
1378 sock->state = SS_UNCONNECTED;
1379 vsock_transport_cancel_pkt(vsk);
1383 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1388 sk->sk_state = TCP_CLOSE;
1389 sock->state = SS_UNCONNECTED;
1395 finish_wait(sk_sleep(sk), &wait);
1401 static int vsock_accept(struct socket *sock, struct socket *newsock, int flags,
1404 struct sock *listener;
1406 struct sock *connected;
1407 struct vsock_sock *vconnected;
1412 listener = sock->sk;
1414 lock_sock(listener);
1416 if (sock->type != SOCK_STREAM) {
1421 if (listener->sk_state != TCP_LISTEN) {
1426 /* Wait for children sockets to appear; these are the new sockets
1427 * created upon connection establishment.
1429 timeout = sock_rcvtimeo(listener, flags & O_NONBLOCK);
1430 prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1432 while ((connected = vsock_dequeue_accept(listener)) == NULL &&
1433 listener->sk_err == 0) {
1434 release_sock(listener);
1435 timeout = schedule_timeout(timeout);
1436 finish_wait(sk_sleep(listener), &wait);
1437 lock_sock(listener);
1439 if (signal_pending(current)) {
1440 err = sock_intr_errno(timeout);
1442 } else if (timeout == 0) {
1447 prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1449 finish_wait(sk_sleep(listener), &wait);
1451 if (listener->sk_err)
1452 err = -listener->sk_err;
1455 sk_acceptq_removed(listener);
1457 lock_sock_nested(connected, SINGLE_DEPTH_NESTING);
1458 vconnected = vsock_sk(connected);
1460 /* If the listener socket has received an error, then we should
1461 * reject this socket and return. Note that we simply mark the
1462 * socket rejected, drop our reference, and let the cleanup
1463 * function handle the cleanup; the fact that we found it in
1464 * the listener's accept queue guarantees that the cleanup
1465 * function hasn't run yet.
1468 vconnected->rejected = true;
1470 newsock->state = SS_CONNECTED;
1471 sock_graft(connected, newsock);
1474 release_sock(connected);
1475 sock_put(connected);
1479 release_sock(listener);
1483 static int vsock_listen(struct socket *sock, int backlog)
1487 struct vsock_sock *vsk;
1493 if (sock->type != SOCK_STREAM) {
1498 if (sock->state != SS_UNCONNECTED) {
1505 if (!vsock_addr_bound(&vsk->local_addr)) {
1510 sk->sk_max_ack_backlog = backlog;
1511 sk->sk_state = TCP_LISTEN;
1520 static void vsock_update_buffer_size(struct vsock_sock *vsk,
1521 const struct vsock_transport *transport,
1524 if (val > vsk->buffer_max_size)
1525 val = vsk->buffer_max_size;
1527 if (val < vsk->buffer_min_size)
1528 val = vsk->buffer_min_size;
1530 if (val != vsk->buffer_size &&
1531 transport && transport->notify_buffer_size)
1532 transport->notify_buffer_size(vsk, &val);
1534 vsk->buffer_size = val;
1537 static int vsock_stream_setsockopt(struct socket *sock,
1541 unsigned int optlen)
1545 struct vsock_sock *vsk;
1546 const struct vsock_transport *transport;
1549 if (level != AF_VSOCK)
1550 return -ENOPROTOOPT;
1552 #define COPY_IN(_v) \
1554 if (optlen < sizeof(_v)) { \
1558 if (copy_from_sockptr(&_v, optval, sizeof(_v)) != 0) { \
1570 transport = vsk->transport;
1573 case SO_VM_SOCKETS_BUFFER_SIZE:
1575 vsock_update_buffer_size(vsk, transport, val);
1578 case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1580 vsk->buffer_max_size = val;
1581 vsock_update_buffer_size(vsk, transport, vsk->buffer_size);
1584 case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1586 vsk->buffer_min_size = val;
1587 vsock_update_buffer_size(vsk, transport, vsk->buffer_size);
1590 case SO_VM_SOCKETS_CONNECT_TIMEOUT: {
1591 struct __kernel_old_timeval tv;
1593 if (tv.tv_sec >= 0 && tv.tv_usec < USEC_PER_SEC &&
1594 tv.tv_sec < (MAX_SCHEDULE_TIMEOUT / HZ - 1)) {
1595 vsk->connect_timeout = tv.tv_sec * HZ +
1596 DIV_ROUND_UP(tv.tv_usec, (1000000 / HZ));
1597 if (vsk->connect_timeout == 0)
1598 vsk->connect_timeout =
1599 VSOCK_DEFAULT_CONNECT_TIMEOUT;
1619 static int vsock_stream_getsockopt(struct socket *sock,
1620 int level, int optname,
1621 char __user *optval,
1627 struct vsock_sock *vsk;
1630 if (level != AF_VSOCK)
1631 return -ENOPROTOOPT;
1633 err = get_user(len, optlen);
1637 #define COPY_OUT(_v) \
1639 if (len < sizeof(_v)) \
1643 if (copy_to_user(optval, &_v, len) != 0) \
1653 case SO_VM_SOCKETS_BUFFER_SIZE:
1654 val = vsk->buffer_size;
1658 case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1659 val = vsk->buffer_max_size;
1663 case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1664 val = vsk->buffer_min_size;
1668 case SO_VM_SOCKETS_CONNECT_TIMEOUT: {
1669 struct __kernel_old_timeval tv;
1670 tv.tv_sec = vsk->connect_timeout / HZ;
1672 (vsk->connect_timeout -
1673 tv.tv_sec * HZ) * (1000000 / HZ);
1678 return -ENOPROTOOPT;
1681 err = put_user(len, optlen);
1690 static int vsock_stream_sendmsg(struct socket *sock, struct msghdr *msg,
1694 struct vsock_sock *vsk;
1695 const struct vsock_transport *transport;
1696 ssize_t total_written;
1699 struct vsock_transport_send_notify_data send_data;
1700 DEFINE_WAIT_FUNC(wait, woken_wake_function);
1707 if (msg->msg_flags & MSG_OOB)
1712 transport = vsk->transport;
1714 /* Callers should not provide a destination with stream sockets. */
1715 if (msg->msg_namelen) {
1716 err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP;
1720 /* Send data only if both sides are not shutdown in the direction. */
1721 if (sk->sk_shutdown & SEND_SHUTDOWN ||
1722 vsk->peer_shutdown & RCV_SHUTDOWN) {
1727 if (!transport || sk->sk_state != TCP_ESTABLISHED ||
1728 !vsock_addr_bound(&vsk->local_addr)) {
1733 if (!vsock_addr_bound(&vsk->remote_addr)) {
1734 err = -EDESTADDRREQ;
1738 /* Wait for room in the produce queue to enqueue our user's data. */
1739 timeout = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1741 err = transport->notify_send_init(vsk, &send_data);
1745 while (total_written < len) {
1748 add_wait_queue(sk_sleep(sk), &wait);
1749 while (vsock_stream_has_space(vsk) == 0 &&
1751 !(sk->sk_shutdown & SEND_SHUTDOWN) &&
1752 !(vsk->peer_shutdown & RCV_SHUTDOWN)) {
1754 /* Don't wait for non-blocking sockets. */
1757 remove_wait_queue(sk_sleep(sk), &wait);
1761 err = transport->notify_send_pre_block(vsk, &send_data);
1763 remove_wait_queue(sk_sleep(sk), &wait);
1768 timeout = wait_woken(&wait, TASK_INTERRUPTIBLE, timeout);
1770 if (signal_pending(current)) {
1771 err = sock_intr_errno(timeout);
1772 remove_wait_queue(sk_sleep(sk), &wait);
1774 } else if (timeout == 0) {
1776 remove_wait_queue(sk_sleep(sk), &wait);
1780 remove_wait_queue(sk_sleep(sk), &wait);
1782 /* These checks occur both as part of and after the loop
1783 * conditional since we need to check before and after
1789 } else if ((sk->sk_shutdown & SEND_SHUTDOWN) ||
1790 (vsk->peer_shutdown & RCV_SHUTDOWN)) {
1795 err = transport->notify_send_pre_enqueue(vsk, &send_data);
1799 /* Note that enqueue will only write as many bytes as are free
1800 * in the produce queue, so we don't need to ensure len is
1801 * smaller than the queue size. It is the caller's
1802 * responsibility to check how many bytes we were able to send.
1805 written = transport->stream_enqueue(
1807 len - total_written);
1813 total_written += written;
1815 err = transport->notify_send_post_enqueue(
1816 vsk, written, &send_data);
1823 if (total_written > 0)
1824 err = total_written;
1832 vsock_stream_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
1836 struct vsock_sock *vsk;
1837 const struct vsock_transport *transport;
1842 struct vsock_transport_recv_notify_data recv_data;
1852 transport = vsk->transport;
1854 if (!transport || sk->sk_state != TCP_ESTABLISHED) {
1855 /* Recvmsg is supposed to return 0 if a peer performs an
1856 * orderly shutdown. Differentiate between that case and when a
1857 * peer has not connected or a local shutdown occured with the
1860 if (sock_flag(sk, SOCK_DONE))
1868 if (flags & MSG_OOB) {
1873 /* We don't check peer_shutdown flag here since peer may actually shut
1874 * down, but there can be data in the queue that a local socket can
1877 if (sk->sk_shutdown & RCV_SHUTDOWN) {
1882 /* It is valid on Linux to pass in a zero-length receive buffer. This
1883 * is not an error. We may as well bail out now.
1890 /* We must not copy less than target bytes into the user's buffer
1891 * before returning successfully, so we wait for the consume queue to
1892 * have that much data to consume before dequeueing. Note that this
1893 * makes it impossible to handle cases where target is greater than the
1896 target = sock_rcvlowat(sk, flags & MSG_WAITALL, len);
1897 if (target >= transport->stream_rcvhiwat(vsk)) {
1901 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1904 err = transport->notify_recv_init(vsk, target, &recv_data);
1912 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1913 ready = vsock_stream_has_data(vsk);
1916 if (sk->sk_err != 0 ||
1917 (sk->sk_shutdown & RCV_SHUTDOWN) ||
1918 (vsk->peer_shutdown & SEND_SHUTDOWN)) {
1919 finish_wait(sk_sleep(sk), &wait);
1922 /* Don't wait for non-blocking sockets. */
1925 finish_wait(sk_sleep(sk), &wait);
1929 err = transport->notify_recv_pre_block(
1930 vsk, target, &recv_data);
1932 finish_wait(sk_sleep(sk), &wait);
1936 timeout = schedule_timeout(timeout);
1939 if (signal_pending(current)) {
1940 err = sock_intr_errno(timeout);
1941 finish_wait(sk_sleep(sk), &wait);
1943 } else if (timeout == 0) {
1945 finish_wait(sk_sleep(sk), &wait);
1951 finish_wait(sk_sleep(sk), &wait);
1954 /* Invalid queue pair content. XXX This should
1955 * be changed to a connection reset in a later
1963 err = transport->notify_recv_pre_dequeue(
1964 vsk, target, &recv_data);
1968 read = transport->stream_dequeue(
1970 len - copied, flags);
1978 err = transport->notify_recv_post_dequeue(
1980 !(flags & MSG_PEEK), &recv_data);
1984 if (read >= target || flags & MSG_PEEK)
1993 else if (sk->sk_shutdown & RCV_SHUTDOWN)
2004 static const struct proto_ops vsock_stream_ops = {
2006 .owner = THIS_MODULE,
2007 .release = vsock_release,
2009 .connect = vsock_stream_connect,
2010 .socketpair = sock_no_socketpair,
2011 .accept = vsock_accept,
2012 .getname = vsock_getname,
2014 .ioctl = sock_no_ioctl,
2015 .listen = vsock_listen,
2016 .shutdown = vsock_shutdown,
2017 .setsockopt = vsock_stream_setsockopt,
2018 .getsockopt = vsock_stream_getsockopt,
2019 .sendmsg = vsock_stream_sendmsg,
2020 .recvmsg = vsock_stream_recvmsg,
2021 .mmap = sock_no_mmap,
2022 .sendpage = sock_no_sendpage,
2025 static int vsock_create(struct net *net, struct socket *sock,
2026 int protocol, int kern)
2028 struct vsock_sock *vsk;
2035 if (protocol && protocol != PF_VSOCK)
2036 return -EPROTONOSUPPORT;
2038 switch (sock->type) {
2040 sock->ops = &vsock_dgram_ops;
2043 sock->ops = &vsock_stream_ops;
2046 return -ESOCKTNOSUPPORT;
2049 sock->state = SS_UNCONNECTED;
2051 sk = __vsock_create(net, sock, NULL, GFP_KERNEL, 0, kern);
2057 if (sock->type == SOCK_DGRAM) {
2058 ret = vsock_assign_transport(vsk, NULL);
2065 vsock_insert_unbound(vsk);
2070 static const struct net_proto_family vsock_family_ops = {
2072 .create = vsock_create,
2073 .owner = THIS_MODULE,
2076 static long vsock_dev_do_ioctl(struct file *filp,
2077 unsigned int cmd, void __user *ptr)
2079 u32 __user *p = ptr;
2080 u32 cid = VMADDR_CID_ANY;
2084 case IOCTL_VM_SOCKETS_GET_LOCAL_CID:
2085 /* To be compatible with the VMCI behavior, we prioritize the
2086 * guest CID instead of well-know host CID (VMADDR_CID_HOST).
2089 cid = transport_g2h->get_local_cid();
2090 else if (transport_h2g)
2091 cid = transport_h2g->get_local_cid();
2093 if (put_user(cid, p) != 0)
2098 pr_err("Unknown ioctl %d\n", cmd);
2105 static long vsock_dev_ioctl(struct file *filp,
2106 unsigned int cmd, unsigned long arg)
2108 return vsock_dev_do_ioctl(filp, cmd, (void __user *)arg);
2111 #ifdef CONFIG_COMPAT
2112 static long vsock_dev_compat_ioctl(struct file *filp,
2113 unsigned int cmd, unsigned long arg)
2115 return vsock_dev_do_ioctl(filp, cmd, compat_ptr(arg));
2119 static const struct file_operations vsock_device_ops = {
2120 .owner = THIS_MODULE,
2121 .unlocked_ioctl = vsock_dev_ioctl,
2122 #ifdef CONFIG_COMPAT
2123 .compat_ioctl = vsock_dev_compat_ioctl,
2125 .open = nonseekable_open,
2128 static struct miscdevice vsock_device = {
2130 .fops = &vsock_device_ops,
2133 static int __init vsock_init(void)
2137 vsock_init_tables();
2139 vsock_proto.owner = THIS_MODULE;
2140 vsock_device.minor = MISC_DYNAMIC_MINOR;
2141 err = misc_register(&vsock_device);
2143 pr_err("Failed to register misc device\n");
2144 goto err_reset_transport;
2147 err = proto_register(&vsock_proto, 1); /* we want our slab */
2149 pr_err("Cannot register vsock protocol\n");
2150 goto err_deregister_misc;
2153 err = sock_register(&vsock_family_ops);
2155 pr_err("could not register af_vsock (%d) address family: %d\n",
2157 goto err_unregister_proto;
2162 err_unregister_proto:
2163 proto_unregister(&vsock_proto);
2164 err_deregister_misc:
2165 misc_deregister(&vsock_device);
2166 err_reset_transport:
2170 static void __exit vsock_exit(void)
2172 misc_deregister(&vsock_device);
2173 sock_unregister(AF_VSOCK);
2174 proto_unregister(&vsock_proto);
2177 const struct vsock_transport *vsock_core_get_transport(struct vsock_sock *vsk)
2179 return vsk->transport;
2181 EXPORT_SYMBOL_GPL(vsock_core_get_transport);
2183 int vsock_core_register(const struct vsock_transport *t, int features)
2185 const struct vsock_transport *t_h2g, *t_g2h, *t_dgram, *t_local;
2186 int err = mutex_lock_interruptible(&vsock_register_mutex);
2191 t_h2g = transport_h2g;
2192 t_g2h = transport_g2h;
2193 t_dgram = transport_dgram;
2194 t_local = transport_local;
2196 if (features & VSOCK_TRANSPORT_F_H2G) {
2204 if (features & VSOCK_TRANSPORT_F_G2H) {
2212 if (features & VSOCK_TRANSPORT_F_DGRAM) {
2220 if (features & VSOCK_TRANSPORT_F_LOCAL) {
2228 transport_h2g = t_h2g;
2229 transport_g2h = t_g2h;
2230 transport_dgram = t_dgram;
2231 transport_local = t_local;
2234 mutex_unlock(&vsock_register_mutex);
2237 EXPORT_SYMBOL_GPL(vsock_core_register);
2239 void vsock_core_unregister(const struct vsock_transport *t)
2241 mutex_lock(&vsock_register_mutex);
2243 if (transport_h2g == t)
2244 transport_h2g = NULL;
2246 if (transport_g2h == t)
2247 transport_g2h = NULL;
2249 if (transport_dgram == t)
2250 transport_dgram = NULL;
2252 if (transport_local == t)
2253 transport_local = NULL;
2255 mutex_unlock(&vsock_register_mutex);
2257 EXPORT_SYMBOL_GPL(vsock_core_unregister);
2259 module_init(vsock_init);
2260 module_exit(vsock_exit);
2262 MODULE_AUTHOR("VMware, Inc.");
2263 MODULE_DESCRIPTION("VMware Virtual Socket Family");
2264 MODULE_VERSION("1.0.2.0-k");
2265 MODULE_LICENSE("GPL v2");