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(void (*fn)(struct sock *sk))
340 spin_lock_bh(&vsock_table_lock);
342 for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) {
343 struct vsock_sock *vsk;
344 list_for_each_entry(vsk, &vsock_connected_table[i],
349 spin_unlock_bh(&vsock_table_lock);
351 EXPORT_SYMBOL_GPL(vsock_for_each_connected_socket);
353 void vsock_add_pending(struct sock *listener, struct sock *pending)
355 struct vsock_sock *vlistener;
356 struct vsock_sock *vpending;
358 vlistener = vsock_sk(listener);
359 vpending = vsock_sk(pending);
363 list_add_tail(&vpending->pending_links, &vlistener->pending_links);
365 EXPORT_SYMBOL_GPL(vsock_add_pending);
367 void vsock_remove_pending(struct sock *listener, struct sock *pending)
369 struct vsock_sock *vpending = vsock_sk(pending);
371 list_del_init(&vpending->pending_links);
375 EXPORT_SYMBOL_GPL(vsock_remove_pending);
377 void vsock_enqueue_accept(struct sock *listener, struct sock *connected)
379 struct vsock_sock *vlistener;
380 struct vsock_sock *vconnected;
382 vlistener = vsock_sk(listener);
383 vconnected = vsock_sk(connected);
385 sock_hold(connected);
387 list_add_tail(&vconnected->accept_queue, &vlistener->accept_queue);
389 EXPORT_SYMBOL_GPL(vsock_enqueue_accept);
391 static bool vsock_use_local_transport(unsigned int remote_cid)
393 if (!transport_local)
396 if (remote_cid == VMADDR_CID_LOCAL)
400 return remote_cid == transport_g2h->get_local_cid();
402 return remote_cid == VMADDR_CID_HOST;
406 static void vsock_deassign_transport(struct vsock_sock *vsk)
411 vsk->transport->destruct(vsk);
412 module_put(vsk->transport->module);
413 vsk->transport = NULL;
416 /* Assign a transport to a socket and call the .init transport callback.
418 * Note: for stream socket this must be called when vsk->remote_addr is set
419 * (e.g. during the connect() or when a connection request on a listener
420 * socket is received).
421 * The vsk->remote_addr is used to decide which transport to use:
422 * - remote CID == VMADDR_CID_LOCAL or g2h->local_cid or VMADDR_CID_HOST if
423 * g2h is not loaded, will use local transport;
424 * - remote CID <= VMADDR_CID_HOST or h2g is not loaded or remote flags field
425 * includes VMADDR_FLAG_TO_HOST flag value, will use guest->host transport;
426 * - remote CID > VMADDR_CID_HOST will use host->guest transport;
428 int vsock_assign_transport(struct vsock_sock *vsk, struct vsock_sock *psk)
430 const struct vsock_transport *new_transport;
431 struct sock *sk = sk_vsock(vsk);
432 unsigned int remote_cid = vsk->remote_addr.svm_cid;
436 /* If the packet is coming with the source and destination CIDs higher
437 * than VMADDR_CID_HOST, then a vsock channel where all the packets are
438 * forwarded to the host should be established. Then the host will
439 * need to forward the packets to the guest.
441 * The flag is set on the (listen) receive path (psk is not NULL). On
442 * the connect path the flag can be set by the user space application.
444 if (psk && vsk->local_addr.svm_cid > VMADDR_CID_HOST &&
445 vsk->remote_addr.svm_cid > VMADDR_CID_HOST)
446 vsk->remote_addr.svm_flags |= VMADDR_FLAG_TO_HOST;
448 remote_flags = vsk->remote_addr.svm_flags;
450 switch (sk->sk_type) {
452 new_transport = transport_dgram;
455 if (vsock_use_local_transport(remote_cid))
456 new_transport = transport_local;
457 else if (remote_cid <= VMADDR_CID_HOST || !transport_h2g ||
458 (remote_flags & VMADDR_FLAG_TO_HOST))
459 new_transport = transport_g2h;
461 new_transport = transport_h2g;
464 return -ESOCKTNOSUPPORT;
467 if (vsk->transport) {
468 if (vsk->transport == new_transport)
471 /* transport->release() must be called with sock lock acquired.
472 * This path can only be taken during vsock_stream_connect(),
473 * where we have already held the sock lock.
474 * In the other cases, this function is called on a new socket
475 * which is not assigned to any transport.
477 vsk->transport->release(vsk);
478 vsock_deassign_transport(vsk);
481 /* We increase the module refcnt to prevent the transport unloading
482 * while there are open sockets assigned to it.
484 if (!new_transport || !try_module_get(new_transport->module))
487 ret = new_transport->init(vsk, psk);
489 module_put(new_transport->module);
493 vsk->transport = new_transport;
497 EXPORT_SYMBOL_GPL(vsock_assign_transport);
499 bool vsock_find_cid(unsigned int cid)
501 if (transport_g2h && cid == transport_g2h->get_local_cid())
504 if (transport_h2g && cid == VMADDR_CID_HOST)
507 if (transport_local && cid == VMADDR_CID_LOCAL)
512 EXPORT_SYMBOL_GPL(vsock_find_cid);
514 static struct sock *vsock_dequeue_accept(struct sock *listener)
516 struct vsock_sock *vlistener;
517 struct vsock_sock *vconnected;
519 vlistener = vsock_sk(listener);
521 if (list_empty(&vlistener->accept_queue))
524 vconnected = list_entry(vlistener->accept_queue.next,
525 struct vsock_sock, accept_queue);
527 list_del_init(&vconnected->accept_queue);
529 /* The caller will need a reference on the connected socket so we let
530 * it call sock_put().
533 return sk_vsock(vconnected);
536 static bool vsock_is_accept_queue_empty(struct sock *sk)
538 struct vsock_sock *vsk = vsock_sk(sk);
539 return list_empty(&vsk->accept_queue);
542 static bool vsock_is_pending(struct sock *sk)
544 struct vsock_sock *vsk = vsock_sk(sk);
545 return !list_empty(&vsk->pending_links);
548 static int vsock_send_shutdown(struct sock *sk, int mode)
550 struct vsock_sock *vsk = vsock_sk(sk);
555 return vsk->transport->shutdown(vsk, mode);
558 static void vsock_pending_work(struct work_struct *work)
561 struct sock *listener;
562 struct vsock_sock *vsk;
565 vsk = container_of(work, struct vsock_sock, pending_work.work);
567 listener = vsk->listener;
571 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
573 if (vsock_is_pending(sk)) {
574 vsock_remove_pending(listener, sk);
576 sk_acceptq_removed(listener);
577 } else if (!vsk->rejected) {
578 /* We are not on the pending list and accept() did not reject
579 * us, so we must have been accepted by our user process. We
580 * just need to drop our references to the sockets and be on
587 /* We need to remove ourself from the global connected sockets list so
588 * incoming packets can't find this socket, and to reduce the reference
591 vsock_remove_connected(vsk);
593 sk->sk_state = TCP_CLOSE;
597 release_sock(listener);
605 /**** SOCKET OPERATIONS ****/
607 static int __vsock_bind_connectible(struct vsock_sock *vsk,
608 struct sockaddr_vm *addr)
611 struct sockaddr_vm new_addr;
614 port = LAST_RESERVED_PORT + 1 +
615 prandom_u32_max(U32_MAX - LAST_RESERVED_PORT);
617 vsock_addr_init(&new_addr, addr->svm_cid, addr->svm_port);
619 if (addr->svm_port == VMADDR_PORT_ANY) {
623 for (i = 0; i < MAX_PORT_RETRIES; i++) {
624 if (port <= LAST_RESERVED_PORT)
625 port = LAST_RESERVED_PORT + 1;
627 new_addr.svm_port = port++;
629 if (!__vsock_find_bound_socket(&new_addr)) {
636 return -EADDRNOTAVAIL;
638 /* If port is in reserved range, ensure caller
639 * has necessary privileges.
641 if (addr->svm_port <= LAST_RESERVED_PORT &&
642 !capable(CAP_NET_BIND_SERVICE)) {
646 if (__vsock_find_bound_socket(&new_addr))
650 vsock_addr_init(&vsk->local_addr, new_addr.svm_cid, new_addr.svm_port);
652 /* Remove stream sockets from the unbound list and add them to the hash
653 * table for easy lookup by its address. The unbound list is simply an
654 * extra entry at the end of the hash table, a trick used by AF_UNIX.
656 __vsock_remove_bound(vsk);
657 __vsock_insert_bound(vsock_bound_sockets(&vsk->local_addr), vsk);
662 static int __vsock_bind_dgram(struct vsock_sock *vsk,
663 struct sockaddr_vm *addr)
665 return vsk->transport->dgram_bind(vsk, addr);
668 static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr)
670 struct vsock_sock *vsk = vsock_sk(sk);
673 /* First ensure this socket isn't already bound. */
674 if (vsock_addr_bound(&vsk->local_addr))
677 /* Now bind to the provided address or select appropriate values if
678 * none are provided (VMADDR_CID_ANY and VMADDR_PORT_ANY). Note that
679 * like AF_INET prevents binding to a non-local IP address (in most
680 * cases), we only allow binding to a local CID.
682 if (addr->svm_cid != VMADDR_CID_ANY && !vsock_find_cid(addr->svm_cid))
683 return -EADDRNOTAVAIL;
685 switch (sk->sk_socket->type) {
687 spin_lock_bh(&vsock_table_lock);
688 retval = __vsock_bind_connectible(vsk, addr);
689 spin_unlock_bh(&vsock_table_lock);
693 retval = __vsock_bind_dgram(vsk, addr);
704 static void vsock_connect_timeout(struct work_struct *work);
706 static struct sock *__vsock_create(struct net *net,
714 struct vsock_sock *psk;
715 struct vsock_sock *vsk;
717 sk = sk_alloc(net, AF_VSOCK, priority, &vsock_proto, kern);
721 sock_init_data(sock, sk);
723 /* sk->sk_type is normally set in sock_init_data, but only if sock is
724 * non-NULL. We make sure that our sockets always have a type by
725 * setting it here if needed.
731 vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
732 vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
734 sk->sk_destruct = vsock_sk_destruct;
735 sk->sk_backlog_rcv = vsock_queue_rcv_skb;
736 sock_reset_flag(sk, SOCK_DONE);
738 INIT_LIST_HEAD(&vsk->bound_table);
739 INIT_LIST_HEAD(&vsk->connected_table);
740 vsk->listener = NULL;
741 INIT_LIST_HEAD(&vsk->pending_links);
742 INIT_LIST_HEAD(&vsk->accept_queue);
743 vsk->rejected = false;
744 vsk->sent_request = false;
745 vsk->ignore_connecting_rst = false;
746 vsk->peer_shutdown = 0;
747 INIT_DELAYED_WORK(&vsk->connect_work, vsock_connect_timeout);
748 INIT_DELAYED_WORK(&vsk->pending_work, vsock_pending_work);
750 psk = parent ? vsock_sk(parent) : NULL;
752 vsk->trusted = psk->trusted;
753 vsk->owner = get_cred(psk->owner);
754 vsk->connect_timeout = psk->connect_timeout;
755 vsk->buffer_size = psk->buffer_size;
756 vsk->buffer_min_size = psk->buffer_min_size;
757 vsk->buffer_max_size = psk->buffer_max_size;
758 security_sk_clone(parent, sk);
760 vsk->trusted = ns_capable_noaudit(&init_user_ns, CAP_NET_ADMIN);
761 vsk->owner = get_current_cred();
762 vsk->connect_timeout = VSOCK_DEFAULT_CONNECT_TIMEOUT;
763 vsk->buffer_size = VSOCK_DEFAULT_BUFFER_SIZE;
764 vsk->buffer_min_size = VSOCK_DEFAULT_BUFFER_MIN_SIZE;
765 vsk->buffer_max_size = VSOCK_DEFAULT_BUFFER_MAX_SIZE;
771 static bool sock_type_connectible(u16 type)
773 return type == SOCK_STREAM;
776 static void __vsock_release(struct sock *sk, int level)
779 struct sock *pending;
780 struct vsock_sock *vsk;
783 pending = NULL; /* Compiler warning. */
785 /* When "level" is SINGLE_DEPTH_NESTING, use the nested
786 * version to avoid the warning "possible recursive locking
787 * detected". When "level" is 0, lock_sock_nested(sk, level)
788 * is the same as lock_sock(sk).
790 lock_sock_nested(sk, level);
793 vsk->transport->release(vsk);
794 else if (sock_type_connectible(sk->sk_type))
795 vsock_remove_sock(vsk);
798 sk->sk_shutdown = SHUTDOWN_MASK;
800 skb_queue_purge(&sk->sk_receive_queue);
802 /* Clean up any sockets that never were accepted. */
803 while ((pending = vsock_dequeue_accept(sk)) != NULL) {
804 __vsock_release(pending, SINGLE_DEPTH_NESTING);
813 static void vsock_sk_destruct(struct sock *sk)
815 struct vsock_sock *vsk = vsock_sk(sk);
817 vsock_deassign_transport(vsk);
819 /* When clearing these addresses, there's no need to set the family and
820 * possibly register the address family with the kernel.
822 vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
823 vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
825 put_cred(vsk->owner);
828 static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
832 err = sock_queue_rcv_skb(sk, skb);
839 struct sock *vsock_create_connected(struct sock *parent)
841 return __vsock_create(sock_net(parent), NULL, parent, GFP_KERNEL,
844 EXPORT_SYMBOL_GPL(vsock_create_connected);
846 s64 vsock_stream_has_data(struct vsock_sock *vsk)
848 return vsk->transport->stream_has_data(vsk);
850 EXPORT_SYMBOL_GPL(vsock_stream_has_data);
852 s64 vsock_stream_has_space(struct vsock_sock *vsk)
854 return vsk->transport->stream_has_space(vsk);
856 EXPORT_SYMBOL_GPL(vsock_stream_has_space);
858 static int vsock_release(struct socket *sock)
860 __vsock_release(sock->sk, 0);
862 sock->state = SS_FREE;
868 vsock_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
872 struct sockaddr_vm *vm_addr;
876 if (vsock_addr_cast(addr, addr_len, &vm_addr) != 0)
880 err = __vsock_bind(sk, vm_addr);
886 static int vsock_getname(struct socket *sock,
887 struct sockaddr *addr, int peer)
891 struct vsock_sock *vsk;
892 struct sockaddr_vm *vm_addr;
901 if (sock->state != SS_CONNECTED) {
905 vm_addr = &vsk->remote_addr;
907 vm_addr = &vsk->local_addr;
915 /* sys_getsockname() and sys_getpeername() pass us a
916 * MAX_SOCK_ADDR-sized buffer and don't set addr_len. Unfortunately
917 * that macro is defined in socket.c instead of .h, so we hardcode its
920 BUILD_BUG_ON(sizeof(*vm_addr) > 128);
921 memcpy(addr, vm_addr, sizeof(*vm_addr));
922 err = sizeof(*vm_addr);
929 static int vsock_shutdown(struct socket *sock, int mode)
934 /* User level uses SHUT_RD (0) and SHUT_WR (1), but the kernel uses
935 * RCV_SHUTDOWN (1) and SEND_SHUTDOWN (2), so we must increment mode
936 * here like the other address families do. Note also that the
937 * increment makes SHUT_RDWR (2) into RCV_SHUTDOWN | SEND_SHUTDOWN (3),
938 * which is what we want.
942 if ((mode & ~SHUTDOWN_MASK) || !mode)
945 /* If this is a STREAM socket and it is not connected then bail out
946 * immediately. If it is a DGRAM socket then we must first kick the
947 * socket so that it wakes up from any sleeping calls, for example
948 * recv(), and then afterwards return the error.
954 if (sock->state == SS_UNCONNECTED) {
956 if (sock_type_connectible(sk->sk_type))
959 sock->state = SS_DISCONNECTING;
963 /* Receive and send shutdowns are treated alike. */
964 mode = mode & (RCV_SHUTDOWN | SEND_SHUTDOWN);
966 sk->sk_shutdown |= mode;
967 sk->sk_state_change(sk);
969 if (sock_type_connectible(sk->sk_type)) {
970 sock_reset_flag(sk, SOCK_DONE);
971 vsock_send_shutdown(sk, mode);
980 static __poll_t vsock_poll(struct file *file, struct socket *sock,
985 struct vsock_sock *vsk;
990 poll_wait(file, sk_sleep(sk), wait);
994 /* Signify that there has been an error on this socket. */
997 /* INET sockets treat local write shutdown and peer write shutdown as a
998 * case of EPOLLHUP set.
1000 if ((sk->sk_shutdown == SHUTDOWN_MASK) ||
1001 ((sk->sk_shutdown & SEND_SHUTDOWN) &&
1002 (vsk->peer_shutdown & SEND_SHUTDOWN))) {
1006 if (sk->sk_shutdown & RCV_SHUTDOWN ||
1007 vsk->peer_shutdown & SEND_SHUTDOWN) {
1011 if (sock->type == SOCK_DGRAM) {
1012 /* For datagram sockets we can read if there is something in
1013 * the queue and write as long as the socket isn't shutdown for
1016 if (!skb_queue_empty_lockless(&sk->sk_receive_queue) ||
1017 (sk->sk_shutdown & RCV_SHUTDOWN)) {
1018 mask |= EPOLLIN | EPOLLRDNORM;
1021 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
1022 mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND;
1024 } else if (sock_type_connectible(sk->sk_type)) {
1025 const struct vsock_transport *transport;
1029 transport = vsk->transport;
1031 /* Listening sockets that have connections in their accept
1032 * queue can be read.
1034 if (sk->sk_state == TCP_LISTEN
1035 && !vsock_is_accept_queue_empty(sk))
1036 mask |= EPOLLIN | EPOLLRDNORM;
1038 /* If there is something in the queue then we can read. */
1039 if (transport && transport->stream_is_active(vsk) &&
1040 !(sk->sk_shutdown & RCV_SHUTDOWN)) {
1041 bool data_ready_now = false;
1042 int ret = transport->notify_poll_in(
1043 vsk, 1, &data_ready_now);
1048 mask |= EPOLLIN | EPOLLRDNORM;
1053 /* Sockets whose connections have been closed, reset, or
1054 * terminated should also be considered read, and we check the
1055 * shutdown flag for that.
1057 if (sk->sk_shutdown & RCV_SHUTDOWN ||
1058 vsk->peer_shutdown & SEND_SHUTDOWN) {
1059 mask |= EPOLLIN | EPOLLRDNORM;
1062 /* Connected sockets that can produce data can be written. */
1063 if (transport && sk->sk_state == TCP_ESTABLISHED) {
1064 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1065 bool space_avail_now = false;
1066 int ret = transport->notify_poll_out(
1067 vsk, 1, &space_avail_now);
1071 if (space_avail_now)
1072 /* Remove EPOLLWRBAND since INET
1073 * sockets are not setting it.
1075 mask |= EPOLLOUT | EPOLLWRNORM;
1081 /* Simulate INET socket poll behaviors, which sets
1082 * EPOLLOUT|EPOLLWRNORM when peer is closed and nothing to read,
1083 * but local send is not shutdown.
1085 if (sk->sk_state == TCP_CLOSE || sk->sk_state == TCP_CLOSING) {
1086 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
1087 mask |= EPOLLOUT | EPOLLWRNORM;
1097 static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg,
1102 struct vsock_sock *vsk;
1103 struct sockaddr_vm *remote_addr;
1104 const struct vsock_transport *transport;
1106 if (msg->msg_flags & MSG_OOB)
1109 /* For now, MSG_DONTWAIT is always assumed... */
1116 transport = vsk->transport;
1118 err = vsock_auto_bind(vsk);
1123 /* If the provided message contains an address, use that. Otherwise
1124 * fall back on the socket's remote handle (if it has been connected).
1126 if (msg->msg_name &&
1127 vsock_addr_cast(msg->msg_name, msg->msg_namelen,
1128 &remote_addr) == 0) {
1129 /* Ensure this address is of the right type and is a valid
1133 if (remote_addr->svm_cid == VMADDR_CID_ANY)
1134 remote_addr->svm_cid = transport->get_local_cid();
1136 if (!vsock_addr_bound(remote_addr)) {
1140 } else if (sock->state == SS_CONNECTED) {
1141 remote_addr = &vsk->remote_addr;
1143 if (remote_addr->svm_cid == VMADDR_CID_ANY)
1144 remote_addr->svm_cid = transport->get_local_cid();
1146 /* XXX Should connect() or this function ensure remote_addr is
1149 if (!vsock_addr_bound(&vsk->remote_addr)) {
1158 if (!transport->dgram_allow(remote_addr->svm_cid,
1159 remote_addr->svm_port)) {
1164 err = transport->dgram_enqueue(vsk, remote_addr, msg, len);
1171 static int vsock_dgram_connect(struct socket *sock,
1172 struct sockaddr *addr, int addr_len, int flags)
1176 struct vsock_sock *vsk;
1177 struct sockaddr_vm *remote_addr;
1182 err = vsock_addr_cast(addr, addr_len, &remote_addr);
1183 if (err == -EAFNOSUPPORT && remote_addr->svm_family == AF_UNSPEC) {
1185 vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY,
1187 sock->state = SS_UNCONNECTED;
1190 } else if (err != 0)
1195 err = vsock_auto_bind(vsk);
1199 if (!vsk->transport->dgram_allow(remote_addr->svm_cid,
1200 remote_addr->svm_port)) {
1205 memcpy(&vsk->remote_addr, remote_addr, sizeof(vsk->remote_addr));
1206 sock->state = SS_CONNECTED;
1213 static int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg,
1214 size_t len, int flags)
1216 struct vsock_sock *vsk = vsock_sk(sock->sk);
1218 return vsk->transport->dgram_dequeue(vsk, msg, len, flags);
1221 static const struct proto_ops vsock_dgram_ops = {
1223 .owner = THIS_MODULE,
1224 .release = vsock_release,
1226 .connect = vsock_dgram_connect,
1227 .socketpair = sock_no_socketpair,
1228 .accept = sock_no_accept,
1229 .getname = vsock_getname,
1231 .ioctl = sock_no_ioctl,
1232 .listen = sock_no_listen,
1233 .shutdown = vsock_shutdown,
1234 .sendmsg = vsock_dgram_sendmsg,
1235 .recvmsg = vsock_dgram_recvmsg,
1236 .mmap = sock_no_mmap,
1237 .sendpage = sock_no_sendpage,
1240 static int vsock_transport_cancel_pkt(struct vsock_sock *vsk)
1242 const struct vsock_transport *transport = vsk->transport;
1244 if (!transport || !transport->cancel_pkt)
1247 return transport->cancel_pkt(vsk);
1250 static void vsock_connect_timeout(struct work_struct *work)
1253 struct vsock_sock *vsk;
1255 vsk = container_of(work, struct vsock_sock, connect_work.work);
1259 if (sk->sk_state == TCP_SYN_SENT &&
1260 (sk->sk_shutdown != SHUTDOWN_MASK)) {
1261 sk->sk_state = TCP_CLOSE;
1262 sk->sk_err = ETIMEDOUT;
1263 sk->sk_error_report(sk);
1264 vsock_transport_cancel_pkt(vsk);
1271 static int vsock_connect(struct socket *sock, struct sockaddr *addr,
1272 int addr_len, int flags)
1276 struct vsock_sock *vsk;
1277 const struct vsock_transport *transport;
1278 struct sockaddr_vm *remote_addr;
1288 /* XXX AF_UNSPEC should make us disconnect like AF_INET. */
1289 switch (sock->state) {
1293 case SS_DISCONNECTING:
1297 /* This continues on so we can move sock into the SS_CONNECTED
1298 * state once the connection has completed (at which point err
1299 * will be set to zero also). Otherwise, we will either wait
1300 * for the connection or return -EALREADY should this be a
1301 * non-blocking call.
1306 if ((sk->sk_state == TCP_LISTEN) ||
1307 vsock_addr_cast(addr, addr_len, &remote_addr) != 0) {
1312 /* Set the remote address that we are connecting to. */
1313 memcpy(&vsk->remote_addr, remote_addr,
1314 sizeof(vsk->remote_addr));
1316 err = vsock_assign_transport(vsk, NULL);
1320 transport = vsk->transport;
1322 /* The hypervisor and well-known contexts do not have socket
1326 !transport->stream_allow(remote_addr->svm_cid,
1327 remote_addr->svm_port)) {
1332 err = vsock_auto_bind(vsk);
1336 sk->sk_state = TCP_SYN_SENT;
1338 err = transport->connect(vsk);
1342 /* Mark sock as connecting and set the error code to in
1343 * progress in case this is a non-blocking connect.
1345 sock->state = SS_CONNECTING;
1349 /* The receive path will handle all communication until we are able to
1350 * enter the connected state. Here we wait for the connection to be
1351 * completed or a notification of an error.
1353 timeout = vsk->connect_timeout;
1354 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1356 while (sk->sk_state != TCP_ESTABLISHED && sk->sk_err == 0) {
1357 if (flags & O_NONBLOCK) {
1358 /* If we're not going to block, we schedule a timeout
1359 * function to generate a timeout on the connection
1360 * attempt, in case the peer doesn't respond in a
1361 * timely manner. We hold on to the socket until the
1365 schedule_delayed_work(&vsk->connect_work, timeout);
1367 /* Skip ahead to preserve error code set above. */
1372 timeout = schedule_timeout(timeout);
1375 if (signal_pending(current)) {
1376 err = sock_intr_errno(timeout);
1377 sk->sk_state = TCP_CLOSE;
1378 sock->state = SS_UNCONNECTED;
1379 vsock_transport_cancel_pkt(vsk);
1381 } else if (timeout == 0) {
1383 sk->sk_state = TCP_CLOSE;
1384 sock->state = SS_UNCONNECTED;
1385 vsock_transport_cancel_pkt(vsk);
1389 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1394 sk->sk_state = TCP_CLOSE;
1395 sock->state = SS_UNCONNECTED;
1401 finish_wait(sk_sleep(sk), &wait);
1407 static int vsock_accept(struct socket *sock, struct socket *newsock, int flags,
1410 struct sock *listener;
1412 struct sock *connected;
1413 struct vsock_sock *vconnected;
1418 listener = sock->sk;
1420 lock_sock(listener);
1422 if (!sock_type_connectible(sock->type)) {
1427 if (listener->sk_state != TCP_LISTEN) {
1432 /* Wait for children sockets to appear; these are the new sockets
1433 * created upon connection establishment.
1435 timeout = sock_rcvtimeo(listener, flags & O_NONBLOCK);
1436 prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1438 while ((connected = vsock_dequeue_accept(listener)) == NULL &&
1439 listener->sk_err == 0) {
1440 release_sock(listener);
1441 timeout = schedule_timeout(timeout);
1442 finish_wait(sk_sleep(listener), &wait);
1443 lock_sock(listener);
1445 if (signal_pending(current)) {
1446 err = sock_intr_errno(timeout);
1448 } else if (timeout == 0) {
1453 prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1455 finish_wait(sk_sleep(listener), &wait);
1457 if (listener->sk_err)
1458 err = -listener->sk_err;
1461 sk_acceptq_removed(listener);
1463 lock_sock_nested(connected, SINGLE_DEPTH_NESTING);
1464 vconnected = vsock_sk(connected);
1466 /* If the listener socket has received an error, then we should
1467 * reject this socket and return. Note that we simply mark the
1468 * socket rejected, drop our reference, and let the cleanup
1469 * function handle the cleanup; the fact that we found it in
1470 * the listener's accept queue guarantees that the cleanup
1471 * function hasn't run yet.
1474 vconnected->rejected = true;
1476 newsock->state = SS_CONNECTED;
1477 sock_graft(connected, newsock);
1480 release_sock(connected);
1481 sock_put(connected);
1485 release_sock(listener);
1489 static int vsock_listen(struct socket *sock, int backlog)
1493 struct vsock_sock *vsk;
1499 if (!sock_type_connectible(sk->sk_type)) {
1504 if (sock->state != SS_UNCONNECTED) {
1511 if (!vsock_addr_bound(&vsk->local_addr)) {
1516 sk->sk_max_ack_backlog = backlog;
1517 sk->sk_state = TCP_LISTEN;
1526 static void vsock_update_buffer_size(struct vsock_sock *vsk,
1527 const struct vsock_transport *transport,
1530 if (val > vsk->buffer_max_size)
1531 val = vsk->buffer_max_size;
1533 if (val < vsk->buffer_min_size)
1534 val = vsk->buffer_min_size;
1536 if (val != vsk->buffer_size &&
1537 transport && transport->notify_buffer_size)
1538 transport->notify_buffer_size(vsk, &val);
1540 vsk->buffer_size = val;
1543 static int vsock_connectible_setsockopt(struct socket *sock,
1547 unsigned int optlen)
1551 struct vsock_sock *vsk;
1552 const struct vsock_transport *transport;
1555 if (level != AF_VSOCK)
1556 return -ENOPROTOOPT;
1558 #define COPY_IN(_v) \
1560 if (optlen < sizeof(_v)) { \
1564 if (copy_from_sockptr(&_v, optval, sizeof(_v)) != 0) { \
1576 transport = vsk->transport;
1579 case SO_VM_SOCKETS_BUFFER_SIZE:
1581 vsock_update_buffer_size(vsk, transport, val);
1584 case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1586 vsk->buffer_max_size = val;
1587 vsock_update_buffer_size(vsk, transport, vsk->buffer_size);
1590 case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1592 vsk->buffer_min_size = val;
1593 vsock_update_buffer_size(vsk, transport, vsk->buffer_size);
1596 case SO_VM_SOCKETS_CONNECT_TIMEOUT: {
1597 struct __kernel_old_timeval tv;
1599 if (tv.tv_sec >= 0 && tv.tv_usec < USEC_PER_SEC &&
1600 tv.tv_sec < (MAX_SCHEDULE_TIMEOUT / HZ - 1)) {
1601 vsk->connect_timeout = tv.tv_sec * HZ +
1602 DIV_ROUND_UP(tv.tv_usec, (1000000 / HZ));
1603 if (vsk->connect_timeout == 0)
1604 vsk->connect_timeout =
1605 VSOCK_DEFAULT_CONNECT_TIMEOUT;
1625 static int vsock_connectible_getsockopt(struct socket *sock,
1626 int level, int optname,
1627 char __user *optval,
1633 struct vsock_sock *vsk;
1636 if (level != AF_VSOCK)
1637 return -ENOPROTOOPT;
1639 err = get_user(len, optlen);
1643 #define COPY_OUT(_v) \
1645 if (len < sizeof(_v)) \
1649 if (copy_to_user(optval, &_v, len) != 0) \
1659 case SO_VM_SOCKETS_BUFFER_SIZE:
1660 val = vsk->buffer_size;
1664 case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1665 val = vsk->buffer_max_size;
1669 case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1670 val = vsk->buffer_min_size;
1674 case SO_VM_SOCKETS_CONNECT_TIMEOUT: {
1675 struct __kernel_old_timeval tv;
1676 tv.tv_sec = vsk->connect_timeout / HZ;
1678 (vsk->connect_timeout -
1679 tv.tv_sec * HZ) * (1000000 / HZ);
1684 return -ENOPROTOOPT;
1687 err = put_user(len, optlen);
1696 static int vsock_connectible_sendmsg(struct socket *sock, struct msghdr *msg,
1700 struct vsock_sock *vsk;
1701 const struct vsock_transport *transport;
1702 ssize_t total_written;
1705 struct vsock_transport_send_notify_data send_data;
1706 DEFINE_WAIT_FUNC(wait, woken_wake_function);
1713 if (msg->msg_flags & MSG_OOB)
1718 transport = vsk->transport;
1720 /* Callers should not provide a destination with stream sockets. */
1721 if (msg->msg_namelen) {
1722 err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP;
1726 /* Send data only if both sides are not shutdown in the direction. */
1727 if (sk->sk_shutdown & SEND_SHUTDOWN ||
1728 vsk->peer_shutdown & RCV_SHUTDOWN) {
1733 if (!transport || sk->sk_state != TCP_ESTABLISHED ||
1734 !vsock_addr_bound(&vsk->local_addr)) {
1739 if (!vsock_addr_bound(&vsk->remote_addr)) {
1740 err = -EDESTADDRREQ;
1744 /* Wait for room in the produce queue to enqueue our user's data. */
1745 timeout = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1747 err = transport->notify_send_init(vsk, &send_data);
1751 while (total_written < len) {
1754 add_wait_queue(sk_sleep(sk), &wait);
1755 while (vsock_stream_has_space(vsk) == 0 &&
1757 !(sk->sk_shutdown & SEND_SHUTDOWN) &&
1758 !(vsk->peer_shutdown & RCV_SHUTDOWN)) {
1760 /* Don't wait for non-blocking sockets. */
1763 remove_wait_queue(sk_sleep(sk), &wait);
1767 err = transport->notify_send_pre_block(vsk, &send_data);
1769 remove_wait_queue(sk_sleep(sk), &wait);
1774 timeout = wait_woken(&wait, TASK_INTERRUPTIBLE, timeout);
1776 if (signal_pending(current)) {
1777 err = sock_intr_errno(timeout);
1778 remove_wait_queue(sk_sleep(sk), &wait);
1780 } else if (timeout == 0) {
1782 remove_wait_queue(sk_sleep(sk), &wait);
1786 remove_wait_queue(sk_sleep(sk), &wait);
1788 /* These checks occur both as part of and after the loop
1789 * conditional since we need to check before and after
1795 } else if ((sk->sk_shutdown & SEND_SHUTDOWN) ||
1796 (vsk->peer_shutdown & RCV_SHUTDOWN)) {
1801 err = transport->notify_send_pre_enqueue(vsk, &send_data);
1805 /* Note that enqueue will only write as many bytes as are free
1806 * in the produce queue, so we don't need to ensure len is
1807 * smaller than the queue size. It is the caller's
1808 * responsibility to check how many bytes we were able to send.
1811 if (sk->sk_type == SOCK_SEQPACKET) {
1812 written = transport->seqpacket_enqueue(vsk,
1813 msg, len - total_written);
1815 written = transport->stream_enqueue(vsk,
1816 msg, len - total_written);
1823 total_written += written;
1825 err = transport->notify_send_post_enqueue(
1826 vsk, written, &send_data);
1833 if (total_written > 0) {
1834 /* Return number of written bytes only if:
1835 * 1) SOCK_STREAM socket.
1836 * 2) SOCK_SEQPACKET socket when whole buffer is sent.
1838 if (sk->sk_type == SOCK_STREAM || total_written == len)
1839 err = total_written;
1846 static int vsock_wait_data(struct sock *sk, struct wait_queue_entry *wait,
1848 struct vsock_transport_recv_notify_data *recv_data,
1851 const struct vsock_transport *transport;
1852 struct vsock_sock *vsk;
1858 transport = vsk->transport;
1860 while ((data = vsock_stream_has_data(vsk)) == 0) {
1861 prepare_to_wait(sk_sleep(sk), wait, TASK_INTERRUPTIBLE);
1863 if (sk->sk_err != 0 ||
1864 (sk->sk_shutdown & RCV_SHUTDOWN) ||
1865 (vsk->peer_shutdown & SEND_SHUTDOWN)) {
1869 /* Don't wait for non-blocking sockets. */
1876 err = transport->notify_recv_pre_block(vsk, target, recv_data);
1882 timeout = schedule_timeout(timeout);
1885 if (signal_pending(current)) {
1886 err = sock_intr_errno(timeout);
1888 } else if (timeout == 0) {
1894 finish_wait(sk_sleep(sk), wait);
1899 /* Internal transport error when checking for available
1900 * data. XXX This should be changed to a connection
1901 * reset in a later change.
1909 static int __vsock_stream_recvmsg(struct sock *sk, struct msghdr *msg,
1910 size_t len, int flags)
1912 struct vsock_transport_recv_notify_data recv_data;
1913 const struct vsock_transport *transport;
1914 struct vsock_sock *vsk;
1923 transport = vsk->transport;
1925 /* We must not copy less than target bytes into the user's buffer
1926 * before returning successfully, so we wait for the consume queue to
1927 * have that much data to consume before dequeueing. Note that this
1928 * makes it impossible to handle cases where target is greater than the
1931 target = sock_rcvlowat(sk, flags & MSG_WAITALL, len);
1932 if (target >= transport->stream_rcvhiwat(vsk)) {
1936 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1939 err = transport->notify_recv_init(vsk, target, &recv_data);
1947 err = vsock_wait_data(sk, &wait, timeout, &recv_data, target);
1951 err = transport->notify_recv_pre_dequeue(vsk, target,
1956 read = transport->stream_dequeue(vsk, msg, len - copied, flags);
1964 err = transport->notify_recv_post_dequeue(vsk, target, read,
1965 !(flags & MSG_PEEK), &recv_data);
1969 if (read >= target || flags & MSG_PEEK)
1977 else if (sk->sk_shutdown & RCV_SHUTDOWN)
1987 static int __vsock_seqpacket_recvmsg(struct sock *sk, struct msghdr *msg,
1988 size_t len, int flags)
1990 const struct vsock_transport *transport;
1991 struct vsock_sock *vsk;
1998 transport = vsk->transport;
2000 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2002 err = vsock_wait_data(sk, &wait, timeout, NULL, 0);
2006 record_len = transport->seqpacket_dequeue(vsk, msg, flags);
2008 if (record_len < 0) {
2015 } else if (sk->sk_shutdown & RCV_SHUTDOWN) {
2018 /* User sets MSG_TRUNC, so return real length of
2021 if (flags & MSG_TRUNC)
2024 err = len - msg_data_left(msg);
2026 /* Always set MSG_TRUNC if real length of packet is
2027 * bigger than user's buffer.
2029 if (record_len > len)
2030 msg->msg_flags |= MSG_TRUNC;
2038 vsock_connectible_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
2042 struct vsock_sock *vsk;
2043 const struct vsock_transport *transport;
2054 transport = vsk->transport;
2056 if (!transport || sk->sk_state != TCP_ESTABLISHED) {
2057 /* Recvmsg is supposed to return 0 if a peer performs an
2058 * orderly shutdown. Differentiate between that case and when a
2059 * peer has not connected or a local shutdown occurred with the
2062 if (sock_flag(sk, SOCK_DONE))
2070 if (flags & MSG_OOB) {
2075 /* We don't check peer_shutdown flag here since peer may actually shut
2076 * down, but there can be data in the queue that a local socket can
2079 if (sk->sk_shutdown & RCV_SHUTDOWN) {
2084 /* It is valid on Linux to pass in a zero-length receive buffer. This
2085 * is not an error. We may as well bail out now.
2092 if (sk->sk_type == SOCK_STREAM)
2093 err = __vsock_stream_recvmsg(sk, msg, len, flags);
2095 err = __vsock_seqpacket_recvmsg(sk, msg, len, flags);
2102 static const struct proto_ops vsock_stream_ops = {
2104 .owner = THIS_MODULE,
2105 .release = vsock_release,
2107 .connect = vsock_connect,
2108 .socketpair = sock_no_socketpair,
2109 .accept = vsock_accept,
2110 .getname = vsock_getname,
2112 .ioctl = sock_no_ioctl,
2113 .listen = vsock_listen,
2114 .shutdown = vsock_shutdown,
2115 .setsockopt = vsock_connectible_setsockopt,
2116 .getsockopt = vsock_connectible_getsockopt,
2117 .sendmsg = vsock_connectible_sendmsg,
2118 .recvmsg = vsock_connectible_recvmsg,
2119 .mmap = sock_no_mmap,
2120 .sendpage = sock_no_sendpage,
2123 static int vsock_create(struct net *net, struct socket *sock,
2124 int protocol, int kern)
2126 struct vsock_sock *vsk;
2133 if (protocol && protocol != PF_VSOCK)
2134 return -EPROTONOSUPPORT;
2136 switch (sock->type) {
2138 sock->ops = &vsock_dgram_ops;
2141 sock->ops = &vsock_stream_ops;
2144 return -ESOCKTNOSUPPORT;
2147 sock->state = SS_UNCONNECTED;
2149 sk = __vsock_create(net, sock, NULL, GFP_KERNEL, 0, kern);
2155 if (sock->type == SOCK_DGRAM) {
2156 ret = vsock_assign_transport(vsk, NULL);
2163 vsock_insert_unbound(vsk);
2168 static const struct net_proto_family vsock_family_ops = {
2170 .create = vsock_create,
2171 .owner = THIS_MODULE,
2174 static long vsock_dev_do_ioctl(struct file *filp,
2175 unsigned int cmd, void __user *ptr)
2177 u32 __user *p = ptr;
2178 u32 cid = VMADDR_CID_ANY;
2182 case IOCTL_VM_SOCKETS_GET_LOCAL_CID:
2183 /* To be compatible with the VMCI behavior, we prioritize the
2184 * guest CID instead of well-know host CID (VMADDR_CID_HOST).
2187 cid = transport_g2h->get_local_cid();
2188 else if (transport_h2g)
2189 cid = transport_h2g->get_local_cid();
2191 if (put_user(cid, p) != 0)
2196 retval = -ENOIOCTLCMD;
2202 static long vsock_dev_ioctl(struct file *filp,
2203 unsigned int cmd, unsigned long arg)
2205 return vsock_dev_do_ioctl(filp, cmd, (void __user *)arg);
2208 #ifdef CONFIG_COMPAT
2209 static long vsock_dev_compat_ioctl(struct file *filp,
2210 unsigned int cmd, unsigned long arg)
2212 return vsock_dev_do_ioctl(filp, cmd, compat_ptr(arg));
2216 static const struct file_operations vsock_device_ops = {
2217 .owner = THIS_MODULE,
2218 .unlocked_ioctl = vsock_dev_ioctl,
2219 #ifdef CONFIG_COMPAT
2220 .compat_ioctl = vsock_dev_compat_ioctl,
2222 .open = nonseekable_open,
2225 static struct miscdevice vsock_device = {
2227 .fops = &vsock_device_ops,
2230 static int __init vsock_init(void)
2234 vsock_init_tables();
2236 vsock_proto.owner = THIS_MODULE;
2237 vsock_device.minor = MISC_DYNAMIC_MINOR;
2238 err = misc_register(&vsock_device);
2240 pr_err("Failed to register misc device\n");
2241 goto err_reset_transport;
2244 err = proto_register(&vsock_proto, 1); /* we want our slab */
2246 pr_err("Cannot register vsock protocol\n");
2247 goto err_deregister_misc;
2250 err = sock_register(&vsock_family_ops);
2252 pr_err("could not register af_vsock (%d) address family: %d\n",
2254 goto err_unregister_proto;
2259 err_unregister_proto:
2260 proto_unregister(&vsock_proto);
2261 err_deregister_misc:
2262 misc_deregister(&vsock_device);
2263 err_reset_transport:
2267 static void __exit vsock_exit(void)
2269 misc_deregister(&vsock_device);
2270 sock_unregister(AF_VSOCK);
2271 proto_unregister(&vsock_proto);
2274 const struct vsock_transport *vsock_core_get_transport(struct vsock_sock *vsk)
2276 return vsk->transport;
2278 EXPORT_SYMBOL_GPL(vsock_core_get_transport);
2280 int vsock_core_register(const struct vsock_transport *t, int features)
2282 const struct vsock_transport *t_h2g, *t_g2h, *t_dgram, *t_local;
2283 int err = mutex_lock_interruptible(&vsock_register_mutex);
2288 t_h2g = transport_h2g;
2289 t_g2h = transport_g2h;
2290 t_dgram = transport_dgram;
2291 t_local = transport_local;
2293 if (features & VSOCK_TRANSPORT_F_H2G) {
2301 if (features & VSOCK_TRANSPORT_F_G2H) {
2309 if (features & VSOCK_TRANSPORT_F_DGRAM) {
2317 if (features & VSOCK_TRANSPORT_F_LOCAL) {
2325 transport_h2g = t_h2g;
2326 transport_g2h = t_g2h;
2327 transport_dgram = t_dgram;
2328 transport_local = t_local;
2331 mutex_unlock(&vsock_register_mutex);
2334 EXPORT_SYMBOL_GPL(vsock_core_register);
2336 void vsock_core_unregister(const struct vsock_transport *t)
2338 mutex_lock(&vsock_register_mutex);
2340 if (transport_h2g == t)
2341 transport_h2g = NULL;
2343 if (transport_g2h == t)
2344 transport_g2h = NULL;
2346 if (transport_dgram == t)
2347 transport_dgram = NULL;
2349 if (transport_local == t)
2350 transport_local = NULL;
2352 mutex_unlock(&vsock_register_mutex);
2354 EXPORT_SYMBOL_GPL(vsock_core_unregister);
2356 module_init(vsock_init);
2357 module_exit(vsock_exit);
2359 MODULE_AUTHOR("VMware, Inc.");
2360 MODULE_DESCRIPTION("VMware Virtual Socket Family");
2361 MODULE_VERSION("1.0.2.0-k");
2362 MODULE_LICENSE("GPL v2");