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 connection oriented socket this must be called when vsk->remote_addr
424 * is set (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 or h2g is not loaded or remote flags field
430 * includes VMADDR_FLAG_TO_HOST flag value, will use guest->host transport;
431 * - remote CID > VMADDR_CID_HOST will use host->guest transport;
433 int vsock_assign_transport(struct vsock_sock *vsk, struct vsock_sock *psk)
435 const struct vsock_transport *new_transport;
436 struct sock *sk = sk_vsock(vsk);
437 unsigned int remote_cid = vsk->remote_addr.svm_cid;
441 /* If the packet is coming with the source and destination CIDs higher
442 * than VMADDR_CID_HOST, then a vsock channel where all the packets are
443 * forwarded to the host should be established. Then the host will
444 * need to forward the packets to the guest.
446 * The flag is set on the (listen) receive path (psk is not NULL). On
447 * the connect path the flag can be set by the user space application.
449 if (psk && vsk->local_addr.svm_cid > VMADDR_CID_HOST &&
450 vsk->remote_addr.svm_cid > VMADDR_CID_HOST)
451 vsk->remote_addr.svm_flags |= VMADDR_FLAG_TO_HOST;
453 remote_flags = vsk->remote_addr.svm_flags;
455 switch (sk->sk_type) {
457 new_transport = transport_dgram;
461 if (vsock_use_local_transport(remote_cid))
462 new_transport = transport_local;
463 else if (remote_cid <= VMADDR_CID_HOST || !transport_h2g ||
464 (remote_flags & VMADDR_FLAG_TO_HOST))
465 new_transport = transport_g2h;
467 new_transport = transport_h2g;
470 return -ESOCKTNOSUPPORT;
473 if (vsk->transport) {
474 if (vsk->transport == new_transport)
477 /* transport->release() must be called with sock lock acquired.
478 * This path can only be taken during vsock_connect(), where we
479 * have already held the sock lock. In the other cases, this
480 * function is called on a new socket which is not assigned to
483 vsk->transport->release(vsk);
484 vsock_deassign_transport(vsk);
487 /* We increase the module refcnt to prevent the transport unloading
488 * while there are open sockets assigned to it.
490 if (!new_transport || !try_module_get(new_transport->module))
493 if (sk->sk_type == SOCK_SEQPACKET) {
494 if (!new_transport->seqpacket_allow ||
495 !new_transport->seqpacket_allow(remote_cid)) {
496 module_put(new_transport->module);
497 return -ESOCKTNOSUPPORT;
501 ret = new_transport->init(vsk, psk);
503 module_put(new_transport->module);
507 vsk->transport = new_transport;
511 EXPORT_SYMBOL_GPL(vsock_assign_transport);
513 bool vsock_find_cid(unsigned int cid)
515 if (transport_g2h && cid == transport_g2h->get_local_cid())
518 if (transport_h2g && cid == VMADDR_CID_HOST)
521 if (transport_local && cid == VMADDR_CID_LOCAL)
526 EXPORT_SYMBOL_GPL(vsock_find_cid);
528 static struct sock *vsock_dequeue_accept(struct sock *listener)
530 struct vsock_sock *vlistener;
531 struct vsock_sock *vconnected;
533 vlistener = vsock_sk(listener);
535 if (list_empty(&vlistener->accept_queue))
538 vconnected = list_entry(vlistener->accept_queue.next,
539 struct vsock_sock, accept_queue);
541 list_del_init(&vconnected->accept_queue);
543 /* The caller will need a reference on the connected socket so we let
544 * it call sock_put().
547 return sk_vsock(vconnected);
550 static bool vsock_is_accept_queue_empty(struct sock *sk)
552 struct vsock_sock *vsk = vsock_sk(sk);
553 return list_empty(&vsk->accept_queue);
556 static bool vsock_is_pending(struct sock *sk)
558 struct vsock_sock *vsk = vsock_sk(sk);
559 return !list_empty(&vsk->pending_links);
562 static int vsock_send_shutdown(struct sock *sk, int mode)
564 struct vsock_sock *vsk = vsock_sk(sk);
569 return vsk->transport->shutdown(vsk, mode);
572 static void vsock_pending_work(struct work_struct *work)
575 struct sock *listener;
576 struct vsock_sock *vsk;
579 vsk = container_of(work, struct vsock_sock, pending_work.work);
581 listener = vsk->listener;
585 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
587 if (vsock_is_pending(sk)) {
588 vsock_remove_pending(listener, sk);
590 sk_acceptq_removed(listener);
591 } else if (!vsk->rejected) {
592 /* We are not on the pending list and accept() did not reject
593 * us, so we must have been accepted by our user process. We
594 * just need to drop our references to the sockets and be on
601 /* We need to remove ourself from the global connected sockets list so
602 * incoming packets can't find this socket, and to reduce the reference
605 vsock_remove_connected(vsk);
607 sk->sk_state = TCP_CLOSE;
611 release_sock(listener);
619 /**** SOCKET OPERATIONS ****/
621 static int __vsock_bind_connectible(struct vsock_sock *vsk,
622 struct sockaddr_vm *addr)
625 struct sockaddr_vm new_addr;
628 port = LAST_RESERVED_PORT + 1 +
629 prandom_u32_max(U32_MAX - LAST_RESERVED_PORT);
631 vsock_addr_init(&new_addr, addr->svm_cid, addr->svm_port);
633 if (addr->svm_port == VMADDR_PORT_ANY) {
637 for (i = 0; i < MAX_PORT_RETRIES; i++) {
638 if (port <= LAST_RESERVED_PORT)
639 port = LAST_RESERVED_PORT + 1;
641 new_addr.svm_port = port++;
643 if (!__vsock_find_bound_socket(&new_addr)) {
650 return -EADDRNOTAVAIL;
652 /* If port is in reserved range, ensure caller
653 * has necessary privileges.
655 if (addr->svm_port <= LAST_RESERVED_PORT &&
656 !capable(CAP_NET_BIND_SERVICE)) {
660 if (__vsock_find_bound_socket(&new_addr))
664 vsock_addr_init(&vsk->local_addr, new_addr.svm_cid, new_addr.svm_port);
666 /* Remove connection oriented sockets from the unbound list and add them
667 * to the hash table for easy lookup by its address. The unbound list
668 * is simply an extra entry at the end of the hash table, a trick used
671 __vsock_remove_bound(vsk);
672 __vsock_insert_bound(vsock_bound_sockets(&vsk->local_addr), vsk);
677 static int __vsock_bind_dgram(struct vsock_sock *vsk,
678 struct sockaddr_vm *addr)
680 return vsk->transport->dgram_bind(vsk, addr);
683 static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr)
685 struct vsock_sock *vsk = vsock_sk(sk);
688 /* First ensure this socket isn't already bound. */
689 if (vsock_addr_bound(&vsk->local_addr))
692 /* Now bind to the provided address or select appropriate values if
693 * none are provided (VMADDR_CID_ANY and VMADDR_PORT_ANY). Note that
694 * like AF_INET prevents binding to a non-local IP address (in most
695 * cases), we only allow binding to a local CID.
697 if (addr->svm_cid != VMADDR_CID_ANY && !vsock_find_cid(addr->svm_cid))
698 return -EADDRNOTAVAIL;
700 switch (sk->sk_socket->type) {
703 spin_lock_bh(&vsock_table_lock);
704 retval = __vsock_bind_connectible(vsk, addr);
705 spin_unlock_bh(&vsock_table_lock);
709 retval = __vsock_bind_dgram(vsk, addr);
720 static void vsock_connect_timeout(struct work_struct *work);
722 static struct sock *__vsock_create(struct net *net,
730 struct vsock_sock *psk;
731 struct vsock_sock *vsk;
733 sk = sk_alloc(net, AF_VSOCK, priority, &vsock_proto, kern);
737 sock_init_data(sock, sk);
739 /* sk->sk_type is normally set in sock_init_data, but only if sock is
740 * non-NULL. We make sure that our sockets always have a type by
741 * setting it here if needed.
747 vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
748 vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
750 sk->sk_destruct = vsock_sk_destruct;
751 sk->sk_backlog_rcv = vsock_queue_rcv_skb;
752 sock_reset_flag(sk, SOCK_DONE);
754 INIT_LIST_HEAD(&vsk->bound_table);
755 INIT_LIST_HEAD(&vsk->connected_table);
756 vsk->listener = NULL;
757 INIT_LIST_HEAD(&vsk->pending_links);
758 INIT_LIST_HEAD(&vsk->accept_queue);
759 vsk->rejected = false;
760 vsk->sent_request = false;
761 vsk->ignore_connecting_rst = false;
762 vsk->peer_shutdown = 0;
763 INIT_DELAYED_WORK(&vsk->connect_work, vsock_connect_timeout);
764 INIT_DELAYED_WORK(&vsk->pending_work, vsock_pending_work);
766 psk = parent ? vsock_sk(parent) : NULL;
768 vsk->trusted = psk->trusted;
769 vsk->owner = get_cred(psk->owner);
770 vsk->connect_timeout = psk->connect_timeout;
771 vsk->buffer_size = psk->buffer_size;
772 vsk->buffer_min_size = psk->buffer_min_size;
773 vsk->buffer_max_size = psk->buffer_max_size;
774 security_sk_clone(parent, sk);
776 vsk->trusted = ns_capable_noaudit(&init_user_ns, CAP_NET_ADMIN);
777 vsk->owner = get_current_cred();
778 vsk->connect_timeout = VSOCK_DEFAULT_CONNECT_TIMEOUT;
779 vsk->buffer_size = VSOCK_DEFAULT_BUFFER_SIZE;
780 vsk->buffer_min_size = VSOCK_DEFAULT_BUFFER_MIN_SIZE;
781 vsk->buffer_max_size = VSOCK_DEFAULT_BUFFER_MAX_SIZE;
787 static bool sock_type_connectible(u16 type)
789 return (type == SOCK_STREAM) || (type == SOCK_SEQPACKET);
792 static void __vsock_release(struct sock *sk, int level)
795 struct sock *pending;
796 struct vsock_sock *vsk;
799 pending = NULL; /* Compiler warning. */
801 /* When "level" is SINGLE_DEPTH_NESTING, use the nested
802 * version to avoid the warning "possible recursive locking
803 * detected". When "level" is 0, lock_sock_nested(sk, level)
804 * is the same as lock_sock(sk).
806 lock_sock_nested(sk, level);
809 vsk->transport->release(vsk);
810 else if (sock_type_connectible(sk->sk_type))
811 vsock_remove_sock(vsk);
814 sk->sk_shutdown = SHUTDOWN_MASK;
816 skb_queue_purge(&sk->sk_receive_queue);
818 /* Clean up any sockets that never were accepted. */
819 while ((pending = vsock_dequeue_accept(sk)) != NULL) {
820 __vsock_release(pending, SINGLE_DEPTH_NESTING);
829 static void vsock_sk_destruct(struct sock *sk)
831 struct vsock_sock *vsk = vsock_sk(sk);
833 vsock_deassign_transport(vsk);
835 /* When clearing these addresses, there's no need to set the family and
836 * possibly register the address family with the kernel.
838 vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
839 vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
841 put_cred(vsk->owner);
844 static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
848 err = sock_queue_rcv_skb(sk, skb);
855 struct sock *vsock_create_connected(struct sock *parent)
857 return __vsock_create(sock_net(parent), NULL, parent, GFP_KERNEL,
860 EXPORT_SYMBOL_GPL(vsock_create_connected);
862 s64 vsock_stream_has_data(struct vsock_sock *vsk)
864 return vsk->transport->stream_has_data(vsk);
866 EXPORT_SYMBOL_GPL(vsock_stream_has_data);
868 static s64 vsock_connectible_has_data(struct vsock_sock *vsk)
870 struct sock *sk = sk_vsock(vsk);
872 if (sk->sk_type == SOCK_SEQPACKET)
873 return vsk->transport->seqpacket_has_data(vsk);
875 return vsock_stream_has_data(vsk);
878 s64 vsock_stream_has_space(struct vsock_sock *vsk)
880 return vsk->transport->stream_has_space(vsk);
882 EXPORT_SYMBOL_GPL(vsock_stream_has_space);
884 static int vsock_release(struct socket *sock)
886 __vsock_release(sock->sk, 0);
888 sock->state = SS_FREE;
894 vsock_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
898 struct sockaddr_vm *vm_addr;
902 if (vsock_addr_cast(addr, addr_len, &vm_addr) != 0)
906 err = __vsock_bind(sk, vm_addr);
912 static int vsock_getname(struct socket *sock,
913 struct sockaddr *addr, int peer)
917 struct vsock_sock *vsk;
918 struct sockaddr_vm *vm_addr;
927 if (sock->state != SS_CONNECTED) {
931 vm_addr = &vsk->remote_addr;
933 vm_addr = &vsk->local_addr;
941 /* sys_getsockname() and sys_getpeername() pass us a
942 * MAX_SOCK_ADDR-sized buffer and don't set addr_len. Unfortunately
943 * that macro is defined in socket.c instead of .h, so we hardcode its
946 BUILD_BUG_ON(sizeof(*vm_addr) > 128);
947 memcpy(addr, vm_addr, sizeof(*vm_addr));
948 err = sizeof(*vm_addr);
955 static int vsock_shutdown(struct socket *sock, int mode)
960 /* User level uses SHUT_RD (0) and SHUT_WR (1), but the kernel uses
961 * RCV_SHUTDOWN (1) and SEND_SHUTDOWN (2), so we must increment mode
962 * here like the other address families do. Note also that the
963 * increment makes SHUT_RDWR (2) into RCV_SHUTDOWN | SEND_SHUTDOWN (3),
964 * which is what we want.
968 if ((mode & ~SHUTDOWN_MASK) || !mode)
971 /* If this is a connection oriented socket and it is not connected then
972 * bail out immediately. If it is a DGRAM socket then we must first
973 * kick the socket so that it wakes up from any sleeping calls, for
974 * example recv(), and then afterwards return the error.
980 if (sock->state == SS_UNCONNECTED) {
982 if (sock_type_connectible(sk->sk_type))
985 sock->state = SS_DISCONNECTING;
989 /* Receive and send shutdowns are treated alike. */
990 mode = mode & (RCV_SHUTDOWN | SEND_SHUTDOWN);
992 sk->sk_shutdown |= mode;
993 sk->sk_state_change(sk);
995 if (sock_type_connectible(sk->sk_type)) {
996 sock_reset_flag(sk, SOCK_DONE);
997 vsock_send_shutdown(sk, mode);
1006 static __poll_t vsock_poll(struct file *file, struct socket *sock,
1011 struct vsock_sock *vsk;
1016 poll_wait(file, sk_sleep(sk), wait);
1020 /* Signify that there has been an error on this socket. */
1023 /* INET sockets treat local write shutdown and peer write shutdown as a
1024 * case of EPOLLHUP set.
1026 if ((sk->sk_shutdown == SHUTDOWN_MASK) ||
1027 ((sk->sk_shutdown & SEND_SHUTDOWN) &&
1028 (vsk->peer_shutdown & SEND_SHUTDOWN))) {
1032 if (sk->sk_shutdown & RCV_SHUTDOWN ||
1033 vsk->peer_shutdown & SEND_SHUTDOWN) {
1037 if (sock->type == SOCK_DGRAM) {
1038 /* For datagram sockets we can read if there is something in
1039 * the queue and write as long as the socket isn't shutdown for
1042 if (!skb_queue_empty_lockless(&sk->sk_receive_queue) ||
1043 (sk->sk_shutdown & RCV_SHUTDOWN)) {
1044 mask |= EPOLLIN | EPOLLRDNORM;
1047 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
1048 mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND;
1050 } else if (sock_type_connectible(sk->sk_type)) {
1051 const struct vsock_transport *transport;
1055 transport = vsk->transport;
1057 /* Listening sockets that have connections in their accept
1058 * queue can be read.
1060 if (sk->sk_state == TCP_LISTEN
1061 && !vsock_is_accept_queue_empty(sk))
1062 mask |= EPOLLIN | EPOLLRDNORM;
1064 /* If there is something in the queue then we can read. */
1065 if (transport && transport->stream_is_active(vsk) &&
1066 !(sk->sk_shutdown & RCV_SHUTDOWN)) {
1067 bool data_ready_now = false;
1068 int ret = transport->notify_poll_in(
1069 vsk, 1, &data_ready_now);
1074 mask |= EPOLLIN | EPOLLRDNORM;
1079 /* Sockets whose connections have been closed, reset, or
1080 * terminated should also be considered read, and we check the
1081 * shutdown flag for that.
1083 if (sk->sk_shutdown & RCV_SHUTDOWN ||
1084 vsk->peer_shutdown & SEND_SHUTDOWN) {
1085 mask |= EPOLLIN | EPOLLRDNORM;
1088 /* Connected sockets that can produce data can be written. */
1089 if (transport && sk->sk_state == TCP_ESTABLISHED) {
1090 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1091 bool space_avail_now = false;
1092 int ret = transport->notify_poll_out(
1093 vsk, 1, &space_avail_now);
1097 if (space_avail_now)
1098 /* Remove EPOLLWRBAND since INET
1099 * sockets are not setting it.
1101 mask |= EPOLLOUT | EPOLLWRNORM;
1107 /* Simulate INET socket poll behaviors, which sets
1108 * EPOLLOUT|EPOLLWRNORM when peer is closed and nothing to read,
1109 * but local send is not shutdown.
1111 if (sk->sk_state == TCP_CLOSE || sk->sk_state == TCP_CLOSING) {
1112 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
1113 mask |= EPOLLOUT | EPOLLWRNORM;
1123 static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg,
1128 struct vsock_sock *vsk;
1129 struct sockaddr_vm *remote_addr;
1130 const struct vsock_transport *transport;
1132 if (msg->msg_flags & MSG_OOB)
1135 /* For now, MSG_DONTWAIT is always assumed... */
1142 transport = vsk->transport;
1144 err = vsock_auto_bind(vsk);
1149 /* If the provided message contains an address, use that. Otherwise
1150 * fall back on the socket's remote handle (if it has been connected).
1152 if (msg->msg_name &&
1153 vsock_addr_cast(msg->msg_name, msg->msg_namelen,
1154 &remote_addr) == 0) {
1155 /* Ensure this address is of the right type and is a valid
1159 if (remote_addr->svm_cid == VMADDR_CID_ANY)
1160 remote_addr->svm_cid = transport->get_local_cid();
1162 if (!vsock_addr_bound(remote_addr)) {
1166 } else if (sock->state == SS_CONNECTED) {
1167 remote_addr = &vsk->remote_addr;
1169 if (remote_addr->svm_cid == VMADDR_CID_ANY)
1170 remote_addr->svm_cid = transport->get_local_cid();
1172 /* XXX Should connect() or this function ensure remote_addr is
1175 if (!vsock_addr_bound(&vsk->remote_addr)) {
1184 if (!transport->dgram_allow(remote_addr->svm_cid,
1185 remote_addr->svm_port)) {
1190 err = transport->dgram_enqueue(vsk, remote_addr, msg, len);
1197 static int vsock_dgram_connect(struct socket *sock,
1198 struct sockaddr *addr, int addr_len, int flags)
1202 struct vsock_sock *vsk;
1203 struct sockaddr_vm *remote_addr;
1208 err = vsock_addr_cast(addr, addr_len, &remote_addr);
1209 if (err == -EAFNOSUPPORT && remote_addr->svm_family == AF_UNSPEC) {
1211 vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY,
1213 sock->state = SS_UNCONNECTED;
1216 } else if (err != 0)
1221 err = vsock_auto_bind(vsk);
1225 if (!vsk->transport->dgram_allow(remote_addr->svm_cid,
1226 remote_addr->svm_port)) {
1231 memcpy(&vsk->remote_addr, remote_addr, sizeof(vsk->remote_addr));
1232 sock->state = SS_CONNECTED;
1239 static int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg,
1240 size_t len, int flags)
1242 struct vsock_sock *vsk = vsock_sk(sock->sk);
1244 return vsk->transport->dgram_dequeue(vsk, msg, len, flags);
1247 static const struct proto_ops vsock_dgram_ops = {
1249 .owner = THIS_MODULE,
1250 .release = vsock_release,
1252 .connect = vsock_dgram_connect,
1253 .socketpair = sock_no_socketpair,
1254 .accept = sock_no_accept,
1255 .getname = vsock_getname,
1257 .ioctl = sock_no_ioctl,
1258 .listen = sock_no_listen,
1259 .shutdown = vsock_shutdown,
1260 .sendmsg = vsock_dgram_sendmsg,
1261 .recvmsg = vsock_dgram_recvmsg,
1262 .mmap = sock_no_mmap,
1263 .sendpage = sock_no_sendpage,
1266 static int vsock_transport_cancel_pkt(struct vsock_sock *vsk)
1268 const struct vsock_transport *transport = vsk->transport;
1270 if (!transport || !transport->cancel_pkt)
1273 return transport->cancel_pkt(vsk);
1276 static void vsock_connect_timeout(struct work_struct *work)
1279 struct vsock_sock *vsk;
1281 vsk = container_of(work, struct vsock_sock, connect_work.work);
1285 if (sk->sk_state == TCP_SYN_SENT &&
1286 (sk->sk_shutdown != SHUTDOWN_MASK)) {
1287 sk->sk_state = TCP_CLOSE;
1288 sk->sk_socket->state = SS_UNCONNECTED;
1289 sk->sk_err = ETIMEDOUT;
1290 sk_error_report(sk);
1291 vsock_transport_cancel_pkt(vsk);
1298 static int vsock_connect(struct socket *sock, struct sockaddr *addr,
1299 int addr_len, int flags)
1303 struct vsock_sock *vsk;
1304 const struct vsock_transport *transport;
1305 struct sockaddr_vm *remote_addr;
1315 /* XXX AF_UNSPEC should make us disconnect like AF_INET. */
1316 switch (sock->state) {
1320 case SS_DISCONNECTING:
1324 /* This continues on so we can move sock into the SS_CONNECTED
1325 * state once the connection has completed (at which point err
1326 * will be set to zero also). Otherwise, we will either wait
1327 * for the connection or return -EALREADY should this be a
1328 * non-blocking call.
1331 if (flags & O_NONBLOCK)
1335 if ((sk->sk_state == TCP_LISTEN) ||
1336 vsock_addr_cast(addr, addr_len, &remote_addr) != 0) {
1341 /* Set the remote address that we are connecting to. */
1342 memcpy(&vsk->remote_addr, remote_addr,
1343 sizeof(vsk->remote_addr));
1345 err = vsock_assign_transport(vsk, NULL);
1349 transport = vsk->transport;
1351 /* The hypervisor and well-known contexts do not have socket
1355 !transport->stream_allow(remote_addr->svm_cid,
1356 remote_addr->svm_port)) {
1361 err = vsock_auto_bind(vsk);
1365 sk->sk_state = TCP_SYN_SENT;
1367 err = transport->connect(vsk);
1371 /* Mark sock as connecting and set the error code to in
1372 * progress in case this is a non-blocking connect.
1374 sock->state = SS_CONNECTING;
1378 /* The receive path will handle all communication until we are able to
1379 * enter the connected state. Here we wait for the connection to be
1380 * completed or a notification of an error.
1382 timeout = vsk->connect_timeout;
1383 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1385 while (sk->sk_state != TCP_ESTABLISHED && sk->sk_err == 0) {
1386 if (flags & O_NONBLOCK) {
1387 /* If we're not going to block, we schedule a timeout
1388 * function to generate a timeout on the connection
1389 * attempt, in case the peer doesn't respond in a
1390 * timely manner. We hold on to the socket until the
1395 /* If the timeout function is already scheduled,
1396 * reschedule it, then ungrab the socket refcount to
1399 if (mod_delayed_work(system_wq, &vsk->connect_work,
1403 /* Skip ahead to preserve error code set above. */
1408 timeout = schedule_timeout(timeout);
1411 if (signal_pending(current)) {
1412 err = sock_intr_errno(timeout);
1413 sk->sk_state = sk->sk_state == TCP_ESTABLISHED ? TCP_CLOSING : TCP_CLOSE;
1414 sock->state = SS_UNCONNECTED;
1415 vsock_transport_cancel_pkt(vsk);
1416 vsock_remove_connected(vsk);
1418 } else if ((sk->sk_state != TCP_ESTABLISHED) && (timeout == 0)) {
1420 sk->sk_state = TCP_CLOSE;
1421 sock->state = SS_UNCONNECTED;
1422 vsock_transport_cancel_pkt(vsk);
1426 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1431 sk->sk_state = TCP_CLOSE;
1432 sock->state = SS_UNCONNECTED;
1438 finish_wait(sk_sleep(sk), &wait);
1444 static int vsock_accept(struct socket *sock, struct socket *newsock, int flags,
1447 struct sock *listener;
1449 struct sock *connected;
1450 struct vsock_sock *vconnected;
1455 listener = sock->sk;
1457 lock_sock(listener);
1459 if (!sock_type_connectible(sock->type)) {
1464 if (listener->sk_state != TCP_LISTEN) {
1469 /* Wait for children sockets to appear; these are the new sockets
1470 * created upon connection establishment.
1472 timeout = sock_rcvtimeo(listener, flags & O_NONBLOCK);
1473 prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1475 while ((connected = vsock_dequeue_accept(listener)) == NULL &&
1476 listener->sk_err == 0) {
1477 release_sock(listener);
1478 timeout = schedule_timeout(timeout);
1479 finish_wait(sk_sleep(listener), &wait);
1480 lock_sock(listener);
1482 if (signal_pending(current)) {
1483 err = sock_intr_errno(timeout);
1485 } else if (timeout == 0) {
1490 prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1492 finish_wait(sk_sleep(listener), &wait);
1494 if (listener->sk_err)
1495 err = -listener->sk_err;
1498 sk_acceptq_removed(listener);
1500 lock_sock_nested(connected, SINGLE_DEPTH_NESTING);
1501 vconnected = vsock_sk(connected);
1503 /* If the listener socket has received an error, then we should
1504 * reject this socket and return. Note that we simply mark the
1505 * socket rejected, drop our reference, and let the cleanup
1506 * function handle the cleanup; the fact that we found it in
1507 * the listener's accept queue guarantees that the cleanup
1508 * function hasn't run yet.
1511 vconnected->rejected = true;
1513 newsock->state = SS_CONNECTED;
1514 sock_graft(connected, newsock);
1517 release_sock(connected);
1518 sock_put(connected);
1522 release_sock(listener);
1526 static int vsock_listen(struct socket *sock, int backlog)
1530 struct vsock_sock *vsk;
1536 if (!sock_type_connectible(sk->sk_type)) {
1541 if (sock->state != SS_UNCONNECTED) {
1548 if (!vsock_addr_bound(&vsk->local_addr)) {
1553 sk->sk_max_ack_backlog = backlog;
1554 sk->sk_state = TCP_LISTEN;
1563 static void vsock_update_buffer_size(struct vsock_sock *vsk,
1564 const struct vsock_transport *transport,
1567 if (val > vsk->buffer_max_size)
1568 val = vsk->buffer_max_size;
1570 if (val < vsk->buffer_min_size)
1571 val = vsk->buffer_min_size;
1573 if (val != vsk->buffer_size &&
1574 transport && transport->notify_buffer_size)
1575 transport->notify_buffer_size(vsk, &val);
1577 vsk->buffer_size = val;
1580 static int vsock_connectible_setsockopt(struct socket *sock,
1584 unsigned int optlen)
1588 struct vsock_sock *vsk;
1589 const struct vsock_transport *transport;
1592 if (level != AF_VSOCK)
1593 return -ENOPROTOOPT;
1595 #define COPY_IN(_v) \
1597 if (optlen < sizeof(_v)) { \
1601 if (copy_from_sockptr(&_v, optval, sizeof(_v)) != 0) { \
1613 transport = vsk->transport;
1616 case SO_VM_SOCKETS_BUFFER_SIZE:
1618 vsock_update_buffer_size(vsk, transport, val);
1621 case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1623 vsk->buffer_max_size = val;
1624 vsock_update_buffer_size(vsk, transport, vsk->buffer_size);
1627 case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1629 vsk->buffer_min_size = val;
1630 vsock_update_buffer_size(vsk, transport, vsk->buffer_size);
1633 case SO_VM_SOCKETS_CONNECT_TIMEOUT: {
1634 struct __kernel_old_timeval tv;
1636 if (tv.tv_sec >= 0 && tv.tv_usec < USEC_PER_SEC &&
1637 tv.tv_sec < (MAX_SCHEDULE_TIMEOUT / HZ - 1)) {
1638 vsk->connect_timeout = tv.tv_sec * HZ +
1639 DIV_ROUND_UP(tv.tv_usec, (1000000 / HZ));
1640 if (vsk->connect_timeout == 0)
1641 vsk->connect_timeout =
1642 VSOCK_DEFAULT_CONNECT_TIMEOUT;
1662 static int vsock_connectible_getsockopt(struct socket *sock,
1663 int level, int optname,
1664 char __user *optval,
1670 struct vsock_sock *vsk;
1673 if (level != AF_VSOCK)
1674 return -ENOPROTOOPT;
1676 err = get_user(len, optlen);
1680 #define COPY_OUT(_v) \
1682 if (len < sizeof(_v)) \
1686 if (copy_to_user(optval, &_v, len) != 0) \
1696 case SO_VM_SOCKETS_BUFFER_SIZE:
1697 val = vsk->buffer_size;
1701 case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1702 val = vsk->buffer_max_size;
1706 case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1707 val = vsk->buffer_min_size;
1711 case SO_VM_SOCKETS_CONNECT_TIMEOUT: {
1712 struct __kernel_old_timeval tv;
1713 tv.tv_sec = vsk->connect_timeout / HZ;
1715 (vsk->connect_timeout -
1716 tv.tv_sec * HZ) * (1000000 / HZ);
1721 return -ENOPROTOOPT;
1724 err = put_user(len, optlen);
1733 static int vsock_connectible_sendmsg(struct socket *sock, struct msghdr *msg,
1737 struct vsock_sock *vsk;
1738 const struct vsock_transport *transport;
1739 ssize_t total_written;
1742 struct vsock_transport_send_notify_data send_data;
1743 DEFINE_WAIT_FUNC(wait, woken_wake_function);
1750 if (msg->msg_flags & MSG_OOB)
1755 transport = vsk->transport;
1757 /* Callers should not provide a destination with connection oriented
1760 if (msg->msg_namelen) {
1761 err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP;
1765 /* Send data only if both sides are not shutdown in the direction. */
1766 if (sk->sk_shutdown & SEND_SHUTDOWN ||
1767 vsk->peer_shutdown & RCV_SHUTDOWN) {
1772 if (!transport || sk->sk_state != TCP_ESTABLISHED ||
1773 !vsock_addr_bound(&vsk->local_addr)) {
1778 if (!vsock_addr_bound(&vsk->remote_addr)) {
1779 err = -EDESTADDRREQ;
1783 /* Wait for room in the produce queue to enqueue our user's data. */
1784 timeout = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1786 err = transport->notify_send_init(vsk, &send_data);
1790 while (total_written < len) {
1793 add_wait_queue(sk_sleep(sk), &wait);
1794 while (vsock_stream_has_space(vsk) == 0 &&
1796 !(sk->sk_shutdown & SEND_SHUTDOWN) &&
1797 !(vsk->peer_shutdown & RCV_SHUTDOWN)) {
1799 /* Don't wait for non-blocking sockets. */
1802 remove_wait_queue(sk_sleep(sk), &wait);
1806 err = transport->notify_send_pre_block(vsk, &send_data);
1808 remove_wait_queue(sk_sleep(sk), &wait);
1813 timeout = wait_woken(&wait, TASK_INTERRUPTIBLE, timeout);
1815 if (signal_pending(current)) {
1816 err = sock_intr_errno(timeout);
1817 remove_wait_queue(sk_sleep(sk), &wait);
1819 } else if (timeout == 0) {
1821 remove_wait_queue(sk_sleep(sk), &wait);
1825 remove_wait_queue(sk_sleep(sk), &wait);
1827 /* These checks occur both as part of and after the loop
1828 * conditional since we need to check before and after
1834 } else if ((sk->sk_shutdown & SEND_SHUTDOWN) ||
1835 (vsk->peer_shutdown & RCV_SHUTDOWN)) {
1840 err = transport->notify_send_pre_enqueue(vsk, &send_data);
1844 /* Note that enqueue will only write as many bytes as are free
1845 * in the produce queue, so we don't need to ensure len is
1846 * smaller than the queue size. It is the caller's
1847 * responsibility to check how many bytes we were able to send.
1850 if (sk->sk_type == SOCK_SEQPACKET) {
1851 written = transport->seqpacket_enqueue(vsk,
1852 msg, len - total_written);
1854 written = transport->stream_enqueue(vsk,
1855 msg, len - total_written);
1862 total_written += written;
1864 err = transport->notify_send_post_enqueue(
1865 vsk, written, &send_data);
1872 if (total_written > 0) {
1873 /* Return number of written bytes only if:
1874 * 1) SOCK_STREAM socket.
1875 * 2) SOCK_SEQPACKET socket when whole buffer is sent.
1877 if (sk->sk_type == SOCK_STREAM || total_written == len)
1878 err = total_written;
1885 static int vsock_connectible_wait_data(struct sock *sk,
1886 struct wait_queue_entry *wait,
1888 struct vsock_transport_recv_notify_data *recv_data,
1891 const struct vsock_transport *transport;
1892 struct vsock_sock *vsk;
1898 transport = vsk->transport;
1901 prepare_to_wait(sk_sleep(sk), wait, TASK_INTERRUPTIBLE);
1902 data = vsock_connectible_has_data(vsk);
1906 if (sk->sk_err != 0 ||
1907 (sk->sk_shutdown & RCV_SHUTDOWN) ||
1908 (vsk->peer_shutdown & SEND_SHUTDOWN)) {
1912 /* Don't wait for non-blocking sockets. */
1919 err = transport->notify_recv_pre_block(vsk, target, recv_data);
1925 timeout = schedule_timeout(timeout);
1928 if (signal_pending(current)) {
1929 err = sock_intr_errno(timeout);
1931 } else if (timeout == 0) {
1937 finish_wait(sk_sleep(sk), wait);
1942 /* Internal transport error when checking for available
1943 * data. XXX This should be changed to a connection
1944 * reset in a later change.
1952 static int __vsock_stream_recvmsg(struct sock *sk, struct msghdr *msg,
1953 size_t len, int flags)
1955 struct vsock_transport_recv_notify_data recv_data;
1956 const struct vsock_transport *transport;
1957 struct vsock_sock *vsk;
1966 transport = vsk->transport;
1968 /* We must not copy less than target bytes into the user's buffer
1969 * before returning successfully, so we wait for the consume queue to
1970 * have that much data to consume before dequeueing. Note that this
1971 * makes it impossible to handle cases where target is greater than the
1974 target = sock_rcvlowat(sk, flags & MSG_WAITALL, len);
1975 if (target >= transport->stream_rcvhiwat(vsk)) {
1979 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1982 err = transport->notify_recv_init(vsk, target, &recv_data);
1990 err = vsock_connectible_wait_data(sk, &wait, timeout,
1991 &recv_data, target);
1995 err = transport->notify_recv_pre_dequeue(vsk, target,
2000 read = transport->stream_dequeue(vsk, msg, len - copied, flags);
2008 err = transport->notify_recv_post_dequeue(vsk, target, read,
2009 !(flags & MSG_PEEK), &recv_data);
2013 if (read >= target || flags & MSG_PEEK)
2021 else if (sk->sk_shutdown & RCV_SHUTDOWN)
2031 static int __vsock_seqpacket_recvmsg(struct sock *sk, struct msghdr *msg,
2032 size_t len, int flags)
2034 const struct vsock_transport *transport;
2035 struct vsock_sock *vsk;
2042 transport = vsk->transport;
2044 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2046 err = vsock_connectible_wait_data(sk, &wait, timeout, NULL, 0);
2050 msg_len = transport->seqpacket_dequeue(vsk, msg, flags);
2059 } else if (sk->sk_shutdown & RCV_SHUTDOWN) {
2062 /* User sets MSG_TRUNC, so return real length of
2065 if (flags & MSG_TRUNC)
2068 err = len - msg_data_left(msg);
2070 /* Always set MSG_TRUNC if real length of packet is
2071 * bigger than user's buffer.
2074 msg->msg_flags |= MSG_TRUNC;
2082 vsock_connectible_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
2086 struct vsock_sock *vsk;
2087 const struct vsock_transport *transport;
2098 transport = vsk->transport;
2100 if (!transport || sk->sk_state != TCP_ESTABLISHED) {
2101 /* Recvmsg is supposed to return 0 if a peer performs an
2102 * orderly shutdown. Differentiate between that case and when a
2103 * peer has not connected or a local shutdown occurred with the
2106 if (sock_flag(sk, SOCK_DONE))
2114 if (flags & MSG_OOB) {
2119 /* We don't check peer_shutdown flag here since peer may actually shut
2120 * down, but there can be data in the queue that a local socket can
2123 if (sk->sk_shutdown & RCV_SHUTDOWN) {
2128 /* It is valid on Linux to pass in a zero-length receive buffer. This
2129 * is not an error. We may as well bail out now.
2136 if (sk->sk_type == SOCK_STREAM)
2137 err = __vsock_stream_recvmsg(sk, msg, len, flags);
2139 err = __vsock_seqpacket_recvmsg(sk, msg, len, flags);
2146 static const struct proto_ops vsock_stream_ops = {
2148 .owner = THIS_MODULE,
2149 .release = vsock_release,
2151 .connect = vsock_connect,
2152 .socketpair = sock_no_socketpair,
2153 .accept = vsock_accept,
2154 .getname = vsock_getname,
2156 .ioctl = sock_no_ioctl,
2157 .listen = vsock_listen,
2158 .shutdown = vsock_shutdown,
2159 .setsockopt = vsock_connectible_setsockopt,
2160 .getsockopt = vsock_connectible_getsockopt,
2161 .sendmsg = vsock_connectible_sendmsg,
2162 .recvmsg = vsock_connectible_recvmsg,
2163 .mmap = sock_no_mmap,
2164 .sendpage = sock_no_sendpage,
2167 static const struct proto_ops vsock_seqpacket_ops = {
2169 .owner = THIS_MODULE,
2170 .release = vsock_release,
2172 .connect = vsock_connect,
2173 .socketpair = sock_no_socketpair,
2174 .accept = vsock_accept,
2175 .getname = vsock_getname,
2177 .ioctl = sock_no_ioctl,
2178 .listen = vsock_listen,
2179 .shutdown = vsock_shutdown,
2180 .setsockopt = vsock_connectible_setsockopt,
2181 .getsockopt = vsock_connectible_getsockopt,
2182 .sendmsg = vsock_connectible_sendmsg,
2183 .recvmsg = vsock_connectible_recvmsg,
2184 .mmap = sock_no_mmap,
2185 .sendpage = sock_no_sendpage,
2188 static int vsock_create(struct net *net, struct socket *sock,
2189 int protocol, int kern)
2191 struct vsock_sock *vsk;
2198 if (protocol && protocol != PF_VSOCK)
2199 return -EPROTONOSUPPORT;
2201 switch (sock->type) {
2203 sock->ops = &vsock_dgram_ops;
2206 sock->ops = &vsock_stream_ops;
2208 case SOCK_SEQPACKET:
2209 sock->ops = &vsock_seqpacket_ops;
2212 return -ESOCKTNOSUPPORT;
2215 sock->state = SS_UNCONNECTED;
2217 sk = __vsock_create(net, sock, NULL, GFP_KERNEL, 0, kern);
2223 if (sock->type == SOCK_DGRAM) {
2224 ret = vsock_assign_transport(vsk, NULL);
2231 vsock_insert_unbound(vsk);
2236 static const struct net_proto_family vsock_family_ops = {
2238 .create = vsock_create,
2239 .owner = THIS_MODULE,
2242 static long vsock_dev_do_ioctl(struct file *filp,
2243 unsigned int cmd, void __user *ptr)
2245 u32 __user *p = ptr;
2246 u32 cid = VMADDR_CID_ANY;
2250 case IOCTL_VM_SOCKETS_GET_LOCAL_CID:
2251 /* To be compatible with the VMCI behavior, we prioritize the
2252 * guest CID instead of well-know host CID (VMADDR_CID_HOST).
2255 cid = transport_g2h->get_local_cid();
2256 else if (transport_h2g)
2257 cid = transport_h2g->get_local_cid();
2259 if (put_user(cid, p) != 0)
2264 retval = -ENOIOCTLCMD;
2270 static long vsock_dev_ioctl(struct file *filp,
2271 unsigned int cmd, unsigned long arg)
2273 return vsock_dev_do_ioctl(filp, cmd, (void __user *)arg);
2276 #ifdef CONFIG_COMPAT
2277 static long vsock_dev_compat_ioctl(struct file *filp,
2278 unsigned int cmd, unsigned long arg)
2280 return vsock_dev_do_ioctl(filp, cmd, compat_ptr(arg));
2284 static const struct file_operations vsock_device_ops = {
2285 .owner = THIS_MODULE,
2286 .unlocked_ioctl = vsock_dev_ioctl,
2287 #ifdef CONFIG_COMPAT
2288 .compat_ioctl = vsock_dev_compat_ioctl,
2290 .open = nonseekable_open,
2293 static struct miscdevice vsock_device = {
2295 .fops = &vsock_device_ops,
2298 static int __init vsock_init(void)
2302 vsock_init_tables();
2304 vsock_proto.owner = THIS_MODULE;
2305 vsock_device.minor = MISC_DYNAMIC_MINOR;
2306 err = misc_register(&vsock_device);
2308 pr_err("Failed to register misc device\n");
2309 goto err_reset_transport;
2312 err = proto_register(&vsock_proto, 1); /* we want our slab */
2314 pr_err("Cannot register vsock protocol\n");
2315 goto err_deregister_misc;
2318 err = sock_register(&vsock_family_ops);
2320 pr_err("could not register af_vsock (%d) address family: %d\n",
2322 goto err_unregister_proto;
2327 err_unregister_proto:
2328 proto_unregister(&vsock_proto);
2329 err_deregister_misc:
2330 misc_deregister(&vsock_device);
2331 err_reset_transport:
2335 static void __exit vsock_exit(void)
2337 misc_deregister(&vsock_device);
2338 sock_unregister(AF_VSOCK);
2339 proto_unregister(&vsock_proto);
2342 const struct vsock_transport *vsock_core_get_transport(struct vsock_sock *vsk)
2344 return vsk->transport;
2346 EXPORT_SYMBOL_GPL(vsock_core_get_transport);
2348 int vsock_core_register(const struct vsock_transport *t, int features)
2350 const struct vsock_transport *t_h2g, *t_g2h, *t_dgram, *t_local;
2351 int err = mutex_lock_interruptible(&vsock_register_mutex);
2356 t_h2g = transport_h2g;
2357 t_g2h = transport_g2h;
2358 t_dgram = transport_dgram;
2359 t_local = transport_local;
2361 if (features & VSOCK_TRANSPORT_F_H2G) {
2369 if (features & VSOCK_TRANSPORT_F_G2H) {
2377 if (features & VSOCK_TRANSPORT_F_DGRAM) {
2385 if (features & VSOCK_TRANSPORT_F_LOCAL) {
2393 transport_h2g = t_h2g;
2394 transport_g2h = t_g2h;
2395 transport_dgram = t_dgram;
2396 transport_local = t_local;
2399 mutex_unlock(&vsock_register_mutex);
2402 EXPORT_SYMBOL_GPL(vsock_core_register);
2404 void vsock_core_unregister(const struct vsock_transport *t)
2406 mutex_lock(&vsock_register_mutex);
2408 if (transport_h2g == t)
2409 transport_h2g = NULL;
2411 if (transport_g2h == t)
2412 transport_g2h = NULL;
2414 if (transport_dgram == t)
2415 transport_dgram = NULL;
2417 if (transport_local == t)
2418 transport_local = NULL;
2420 mutex_unlock(&vsock_register_mutex);
2422 EXPORT_SYMBOL_GPL(vsock_core_unregister);
2424 module_init(vsock_init);
2425 module_exit(vsock_exit);
2427 MODULE_AUTHOR("VMware, Inc.");
2428 MODULE_DESCRIPTION("VMware Virtual Socket Family");
2429 MODULE_VERSION("1.0.2.0-k");
2430 MODULE_LICENSE("GPL v2");