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/compat.h>
89 #include <linux/types.h>
90 #include <linux/bitops.h>
91 #include <linux/cred.h>
92 #include <linux/init.h>
94 #include <linux/kernel.h>
95 #include <linux/sched/signal.h>
96 #include <linux/kmod.h>
97 #include <linux/list.h>
98 #include <linux/miscdevice.h>
99 #include <linux/module.h>
100 #include <linux/mutex.h>
101 #include <linux/net.h>
102 #include <linux/poll.h>
103 #include <linux/random.h>
104 #include <linux/skbuff.h>
105 #include <linux/smp.h>
106 #include <linux/socket.h>
107 #include <linux/stddef.h>
108 #include <linux/unistd.h>
109 #include <linux/wait.h>
110 #include <linux/workqueue.h>
111 #include <net/sock.h>
112 #include <net/af_vsock.h>
114 static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr);
115 static void vsock_sk_destruct(struct sock *sk);
116 static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
118 /* Protocol family. */
119 static struct proto vsock_proto = {
121 .owner = THIS_MODULE,
122 .obj_size = sizeof(struct vsock_sock),
125 /* The default peer timeout indicates how long we will wait for a peer response
126 * to a control message.
128 #define VSOCK_DEFAULT_CONNECT_TIMEOUT (2 * HZ)
130 #define VSOCK_DEFAULT_BUFFER_SIZE (1024 * 256)
131 #define VSOCK_DEFAULT_BUFFER_MAX_SIZE (1024 * 256)
132 #define VSOCK_DEFAULT_BUFFER_MIN_SIZE 128
134 /* Transport used for host->guest communication */
135 static const struct vsock_transport *transport_h2g;
136 /* Transport used for guest->host communication */
137 static const struct vsock_transport *transport_g2h;
138 /* Transport used for DGRAM communication */
139 static const struct vsock_transport *transport_dgram;
140 /* Transport used for local communication */
141 static const struct vsock_transport *transport_local;
142 static DEFINE_MUTEX(vsock_register_mutex);
146 /* Each bound VSocket is stored in the bind hash table and each connected
147 * VSocket is stored in the connected hash table.
149 * Unbound sockets are all put on the same list attached to the end of the hash
150 * table (vsock_unbound_sockets). Bound sockets are added to the hash table in
151 * the bucket that their local address hashes to (vsock_bound_sockets(addr)
152 * represents the list that addr hashes to).
154 * Specifically, we initialize the vsock_bind_table array to a size of
155 * VSOCK_HASH_SIZE + 1 so that vsock_bind_table[0] through
156 * vsock_bind_table[VSOCK_HASH_SIZE - 1] are for bound sockets and
157 * vsock_bind_table[VSOCK_HASH_SIZE] is for unbound sockets. The hash function
158 * mods with VSOCK_HASH_SIZE to ensure this.
160 #define MAX_PORT_RETRIES 24
162 #define VSOCK_HASH(addr) ((addr)->svm_port % VSOCK_HASH_SIZE)
163 #define vsock_bound_sockets(addr) (&vsock_bind_table[VSOCK_HASH(addr)])
164 #define vsock_unbound_sockets (&vsock_bind_table[VSOCK_HASH_SIZE])
166 /* XXX This can probably be implemented in a better way. */
167 #define VSOCK_CONN_HASH(src, dst) \
168 (((src)->svm_cid ^ (dst)->svm_port) % VSOCK_HASH_SIZE)
169 #define vsock_connected_sockets(src, dst) \
170 (&vsock_connected_table[VSOCK_CONN_HASH(src, dst)])
171 #define vsock_connected_sockets_vsk(vsk) \
172 vsock_connected_sockets(&(vsk)->remote_addr, &(vsk)->local_addr)
174 struct list_head vsock_bind_table[VSOCK_HASH_SIZE + 1];
175 EXPORT_SYMBOL_GPL(vsock_bind_table);
176 struct list_head vsock_connected_table[VSOCK_HASH_SIZE];
177 EXPORT_SYMBOL_GPL(vsock_connected_table);
178 DEFINE_SPINLOCK(vsock_table_lock);
179 EXPORT_SYMBOL_GPL(vsock_table_lock);
181 /* Autobind this socket to the local address if necessary. */
182 static int vsock_auto_bind(struct vsock_sock *vsk)
184 struct sock *sk = sk_vsock(vsk);
185 struct sockaddr_vm local_addr;
187 if (vsock_addr_bound(&vsk->local_addr))
189 vsock_addr_init(&local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
190 return __vsock_bind(sk, &local_addr);
193 static void vsock_init_tables(void)
197 for (i = 0; i < ARRAY_SIZE(vsock_bind_table); i++)
198 INIT_LIST_HEAD(&vsock_bind_table[i]);
200 for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++)
201 INIT_LIST_HEAD(&vsock_connected_table[i]);
204 static void __vsock_insert_bound(struct list_head *list,
205 struct vsock_sock *vsk)
208 list_add(&vsk->bound_table, list);
211 static void __vsock_insert_connected(struct list_head *list,
212 struct vsock_sock *vsk)
215 list_add(&vsk->connected_table, list);
218 static void __vsock_remove_bound(struct vsock_sock *vsk)
220 list_del_init(&vsk->bound_table);
224 static void __vsock_remove_connected(struct vsock_sock *vsk)
226 list_del_init(&vsk->connected_table);
230 static struct sock *__vsock_find_bound_socket(struct sockaddr_vm *addr)
232 struct vsock_sock *vsk;
234 list_for_each_entry(vsk, vsock_bound_sockets(addr), bound_table) {
235 if (vsock_addr_equals_addr(addr, &vsk->local_addr))
236 return sk_vsock(vsk);
238 if (addr->svm_port == vsk->local_addr.svm_port &&
239 (vsk->local_addr.svm_cid == VMADDR_CID_ANY ||
240 addr->svm_cid == VMADDR_CID_ANY))
241 return sk_vsock(vsk);
247 static struct sock *__vsock_find_connected_socket(struct sockaddr_vm *src,
248 struct sockaddr_vm *dst)
250 struct vsock_sock *vsk;
252 list_for_each_entry(vsk, vsock_connected_sockets(src, dst),
254 if (vsock_addr_equals_addr(src, &vsk->remote_addr) &&
255 dst->svm_port == vsk->local_addr.svm_port) {
256 return sk_vsock(vsk);
263 static void vsock_insert_unbound(struct vsock_sock *vsk)
265 spin_lock_bh(&vsock_table_lock);
266 __vsock_insert_bound(vsock_unbound_sockets, vsk);
267 spin_unlock_bh(&vsock_table_lock);
270 void vsock_insert_connected(struct vsock_sock *vsk)
272 struct list_head *list = vsock_connected_sockets(
273 &vsk->remote_addr, &vsk->local_addr);
275 spin_lock_bh(&vsock_table_lock);
276 __vsock_insert_connected(list, vsk);
277 spin_unlock_bh(&vsock_table_lock);
279 EXPORT_SYMBOL_GPL(vsock_insert_connected);
281 void vsock_remove_bound(struct vsock_sock *vsk)
283 spin_lock_bh(&vsock_table_lock);
284 if (__vsock_in_bound_table(vsk))
285 __vsock_remove_bound(vsk);
286 spin_unlock_bh(&vsock_table_lock);
288 EXPORT_SYMBOL_GPL(vsock_remove_bound);
290 void vsock_remove_connected(struct vsock_sock *vsk)
292 spin_lock_bh(&vsock_table_lock);
293 if (__vsock_in_connected_table(vsk))
294 __vsock_remove_connected(vsk);
295 spin_unlock_bh(&vsock_table_lock);
297 EXPORT_SYMBOL_GPL(vsock_remove_connected);
299 struct sock *vsock_find_bound_socket(struct sockaddr_vm *addr)
303 spin_lock_bh(&vsock_table_lock);
304 sk = __vsock_find_bound_socket(addr);
308 spin_unlock_bh(&vsock_table_lock);
312 EXPORT_SYMBOL_GPL(vsock_find_bound_socket);
314 struct sock *vsock_find_connected_socket(struct sockaddr_vm *src,
315 struct sockaddr_vm *dst)
319 spin_lock_bh(&vsock_table_lock);
320 sk = __vsock_find_connected_socket(src, dst);
324 spin_unlock_bh(&vsock_table_lock);
328 EXPORT_SYMBOL_GPL(vsock_find_connected_socket);
330 void vsock_remove_sock(struct vsock_sock *vsk)
332 vsock_remove_bound(vsk);
333 vsock_remove_connected(vsk);
335 EXPORT_SYMBOL_GPL(vsock_remove_sock);
337 void vsock_for_each_connected_socket(struct vsock_transport *transport,
338 void (*fn)(struct sock *sk))
342 spin_lock_bh(&vsock_table_lock);
344 for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) {
345 struct vsock_sock *vsk;
346 list_for_each_entry(vsk, &vsock_connected_table[i],
348 if (vsk->transport != transport)
355 spin_unlock_bh(&vsock_table_lock);
357 EXPORT_SYMBOL_GPL(vsock_for_each_connected_socket);
359 void vsock_add_pending(struct sock *listener, struct sock *pending)
361 struct vsock_sock *vlistener;
362 struct vsock_sock *vpending;
364 vlistener = vsock_sk(listener);
365 vpending = vsock_sk(pending);
369 list_add_tail(&vpending->pending_links, &vlistener->pending_links);
371 EXPORT_SYMBOL_GPL(vsock_add_pending);
373 void vsock_remove_pending(struct sock *listener, struct sock *pending)
375 struct vsock_sock *vpending = vsock_sk(pending);
377 list_del_init(&vpending->pending_links);
381 EXPORT_SYMBOL_GPL(vsock_remove_pending);
383 void vsock_enqueue_accept(struct sock *listener, struct sock *connected)
385 struct vsock_sock *vlistener;
386 struct vsock_sock *vconnected;
388 vlistener = vsock_sk(listener);
389 vconnected = vsock_sk(connected);
391 sock_hold(connected);
393 list_add_tail(&vconnected->accept_queue, &vlistener->accept_queue);
395 EXPORT_SYMBOL_GPL(vsock_enqueue_accept);
397 static bool vsock_use_local_transport(unsigned int remote_cid)
399 if (!transport_local)
402 if (remote_cid == VMADDR_CID_LOCAL)
406 return remote_cid == transport_g2h->get_local_cid();
408 return remote_cid == VMADDR_CID_HOST;
412 static void vsock_deassign_transport(struct vsock_sock *vsk)
417 vsk->transport->destruct(vsk);
418 module_put(vsk->transport->module);
419 vsk->transport = NULL;
422 /* Assign a transport to a socket and call the .init transport callback.
424 * Note: for connection oriented socket this must be called when vsk->remote_addr
425 * is set (e.g. during the connect() or when a connection request on a listener
426 * socket is received).
427 * The vsk->remote_addr is used to decide which transport to use:
428 * - remote CID == VMADDR_CID_LOCAL or g2h->local_cid or VMADDR_CID_HOST if
429 * g2h is not loaded, will use local transport;
430 * - remote CID <= VMADDR_CID_HOST or h2g is not loaded or remote flags field
431 * includes VMADDR_FLAG_TO_HOST flag value, will use guest->host transport;
432 * - remote CID > VMADDR_CID_HOST will use host->guest transport;
434 int vsock_assign_transport(struct vsock_sock *vsk, struct vsock_sock *psk)
436 const struct vsock_transport *new_transport;
437 struct sock *sk = sk_vsock(vsk);
438 unsigned int remote_cid = vsk->remote_addr.svm_cid;
442 /* If the packet is coming with the source and destination CIDs higher
443 * than VMADDR_CID_HOST, then a vsock channel where all the packets are
444 * forwarded to the host should be established. Then the host will
445 * need to forward the packets to the guest.
447 * The flag is set on the (listen) receive path (psk is not NULL). On
448 * the connect path the flag can be set by the user space application.
450 if (psk && vsk->local_addr.svm_cid > VMADDR_CID_HOST &&
451 vsk->remote_addr.svm_cid > VMADDR_CID_HOST)
452 vsk->remote_addr.svm_flags |= VMADDR_FLAG_TO_HOST;
454 remote_flags = vsk->remote_addr.svm_flags;
456 switch (sk->sk_type) {
458 new_transport = transport_dgram;
462 if (vsock_use_local_transport(remote_cid))
463 new_transport = transport_local;
464 else if (remote_cid <= VMADDR_CID_HOST || !transport_h2g ||
465 (remote_flags & VMADDR_FLAG_TO_HOST))
466 new_transport = transport_g2h;
468 new_transport = transport_h2g;
471 return -ESOCKTNOSUPPORT;
474 if (vsk->transport) {
475 if (vsk->transport == new_transport)
478 /* transport->release() must be called with sock lock acquired.
479 * This path can only be taken during vsock_connect(), where we
480 * have already held the sock lock. In the other cases, this
481 * function is called on a new socket which is not assigned to
484 vsk->transport->release(vsk);
485 vsock_deassign_transport(vsk);
488 /* We increase the module refcnt to prevent the transport unloading
489 * while there are open sockets assigned to it.
491 if (!new_transport || !try_module_get(new_transport->module))
494 if (sk->sk_type == SOCK_SEQPACKET) {
495 if (!new_transport->seqpacket_allow ||
496 !new_transport->seqpacket_allow(remote_cid)) {
497 module_put(new_transport->module);
498 return -ESOCKTNOSUPPORT;
502 ret = new_transport->init(vsk, psk);
504 module_put(new_transport->module);
508 vsk->transport = new_transport;
512 EXPORT_SYMBOL_GPL(vsock_assign_transport);
514 bool vsock_find_cid(unsigned int cid)
516 if (transport_g2h && cid == transport_g2h->get_local_cid())
519 if (transport_h2g && cid == VMADDR_CID_HOST)
522 if (transport_local && cid == VMADDR_CID_LOCAL)
527 EXPORT_SYMBOL_GPL(vsock_find_cid);
529 static struct sock *vsock_dequeue_accept(struct sock *listener)
531 struct vsock_sock *vlistener;
532 struct vsock_sock *vconnected;
534 vlistener = vsock_sk(listener);
536 if (list_empty(&vlistener->accept_queue))
539 vconnected = list_entry(vlistener->accept_queue.next,
540 struct vsock_sock, accept_queue);
542 list_del_init(&vconnected->accept_queue);
544 /* The caller will need a reference on the connected socket so we let
545 * it call sock_put().
548 return sk_vsock(vconnected);
551 static bool vsock_is_accept_queue_empty(struct sock *sk)
553 struct vsock_sock *vsk = vsock_sk(sk);
554 return list_empty(&vsk->accept_queue);
557 static bool vsock_is_pending(struct sock *sk)
559 struct vsock_sock *vsk = vsock_sk(sk);
560 return !list_empty(&vsk->pending_links);
563 static int vsock_send_shutdown(struct sock *sk, int mode)
565 struct vsock_sock *vsk = vsock_sk(sk);
570 return vsk->transport->shutdown(vsk, mode);
573 static void vsock_pending_work(struct work_struct *work)
576 struct sock *listener;
577 struct vsock_sock *vsk;
580 vsk = container_of(work, struct vsock_sock, pending_work.work);
582 listener = vsk->listener;
586 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
588 if (vsock_is_pending(sk)) {
589 vsock_remove_pending(listener, sk);
591 sk_acceptq_removed(listener);
592 } else if (!vsk->rejected) {
593 /* We are not on the pending list and accept() did not reject
594 * us, so we must have been accepted by our user process. We
595 * just need to drop our references to the sockets and be on
602 /* We need to remove ourself from the global connected sockets list so
603 * incoming packets can't find this socket, and to reduce the reference
606 vsock_remove_connected(vsk);
608 sk->sk_state = TCP_CLOSE;
612 release_sock(listener);
620 /**** SOCKET OPERATIONS ****/
622 static int __vsock_bind_connectible(struct vsock_sock *vsk,
623 struct sockaddr_vm *addr)
626 struct sockaddr_vm new_addr;
629 port = LAST_RESERVED_PORT + 1 +
630 prandom_u32_max(U32_MAX - LAST_RESERVED_PORT);
632 vsock_addr_init(&new_addr, addr->svm_cid, addr->svm_port);
634 if (addr->svm_port == VMADDR_PORT_ANY) {
638 for (i = 0; i < MAX_PORT_RETRIES; i++) {
639 if (port <= LAST_RESERVED_PORT)
640 port = LAST_RESERVED_PORT + 1;
642 new_addr.svm_port = port++;
644 if (!__vsock_find_bound_socket(&new_addr)) {
651 return -EADDRNOTAVAIL;
653 /* If port is in reserved range, ensure caller
654 * has necessary privileges.
656 if (addr->svm_port <= LAST_RESERVED_PORT &&
657 !capable(CAP_NET_BIND_SERVICE)) {
661 if (__vsock_find_bound_socket(&new_addr))
665 vsock_addr_init(&vsk->local_addr, new_addr.svm_cid, new_addr.svm_port);
667 /* Remove connection oriented sockets from the unbound list and add them
668 * to the hash table for easy lookup by its address. The unbound list
669 * is simply an extra entry at the end of the hash table, a trick used
672 __vsock_remove_bound(vsk);
673 __vsock_insert_bound(vsock_bound_sockets(&vsk->local_addr), vsk);
678 static int __vsock_bind_dgram(struct vsock_sock *vsk,
679 struct sockaddr_vm *addr)
681 return vsk->transport->dgram_bind(vsk, addr);
684 static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr)
686 struct vsock_sock *vsk = vsock_sk(sk);
689 /* First ensure this socket isn't already bound. */
690 if (vsock_addr_bound(&vsk->local_addr))
693 /* Now bind to the provided address or select appropriate values if
694 * none are provided (VMADDR_CID_ANY and VMADDR_PORT_ANY). Note that
695 * like AF_INET prevents binding to a non-local IP address (in most
696 * cases), we only allow binding to a local CID.
698 if (addr->svm_cid != VMADDR_CID_ANY && !vsock_find_cid(addr->svm_cid))
699 return -EADDRNOTAVAIL;
701 switch (sk->sk_socket->type) {
704 spin_lock_bh(&vsock_table_lock);
705 retval = __vsock_bind_connectible(vsk, addr);
706 spin_unlock_bh(&vsock_table_lock);
710 retval = __vsock_bind_dgram(vsk, addr);
721 static void vsock_connect_timeout(struct work_struct *work);
723 static struct sock *__vsock_create(struct net *net,
731 struct vsock_sock *psk;
732 struct vsock_sock *vsk;
734 sk = sk_alloc(net, AF_VSOCK, priority, &vsock_proto, kern);
738 sock_init_data(sock, sk);
740 /* sk->sk_type is normally set in sock_init_data, but only if sock is
741 * non-NULL. We make sure that our sockets always have a type by
742 * setting it here if needed.
748 vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
749 vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
751 sk->sk_destruct = vsock_sk_destruct;
752 sk->sk_backlog_rcv = vsock_queue_rcv_skb;
753 sock_reset_flag(sk, SOCK_DONE);
755 INIT_LIST_HEAD(&vsk->bound_table);
756 INIT_LIST_HEAD(&vsk->connected_table);
757 vsk->listener = NULL;
758 INIT_LIST_HEAD(&vsk->pending_links);
759 INIT_LIST_HEAD(&vsk->accept_queue);
760 vsk->rejected = false;
761 vsk->sent_request = false;
762 vsk->ignore_connecting_rst = false;
763 vsk->peer_shutdown = 0;
764 INIT_DELAYED_WORK(&vsk->connect_work, vsock_connect_timeout);
765 INIT_DELAYED_WORK(&vsk->pending_work, vsock_pending_work);
767 psk = parent ? vsock_sk(parent) : NULL;
769 vsk->trusted = psk->trusted;
770 vsk->owner = get_cred(psk->owner);
771 vsk->connect_timeout = psk->connect_timeout;
772 vsk->buffer_size = psk->buffer_size;
773 vsk->buffer_min_size = psk->buffer_min_size;
774 vsk->buffer_max_size = psk->buffer_max_size;
775 security_sk_clone(parent, sk);
777 vsk->trusted = ns_capable_noaudit(&init_user_ns, CAP_NET_ADMIN);
778 vsk->owner = get_current_cred();
779 vsk->connect_timeout = VSOCK_DEFAULT_CONNECT_TIMEOUT;
780 vsk->buffer_size = VSOCK_DEFAULT_BUFFER_SIZE;
781 vsk->buffer_min_size = VSOCK_DEFAULT_BUFFER_MIN_SIZE;
782 vsk->buffer_max_size = VSOCK_DEFAULT_BUFFER_MAX_SIZE;
788 static bool sock_type_connectible(u16 type)
790 return (type == SOCK_STREAM) || (type == SOCK_SEQPACKET);
793 static void __vsock_release(struct sock *sk, int level)
796 struct sock *pending;
797 struct vsock_sock *vsk;
800 pending = NULL; /* Compiler warning. */
802 /* When "level" is SINGLE_DEPTH_NESTING, use the nested
803 * version to avoid the warning "possible recursive locking
804 * detected". When "level" is 0, lock_sock_nested(sk, level)
805 * is the same as lock_sock(sk).
807 lock_sock_nested(sk, level);
810 vsk->transport->release(vsk);
811 else if (sock_type_connectible(sk->sk_type))
812 vsock_remove_sock(vsk);
815 sk->sk_shutdown = SHUTDOWN_MASK;
817 skb_queue_purge(&sk->sk_receive_queue);
819 /* Clean up any sockets that never were accepted. */
820 while ((pending = vsock_dequeue_accept(sk)) != NULL) {
821 __vsock_release(pending, SINGLE_DEPTH_NESTING);
830 static void vsock_sk_destruct(struct sock *sk)
832 struct vsock_sock *vsk = vsock_sk(sk);
834 vsock_deassign_transport(vsk);
836 /* When clearing these addresses, there's no need to set the family and
837 * possibly register the address family with the kernel.
839 vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
840 vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
842 put_cred(vsk->owner);
845 static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
849 err = sock_queue_rcv_skb(sk, skb);
856 struct sock *vsock_create_connected(struct sock *parent)
858 return __vsock_create(sock_net(parent), NULL, parent, GFP_KERNEL,
861 EXPORT_SYMBOL_GPL(vsock_create_connected);
863 s64 vsock_stream_has_data(struct vsock_sock *vsk)
865 return vsk->transport->stream_has_data(vsk);
867 EXPORT_SYMBOL_GPL(vsock_stream_has_data);
869 static s64 vsock_connectible_has_data(struct vsock_sock *vsk)
871 struct sock *sk = sk_vsock(vsk);
873 if (sk->sk_type == SOCK_SEQPACKET)
874 return vsk->transport->seqpacket_has_data(vsk);
876 return vsock_stream_has_data(vsk);
879 s64 vsock_stream_has_space(struct vsock_sock *vsk)
881 return vsk->transport->stream_has_space(vsk);
883 EXPORT_SYMBOL_GPL(vsock_stream_has_space);
885 static int vsock_release(struct socket *sock)
887 __vsock_release(sock->sk, 0);
889 sock->state = SS_FREE;
895 vsock_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
899 struct sockaddr_vm *vm_addr;
903 if (vsock_addr_cast(addr, addr_len, &vm_addr) != 0)
907 err = __vsock_bind(sk, vm_addr);
913 static int vsock_getname(struct socket *sock,
914 struct sockaddr *addr, int peer)
918 struct vsock_sock *vsk;
919 struct sockaddr_vm *vm_addr;
928 if (sock->state != SS_CONNECTED) {
932 vm_addr = &vsk->remote_addr;
934 vm_addr = &vsk->local_addr;
942 /* sys_getsockname() and sys_getpeername() pass us a
943 * MAX_SOCK_ADDR-sized buffer and don't set addr_len. Unfortunately
944 * that macro is defined in socket.c instead of .h, so we hardcode its
947 BUILD_BUG_ON(sizeof(*vm_addr) > 128);
948 memcpy(addr, vm_addr, sizeof(*vm_addr));
949 err = sizeof(*vm_addr);
956 static int vsock_shutdown(struct socket *sock, int mode)
961 /* User level uses SHUT_RD (0) and SHUT_WR (1), but the kernel uses
962 * RCV_SHUTDOWN (1) and SEND_SHUTDOWN (2), so we must increment mode
963 * here like the other address families do. Note also that the
964 * increment makes SHUT_RDWR (2) into RCV_SHUTDOWN | SEND_SHUTDOWN (3),
965 * which is what we want.
969 if ((mode & ~SHUTDOWN_MASK) || !mode)
972 /* If this is a connection oriented socket and it is not connected then
973 * bail out immediately. If it is a DGRAM socket then we must first
974 * kick the socket so that it wakes up from any sleeping calls, for
975 * example recv(), and then afterwards return the error.
981 if (sock->state == SS_UNCONNECTED) {
983 if (sock_type_connectible(sk->sk_type))
986 sock->state = SS_DISCONNECTING;
990 /* Receive and send shutdowns are treated alike. */
991 mode = mode & (RCV_SHUTDOWN | SEND_SHUTDOWN);
993 sk->sk_shutdown |= mode;
994 sk->sk_state_change(sk);
996 if (sock_type_connectible(sk->sk_type)) {
997 sock_reset_flag(sk, SOCK_DONE);
998 vsock_send_shutdown(sk, mode);
1007 static __poll_t vsock_poll(struct file *file, struct socket *sock,
1012 struct vsock_sock *vsk;
1017 poll_wait(file, sk_sleep(sk), wait);
1021 /* Signify that there has been an error on this socket. */
1024 /* INET sockets treat local write shutdown and peer write shutdown as a
1025 * case of EPOLLHUP set.
1027 if ((sk->sk_shutdown == SHUTDOWN_MASK) ||
1028 ((sk->sk_shutdown & SEND_SHUTDOWN) &&
1029 (vsk->peer_shutdown & SEND_SHUTDOWN))) {
1033 if (sk->sk_shutdown & RCV_SHUTDOWN ||
1034 vsk->peer_shutdown & SEND_SHUTDOWN) {
1038 if (sock->type == SOCK_DGRAM) {
1039 /* For datagram sockets we can read if there is something in
1040 * the queue and write as long as the socket isn't shutdown for
1043 if (!skb_queue_empty_lockless(&sk->sk_receive_queue) ||
1044 (sk->sk_shutdown & RCV_SHUTDOWN)) {
1045 mask |= EPOLLIN | EPOLLRDNORM;
1048 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
1049 mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND;
1051 } else if (sock_type_connectible(sk->sk_type)) {
1052 const struct vsock_transport *transport;
1056 transport = vsk->transport;
1058 /* Listening sockets that have connections in their accept
1059 * queue can be read.
1061 if (sk->sk_state == TCP_LISTEN
1062 && !vsock_is_accept_queue_empty(sk))
1063 mask |= EPOLLIN | EPOLLRDNORM;
1065 /* If there is something in the queue then we can read. */
1066 if (transport && transport->stream_is_active(vsk) &&
1067 !(sk->sk_shutdown & RCV_SHUTDOWN)) {
1068 bool data_ready_now = false;
1069 int ret = transport->notify_poll_in(
1070 vsk, 1, &data_ready_now);
1075 mask |= EPOLLIN | EPOLLRDNORM;
1080 /* Sockets whose connections have been closed, reset, or
1081 * terminated should also be considered read, and we check the
1082 * shutdown flag for that.
1084 if (sk->sk_shutdown & RCV_SHUTDOWN ||
1085 vsk->peer_shutdown & SEND_SHUTDOWN) {
1086 mask |= EPOLLIN | EPOLLRDNORM;
1089 /* Connected sockets that can produce data can be written. */
1090 if (transport && sk->sk_state == TCP_ESTABLISHED) {
1091 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1092 bool space_avail_now = false;
1093 int ret = transport->notify_poll_out(
1094 vsk, 1, &space_avail_now);
1098 if (space_avail_now)
1099 /* Remove EPOLLWRBAND since INET
1100 * sockets are not setting it.
1102 mask |= EPOLLOUT | EPOLLWRNORM;
1108 /* Simulate INET socket poll behaviors, which sets
1109 * EPOLLOUT|EPOLLWRNORM when peer is closed and nothing to read,
1110 * but local send is not shutdown.
1112 if (sk->sk_state == TCP_CLOSE || sk->sk_state == TCP_CLOSING) {
1113 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
1114 mask |= EPOLLOUT | EPOLLWRNORM;
1124 static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg,
1129 struct vsock_sock *vsk;
1130 struct sockaddr_vm *remote_addr;
1131 const struct vsock_transport *transport;
1133 if (msg->msg_flags & MSG_OOB)
1136 /* For now, MSG_DONTWAIT is always assumed... */
1143 transport = vsk->transport;
1145 err = vsock_auto_bind(vsk);
1150 /* If the provided message contains an address, use that. Otherwise
1151 * fall back on the socket's remote handle (if it has been connected).
1153 if (msg->msg_name &&
1154 vsock_addr_cast(msg->msg_name, msg->msg_namelen,
1155 &remote_addr) == 0) {
1156 /* Ensure this address is of the right type and is a valid
1160 if (remote_addr->svm_cid == VMADDR_CID_ANY)
1161 remote_addr->svm_cid = transport->get_local_cid();
1163 if (!vsock_addr_bound(remote_addr)) {
1167 } else if (sock->state == SS_CONNECTED) {
1168 remote_addr = &vsk->remote_addr;
1170 if (remote_addr->svm_cid == VMADDR_CID_ANY)
1171 remote_addr->svm_cid = transport->get_local_cid();
1173 /* XXX Should connect() or this function ensure remote_addr is
1176 if (!vsock_addr_bound(&vsk->remote_addr)) {
1185 if (!transport->dgram_allow(remote_addr->svm_cid,
1186 remote_addr->svm_port)) {
1191 err = transport->dgram_enqueue(vsk, remote_addr, msg, len);
1198 static int vsock_dgram_connect(struct socket *sock,
1199 struct sockaddr *addr, int addr_len, int flags)
1203 struct vsock_sock *vsk;
1204 struct sockaddr_vm *remote_addr;
1209 err = vsock_addr_cast(addr, addr_len, &remote_addr);
1210 if (err == -EAFNOSUPPORT && remote_addr->svm_family == AF_UNSPEC) {
1212 vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY,
1214 sock->state = SS_UNCONNECTED;
1217 } else if (err != 0)
1222 err = vsock_auto_bind(vsk);
1226 if (!vsk->transport->dgram_allow(remote_addr->svm_cid,
1227 remote_addr->svm_port)) {
1232 memcpy(&vsk->remote_addr, remote_addr, sizeof(vsk->remote_addr));
1233 sock->state = SS_CONNECTED;
1240 static int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg,
1241 size_t len, int flags)
1243 struct vsock_sock *vsk = vsock_sk(sock->sk);
1245 return vsk->transport->dgram_dequeue(vsk, msg, len, flags);
1248 static const struct proto_ops vsock_dgram_ops = {
1250 .owner = THIS_MODULE,
1251 .release = vsock_release,
1253 .connect = vsock_dgram_connect,
1254 .socketpair = sock_no_socketpair,
1255 .accept = sock_no_accept,
1256 .getname = vsock_getname,
1258 .ioctl = sock_no_ioctl,
1259 .listen = sock_no_listen,
1260 .shutdown = vsock_shutdown,
1261 .sendmsg = vsock_dgram_sendmsg,
1262 .recvmsg = vsock_dgram_recvmsg,
1263 .mmap = sock_no_mmap,
1264 .sendpage = sock_no_sendpage,
1267 static int vsock_transport_cancel_pkt(struct vsock_sock *vsk)
1269 const struct vsock_transport *transport = vsk->transport;
1271 if (!transport || !transport->cancel_pkt)
1274 return transport->cancel_pkt(vsk);
1277 static void vsock_connect_timeout(struct work_struct *work)
1280 struct vsock_sock *vsk;
1282 vsk = container_of(work, struct vsock_sock, connect_work.work);
1286 if (sk->sk_state == TCP_SYN_SENT &&
1287 (sk->sk_shutdown != SHUTDOWN_MASK)) {
1288 sk->sk_state = TCP_CLOSE;
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
1394 schedule_delayed_work(&vsk->connect_work, timeout);
1396 /* Skip ahead to preserve error code set above. */
1401 timeout = schedule_timeout(timeout);
1404 if (signal_pending(current)) {
1405 err = sock_intr_errno(timeout);
1406 sk->sk_state = sk->sk_state == TCP_ESTABLISHED ? TCP_CLOSING : TCP_CLOSE;
1407 sock->state = SS_UNCONNECTED;
1408 vsock_transport_cancel_pkt(vsk);
1409 vsock_remove_connected(vsk);
1411 } else if (timeout == 0) {
1413 sk->sk_state = TCP_CLOSE;
1414 sock->state = SS_UNCONNECTED;
1415 vsock_transport_cancel_pkt(vsk);
1419 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1424 sk->sk_state = TCP_CLOSE;
1425 sock->state = SS_UNCONNECTED;
1431 finish_wait(sk_sleep(sk), &wait);
1437 static int vsock_accept(struct socket *sock, struct socket *newsock, int flags,
1440 struct sock *listener;
1442 struct sock *connected;
1443 struct vsock_sock *vconnected;
1448 listener = sock->sk;
1450 lock_sock(listener);
1452 if (!sock_type_connectible(sock->type)) {
1457 if (listener->sk_state != TCP_LISTEN) {
1462 /* Wait for children sockets to appear; these are the new sockets
1463 * created upon connection establishment.
1465 timeout = sock_rcvtimeo(listener, flags & O_NONBLOCK);
1466 prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1468 while ((connected = vsock_dequeue_accept(listener)) == NULL &&
1469 listener->sk_err == 0) {
1470 release_sock(listener);
1471 timeout = schedule_timeout(timeout);
1472 finish_wait(sk_sleep(listener), &wait);
1473 lock_sock(listener);
1475 if (signal_pending(current)) {
1476 err = sock_intr_errno(timeout);
1478 } else if (timeout == 0) {
1483 prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1485 finish_wait(sk_sleep(listener), &wait);
1487 if (listener->sk_err)
1488 err = -listener->sk_err;
1491 sk_acceptq_removed(listener);
1493 lock_sock_nested(connected, SINGLE_DEPTH_NESTING);
1494 vconnected = vsock_sk(connected);
1496 /* If the listener socket has received an error, then we should
1497 * reject this socket and return. Note that we simply mark the
1498 * socket rejected, drop our reference, and let the cleanup
1499 * function handle the cleanup; the fact that we found it in
1500 * the listener's accept queue guarantees that the cleanup
1501 * function hasn't run yet.
1504 vconnected->rejected = true;
1506 newsock->state = SS_CONNECTED;
1507 sock_graft(connected, newsock);
1510 release_sock(connected);
1511 sock_put(connected);
1515 release_sock(listener);
1519 static int vsock_listen(struct socket *sock, int backlog)
1523 struct vsock_sock *vsk;
1529 if (!sock_type_connectible(sk->sk_type)) {
1534 if (sock->state != SS_UNCONNECTED) {
1541 if (!vsock_addr_bound(&vsk->local_addr)) {
1546 sk->sk_max_ack_backlog = backlog;
1547 sk->sk_state = TCP_LISTEN;
1556 static void vsock_update_buffer_size(struct vsock_sock *vsk,
1557 const struct vsock_transport *transport,
1560 if (val > vsk->buffer_max_size)
1561 val = vsk->buffer_max_size;
1563 if (val < vsk->buffer_min_size)
1564 val = vsk->buffer_min_size;
1566 if (val != vsk->buffer_size &&
1567 transport && transport->notify_buffer_size)
1568 transport->notify_buffer_size(vsk, &val);
1570 vsk->buffer_size = val;
1573 static int vsock_connectible_setsockopt(struct socket *sock,
1577 unsigned int optlen)
1581 struct vsock_sock *vsk;
1582 const struct vsock_transport *transport;
1585 if (level != AF_VSOCK)
1586 return -ENOPROTOOPT;
1588 #define COPY_IN(_v) \
1590 if (optlen < sizeof(_v)) { \
1594 if (copy_from_sockptr(&_v, optval, sizeof(_v)) != 0) { \
1606 transport = vsk->transport;
1609 case SO_VM_SOCKETS_BUFFER_SIZE:
1611 vsock_update_buffer_size(vsk, transport, val);
1614 case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1616 vsk->buffer_max_size = val;
1617 vsock_update_buffer_size(vsk, transport, vsk->buffer_size);
1620 case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1622 vsk->buffer_min_size = val;
1623 vsock_update_buffer_size(vsk, transport, vsk->buffer_size);
1626 case SO_VM_SOCKETS_CONNECT_TIMEOUT_NEW:
1627 case SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD: {
1628 struct __kernel_sock_timeval tv;
1630 err = sock_copy_user_timeval(&tv, optval, optlen,
1631 optname == SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD);
1634 if (tv.tv_sec >= 0 && tv.tv_usec < USEC_PER_SEC &&
1635 tv.tv_sec < (MAX_SCHEDULE_TIMEOUT / HZ - 1)) {
1636 vsk->connect_timeout = tv.tv_sec * HZ +
1637 DIV_ROUND_UP((unsigned long)tv.tv_usec, (USEC_PER_SEC / HZ));
1638 if (vsk->connect_timeout == 0)
1639 vsk->connect_timeout =
1640 VSOCK_DEFAULT_CONNECT_TIMEOUT;
1660 static int vsock_connectible_getsockopt(struct socket *sock,
1661 int level, int optname,
1662 char __user *optval,
1665 struct sock *sk = sock->sk;
1666 struct vsock_sock *vsk = vsock_sk(sk);
1670 struct old_timeval32 tm32;
1671 struct __kernel_old_timeval tm;
1672 struct __kernel_sock_timeval stm;
1675 int lv = sizeof(v.val64);
1678 if (level != AF_VSOCK)
1679 return -ENOPROTOOPT;
1681 if (get_user(len, optlen))
1684 memset(&v, 0, sizeof(v));
1687 case SO_VM_SOCKETS_BUFFER_SIZE:
1688 v.val64 = vsk->buffer_size;
1691 case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1692 v.val64 = vsk->buffer_max_size;
1695 case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1696 v.val64 = vsk->buffer_min_size;
1699 case SO_VM_SOCKETS_CONNECT_TIMEOUT_NEW:
1700 case SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD:
1701 lv = sock_get_timeout(vsk->connect_timeout, &v,
1702 optname == SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD);
1706 return -ENOPROTOOPT;
1713 if (copy_to_user(optval, &v, len))
1716 if (put_user(len, optlen))
1722 static int vsock_connectible_sendmsg(struct socket *sock, struct msghdr *msg,
1726 struct vsock_sock *vsk;
1727 const struct vsock_transport *transport;
1728 ssize_t total_written;
1731 struct vsock_transport_send_notify_data send_data;
1732 DEFINE_WAIT_FUNC(wait, woken_wake_function);
1739 if (msg->msg_flags & MSG_OOB)
1744 transport = vsk->transport;
1746 /* Callers should not provide a destination with connection oriented
1749 if (msg->msg_namelen) {
1750 err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP;
1754 /* Send data only if both sides are not shutdown in the direction. */
1755 if (sk->sk_shutdown & SEND_SHUTDOWN ||
1756 vsk->peer_shutdown & RCV_SHUTDOWN) {
1761 if (!transport || sk->sk_state != TCP_ESTABLISHED ||
1762 !vsock_addr_bound(&vsk->local_addr)) {
1767 if (!vsock_addr_bound(&vsk->remote_addr)) {
1768 err = -EDESTADDRREQ;
1772 /* Wait for room in the produce queue to enqueue our user's data. */
1773 timeout = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1775 err = transport->notify_send_init(vsk, &send_data);
1779 while (total_written < len) {
1782 add_wait_queue(sk_sleep(sk), &wait);
1783 while (vsock_stream_has_space(vsk) == 0 &&
1785 !(sk->sk_shutdown & SEND_SHUTDOWN) &&
1786 !(vsk->peer_shutdown & RCV_SHUTDOWN)) {
1788 /* Don't wait for non-blocking sockets. */
1791 remove_wait_queue(sk_sleep(sk), &wait);
1795 err = transport->notify_send_pre_block(vsk, &send_data);
1797 remove_wait_queue(sk_sleep(sk), &wait);
1802 timeout = wait_woken(&wait, TASK_INTERRUPTIBLE, timeout);
1804 if (signal_pending(current)) {
1805 err = sock_intr_errno(timeout);
1806 remove_wait_queue(sk_sleep(sk), &wait);
1808 } else if (timeout == 0) {
1810 remove_wait_queue(sk_sleep(sk), &wait);
1814 remove_wait_queue(sk_sleep(sk), &wait);
1816 /* These checks occur both as part of and after the loop
1817 * conditional since we need to check before and after
1823 } else if ((sk->sk_shutdown & SEND_SHUTDOWN) ||
1824 (vsk->peer_shutdown & RCV_SHUTDOWN)) {
1829 err = transport->notify_send_pre_enqueue(vsk, &send_data);
1833 /* Note that enqueue will only write as many bytes as are free
1834 * in the produce queue, so we don't need to ensure len is
1835 * smaller than the queue size. It is the caller's
1836 * responsibility to check how many bytes we were able to send.
1839 if (sk->sk_type == SOCK_SEQPACKET) {
1840 written = transport->seqpacket_enqueue(vsk,
1841 msg, len - total_written);
1843 written = transport->stream_enqueue(vsk,
1844 msg, len - total_written);
1851 total_written += written;
1853 err = transport->notify_send_post_enqueue(
1854 vsk, written, &send_data);
1861 if (total_written > 0) {
1862 /* Return number of written bytes only if:
1863 * 1) SOCK_STREAM socket.
1864 * 2) SOCK_SEQPACKET socket when whole buffer is sent.
1866 if (sk->sk_type == SOCK_STREAM || total_written == len)
1867 err = total_written;
1874 static int vsock_connectible_wait_data(struct sock *sk,
1875 struct wait_queue_entry *wait,
1877 struct vsock_transport_recv_notify_data *recv_data,
1880 const struct vsock_transport *transport;
1881 struct vsock_sock *vsk;
1887 transport = vsk->transport;
1889 while ((data = vsock_connectible_has_data(vsk)) == 0) {
1890 prepare_to_wait(sk_sleep(sk), wait, TASK_INTERRUPTIBLE);
1892 if (sk->sk_err != 0 ||
1893 (sk->sk_shutdown & RCV_SHUTDOWN) ||
1894 (vsk->peer_shutdown & SEND_SHUTDOWN)) {
1898 /* Don't wait for non-blocking sockets. */
1905 err = transport->notify_recv_pre_block(vsk, target, recv_data);
1911 timeout = schedule_timeout(timeout);
1914 if (signal_pending(current)) {
1915 err = sock_intr_errno(timeout);
1917 } else if (timeout == 0) {
1923 finish_wait(sk_sleep(sk), wait);
1928 /* Internal transport error when checking for available
1929 * data. XXX This should be changed to a connection
1930 * reset in a later change.
1938 static int __vsock_stream_recvmsg(struct sock *sk, struct msghdr *msg,
1939 size_t len, int flags)
1941 struct vsock_transport_recv_notify_data recv_data;
1942 const struct vsock_transport *transport;
1943 struct vsock_sock *vsk;
1952 transport = vsk->transport;
1954 /* We must not copy less than target bytes into the user's buffer
1955 * before returning successfully, so we wait for the consume queue to
1956 * have that much data to consume before dequeueing. Note that this
1957 * makes it impossible to handle cases where target is greater than the
1960 target = sock_rcvlowat(sk, flags & MSG_WAITALL, len);
1961 if (target >= transport->stream_rcvhiwat(vsk)) {
1965 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1968 err = transport->notify_recv_init(vsk, target, &recv_data);
1976 err = vsock_connectible_wait_data(sk, &wait, timeout,
1977 &recv_data, target);
1981 err = transport->notify_recv_pre_dequeue(vsk, target,
1986 read = transport->stream_dequeue(vsk, msg, len - copied, flags);
1994 err = transport->notify_recv_post_dequeue(vsk, target, read,
1995 !(flags & MSG_PEEK), &recv_data);
1999 if (read >= target || flags & MSG_PEEK)
2007 else if (sk->sk_shutdown & RCV_SHUTDOWN)
2017 static int __vsock_seqpacket_recvmsg(struct sock *sk, struct msghdr *msg,
2018 size_t len, int flags)
2020 const struct vsock_transport *transport;
2021 struct vsock_sock *vsk;
2028 transport = vsk->transport;
2030 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2032 err = vsock_connectible_wait_data(sk, &wait, timeout, NULL, 0);
2036 msg_len = transport->seqpacket_dequeue(vsk, msg, flags);
2045 } else if (sk->sk_shutdown & RCV_SHUTDOWN) {
2048 /* User sets MSG_TRUNC, so return real length of
2051 if (flags & MSG_TRUNC)
2054 err = len - msg_data_left(msg);
2056 /* Always set MSG_TRUNC if real length of packet is
2057 * bigger than user's buffer.
2060 msg->msg_flags |= MSG_TRUNC;
2068 vsock_connectible_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
2072 struct vsock_sock *vsk;
2073 const struct vsock_transport *transport;
2084 transport = vsk->transport;
2086 if (!transport || sk->sk_state != TCP_ESTABLISHED) {
2087 /* Recvmsg is supposed to return 0 if a peer performs an
2088 * orderly shutdown. Differentiate between that case and when a
2089 * peer has not connected or a local shutdown occurred with the
2092 if (sock_flag(sk, SOCK_DONE))
2100 if (flags & MSG_OOB) {
2105 /* We don't check peer_shutdown flag here since peer may actually shut
2106 * down, but there can be data in the queue that a local socket can
2109 if (sk->sk_shutdown & RCV_SHUTDOWN) {
2114 /* It is valid on Linux to pass in a zero-length receive buffer. This
2115 * is not an error. We may as well bail out now.
2122 if (sk->sk_type == SOCK_STREAM)
2123 err = __vsock_stream_recvmsg(sk, msg, len, flags);
2125 err = __vsock_seqpacket_recvmsg(sk, msg, len, flags);
2132 static const struct proto_ops vsock_stream_ops = {
2134 .owner = THIS_MODULE,
2135 .release = vsock_release,
2137 .connect = vsock_connect,
2138 .socketpair = sock_no_socketpair,
2139 .accept = vsock_accept,
2140 .getname = vsock_getname,
2142 .ioctl = sock_no_ioctl,
2143 .listen = vsock_listen,
2144 .shutdown = vsock_shutdown,
2145 .setsockopt = vsock_connectible_setsockopt,
2146 .getsockopt = vsock_connectible_getsockopt,
2147 .sendmsg = vsock_connectible_sendmsg,
2148 .recvmsg = vsock_connectible_recvmsg,
2149 .mmap = sock_no_mmap,
2150 .sendpage = sock_no_sendpage,
2153 static const struct proto_ops vsock_seqpacket_ops = {
2155 .owner = THIS_MODULE,
2156 .release = vsock_release,
2158 .connect = vsock_connect,
2159 .socketpair = sock_no_socketpair,
2160 .accept = vsock_accept,
2161 .getname = vsock_getname,
2163 .ioctl = sock_no_ioctl,
2164 .listen = vsock_listen,
2165 .shutdown = vsock_shutdown,
2166 .setsockopt = vsock_connectible_setsockopt,
2167 .getsockopt = vsock_connectible_getsockopt,
2168 .sendmsg = vsock_connectible_sendmsg,
2169 .recvmsg = vsock_connectible_recvmsg,
2170 .mmap = sock_no_mmap,
2171 .sendpage = sock_no_sendpage,
2174 static int vsock_create(struct net *net, struct socket *sock,
2175 int protocol, int kern)
2177 struct vsock_sock *vsk;
2184 if (protocol && protocol != PF_VSOCK)
2185 return -EPROTONOSUPPORT;
2187 switch (sock->type) {
2189 sock->ops = &vsock_dgram_ops;
2192 sock->ops = &vsock_stream_ops;
2194 case SOCK_SEQPACKET:
2195 sock->ops = &vsock_seqpacket_ops;
2198 return -ESOCKTNOSUPPORT;
2201 sock->state = SS_UNCONNECTED;
2203 sk = __vsock_create(net, sock, NULL, GFP_KERNEL, 0, kern);
2209 if (sock->type == SOCK_DGRAM) {
2210 ret = vsock_assign_transport(vsk, NULL);
2217 vsock_insert_unbound(vsk);
2222 static const struct net_proto_family vsock_family_ops = {
2224 .create = vsock_create,
2225 .owner = THIS_MODULE,
2228 static long vsock_dev_do_ioctl(struct file *filp,
2229 unsigned int cmd, void __user *ptr)
2231 u32 __user *p = ptr;
2232 u32 cid = VMADDR_CID_ANY;
2236 case IOCTL_VM_SOCKETS_GET_LOCAL_CID:
2237 /* To be compatible with the VMCI behavior, we prioritize the
2238 * guest CID instead of well-know host CID (VMADDR_CID_HOST).
2241 cid = transport_g2h->get_local_cid();
2242 else if (transport_h2g)
2243 cid = transport_h2g->get_local_cid();
2245 if (put_user(cid, p) != 0)
2250 retval = -ENOIOCTLCMD;
2256 static long vsock_dev_ioctl(struct file *filp,
2257 unsigned int cmd, unsigned long arg)
2259 return vsock_dev_do_ioctl(filp, cmd, (void __user *)arg);
2262 #ifdef CONFIG_COMPAT
2263 static long vsock_dev_compat_ioctl(struct file *filp,
2264 unsigned int cmd, unsigned long arg)
2266 return vsock_dev_do_ioctl(filp, cmd, compat_ptr(arg));
2270 static const struct file_operations vsock_device_ops = {
2271 .owner = THIS_MODULE,
2272 .unlocked_ioctl = vsock_dev_ioctl,
2273 #ifdef CONFIG_COMPAT
2274 .compat_ioctl = vsock_dev_compat_ioctl,
2276 .open = nonseekable_open,
2279 static struct miscdevice vsock_device = {
2281 .fops = &vsock_device_ops,
2284 static int __init vsock_init(void)
2288 vsock_init_tables();
2290 vsock_proto.owner = THIS_MODULE;
2291 vsock_device.minor = MISC_DYNAMIC_MINOR;
2292 err = misc_register(&vsock_device);
2294 pr_err("Failed to register misc device\n");
2295 goto err_reset_transport;
2298 err = proto_register(&vsock_proto, 1); /* we want our slab */
2300 pr_err("Cannot register vsock protocol\n");
2301 goto err_deregister_misc;
2304 err = sock_register(&vsock_family_ops);
2306 pr_err("could not register af_vsock (%d) address family: %d\n",
2308 goto err_unregister_proto;
2313 err_unregister_proto:
2314 proto_unregister(&vsock_proto);
2315 err_deregister_misc:
2316 misc_deregister(&vsock_device);
2317 err_reset_transport:
2321 static void __exit vsock_exit(void)
2323 misc_deregister(&vsock_device);
2324 sock_unregister(AF_VSOCK);
2325 proto_unregister(&vsock_proto);
2328 const struct vsock_transport *vsock_core_get_transport(struct vsock_sock *vsk)
2330 return vsk->transport;
2332 EXPORT_SYMBOL_GPL(vsock_core_get_transport);
2334 int vsock_core_register(const struct vsock_transport *t, int features)
2336 const struct vsock_transport *t_h2g, *t_g2h, *t_dgram, *t_local;
2337 int err = mutex_lock_interruptible(&vsock_register_mutex);
2342 t_h2g = transport_h2g;
2343 t_g2h = transport_g2h;
2344 t_dgram = transport_dgram;
2345 t_local = transport_local;
2347 if (features & VSOCK_TRANSPORT_F_H2G) {
2355 if (features & VSOCK_TRANSPORT_F_G2H) {
2363 if (features & VSOCK_TRANSPORT_F_DGRAM) {
2371 if (features & VSOCK_TRANSPORT_F_LOCAL) {
2379 transport_h2g = t_h2g;
2380 transport_g2h = t_g2h;
2381 transport_dgram = t_dgram;
2382 transport_local = t_local;
2385 mutex_unlock(&vsock_register_mutex);
2388 EXPORT_SYMBOL_GPL(vsock_core_register);
2390 void vsock_core_unregister(const struct vsock_transport *t)
2392 mutex_lock(&vsock_register_mutex);
2394 if (transport_h2g == t)
2395 transport_h2g = NULL;
2397 if (transport_g2h == t)
2398 transport_g2h = NULL;
2400 if (transport_dgram == t)
2401 transport_dgram = NULL;
2403 if (transport_local == t)
2404 transport_local = NULL;
2406 mutex_unlock(&vsock_register_mutex);
2408 EXPORT_SYMBOL_GPL(vsock_core_unregister);
2410 module_init(vsock_init);
2411 module_exit(vsock_exit);
2413 MODULE_AUTHOR("VMware, Inc.");
2414 MODULE_DESCRIPTION("VMware Virtual Socket Family");
2415 MODULE_VERSION("1.0.2.0-k");
2416 MODULE_LICENSE("GPL v2");