xref: /openbmc/linux/net/vmw_vsock/af_vsock.c (revision d37cf9b63113f13d742713881ce691fc615d8b3b)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * VMware vSockets Driver
4  *
5  * Copyright (C) 2007-2013 VMware, Inc. All rights reserved.
6  */
7 
8 /* Implementation notes:
9  *
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.
12  *
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.
25  *
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).
29  *
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.
45  *
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.
55  *
56  * - Lock ordering for pending or accept queue sockets is:
57  *
58  *     lock_sock(listener);
59  *     lock_sock_nested(pending, SINGLE_DEPTH_NESTING);
60  *
61  * Using explicit nested locking keeps lockdep happy since normally only one
62  * lock of a given class may be taken at a time.
63  *
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.
69  *
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.
77  *
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):
80  *
81  *   TCP_CLOSE - unconnected
82  *   TCP_SYN_SENT - connecting
83  *   TCP_ESTABLISHED - connected
84  *   TCP_CLOSING - disconnecting
85  *   TCP_LISTEN - listening
86  */
87 
88 #include <linux/compat.h>
89 #include <linux/types.h>
90 #include <linux/bitops.h>
91 #include <linux/cred.h>
92 #include <linux/errqueue.h>
93 #include <linux/init.h>
94 #include <linux/io.h>
95 #include <linux/kernel.h>
96 #include <linux/sched/signal.h>
97 #include <linux/kmod.h>
98 #include <linux/list.h>
99 #include <linux/miscdevice.h>
100 #include <linux/module.h>
101 #include <linux/mutex.h>
102 #include <linux/net.h>
103 #include <linux/poll.h>
104 #include <linux/random.h>
105 #include <linux/skbuff.h>
106 #include <linux/smp.h>
107 #include <linux/socket.h>
108 #include <linux/stddef.h>
109 #include <linux/unistd.h>
110 #include <linux/wait.h>
111 #include <linux/workqueue.h>
112 #include <net/sock.h>
113 #include <net/af_vsock.h>
114 #include <uapi/linux/vm_sockets.h>
115 
116 static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr);
117 static void vsock_sk_destruct(struct sock *sk);
118 static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
119 static void vsock_close(struct sock *sk, long timeout);
120 
121 /* Protocol family. */
122 struct proto vsock_proto = {
123 	.name = "AF_VSOCK",
124 	.owner = THIS_MODULE,
125 	.obj_size = sizeof(struct vsock_sock),
126 	.close = vsock_close,
127 #ifdef CONFIG_BPF_SYSCALL
128 	.psock_update_sk_prot = vsock_bpf_update_proto,
129 #endif
130 };
131 
132 /* The default peer timeout indicates how long we will wait for a peer response
133  * to a control message.
134  */
135 #define VSOCK_DEFAULT_CONNECT_TIMEOUT (2 * HZ)
136 
137 #define VSOCK_DEFAULT_BUFFER_SIZE     (1024 * 256)
138 #define VSOCK_DEFAULT_BUFFER_MAX_SIZE (1024 * 256)
139 #define VSOCK_DEFAULT_BUFFER_MIN_SIZE 128
140 
141 /* Transport used for host->guest communication */
142 static const struct vsock_transport *transport_h2g;
143 /* Transport used for guest->host communication */
144 static const struct vsock_transport *transport_g2h;
145 /* Transport used for DGRAM communication */
146 static const struct vsock_transport *transport_dgram;
147 /* Transport used for local communication */
148 static const struct vsock_transport *transport_local;
149 static DEFINE_MUTEX(vsock_register_mutex);
150 
151 /**** UTILS ****/
152 
153 /* Each bound VSocket is stored in the bind hash table and each connected
154  * VSocket is stored in the connected hash table.
155  *
156  * Unbound sockets are all put on the same list attached to the end of the hash
157  * table (vsock_unbound_sockets).  Bound sockets are added to the hash table in
158  * the bucket that their local address hashes to (vsock_bound_sockets(addr)
159  * represents the list that addr hashes to).
160  *
161  * Specifically, we initialize the vsock_bind_table array to a size of
162  * VSOCK_HASH_SIZE + 1 so that vsock_bind_table[0] through
163  * vsock_bind_table[VSOCK_HASH_SIZE - 1] are for bound sockets and
164  * vsock_bind_table[VSOCK_HASH_SIZE] is for unbound sockets.  The hash function
165  * mods with VSOCK_HASH_SIZE to ensure this.
166  */
167 #define MAX_PORT_RETRIES        24
168 
169 #define VSOCK_HASH(addr)        ((addr)->svm_port % VSOCK_HASH_SIZE)
170 #define vsock_bound_sockets(addr) (&vsock_bind_table[VSOCK_HASH(addr)])
171 #define vsock_unbound_sockets     (&vsock_bind_table[VSOCK_HASH_SIZE])
172 
173 /* XXX This can probably be implemented in a better way. */
174 #define VSOCK_CONN_HASH(src, dst)				\
175 	(((src)->svm_cid ^ (dst)->svm_port) % VSOCK_HASH_SIZE)
176 #define vsock_connected_sockets(src, dst)		\
177 	(&vsock_connected_table[VSOCK_CONN_HASH(src, dst)])
178 #define vsock_connected_sockets_vsk(vsk)				\
179 	vsock_connected_sockets(&(vsk)->remote_addr, &(vsk)->local_addr)
180 
181 struct list_head vsock_bind_table[VSOCK_HASH_SIZE + 1];
182 EXPORT_SYMBOL_GPL(vsock_bind_table);
183 struct list_head vsock_connected_table[VSOCK_HASH_SIZE];
184 EXPORT_SYMBOL_GPL(vsock_connected_table);
185 DEFINE_SPINLOCK(vsock_table_lock);
186 EXPORT_SYMBOL_GPL(vsock_table_lock);
187 
188 /* Autobind this socket to the local address if necessary. */
vsock_auto_bind(struct vsock_sock * vsk)189 static int vsock_auto_bind(struct vsock_sock *vsk)
190 {
191 	struct sock *sk = sk_vsock(vsk);
192 	struct sockaddr_vm local_addr;
193 
194 	if (vsock_addr_bound(&vsk->local_addr))
195 		return 0;
196 	vsock_addr_init(&local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
197 	return __vsock_bind(sk, &local_addr);
198 }
199 
vsock_init_tables(void)200 static void vsock_init_tables(void)
201 {
202 	int i;
203 
204 	for (i = 0; i < ARRAY_SIZE(vsock_bind_table); i++)
205 		INIT_LIST_HEAD(&vsock_bind_table[i]);
206 
207 	for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++)
208 		INIT_LIST_HEAD(&vsock_connected_table[i]);
209 }
210 
__vsock_insert_bound(struct list_head * list,struct vsock_sock * vsk)211 static void __vsock_insert_bound(struct list_head *list,
212 				 struct vsock_sock *vsk)
213 {
214 	sock_hold(&vsk->sk);
215 	list_add(&vsk->bound_table, list);
216 }
217 
__vsock_insert_connected(struct list_head * list,struct vsock_sock * vsk)218 static void __vsock_insert_connected(struct list_head *list,
219 				     struct vsock_sock *vsk)
220 {
221 	sock_hold(&vsk->sk);
222 	list_add(&vsk->connected_table, list);
223 }
224 
__vsock_remove_bound(struct vsock_sock * vsk)225 static void __vsock_remove_bound(struct vsock_sock *vsk)
226 {
227 	list_del_init(&vsk->bound_table);
228 	sock_put(&vsk->sk);
229 }
230 
__vsock_remove_connected(struct vsock_sock * vsk)231 static void __vsock_remove_connected(struct vsock_sock *vsk)
232 {
233 	list_del_init(&vsk->connected_table);
234 	sock_put(&vsk->sk);
235 }
236 
__vsock_find_bound_socket(struct sockaddr_vm * addr)237 static struct sock *__vsock_find_bound_socket(struct sockaddr_vm *addr)
238 {
239 	struct vsock_sock *vsk;
240 
241 	list_for_each_entry(vsk, vsock_bound_sockets(addr), bound_table) {
242 		if (vsock_addr_equals_addr(addr, &vsk->local_addr))
243 			return sk_vsock(vsk);
244 
245 		if (addr->svm_port == vsk->local_addr.svm_port &&
246 		    (vsk->local_addr.svm_cid == VMADDR_CID_ANY ||
247 		     addr->svm_cid == VMADDR_CID_ANY))
248 			return sk_vsock(vsk);
249 	}
250 
251 	return NULL;
252 }
253 
__vsock_find_connected_socket(struct sockaddr_vm * src,struct sockaddr_vm * dst)254 static struct sock *__vsock_find_connected_socket(struct sockaddr_vm *src,
255 						  struct sockaddr_vm *dst)
256 {
257 	struct vsock_sock *vsk;
258 
259 	list_for_each_entry(vsk, vsock_connected_sockets(src, dst),
260 			    connected_table) {
261 		if (vsock_addr_equals_addr(src, &vsk->remote_addr) &&
262 		    dst->svm_port == vsk->local_addr.svm_port) {
263 			return sk_vsock(vsk);
264 		}
265 	}
266 
267 	return NULL;
268 }
269 
vsock_insert_unbound(struct vsock_sock * vsk)270 static void vsock_insert_unbound(struct vsock_sock *vsk)
271 {
272 	spin_lock_bh(&vsock_table_lock);
273 	__vsock_insert_bound(vsock_unbound_sockets, vsk);
274 	spin_unlock_bh(&vsock_table_lock);
275 }
276 
vsock_insert_connected(struct vsock_sock * vsk)277 void vsock_insert_connected(struct vsock_sock *vsk)
278 {
279 	struct list_head *list = vsock_connected_sockets(
280 		&vsk->remote_addr, &vsk->local_addr);
281 
282 	spin_lock_bh(&vsock_table_lock);
283 	__vsock_insert_connected(list, vsk);
284 	spin_unlock_bh(&vsock_table_lock);
285 }
286 EXPORT_SYMBOL_GPL(vsock_insert_connected);
287 
vsock_remove_bound(struct vsock_sock * vsk)288 void vsock_remove_bound(struct vsock_sock *vsk)
289 {
290 	spin_lock_bh(&vsock_table_lock);
291 	if (__vsock_in_bound_table(vsk))
292 		__vsock_remove_bound(vsk);
293 	spin_unlock_bh(&vsock_table_lock);
294 }
295 EXPORT_SYMBOL_GPL(vsock_remove_bound);
296 
vsock_remove_connected(struct vsock_sock * vsk)297 void vsock_remove_connected(struct vsock_sock *vsk)
298 {
299 	spin_lock_bh(&vsock_table_lock);
300 	if (__vsock_in_connected_table(vsk))
301 		__vsock_remove_connected(vsk);
302 	spin_unlock_bh(&vsock_table_lock);
303 }
304 EXPORT_SYMBOL_GPL(vsock_remove_connected);
305 
vsock_find_bound_socket(struct sockaddr_vm * addr)306 struct sock *vsock_find_bound_socket(struct sockaddr_vm *addr)
307 {
308 	struct sock *sk;
309 
310 	spin_lock_bh(&vsock_table_lock);
311 	sk = __vsock_find_bound_socket(addr);
312 	if (sk)
313 		sock_hold(sk);
314 
315 	spin_unlock_bh(&vsock_table_lock);
316 
317 	return sk;
318 }
319 EXPORT_SYMBOL_GPL(vsock_find_bound_socket);
320 
vsock_find_connected_socket(struct sockaddr_vm * src,struct sockaddr_vm * dst)321 struct sock *vsock_find_connected_socket(struct sockaddr_vm *src,
322 					 struct sockaddr_vm *dst)
323 {
324 	struct sock *sk;
325 
326 	spin_lock_bh(&vsock_table_lock);
327 	sk = __vsock_find_connected_socket(src, dst);
328 	if (sk)
329 		sock_hold(sk);
330 
331 	spin_unlock_bh(&vsock_table_lock);
332 
333 	return sk;
334 }
335 EXPORT_SYMBOL_GPL(vsock_find_connected_socket);
336 
vsock_remove_sock(struct vsock_sock * vsk)337 void vsock_remove_sock(struct vsock_sock *vsk)
338 {
339 	/* Transport reassignment must not remove the binding. */
340 	if (sock_flag(sk_vsock(vsk), SOCK_DEAD))
341 		vsock_remove_bound(vsk);
342 
343 	vsock_remove_connected(vsk);
344 }
345 EXPORT_SYMBOL_GPL(vsock_remove_sock);
346 
vsock_for_each_connected_socket(struct vsock_transport * transport,void (* fn)(struct sock * sk))347 void vsock_for_each_connected_socket(struct vsock_transport *transport,
348 				     void (*fn)(struct sock *sk))
349 {
350 	int i;
351 
352 	spin_lock_bh(&vsock_table_lock);
353 
354 	for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) {
355 		struct vsock_sock *vsk;
356 		list_for_each_entry(vsk, &vsock_connected_table[i],
357 				    connected_table) {
358 			if (vsk->transport != transport)
359 				continue;
360 
361 			fn(sk_vsock(vsk));
362 		}
363 	}
364 
365 	spin_unlock_bh(&vsock_table_lock);
366 }
367 EXPORT_SYMBOL_GPL(vsock_for_each_connected_socket);
368 
vsock_add_pending(struct sock * listener,struct sock * pending)369 void vsock_add_pending(struct sock *listener, struct sock *pending)
370 {
371 	struct vsock_sock *vlistener;
372 	struct vsock_sock *vpending;
373 
374 	vlistener = vsock_sk(listener);
375 	vpending = vsock_sk(pending);
376 
377 	sock_hold(pending);
378 	sock_hold(listener);
379 	list_add_tail(&vpending->pending_links, &vlistener->pending_links);
380 }
381 EXPORT_SYMBOL_GPL(vsock_add_pending);
382 
vsock_remove_pending(struct sock * listener,struct sock * pending)383 void vsock_remove_pending(struct sock *listener, struct sock *pending)
384 {
385 	struct vsock_sock *vpending = vsock_sk(pending);
386 
387 	list_del_init(&vpending->pending_links);
388 	sock_put(listener);
389 	sock_put(pending);
390 }
391 EXPORT_SYMBOL_GPL(vsock_remove_pending);
392 
vsock_enqueue_accept(struct sock * listener,struct sock * connected)393 void vsock_enqueue_accept(struct sock *listener, struct sock *connected)
394 {
395 	struct vsock_sock *vlistener;
396 	struct vsock_sock *vconnected;
397 
398 	vlistener = vsock_sk(listener);
399 	vconnected = vsock_sk(connected);
400 
401 	sock_hold(connected);
402 	sock_hold(listener);
403 	list_add_tail(&vconnected->accept_queue, &vlistener->accept_queue);
404 }
405 EXPORT_SYMBOL_GPL(vsock_enqueue_accept);
406 
vsock_use_local_transport(unsigned int remote_cid)407 static bool vsock_use_local_transport(unsigned int remote_cid)
408 {
409 	if (!transport_local)
410 		return false;
411 
412 	if (remote_cid == VMADDR_CID_LOCAL)
413 		return true;
414 
415 	if (transport_g2h) {
416 		return remote_cid == transport_g2h->get_local_cid();
417 	} else {
418 		return remote_cid == VMADDR_CID_HOST;
419 	}
420 }
421 
vsock_deassign_transport(struct vsock_sock * vsk)422 static void vsock_deassign_transport(struct vsock_sock *vsk)
423 {
424 	if (!vsk->transport)
425 		return;
426 
427 	vsk->transport->destruct(vsk);
428 	module_put(vsk->transport->module);
429 	vsk->transport = NULL;
430 }
431 
432 /* Assign a transport to a socket and call the .init transport callback.
433  *
434  * Note: for connection oriented socket this must be called when vsk->remote_addr
435  * is set (e.g. during the connect() or when a connection request on a listener
436  * socket is received).
437  * The vsk->remote_addr is used to decide which transport to use:
438  *  - remote CID == VMADDR_CID_LOCAL or g2h->local_cid or VMADDR_CID_HOST if
439  *    g2h is not loaded, will use local transport;
440  *  - remote CID <= VMADDR_CID_HOST or h2g is not loaded or remote flags field
441  *    includes VMADDR_FLAG_TO_HOST flag value, will use guest->host transport;
442  *  - remote CID > VMADDR_CID_HOST will use host->guest transport;
443  */
vsock_assign_transport(struct vsock_sock * vsk,struct vsock_sock * psk)444 int vsock_assign_transport(struct vsock_sock *vsk, struct vsock_sock *psk)
445 {
446 	const struct vsock_transport *new_transport;
447 	struct sock *sk = sk_vsock(vsk);
448 	unsigned int remote_cid = vsk->remote_addr.svm_cid;
449 	__u8 remote_flags;
450 	int ret;
451 
452 	/* If the packet is coming with the source and destination CIDs higher
453 	 * than VMADDR_CID_HOST, then a vsock channel where all the packets are
454 	 * forwarded to the host should be established. Then the host will
455 	 * need to forward the packets to the guest.
456 	 *
457 	 * The flag is set on the (listen) receive path (psk is not NULL). On
458 	 * the connect path the flag can be set by the user space application.
459 	 */
460 	if (psk && vsk->local_addr.svm_cid > VMADDR_CID_HOST &&
461 	    vsk->remote_addr.svm_cid > VMADDR_CID_HOST)
462 		vsk->remote_addr.svm_flags |= VMADDR_FLAG_TO_HOST;
463 
464 	remote_flags = vsk->remote_addr.svm_flags;
465 
466 	switch (sk->sk_type) {
467 	case SOCK_DGRAM:
468 		new_transport = transport_dgram;
469 		break;
470 	case SOCK_STREAM:
471 	case SOCK_SEQPACKET:
472 		if (vsock_use_local_transport(remote_cid))
473 			new_transport = transport_local;
474 		else if (remote_cid <= VMADDR_CID_HOST || !transport_h2g ||
475 			 (remote_flags & VMADDR_FLAG_TO_HOST))
476 			new_transport = transport_g2h;
477 		else
478 			new_transport = transport_h2g;
479 		break;
480 	default:
481 		return -ESOCKTNOSUPPORT;
482 	}
483 
484 	if (vsk->transport) {
485 		if (vsk->transport == new_transport)
486 			return 0;
487 
488 		/* transport->release() must be called with sock lock acquired.
489 		 * This path can only be taken during vsock_connect(), where we
490 		 * have already held the sock lock. In the other cases, this
491 		 * function is called on a new socket which is not assigned to
492 		 * any transport.
493 		 */
494 		vsk->transport->release(vsk);
495 		vsock_deassign_transport(vsk);
496 
497 		/* transport's release() and destruct() can touch some socket
498 		 * state, since we are reassigning the socket to a new transport
499 		 * during vsock_connect(), let's reset these fields to have a
500 		 * clean state.
501 		 */
502 		sock_reset_flag(sk, SOCK_DONE);
503 		sk->sk_state = TCP_CLOSE;
504 		vsk->peer_shutdown = 0;
505 	}
506 
507 	/* We increase the module refcnt to prevent the transport unloading
508 	 * while there are open sockets assigned to it.
509 	 */
510 	if (!new_transport || !try_module_get(new_transport->module))
511 		return -ENODEV;
512 
513 	if (sk->sk_type == SOCK_SEQPACKET) {
514 		if (!new_transport->seqpacket_allow ||
515 		    !new_transport->seqpacket_allow(remote_cid)) {
516 			module_put(new_transport->module);
517 			return -ESOCKTNOSUPPORT;
518 		}
519 	}
520 
521 	ret = new_transport->init(vsk, psk);
522 	if (ret) {
523 		module_put(new_transport->module);
524 		return ret;
525 	}
526 
527 	vsk->transport = new_transport;
528 
529 	return 0;
530 }
531 EXPORT_SYMBOL_GPL(vsock_assign_transport);
532 
vsock_find_cid(unsigned int cid)533 bool vsock_find_cid(unsigned int cid)
534 {
535 	if (transport_g2h && cid == transport_g2h->get_local_cid())
536 		return true;
537 
538 	if (transport_h2g && cid == VMADDR_CID_HOST)
539 		return true;
540 
541 	if (transport_local && cid == VMADDR_CID_LOCAL)
542 		return true;
543 
544 	return false;
545 }
546 EXPORT_SYMBOL_GPL(vsock_find_cid);
547 
vsock_dequeue_accept(struct sock * listener)548 static struct sock *vsock_dequeue_accept(struct sock *listener)
549 {
550 	struct vsock_sock *vlistener;
551 	struct vsock_sock *vconnected;
552 
553 	vlistener = vsock_sk(listener);
554 
555 	if (list_empty(&vlistener->accept_queue))
556 		return NULL;
557 
558 	vconnected = list_entry(vlistener->accept_queue.next,
559 				struct vsock_sock, accept_queue);
560 
561 	list_del_init(&vconnected->accept_queue);
562 	sock_put(listener);
563 	/* The caller will need a reference on the connected socket so we let
564 	 * it call sock_put().
565 	 */
566 
567 	return sk_vsock(vconnected);
568 }
569 
vsock_is_accept_queue_empty(struct sock * sk)570 static bool vsock_is_accept_queue_empty(struct sock *sk)
571 {
572 	struct vsock_sock *vsk = vsock_sk(sk);
573 	return list_empty(&vsk->accept_queue);
574 }
575 
vsock_is_pending(struct sock * sk)576 static bool vsock_is_pending(struct sock *sk)
577 {
578 	struct vsock_sock *vsk = vsock_sk(sk);
579 	return !list_empty(&vsk->pending_links);
580 }
581 
vsock_send_shutdown(struct sock * sk,int mode)582 static int vsock_send_shutdown(struct sock *sk, int mode)
583 {
584 	struct vsock_sock *vsk = vsock_sk(sk);
585 
586 	if (!vsk->transport)
587 		return -ENODEV;
588 
589 	return vsk->transport->shutdown(vsk, mode);
590 }
591 
vsock_pending_work(struct work_struct * work)592 static void vsock_pending_work(struct work_struct *work)
593 {
594 	struct sock *sk;
595 	struct sock *listener;
596 	struct vsock_sock *vsk;
597 	bool cleanup;
598 
599 	vsk = container_of(work, struct vsock_sock, pending_work.work);
600 	sk = sk_vsock(vsk);
601 	listener = vsk->listener;
602 	cleanup = true;
603 
604 	lock_sock(listener);
605 	lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
606 
607 	if (vsock_is_pending(sk)) {
608 		vsock_remove_pending(listener, sk);
609 
610 		sk_acceptq_removed(listener);
611 	} else if (!vsk->rejected) {
612 		/* We are not on the pending list and accept() did not reject
613 		 * us, so we must have been accepted by our user process.  We
614 		 * just need to drop our references to the sockets and be on
615 		 * our way.
616 		 */
617 		cleanup = false;
618 		goto out;
619 	}
620 
621 	/* We need to remove ourself from the global connected sockets list so
622 	 * incoming packets can't find this socket, and to reduce the reference
623 	 * count.
624 	 */
625 	vsock_remove_connected(vsk);
626 
627 	sk->sk_state = TCP_CLOSE;
628 
629 out:
630 	release_sock(sk);
631 	release_sock(listener);
632 	if (cleanup)
633 		sock_put(sk);
634 
635 	sock_put(sk);
636 	sock_put(listener);
637 }
638 
639 /**** SOCKET OPERATIONS ****/
640 
__vsock_bind_connectible(struct vsock_sock * vsk,struct sockaddr_vm * addr)641 static int __vsock_bind_connectible(struct vsock_sock *vsk,
642 				    struct sockaddr_vm *addr)
643 {
644 	static u32 port;
645 	struct sockaddr_vm new_addr;
646 
647 	if (!port)
648 		port = get_random_u32_above(LAST_RESERVED_PORT);
649 
650 	vsock_addr_init(&new_addr, addr->svm_cid, addr->svm_port);
651 
652 	if (addr->svm_port == VMADDR_PORT_ANY) {
653 		bool found = false;
654 		unsigned int i;
655 
656 		for (i = 0; i < MAX_PORT_RETRIES; i++) {
657 			if (port <= LAST_RESERVED_PORT)
658 				port = LAST_RESERVED_PORT + 1;
659 
660 			new_addr.svm_port = port++;
661 
662 			if (!__vsock_find_bound_socket(&new_addr)) {
663 				found = true;
664 				break;
665 			}
666 		}
667 
668 		if (!found)
669 			return -EADDRNOTAVAIL;
670 	} else {
671 		/* If port is in reserved range, ensure caller
672 		 * has necessary privileges.
673 		 */
674 		if (addr->svm_port <= LAST_RESERVED_PORT &&
675 		    !capable(CAP_NET_BIND_SERVICE)) {
676 			return -EACCES;
677 		}
678 
679 		if (__vsock_find_bound_socket(&new_addr))
680 			return -EADDRINUSE;
681 	}
682 
683 	vsock_addr_init(&vsk->local_addr, new_addr.svm_cid, new_addr.svm_port);
684 
685 	/* Remove connection oriented sockets from the unbound list and add them
686 	 * to the hash table for easy lookup by its address.  The unbound list
687 	 * is simply an extra entry at the end of the hash table, a trick used
688 	 * by AF_UNIX.
689 	 */
690 	__vsock_remove_bound(vsk);
691 	__vsock_insert_bound(vsock_bound_sockets(&vsk->local_addr), vsk);
692 
693 	return 0;
694 }
695 
__vsock_bind_dgram(struct vsock_sock * vsk,struct sockaddr_vm * addr)696 static int __vsock_bind_dgram(struct vsock_sock *vsk,
697 			      struct sockaddr_vm *addr)
698 {
699 	return vsk->transport->dgram_bind(vsk, addr);
700 }
701 
__vsock_bind(struct sock * sk,struct sockaddr_vm * addr)702 static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr)
703 {
704 	struct vsock_sock *vsk = vsock_sk(sk);
705 	int retval;
706 
707 	/* First ensure this socket isn't already bound. */
708 	if (vsock_addr_bound(&vsk->local_addr))
709 		return -EINVAL;
710 
711 	/* Now bind to the provided address or select appropriate values if
712 	 * none are provided (VMADDR_CID_ANY and VMADDR_PORT_ANY).  Note that
713 	 * like AF_INET prevents binding to a non-local IP address (in most
714 	 * cases), we only allow binding to a local CID.
715 	 */
716 	if (addr->svm_cid != VMADDR_CID_ANY && !vsock_find_cid(addr->svm_cid))
717 		return -EADDRNOTAVAIL;
718 
719 	switch (sk->sk_socket->type) {
720 	case SOCK_STREAM:
721 	case SOCK_SEQPACKET:
722 		spin_lock_bh(&vsock_table_lock);
723 		retval = __vsock_bind_connectible(vsk, addr);
724 		spin_unlock_bh(&vsock_table_lock);
725 		break;
726 
727 	case SOCK_DGRAM:
728 		retval = __vsock_bind_dgram(vsk, addr);
729 		break;
730 
731 	default:
732 		retval = -EINVAL;
733 		break;
734 	}
735 
736 	return retval;
737 }
738 
739 static void vsock_connect_timeout(struct work_struct *work);
740 
__vsock_create(struct net * net,struct socket * sock,struct sock * parent,gfp_t priority,unsigned short type,int kern)741 static struct sock *__vsock_create(struct net *net,
742 				   struct socket *sock,
743 				   struct sock *parent,
744 				   gfp_t priority,
745 				   unsigned short type,
746 				   int kern)
747 {
748 	struct sock *sk;
749 	struct vsock_sock *psk;
750 	struct vsock_sock *vsk;
751 
752 	sk = sk_alloc(net, AF_VSOCK, priority, &vsock_proto, kern);
753 	if (!sk)
754 		return NULL;
755 
756 	sock_init_data(sock, sk);
757 
758 	/* sk->sk_type is normally set in sock_init_data, but only if sock is
759 	 * non-NULL. We make sure that our sockets always have a type by
760 	 * setting it here if needed.
761 	 */
762 	if (!sock)
763 		sk->sk_type = type;
764 
765 	vsk = vsock_sk(sk);
766 	vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
767 	vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
768 
769 	sk->sk_destruct = vsock_sk_destruct;
770 	sk->sk_backlog_rcv = vsock_queue_rcv_skb;
771 	sock_reset_flag(sk, SOCK_DONE);
772 
773 	INIT_LIST_HEAD(&vsk->bound_table);
774 	INIT_LIST_HEAD(&vsk->connected_table);
775 	vsk->listener = NULL;
776 	INIT_LIST_HEAD(&vsk->pending_links);
777 	INIT_LIST_HEAD(&vsk->accept_queue);
778 	vsk->rejected = false;
779 	vsk->sent_request = false;
780 	vsk->ignore_connecting_rst = false;
781 	vsk->peer_shutdown = 0;
782 	INIT_DELAYED_WORK(&vsk->connect_work, vsock_connect_timeout);
783 	INIT_DELAYED_WORK(&vsk->pending_work, vsock_pending_work);
784 
785 	psk = parent ? vsock_sk(parent) : NULL;
786 	if (parent) {
787 		vsk->trusted = psk->trusted;
788 		vsk->owner = get_cred(psk->owner);
789 		vsk->connect_timeout = psk->connect_timeout;
790 		vsk->buffer_size = psk->buffer_size;
791 		vsk->buffer_min_size = psk->buffer_min_size;
792 		vsk->buffer_max_size = psk->buffer_max_size;
793 		security_sk_clone(parent, sk);
794 	} else {
795 		vsk->trusted = ns_capable_noaudit(&init_user_ns, CAP_NET_ADMIN);
796 		vsk->owner = get_current_cred();
797 		vsk->connect_timeout = VSOCK_DEFAULT_CONNECT_TIMEOUT;
798 		vsk->buffer_size = VSOCK_DEFAULT_BUFFER_SIZE;
799 		vsk->buffer_min_size = VSOCK_DEFAULT_BUFFER_MIN_SIZE;
800 		vsk->buffer_max_size = VSOCK_DEFAULT_BUFFER_MAX_SIZE;
801 	}
802 
803 	return sk;
804 }
805 
sock_type_connectible(u16 type)806 static bool sock_type_connectible(u16 type)
807 {
808 	return (type == SOCK_STREAM) || (type == SOCK_SEQPACKET);
809 }
810 
__vsock_release(struct sock * sk,int level)811 static void __vsock_release(struct sock *sk, int level)
812 {
813 	struct vsock_sock *vsk;
814 	struct sock *pending;
815 
816 	vsk = vsock_sk(sk);
817 	pending = NULL;	/* Compiler warning. */
818 
819 	/* When "level" is SINGLE_DEPTH_NESTING, use the nested
820 	 * version to avoid the warning "possible recursive locking
821 	 * detected". When "level" is 0, lock_sock_nested(sk, level)
822 	 * is the same as lock_sock(sk).
823 	 */
824 	lock_sock_nested(sk, level);
825 
826 	/* Indicate to vsock_remove_sock() that the socket is being released and
827 	 * can be removed from the bound_table. Unlike transport reassignment
828 	 * case, where the socket must remain bound despite vsock_remove_sock()
829 	 * being called from the transport release() callback.
830 	 */
831 	sock_set_flag(sk, SOCK_DEAD);
832 
833 	if (vsk->transport)
834 		vsk->transport->release(vsk);
835 	else if (sock_type_connectible(sk->sk_type))
836 		vsock_remove_sock(vsk);
837 
838 	sock_orphan(sk);
839 	sk->sk_shutdown = SHUTDOWN_MASK;
840 
841 	skb_queue_purge(&sk->sk_receive_queue);
842 
843 	/* Clean up any sockets that never were accepted. */
844 	while ((pending = vsock_dequeue_accept(sk)) != NULL) {
845 		__vsock_release(pending, SINGLE_DEPTH_NESTING);
846 		sock_put(pending);
847 	}
848 
849 	release_sock(sk);
850 	sock_put(sk);
851 }
852 
vsock_sk_destruct(struct sock * sk)853 static void vsock_sk_destruct(struct sock *sk)
854 {
855 	struct vsock_sock *vsk = vsock_sk(sk);
856 
857 	vsock_deassign_transport(vsk);
858 
859 	/* When clearing these addresses, there's no need to set the family and
860 	 * possibly register the address family with the kernel.
861 	 */
862 	vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
863 	vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
864 
865 	put_cred(vsk->owner);
866 }
867 
vsock_queue_rcv_skb(struct sock * sk,struct sk_buff * skb)868 static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
869 {
870 	int err;
871 
872 	err = sock_queue_rcv_skb(sk, skb);
873 	if (err)
874 		kfree_skb(skb);
875 
876 	return err;
877 }
878 
vsock_create_connected(struct sock * parent)879 struct sock *vsock_create_connected(struct sock *parent)
880 {
881 	return __vsock_create(sock_net(parent), NULL, parent, GFP_KERNEL,
882 			      parent->sk_type, 0);
883 }
884 EXPORT_SYMBOL_GPL(vsock_create_connected);
885 
vsock_stream_has_data(struct vsock_sock * vsk)886 s64 vsock_stream_has_data(struct vsock_sock *vsk)
887 {
888 	if (WARN_ON(!vsk->transport))
889 		return 0;
890 
891 	return vsk->transport->stream_has_data(vsk);
892 }
893 EXPORT_SYMBOL_GPL(vsock_stream_has_data);
894 
vsock_connectible_has_data(struct vsock_sock * vsk)895 s64 vsock_connectible_has_data(struct vsock_sock *vsk)
896 {
897 	struct sock *sk = sk_vsock(vsk);
898 
899 	if (WARN_ON(!vsk->transport))
900 		return 0;
901 
902 	if (sk->sk_type == SOCK_SEQPACKET)
903 		return vsk->transport->seqpacket_has_data(vsk);
904 	else
905 		return vsock_stream_has_data(vsk);
906 }
907 EXPORT_SYMBOL_GPL(vsock_connectible_has_data);
908 
vsock_stream_has_space(struct vsock_sock * vsk)909 s64 vsock_stream_has_space(struct vsock_sock *vsk)
910 {
911 	if (WARN_ON(!vsk->transport))
912 		return 0;
913 
914 	return vsk->transport->stream_has_space(vsk);
915 }
916 EXPORT_SYMBOL_GPL(vsock_stream_has_space);
917 
vsock_data_ready(struct sock * sk)918 void vsock_data_ready(struct sock *sk)
919 {
920 	struct vsock_sock *vsk = vsock_sk(sk);
921 
922 	if (vsock_stream_has_data(vsk) >= sk->sk_rcvlowat ||
923 	    sock_flag(sk, SOCK_DONE))
924 		sk->sk_data_ready(sk);
925 }
926 EXPORT_SYMBOL_GPL(vsock_data_ready);
927 
928 /* Dummy callback required by sockmap.
929  * See unconditional call of saved_close() in sock_map_close().
930  */
vsock_close(struct sock * sk,long timeout)931 static void vsock_close(struct sock *sk, long timeout)
932 {
933 }
934 
vsock_release(struct socket * sock)935 static int vsock_release(struct socket *sock)
936 {
937 	struct sock *sk = sock->sk;
938 
939 	if (!sk)
940 		return 0;
941 
942 	sk->sk_prot->close(sk, 0);
943 	__vsock_release(sk, 0);
944 	sock->sk = NULL;
945 	sock->state = SS_FREE;
946 
947 	return 0;
948 }
949 
950 static int
vsock_bind(struct socket * sock,struct sockaddr * addr,int addr_len)951 vsock_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
952 {
953 	int err;
954 	struct sock *sk;
955 	struct sockaddr_vm *vm_addr;
956 
957 	sk = sock->sk;
958 
959 	if (vsock_addr_cast(addr, addr_len, &vm_addr) != 0)
960 		return -EINVAL;
961 
962 	lock_sock(sk);
963 	err = __vsock_bind(sk, vm_addr);
964 	release_sock(sk);
965 
966 	return err;
967 }
968 
vsock_getname(struct socket * sock,struct sockaddr * addr,int peer)969 static int vsock_getname(struct socket *sock,
970 			 struct sockaddr *addr, int peer)
971 {
972 	int err;
973 	struct sock *sk;
974 	struct vsock_sock *vsk;
975 	struct sockaddr_vm *vm_addr;
976 
977 	sk = sock->sk;
978 	vsk = vsock_sk(sk);
979 	err = 0;
980 
981 	lock_sock(sk);
982 
983 	if (peer) {
984 		if (sock->state != SS_CONNECTED) {
985 			err = -ENOTCONN;
986 			goto out;
987 		}
988 		vm_addr = &vsk->remote_addr;
989 	} else {
990 		vm_addr = &vsk->local_addr;
991 	}
992 
993 	if (!vm_addr) {
994 		err = -EINVAL;
995 		goto out;
996 	}
997 
998 	/* sys_getsockname() and sys_getpeername() pass us a
999 	 * MAX_SOCK_ADDR-sized buffer and don't set addr_len.  Unfortunately
1000 	 * that macro is defined in socket.c instead of .h, so we hardcode its
1001 	 * value here.
1002 	 */
1003 	BUILD_BUG_ON(sizeof(*vm_addr) > 128);
1004 	memcpy(addr, vm_addr, sizeof(*vm_addr));
1005 	err = sizeof(*vm_addr);
1006 
1007 out:
1008 	release_sock(sk);
1009 	return err;
1010 }
1011 
vsock_shutdown(struct socket * sock,int mode)1012 static int vsock_shutdown(struct socket *sock, int mode)
1013 {
1014 	int err;
1015 	struct sock *sk;
1016 
1017 	/* User level uses SHUT_RD (0) and SHUT_WR (1), but the kernel uses
1018 	 * RCV_SHUTDOWN (1) and SEND_SHUTDOWN (2), so we must increment mode
1019 	 * here like the other address families do.  Note also that the
1020 	 * increment makes SHUT_RDWR (2) into RCV_SHUTDOWN | SEND_SHUTDOWN (3),
1021 	 * which is what we want.
1022 	 */
1023 	mode++;
1024 
1025 	if ((mode & ~SHUTDOWN_MASK) || !mode)
1026 		return -EINVAL;
1027 
1028 	/* If this is a connection oriented socket and it is not connected then
1029 	 * bail out immediately.  If it is a DGRAM socket then we must first
1030 	 * kick the socket so that it wakes up from any sleeping calls, for
1031 	 * example recv(), and then afterwards return the error.
1032 	 */
1033 
1034 	sk = sock->sk;
1035 
1036 	lock_sock(sk);
1037 	if (sock->state == SS_UNCONNECTED) {
1038 		err = -ENOTCONN;
1039 		if (sock_type_connectible(sk->sk_type))
1040 			goto out;
1041 	} else {
1042 		sock->state = SS_DISCONNECTING;
1043 		err = 0;
1044 	}
1045 
1046 	/* Receive and send shutdowns are treated alike. */
1047 	mode = mode & (RCV_SHUTDOWN | SEND_SHUTDOWN);
1048 	if (mode) {
1049 		sk->sk_shutdown |= mode;
1050 		sk->sk_state_change(sk);
1051 
1052 		if (sock_type_connectible(sk->sk_type)) {
1053 			sock_reset_flag(sk, SOCK_DONE);
1054 			vsock_send_shutdown(sk, mode);
1055 		}
1056 	}
1057 
1058 out:
1059 	release_sock(sk);
1060 	return err;
1061 }
1062 
vsock_poll(struct file * file,struct socket * sock,poll_table * wait)1063 static __poll_t vsock_poll(struct file *file, struct socket *sock,
1064 			       poll_table *wait)
1065 {
1066 	struct sock *sk;
1067 	__poll_t mask;
1068 	struct vsock_sock *vsk;
1069 
1070 	sk = sock->sk;
1071 	vsk = vsock_sk(sk);
1072 
1073 	poll_wait(file, sk_sleep(sk), wait);
1074 	mask = 0;
1075 
1076 	if (sk->sk_err)
1077 		/* Signify that there has been an error on this socket. */
1078 		mask |= EPOLLERR;
1079 
1080 	/* INET sockets treat local write shutdown and peer write shutdown as a
1081 	 * case of EPOLLHUP set.
1082 	 */
1083 	if ((sk->sk_shutdown == SHUTDOWN_MASK) ||
1084 	    ((sk->sk_shutdown & SEND_SHUTDOWN) &&
1085 	     (vsk->peer_shutdown & SEND_SHUTDOWN))) {
1086 		mask |= EPOLLHUP;
1087 	}
1088 
1089 	if (sk->sk_shutdown & RCV_SHUTDOWN ||
1090 	    vsk->peer_shutdown & SEND_SHUTDOWN) {
1091 		mask |= EPOLLRDHUP;
1092 	}
1093 
1094 	if (sk_is_readable(sk))
1095 		mask |= EPOLLIN | EPOLLRDNORM;
1096 
1097 	if (sock->type == SOCK_DGRAM) {
1098 		/* For datagram sockets we can read if there is something in
1099 		 * the queue and write as long as the socket isn't shutdown for
1100 		 * sending.
1101 		 */
1102 		if (!skb_queue_empty_lockless(&sk->sk_receive_queue) ||
1103 		    (sk->sk_shutdown & RCV_SHUTDOWN)) {
1104 			mask |= EPOLLIN | EPOLLRDNORM;
1105 		}
1106 
1107 		if (!(sk->sk_shutdown & SEND_SHUTDOWN))
1108 			mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND;
1109 
1110 	} else if (sock_type_connectible(sk->sk_type)) {
1111 		const struct vsock_transport *transport;
1112 
1113 		lock_sock(sk);
1114 
1115 		transport = vsk->transport;
1116 
1117 		/* Listening sockets that have connections in their accept
1118 		 * queue can be read.
1119 		 */
1120 		if (sk->sk_state == TCP_LISTEN
1121 		    && !vsock_is_accept_queue_empty(sk))
1122 			mask |= EPOLLIN | EPOLLRDNORM;
1123 
1124 		/* If there is something in the queue then we can read. */
1125 		if (transport && transport->stream_is_active(vsk) &&
1126 		    !(sk->sk_shutdown & RCV_SHUTDOWN)) {
1127 			bool data_ready_now = false;
1128 			int target = sock_rcvlowat(sk, 0, INT_MAX);
1129 			int ret = transport->notify_poll_in(
1130 					vsk, target, &data_ready_now);
1131 			if (ret < 0) {
1132 				mask |= EPOLLERR;
1133 			} else {
1134 				if (data_ready_now)
1135 					mask |= EPOLLIN | EPOLLRDNORM;
1136 
1137 			}
1138 		}
1139 
1140 		/* Sockets whose connections have been closed, reset, or
1141 		 * terminated should also be considered read, and we check the
1142 		 * shutdown flag for that.
1143 		 */
1144 		if (sk->sk_shutdown & RCV_SHUTDOWN ||
1145 		    vsk->peer_shutdown & SEND_SHUTDOWN) {
1146 			mask |= EPOLLIN | EPOLLRDNORM;
1147 		}
1148 
1149 		/* Connected sockets that can produce data can be written. */
1150 		if (transport && sk->sk_state == TCP_ESTABLISHED) {
1151 			if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1152 				bool space_avail_now = false;
1153 				int ret = transport->notify_poll_out(
1154 						vsk, 1, &space_avail_now);
1155 				if (ret < 0) {
1156 					mask |= EPOLLERR;
1157 				} else {
1158 					if (space_avail_now)
1159 						/* Remove EPOLLWRBAND since INET
1160 						 * sockets are not setting it.
1161 						 */
1162 						mask |= EPOLLOUT | EPOLLWRNORM;
1163 
1164 				}
1165 			}
1166 		}
1167 
1168 		/* Simulate INET socket poll behaviors, which sets
1169 		 * EPOLLOUT|EPOLLWRNORM when peer is closed and nothing to read,
1170 		 * but local send is not shutdown.
1171 		 */
1172 		if (sk->sk_state == TCP_CLOSE || sk->sk_state == TCP_CLOSING) {
1173 			if (!(sk->sk_shutdown & SEND_SHUTDOWN))
1174 				mask |= EPOLLOUT | EPOLLWRNORM;
1175 
1176 		}
1177 
1178 		release_sock(sk);
1179 	}
1180 
1181 	return mask;
1182 }
1183 
vsock_read_skb(struct sock * sk,skb_read_actor_t read_actor)1184 static int vsock_read_skb(struct sock *sk, skb_read_actor_t read_actor)
1185 {
1186 	struct vsock_sock *vsk = vsock_sk(sk);
1187 
1188 	if (WARN_ON_ONCE(!vsk->transport))
1189 		return -ENODEV;
1190 
1191 	return vsk->transport->read_skb(vsk, read_actor);
1192 }
1193 
vsock_dgram_sendmsg(struct socket * sock,struct msghdr * msg,size_t len)1194 static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg,
1195 			       size_t len)
1196 {
1197 	int err;
1198 	struct sock *sk;
1199 	struct vsock_sock *vsk;
1200 	struct sockaddr_vm *remote_addr;
1201 	const struct vsock_transport *transport;
1202 
1203 	if (msg->msg_flags & MSG_OOB)
1204 		return -EOPNOTSUPP;
1205 
1206 	/* For now, MSG_DONTWAIT is always assumed... */
1207 	err = 0;
1208 	sk = sock->sk;
1209 	vsk = vsock_sk(sk);
1210 
1211 	lock_sock(sk);
1212 
1213 	transport = vsk->transport;
1214 
1215 	err = vsock_auto_bind(vsk);
1216 	if (err)
1217 		goto out;
1218 
1219 
1220 	/* If the provided message contains an address, use that.  Otherwise
1221 	 * fall back on the socket's remote handle (if it has been connected).
1222 	 */
1223 	if (msg->msg_name &&
1224 	    vsock_addr_cast(msg->msg_name, msg->msg_namelen,
1225 			    &remote_addr) == 0) {
1226 		/* Ensure this address is of the right type and is a valid
1227 		 * destination.
1228 		 */
1229 
1230 		if (remote_addr->svm_cid == VMADDR_CID_ANY)
1231 			remote_addr->svm_cid = transport->get_local_cid();
1232 
1233 		if (!vsock_addr_bound(remote_addr)) {
1234 			err = -EINVAL;
1235 			goto out;
1236 		}
1237 	} else if (sock->state == SS_CONNECTED) {
1238 		remote_addr = &vsk->remote_addr;
1239 
1240 		if (remote_addr->svm_cid == VMADDR_CID_ANY)
1241 			remote_addr->svm_cid = transport->get_local_cid();
1242 
1243 		/* XXX Should connect() or this function ensure remote_addr is
1244 		 * bound?
1245 		 */
1246 		if (!vsock_addr_bound(&vsk->remote_addr)) {
1247 			err = -EINVAL;
1248 			goto out;
1249 		}
1250 	} else {
1251 		err = -EINVAL;
1252 		goto out;
1253 	}
1254 
1255 	if (!transport->dgram_allow(remote_addr->svm_cid,
1256 				    remote_addr->svm_port)) {
1257 		err = -EINVAL;
1258 		goto out;
1259 	}
1260 
1261 	err = transport->dgram_enqueue(vsk, remote_addr, msg, len);
1262 
1263 out:
1264 	release_sock(sk);
1265 	return err;
1266 }
1267 
vsock_dgram_connect(struct socket * sock,struct sockaddr * addr,int addr_len,int flags)1268 static int vsock_dgram_connect(struct socket *sock,
1269 			       struct sockaddr *addr, int addr_len, int flags)
1270 {
1271 	int err;
1272 	struct sock *sk;
1273 	struct vsock_sock *vsk;
1274 	struct sockaddr_vm *remote_addr;
1275 
1276 	sk = sock->sk;
1277 	vsk = vsock_sk(sk);
1278 
1279 	err = vsock_addr_cast(addr, addr_len, &remote_addr);
1280 	if (err == -EAFNOSUPPORT && remote_addr->svm_family == AF_UNSPEC) {
1281 		lock_sock(sk);
1282 		vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY,
1283 				VMADDR_PORT_ANY);
1284 		sock->state = SS_UNCONNECTED;
1285 		release_sock(sk);
1286 		return 0;
1287 	} else if (err != 0)
1288 		return -EINVAL;
1289 
1290 	lock_sock(sk);
1291 
1292 	err = vsock_auto_bind(vsk);
1293 	if (err)
1294 		goto out;
1295 
1296 	if (!vsk->transport->dgram_allow(remote_addr->svm_cid,
1297 					 remote_addr->svm_port)) {
1298 		err = -EINVAL;
1299 		goto out;
1300 	}
1301 
1302 	memcpy(&vsk->remote_addr, remote_addr, sizeof(vsk->remote_addr));
1303 	sock->state = SS_CONNECTED;
1304 
1305 	/* sock map disallows redirection of non-TCP sockets with sk_state !=
1306 	 * TCP_ESTABLISHED (see sock_map_redirect_allowed()), so we set
1307 	 * TCP_ESTABLISHED here to allow redirection of connected vsock dgrams.
1308 	 *
1309 	 * This doesn't seem to be abnormal state for datagram sockets, as the
1310 	 * same approach can be see in other datagram socket types as well
1311 	 * (such as unix sockets).
1312 	 */
1313 	sk->sk_state = TCP_ESTABLISHED;
1314 
1315 out:
1316 	release_sock(sk);
1317 	return err;
1318 }
1319 
__vsock_dgram_recvmsg(struct socket * sock,struct msghdr * msg,size_t len,int flags)1320 int __vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg,
1321 			  size_t len, int flags)
1322 {
1323 	struct sock *sk = sock->sk;
1324 	struct vsock_sock *vsk = vsock_sk(sk);
1325 
1326 	return vsk->transport->dgram_dequeue(vsk, msg, len, flags);
1327 }
1328 
vsock_dgram_recvmsg(struct socket * sock,struct msghdr * msg,size_t len,int flags)1329 int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg,
1330 			size_t len, int flags)
1331 {
1332 #ifdef CONFIG_BPF_SYSCALL
1333 	struct sock *sk = sock->sk;
1334 	const struct proto *prot;
1335 
1336 	prot = READ_ONCE(sk->sk_prot);
1337 	if (prot != &vsock_proto)
1338 		return prot->recvmsg(sk, msg, len, flags, NULL);
1339 #endif
1340 
1341 	return __vsock_dgram_recvmsg(sock, msg, len, flags);
1342 }
1343 EXPORT_SYMBOL_GPL(vsock_dgram_recvmsg);
1344 
1345 static const struct proto_ops vsock_dgram_ops = {
1346 	.family = PF_VSOCK,
1347 	.owner = THIS_MODULE,
1348 	.release = vsock_release,
1349 	.bind = vsock_bind,
1350 	.connect = vsock_dgram_connect,
1351 	.socketpair = sock_no_socketpair,
1352 	.accept = sock_no_accept,
1353 	.getname = vsock_getname,
1354 	.poll = vsock_poll,
1355 	.ioctl = sock_no_ioctl,
1356 	.listen = sock_no_listen,
1357 	.shutdown = vsock_shutdown,
1358 	.sendmsg = vsock_dgram_sendmsg,
1359 	.recvmsg = vsock_dgram_recvmsg,
1360 	.mmap = sock_no_mmap,
1361 	.read_skb = vsock_read_skb,
1362 };
1363 
vsock_transport_cancel_pkt(struct vsock_sock * vsk)1364 static int vsock_transport_cancel_pkt(struct vsock_sock *vsk)
1365 {
1366 	const struct vsock_transport *transport = vsk->transport;
1367 
1368 	if (!transport || !transport->cancel_pkt)
1369 		return -EOPNOTSUPP;
1370 
1371 	return transport->cancel_pkt(vsk);
1372 }
1373 
vsock_connect_timeout(struct work_struct * work)1374 static void vsock_connect_timeout(struct work_struct *work)
1375 {
1376 	struct sock *sk;
1377 	struct vsock_sock *vsk;
1378 
1379 	vsk = container_of(work, struct vsock_sock, connect_work.work);
1380 	sk = sk_vsock(vsk);
1381 
1382 	lock_sock(sk);
1383 	if (sk->sk_state == TCP_SYN_SENT &&
1384 	    (sk->sk_shutdown != SHUTDOWN_MASK)) {
1385 		sk->sk_state = TCP_CLOSE;
1386 		sk->sk_socket->state = SS_UNCONNECTED;
1387 		sk->sk_err = ETIMEDOUT;
1388 		sk_error_report(sk);
1389 		vsock_transport_cancel_pkt(vsk);
1390 	}
1391 	release_sock(sk);
1392 
1393 	sock_put(sk);
1394 }
1395 
vsock_connect(struct socket * sock,struct sockaddr * addr,int addr_len,int flags)1396 static int vsock_connect(struct socket *sock, struct sockaddr *addr,
1397 			 int addr_len, int flags)
1398 {
1399 	int err;
1400 	struct sock *sk;
1401 	struct vsock_sock *vsk;
1402 	const struct vsock_transport *transport;
1403 	struct sockaddr_vm *remote_addr;
1404 	long timeout;
1405 	DEFINE_WAIT(wait);
1406 
1407 	err = 0;
1408 	sk = sock->sk;
1409 	vsk = vsock_sk(sk);
1410 
1411 	lock_sock(sk);
1412 
1413 	/* XXX AF_UNSPEC should make us disconnect like AF_INET. */
1414 	switch (sock->state) {
1415 	case SS_CONNECTED:
1416 		err = -EISCONN;
1417 		goto out;
1418 	case SS_DISCONNECTING:
1419 		err = -EINVAL;
1420 		goto out;
1421 	case SS_CONNECTING:
1422 		/* This continues on so we can move sock into the SS_CONNECTED
1423 		 * state once the connection has completed (at which point err
1424 		 * will be set to zero also).  Otherwise, we will either wait
1425 		 * for the connection or return -EALREADY should this be a
1426 		 * non-blocking call.
1427 		 */
1428 		err = -EALREADY;
1429 		if (flags & O_NONBLOCK)
1430 			goto out;
1431 		break;
1432 	default:
1433 		if ((sk->sk_state == TCP_LISTEN) ||
1434 		    vsock_addr_cast(addr, addr_len, &remote_addr) != 0) {
1435 			err = -EINVAL;
1436 			goto out;
1437 		}
1438 
1439 		/* Set the remote address that we are connecting to. */
1440 		memcpy(&vsk->remote_addr, remote_addr,
1441 		       sizeof(vsk->remote_addr));
1442 
1443 		err = vsock_assign_transport(vsk, NULL);
1444 		if (err)
1445 			goto out;
1446 
1447 		transport = vsk->transport;
1448 
1449 		/* The hypervisor and well-known contexts do not have socket
1450 		 * endpoints.
1451 		 */
1452 		if (!transport ||
1453 		    !transport->stream_allow(remote_addr->svm_cid,
1454 					     remote_addr->svm_port)) {
1455 			err = -ENETUNREACH;
1456 			goto out;
1457 		}
1458 
1459 		err = vsock_auto_bind(vsk);
1460 		if (err)
1461 			goto out;
1462 
1463 		sk->sk_state = TCP_SYN_SENT;
1464 
1465 		err = transport->connect(vsk);
1466 		if (err < 0)
1467 			goto out;
1468 
1469 		/* sk_err might have been set as a result of an earlier
1470 		 * (failed) connect attempt.
1471 		 */
1472 		sk->sk_err = 0;
1473 
1474 		/* Mark sock as connecting and set the error code to in
1475 		 * progress in case this is a non-blocking connect.
1476 		 */
1477 		sock->state = SS_CONNECTING;
1478 		err = -EINPROGRESS;
1479 	}
1480 
1481 	/* The receive path will handle all communication until we are able to
1482 	 * enter the connected state.  Here we wait for the connection to be
1483 	 * completed or a notification of an error.
1484 	 */
1485 	timeout = vsk->connect_timeout;
1486 	prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1487 
1488 	while (sk->sk_state != TCP_ESTABLISHED && sk->sk_err == 0) {
1489 		if (flags & O_NONBLOCK) {
1490 			/* If we're not going to block, we schedule a timeout
1491 			 * function to generate a timeout on the connection
1492 			 * attempt, in case the peer doesn't respond in a
1493 			 * timely manner. We hold on to the socket until the
1494 			 * timeout fires.
1495 			 */
1496 			sock_hold(sk);
1497 
1498 			/* If the timeout function is already scheduled,
1499 			 * reschedule it, then ungrab the socket refcount to
1500 			 * keep it balanced.
1501 			 */
1502 			if (mod_delayed_work(system_wq, &vsk->connect_work,
1503 					     timeout))
1504 				sock_put(sk);
1505 
1506 			/* Skip ahead to preserve error code set above. */
1507 			goto out_wait;
1508 		}
1509 
1510 		release_sock(sk);
1511 		timeout = schedule_timeout(timeout);
1512 		lock_sock(sk);
1513 
1514 		if (signal_pending(current)) {
1515 			err = sock_intr_errno(timeout);
1516 			sk->sk_state = sk->sk_state == TCP_ESTABLISHED ? TCP_CLOSING : TCP_CLOSE;
1517 			sock->state = SS_UNCONNECTED;
1518 			vsock_transport_cancel_pkt(vsk);
1519 			vsock_remove_connected(vsk);
1520 			goto out_wait;
1521 		} else if ((sk->sk_state != TCP_ESTABLISHED) && (timeout == 0)) {
1522 			err = -ETIMEDOUT;
1523 			sk->sk_state = TCP_CLOSE;
1524 			sock->state = SS_UNCONNECTED;
1525 			vsock_transport_cancel_pkt(vsk);
1526 			goto out_wait;
1527 		}
1528 
1529 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1530 	}
1531 
1532 	if (sk->sk_err) {
1533 		err = -sk->sk_err;
1534 		sk->sk_state = TCP_CLOSE;
1535 		sock->state = SS_UNCONNECTED;
1536 	} else {
1537 		err = 0;
1538 	}
1539 
1540 out_wait:
1541 	finish_wait(sk_sleep(sk), &wait);
1542 out:
1543 	release_sock(sk);
1544 	return err;
1545 }
1546 
vsock_accept(struct socket * sock,struct socket * newsock,int flags,bool kern)1547 static int vsock_accept(struct socket *sock, struct socket *newsock, int flags,
1548 			bool kern)
1549 {
1550 	struct sock *listener;
1551 	int err;
1552 	struct sock *connected;
1553 	struct vsock_sock *vconnected;
1554 	long timeout;
1555 	DEFINE_WAIT(wait);
1556 
1557 	err = 0;
1558 	listener = sock->sk;
1559 
1560 	lock_sock(listener);
1561 
1562 	if (!sock_type_connectible(sock->type)) {
1563 		err = -EOPNOTSUPP;
1564 		goto out;
1565 	}
1566 
1567 	if (listener->sk_state != TCP_LISTEN) {
1568 		err = -EINVAL;
1569 		goto out;
1570 	}
1571 
1572 	/* Wait for children sockets to appear; these are the new sockets
1573 	 * created upon connection establishment.
1574 	 */
1575 	timeout = sock_rcvtimeo(listener, flags & O_NONBLOCK);
1576 	prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1577 
1578 	while ((connected = vsock_dequeue_accept(listener)) == NULL &&
1579 	       listener->sk_err == 0) {
1580 		release_sock(listener);
1581 		timeout = schedule_timeout(timeout);
1582 		finish_wait(sk_sleep(listener), &wait);
1583 		lock_sock(listener);
1584 
1585 		if (signal_pending(current)) {
1586 			err = sock_intr_errno(timeout);
1587 			goto out;
1588 		} else if (timeout == 0) {
1589 			err = -EAGAIN;
1590 			goto out;
1591 		}
1592 
1593 		prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1594 	}
1595 	finish_wait(sk_sleep(listener), &wait);
1596 
1597 	if (listener->sk_err)
1598 		err = -listener->sk_err;
1599 
1600 	if (connected) {
1601 		sk_acceptq_removed(listener);
1602 
1603 		lock_sock_nested(connected, SINGLE_DEPTH_NESTING);
1604 		vconnected = vsock_sk(connected);
1605 
1606 		/* If the listener socket has received an error, then we should
1607 		 * reject this socket and return.  Note that we simply mark the
1608 		 * socket rejected, drop our reference, and let the cleanup
1609 		 * function handle the cleanup; the fact that we found it in
1610 		 * the listener's accept queue guarantees that the cleanup
1611 		 * function hasn't run yet.
1612 		 */
1613 		if (err) {
1614 			vconnected->rejected = true;
1615 		} else {
1616 			newsock->state = SS_CONNECTED;
1617 			sock_graft(connected, newsock);
1618 		}
1619 
1620 		release_sock(connected);
1621 		sock_put(connected);
1622 	}
1623 
1624 out:
1625 	release_sock(listener);
1626 	return err;
1627 }
1628 
vsock_listen(struct socket * sock,int backlog)1629 static int vsock_listen(struct socket *sock, int backlog)
1630 {
1631 	int err;
1632 	struct sock *sk;
1633 	struct vsock_sock *vsk;
1634 
1635 	sk = sock->sk;
1636 
1637 	lock_sock(sk);
1638 
1639 	if (!sock_type_connectible(sk->sk_type)) {
1640 		err = -EOPNOTSUPP;
1641 		goto out;
1642 	}
1643 
1644 	if (sock->state != SS_UNCONNECTED) {
1645 		err = -EINVAL;
1646 		goto out;
1647 	}
1648 
1649 	vsk = vsock_sk(sk);
1650 
1651 	if (!vsock_addr_bound(&vsk->local_addr)) {
1652 		err = -EINVAL;
1653 		goto out;
1654 	}
1655 
1656 	sk->sk_max_ack_backlog = backlog;
1657 	sk->sk_state = TCP_LISTEN;
1658 
1659 	err = 0;
1660 
1661 out:
1662 	release_sock(sk);
1663 	return err;
1664 }
1665 
vsock_update_buffer_size(struct vsock_sock * vsk,const struct vsock_transport * transport,u64 val)1666 static void vsock_update_buffer_size(struct vsock_sock *vsk,
1667 				     const struct vsock_transport *transport,
1668 				     u64 val)
1669 {
1670 	if (val > vsk->buffer_max_size)
1671 		val = vsk->buffer_max_size;
1672 
1673 	if (val < vsk->buffer_min_size)
1674 		val = vsk->buffer_min_size;
1675 
1676 	if (val != vsk->buffer_size &&
1677 	    transport && transport->notify_buffer_size)
1678 		transport->notify_buffer_size(vsk, &val);
1679 
1680 	vsk->buffer_size = val;
1681 }
1682 
vsock_connectible_setsockopt(struct socket * sock,int level,int optname,sockptr_t optval,unsigned int optlen)1683 static int vsock_connectible_setsockopt(struct socket *sock,
1684 					int level,
1685 					int optname,
1686 					sockptr_t optval,
1687 					unsigned int optlen)
1688 {
1689 	int err;
1690 	struct sock *sk;
1691 	struct vsock_sock *vsk;
1692 	const struct vsock_transport *transport;
1693 	u64 val;
1694 
1695 	if (level != AF_VSOCK)
1696 		return -ENOPROTOOPT;
1697 
1698 #define COPY_IN(_v)                                       \
1699 	do {						  \
1700 		if (optlen < sizeof(_v)) {		  \
1701 			err = -EINVAL;			  \
1702 			goto exit;			  \
1703 		}					  \
1704 		if (copy_from_sockptr(&_v, optval, sizeof(_v)) != 0) {	\
1705 			err = -EFAULT;					\
1706 			goto exit;					\
1707 		}							\
1708 	} while (0)
1709 
1710 	err = 0;
1711 	sk = sock->sk;
1712 	vsk = vsock_sk(sk);
1713 
1714 	lock_sock(sk);
1715 
1716 	transport = vsk->transport;
1717 
1718 	switch (optname) {
1719 	case SO_VM_SOCKETS_BUFFER_SIZE:
1720 		COPY_IN(val);
1721 		vsock_update_buffer_size(vsk, transport, val);
1722 		break;
1723 
1724 	case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1725 		COPY_IN(val);
1726 		vsk->buffer_max_size = val;
1727 		vsock_update_buffer_size(vsk, transport, vsk->buffer_size);
1728 		break;
1729 
1730 	case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1731 		COPY_IN(val);
1732 		vsk->buffer_min_size = val;
1733 		vsock_update_buffer_size(vsk, transport, vsk->buffer_size);
1734 		break;
1735 
1736 	case SO_VM_SOCKETS_CONNECT_TIMEOUT_NEW:
1737 	case SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD: {
1738 		struct __kernel_sock_timeval tv;
1739 
1740 		err = sock_copy_user_timeval(&tv, optval, optlen,
1741 					     optname == SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD);
1742 		if (err)
1743 			break;
1744 		if (tv.tv_sec >= 0 && tv.tv_usec < USEC_PER_SEC &&
1745 		    tv.tv_sec < (MAX_SCHEDULE_TIMEOUT / HZ - 1)) {
1746 			vsk->connect_timeout = tv.tv_sec * HZ +
1747 				DIV_ROUND_UP((unsigned long)tv.tv_usec, (USEC_PER_SEC / HZ));
1748 			if (vsk->connect_timeout == 0)
1749 				vsk->connect_timeout =
1750 				    VSOCK_DEFAULT_CONNECT_TIMEOUT;
1751 
1752 		} else {
1753 			err = -ERANGE;
1754 		}
1755 		break;
1756 	}
1757 
1758 	default:
1759 		err = -ENOPROTOOPT;
1760 		break;
1761 	}
1762 
1763 #undef COPY_IN
1764 
1765 exit:
1766 	release_sock(sk);
1767 	return err;
1768 }
1769 
vsock_connectible_getsockopt(struct socket * sock,int level,int optname,char __user * optval,int __user * optlen)1770 static int vsock_connectible_getsockopt(struct socket *sock,
1771 					int level, int optname,
1772 					char __user *optval,
1773 					int __user *optlen)
1774 {
1775 	struct sock *sk = sock->sk;
1776 	struct vsock_sock *vsk = vsock_sk(sk);
1777 
1778 	union {
1779 		u64 val64;
1780 		struct old_timeval32 tm32;
1781 		struct __kernel_old_timeval tm;
1782 		struct  __kernel_sock_timeval stm;
1783 	} v;
1784 
1785 	int lv = sizeof(v.val64);
1786 	int len;
1787 
1788 	if (level != AF_VSOCK)
1789 		return -ENOPROTOOPT;
1790 
1791 	if (get_user(len, optlen))
1792 		return -EFAULT;
1793 
1794 	memset(&v, 0, sizeof(v));
1795 
1796 	switch (optname) {
1797 	case SO_VM_SOCKETS_BUFFER_SIZE:
1798 		v.val64 = vsk->buffer_size;
1799 		break;
1800 
1801 	case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1802 		v.val64 = vsk->buffer_max_size;
1803 		break;
1804 
1805 	case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1806 		v.val64 = vsk->buffer_min_size;
1807 		break;
1808 
1809 	case SO_VM_SOCKETS_CONNECT_TIMEOUT_NEW:
1810 	case SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD:
1811 		lv = sock_get_timeout(vsk->connect_timeout, &v,
1812 				      optname == SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD);
1813 		break;
1814 
1815 	default:
1816 		return -ENOPROTOOPT;
1817 	}
1818 
1819 	if (len < lv)
1820 		return -EINVAL;
1821 	if (len > lv)
1822 		len = lv;
1823 	if (copy_to_user(optval, &v, len))
1824 		return -EFAULT;
1825 
1826 	if (put_user(len, optlen))
1827 		return -EFAULT;
1828 
1829 	return 0;
1830 }
1831 
vsock_connectible_sendmsg(struct socket * sock,struct msghdr * msg,size_t len)1832 static int vsock_connectible_sendmsg(struct socket *sock, struct msghdr *msg,
1833 				     size_t len)
1834 {
1835 	struct sock *sk;
1836 	struct vsock_sock *vsk;
1837 	const struct vsock_transport *transport;
1838 	ssize_t total_written;
1839 	long timeout;
1840 	int err;
1841 	struct vsock_transport_send_notify_data send_data;
1842 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
1843 
1844 	sk = sock->sk;
1845 	vsk = vsock_sk(sk);
1846 	total_written = 0;
1847 	err = 0;
1848 
1849 	if (msg->msg_flags & MSG_OOB)
1850 		return -EOPNOTSUPP;
1851 
1852 	lock_sock(sk);
1853 
1854 	transport = vsk->transport;
1855 
1856 	/* Callers should not provide a destination with connection oriented
1857 	 * sockets.
1858 	 */
1859 	if (msg->msg_namelen) {
1860 		err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP;
1861 		goto out;
1862 	}
1863 
1864 	/* Send data only if both sides are not shutdown in the direction. */
1865 	if (sk->sk_shutdown & SEND_SHUTDOWN ||
1866 	    vsk->peer_shutdown & RCV_SHUTDOWN) {
1867 		err = -EPIPE;
1868 		goto out;
1869 	}
1870 
1871 	if (!transport || sk->sk_state != TCP_ESTABLISHED ||
1872 	    !vsock_addr_bound(&vsk->local_addr)) {
1873 		err = -ENOTCONN;
1874 		goto out;
1875 	}
1876 
1877 	if (!vsock_addr_bound(&vsk->remote_addr)) {
1878 		err = -EDESTADDRREQ;
1879 		goto out;
1880 	}
1881 
1882 	/* Wait for room in the produce queue to enqueue our user's data. */
1883 	timeout = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1884 
1885 	err = transport->notify_send_init(vsk, &send_data);
1886 	if (err < 0)
1887 		goto out;
1888 
1889 	while (total_written < len) {
1890 		ssize_t written;
1891 
1892 		add_wait_queue(sk_sleep(sk), &wait);
1893 		while (vsock_stream_has_space(vsk) == 0 &&
1894 		       sk->sk_err == 0 &&
1895 		       !(sk->sk_shutdown & SEND_SHUTDOWN) &&
1896 		       !(vsk->peer_shutdown & RCV_SHUTDOWN)) {
1897 
1898 			/* Don't wait for non-blocking sockets. */
1899 			if (timeout == 0) {
1900 				err = -EAGAIN;
1901 				remove_wait_queue(sk_sleep(sk), &wait);
1902 				goto out_err;
1903 			}
1904 
1905 			err = transport->notify_send_pre_block(vsk, &send_data);
1906 			if (err < 0) {
1907 				remove_wait_queue(sk_sleep(sk), &wait);
1908 				goto out_err;
1909 			}
1910 
1911 			release_sock(sk);
1912 			timeout = wait_woken(&wait, TASK_INTERRUPTIBLE, timeout);
1913 			lock_sock(sk);
1914 			if (signal_pending(current)) {
1915 				err = sock_intr_errno(timeout);
1916 				remove_wait_queue(sk_sleep(sk), &wait);
1917 				goto out_err;
1918 			} else if (timeout == 0) {
1919 				err = -EAGAIN;
1920 				remove_wait_queue(sk_sleep(sk), &wait);
1921 				goto out_err;
1922 			}
1923 		}
1924 		remove_wait_queue(sk_sleep(sk), &wait);
1925 
1926 		/* These checks occur both as part of and after the loop
1927 		 * conditional since we need to check before and after
1928 		 * sleeping.
1929 		 */
1930 		if (sk->sk_err) {
1931 			err = -sk->sk_err;
1932 			goto out_err;
1933 		} else if ((sk->sk_shutdown & SEND_SHUTDOWN) ||
1934 			   (vsk->peer_shutdown & RCV_SHUTDOWN)) {
1935 			err = -EPIPE;
1936 			goto out_err;
1937 		}
1938 
1939 		err = transport->notify_send_pre_enqueue(vsk, &send_data);
1940 		if (err < 0)
1941 			goto out_err;
1942 
1943 		/* Note that enqueue will only write as many bytes as are free
1944 		 * in the produce queue, so we don't need to ensure len is
1945 		 * smaller than the queue size.  It is the caller's
1946 		 * responsibility to check how many bytes we were able to send.
1947 		 */
1948 
1949 		if (sk->sk_type == SOCK_SEQPACKET) {
1950 			written = transport->seqpacket_enqueue(vsk,
1951 						msg, len - total_written);
1952 		} else {
1953 			written = transport->stream_enqueue(vsk,
1954 					msg, len - total_written);
1955 		}
1956 
1957 		if (written < 0) {
1958 			err = written;
1959 			goto out_err;
1960 		}
1961 
1962 		total_written += written;
1963 
1964 		err = transport->notify_send_post_enqueue(
1965 				vsk, written, &send_data);
1966 		if (err < 0)
1967 			goto out_err;
1968 
1969 	}
1970 
1971 out_err:
1972 	if (total_written > 0) {
1973 		/* Return number of written bytes only if:
1974 		 * 1) SOCK_STREAM socket.
1975 		 * 2) SOCK_SEQPACKET socket when whole buffer is sent.
1976 		 */
1977 		if (sk->sk_type == SOCK_STREAM || total_written == len)
1978 			err = total_written;
1979 	}
1980 out:
1981 	release_sock(sk);
1982 	return err;
1983 }
1984 
vsock_connectible_wait_data(struct sock * sk,struct wait_queue_entry * wait,long timeout,struct vsock_transport_recv_notify_data * recv_data,size_t target)1985 static int vsock_connectible_wait_data(struct sock *sk,
1986 				       struct wait_queue_entry *wait,
1987 				       long timeout,
1988 				       struct vsock_transport_recv_notify_data *recv_data,
1989 				       size_t target)
1990 {
1991 	const struct vsock_transport *transport;
1992 	struct vsock_sock *vsk;
1993 	s64 data;
1994 	int err;
1995 
1996 	vsk = vsock_sk(sk);
1997 	err = 0;
1998 	transport = vsk->transport;
1999 
2000 	while (1) {
2001 		prepare_to_wait(sk_sleep(sk), wait, TASK_INTERRUPTIBLE);
2002 		data = vsock_connectible_has_data(vsk);
2003 		if (data != 0)
2004 			break;
2005 
2006 		if (sk->sk_err != 0 ||
2007 		    (sk->sk_shutdown & RCV_SHUTDOWN) ||
2008 		    (vsk->peer_shutdown & SEND_SHUTDOWN)) {
2009 			break;
2010 		}
2011 
2012 		/* Don't wait for non-blocking sockets. */
2013 		if (timeout == 0) {
2014 			err = -EAGAIN;
2015 			break;
2016 		}
2017 
2018 		if (recv_data) {
2019 			err = transport->notify_recv_pre_block(vsk, target, recv_data);
2020 			if (err < 0)
2021 				break;
2022 		}
2023 
2024 		release_sock(sk);
2025 		timeout = schedule_timeout(timeout);
2026 		lock_sock(sk);
2027 
2028 		if (signal_pending(current)) {
2029 			err = sock_intr_errno(timeout);
2030 			break;
2031 		} else if (timeout == 0) {
2032 			err = -EAGAIN;
2033 			break;
2034 		}
2035 	}
2036 
2037 	finish_wait(sk_sleep(sk), wait);
2038 
2039 	if (err)
2040 		return err;
2041 
2042 	/* Internal transport error when checking for available
2043 	 * data. XXX This should be changed to a connection
2044 	 * reset in a later change.
2045 	 */
2046 	if (data < 0)
2047 		return -ENOMEM;
2048 
2049 	return data;
2050 }
2051 
__vsock_stream_recvmsg(struct sock * sk,struct msghdr * msg,size_t len,int flags)2052 static int __vsock_stream_recvmsg(struct sock *sk, struct msghdr *msg,
2053 				  size_t len, int flags)
2054 {
2055 	struct vsock_transport_recv_notify_data recv_data;
2056 	const struct vsock_transport *transport;
2057 	struct vsock_sock *vsk;
2058 	ssize_t copied;
2059 	size_t target;
2060 	long timeout;
2061 	int err;
2062 
2063 	DEFINE_WAIT(wait);
2064 
2065 	vsk = vsock_sk(sk);
2066 	transport = vsk->transport;
2067 
2068 	/* We must not copy less than target bytes into the user's buffer
2069 	 * before returning successfully, so we wait for the consume queue to
2070 	 * have that much data to consume before dequeueing.  Note that this
2071 	 * makes it impossible to handle cases where target is greater than the
2072 	 * queue size.
2073 	 */
2074 	target = sock_rcvlowat(sk, flags & MSG_WAITALL, len);
2075 	if (target >= transport->stream_rcvhiwat(vsk)) {
2076 		err = -ENOMEM;
2077 		goto out;
2078 	}
2079 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2080 	copied = 0;
2081 
2082 	err = transport->notify_recv_init(vsk, target, &recv_data);
2083 	if (err < 0)
2084 		goto out;
2085 
2086 
2087 	while (1) {
2088 		ssize_t read;
2089 
2090 		err = vsock_connectible_wait_data(sk, &wait, timeout,
2091 						  &recv_data, target);
2092 		if (err <= 0)
2093 			break;
2094 
2095 		err = transport->notify_recv_pre_dequeue(vsk, target,
2096 							 &recv_data);
2097 		if (err < 0)
2098 			break;
2099 
2100 		read = transport->stream_dequeue(vsk, msg, len - copied, flags);
2101 		if (read < 0) {
2102 			err = read;
2103 			break;
2104 		}
2105 
2106 		copied += read;
2107 
2108 		err = transport->notify_recv_post_dequeue(vsk, target, read,
2109 						!(flags & MSG_PEEK), &recv_data);
2110 		if (err < 0)
2111 			goto out;
2112 
2113 		if (read >= target || flags & MSG_PEEK)
2114 			break;
2115 
2116 		target -= read;
2117 	}
2118 
2119 	if (sk->sk_err)
2120 		err = -sk->sk_err;
2121 	else if (sk->sk_shutdown & RCV_SHUTDOWN)
2122 		err = 0;
2123 
2124 	if (copied > 0)
2125 		err = copied;
2126 
2127 out:
2128 	return err;
2129 }
2130 
__vsock_seqpacket_recvmsg(struct sock * sk,struct msghdr * msg,size_t len,int flags)2131 static int __vsock_seqpacket_recvmsg(struct sock *sk, struct msghdr *msg,
2132 				     size_t len, int flags)
2133 {
2134 	const struct vsock_transport *transport;
2135 	struct vsock_sock *vsk;
2136 	ssize_t msg_len;
2137 	long timeout;
2138 	int err = 0;
2139 	DEFINE_WAIT(wait);
2140 
2141 	vsk = vsock_sk(sk);
2142 	transport = vsk->transport;
2143 
2144 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2145 
2146 	err = vsock_connectible_wait_data(sk, &wait, timeout, NULL, 0);
2147 	if (err <= 0)
2148 		goto out;
2149 
2150 	msg_len = transport->seqpacket_dequeue(vsk, msg, flags);
2151 
2152 	if (msg_len < 0) {
2153 		err = msg_len;
2154 		goto out;
2155 	}
2156 
2157 	if (sk->sk_err) {
2158 		err = -sk->sk_err;
2159 	} else if (sk->sk_shutdown & RCV_SHUTDOWN) {
2160 		err = 0;
2161 	} else {
2162 		/* User sets MSG_TRUNC, so return real length of
2163 		 * packet.
2164 		 */
2165 		if (flags & MSG_TRUNC)
2166 			err = msg_len;
2167 		else
2168 			err = len - msg_data_left(msg);
2169 
2170 		/* Always set MSG_TRUNC if real length of packet is
2171 		 * bigger than user's buffer.
2172 		 */
2173 		if (msg_len > len)
2174 			msg->msg_flags |= MSG_TRUNC;
2175 	}
2176 
2177 out:
2178 	return err;
2179 }
2180 
2181 int
__vsock_connectible_recvmsg(struct socket * sock,struct msghdr * msg,size_t len,int flags)2182 __vsock_connectible_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
2183 			    int flags)
2184 {
2185 	struct sock *sk;
2186 	struct vsock_sock *vsk;
2187 	const struct vsock_transport *transport;
2188 	int err;
2189 
2190 	sk = sock->sk;
2191 
2192 	if (unlikely(flags & MSG_ERRQUEUE))
2193 		return sock_recv_errqueue(sk, msg, len, SOL_VSOCK, VSOCK_RECVERR);
2194 
2195 	vsk = vsock_sk(sk);
2196 	err = 0;
2197 
2198 	lock_sock(sk);
2199 
2200 	transport = vsk->transport;
2201 
2202 	if (!transport || sk->sk_state != TCP_ESTABLISHED) {
2203 		/* Recvmsg is supposed to return 0 if a peer performs an
2204 		 * orderly shutdown. Differentiate between that case and when a
2205 		 * peer has not connected or a local shutdown occurred with the
2206 		 * SOCK_DONE flag.
2207 		 */
2208 		if (sock_flag(sk, SOCK_DONE))
2209 			err = 0;
2210 		else
2211 			err = -ENOTCONN;
2212 
2213 		goto out;
2214 	}
2215 
2216 	if (flags & MSG_OOB) {
2217 		err = -EOPNOTSUPP;
2218 		goto out;
2219 	}
2220 
2221 	/* We don't check peer_shutdown flag here since peer may actually shut
2222 	 * down, but there can be data in the queue that a local socket can
2223 	 * receive.
2224 	 */
2225 	if (sk->sk_shutdown & RCV_SHUTDOWN) {
2226 		err = 0;
2227 		goto out;
2228 	}
2229 
2230 	/* It is valid on Linux to pass in a zero-length receive buffer.  This
2231 	 * is not an error.  We may as well bail out now.
2232 	 */
2233 	if (!len) {
2234 		err = 0;
2235 		goto out;
2236 	}
2237 
2238 	if (sk->sk_type == SOCK_STREAM)
2239 		err = __vsock_stream_recvmsg(sk, msg, len, flags);
2240 	else
2241 		err = __vsock_seqpacket_recvmsg(sk, msg, len, flags);
2242 
2243 out:
2244 	release_sock(sk);
2245 	return err;
2246 }
2247 
2248 int
vsock_connectible_recvmsg(struct socket * sock,struct msghdr * msg,size_t len,int flags)2249 vsock_connectible_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
2250 			  int flags)
2251 {
2252 #ifdef CONFIG_BPF_SYSCALL
2253 	struct sock *sk = sock->sk;
2254 	const struct proto *prot;
2255 
2256 	prot = READ_ONCE(sk->sk_prot);
2257 	if (prot != &vsock_proto)
2258 		return prot->recvmsg(sk, msg, len, flags, NULL);
2259 #endif
2260 
2261 	return __vsock_connectible_recvmsg(sock, msg, len, flags);
2262 }
2263 EXPORT_SYMBOL_GPL(vsock_connectible_recvmsg);
2264 
vsock_set_rcvlowat(struct sock * sk,int val)2265 static int vsock_set_rcvlowat(struct sock *sk, int val)
2266 {
2267 	const struct vsock_transport *transport;
2268 	struct vsock_sock *vsk;
2269 
2270 	vsk = vsock_sk(sk);
2271 
2272 	if (val > vsk->buffer_size)
2273 		return -EINVAL;
2274 
2275 	transport = vsk->transport;
2276 
2277 	if (transport && transport->notify_set_rcvlowat) {
2278 		int err;
2279 
2280 		err = transport->notify_set_rcvlowat(vsk, val);
2281 		if (err)
2282 			return err;
2283 	}
2284 
2285 	WRITE_ONCE(sk->sk_rcvlowat, val ? : 1);
2286 	return 0;
2287 }
2288 
2289 static const struct proto_ops vsock_stream_ops = {
2290 	.family = PF_VSOCK,
2291 	.owner = THIS_MODULE,
2292 	.release = vsock_release,
2293 	.bind = vsock_bind,
2294 	.connect = vsock_connect,
2295 	.socketpair = sock_no_socketpair,
2296 	.accept = vsock_accept,
2297 	.getname = vsock_getname,
2298 	.poll = vsock_poll,
2299 	.ioctl = sock_no_ioctl,
2300 	.listen = vsock_listen,
2301 	.shutdown = vsock_shutdown,
2302 	.setsockopt = vsock_connectible_setsockopt,
2303 	.getsockopt = vsock_connectible_getsockopt,
2304 	.sendmsg = vsock_connectible_sendmsg,
2305 	.recvmsg = vsock_connectible_recvmsg,
2306 	.mmap = sock_no_mmap,
2307 	.set_rcvlowat = vsock_set_rcvlowat,
2308 	.read_skb = vsock_read_skb,
2309 };
2310 
2311 static const struct proto_ops vsock_seqpacket_ops = {
2312 	.family = PF_VSOCK,
2313 	.owner = THIS_MODULE,
2314 	.release = vsock_release,
2315 	.bind = vsock_bind,
2316 	.connect = vsock_connect,
2317 	.socketpair = sock_no_socketpair,
2318 	.accept = vsock_accept,
2319 	.getname = vsock_getname,
2320 	.poll = vsock_poll,
2321 	.ioctl = sock_no_ioctl,
2322 	.listen = vsock_listen,
2323 	.shutdown = vsock_shutdown,
2324 	.setsockopt = vsock_connectible_setsockopt,
2325 	.getsockopt = vsock_connectible_getsockopt,
2326 	.sendmsg = vsock_connectible_sendmsg,
2327 	.recvmsg = vsock_connectible_recvmsg,
2328 	.mmap = sock_no_mmap,
2329 	.read_skb = vsock_read_skb,
2330 };
2331 
vsock_create(struct net * net,struct socket * sock,int protocol,int kern)2332 static int vsock_create(struct net *net, struct socket *sock,
2333 			int protocol, int kern)
2334 {
2335 	struct vsock_sock *vsk;
2336 	struct sock *sk;
2337 	int ret;
2338 
2339 	if (!sock)
2340 		return -EINVAL;
2341 
2342 	if (protocol && protocol != PF_VSOCK)
2343 		return -EPROTONOSUPPORT;
2344 
2345 	switch (sock->type) {
2346 	case SOCK_DGRAM:
2347 		sock->ops = &vsock_dgram_ops;
2348 		break;
2349 	case SOCK_STREAM:
2350 		sock->ops = &vsock_stream_ops;
2351 		break;
2352 	case SOCK_SEQPACKET:
2353 		sock->ops = &vsock_seqpacket_ops;
2354 		break;
2355 	default:
2356 		return -ESOCKTNOSUPPORT;
2357 	}
2358 
2359 	sock->state = SS_UNCONNECTED;
2360 
2361 	sk = __vsock_create(net, sock, NULL, GFP_KERNEL, 0, kern);
2362 	if (!sk)
2363 		return -ENOMEM;
2364 
2365 	vsk = vsock_sk(sk);
2366 
2367 	if (sock->type == SOCK_DGRAM) {
2368 		ret = vsock_assign_transport(vsk, NULL);
2369 		if (ret < 0) {
2370 			sock_put(sk);
2371 			return ret;
2372 		}
2373 	}
2374 
2375 	vsock_insert_unbound(vsk);
2376 
2377 	return 0;
2378 }
2379 
2380 static const struct net_proto_family vsock_family_ops = {
2381 	.family = AF_VSOCK,
2382 	.create = vsock_create,
2383 	.owner = THIS_MODULE,
2384 };
2385 
vsock_dev_do_ioctl(struct file * filp,unsigned int cmd,void __user * ptr)2386 static long vsock_dev_do_ioctl(struct file *filp,
2387 			       unsigned int cmd, void __user *ptr)
2388 {
2389 	u32 __user *p = ptr;
2390 	u32 cid = VMADDR_CID_ANY;
2391 	int retval = 0;
2392 
2393 	switch (cmd) {
2394 	case IOCTL_VM_SOCKETS_GET_LOCAL_CID:
2395 		/* To be compatible with the VMCI behavior, we prioritize the
2396 		 * guest CID instead of well-know host CID (VMADDR_CID_HOST).
2397 		 */
2398 		if (transport_g2h)
2399 			cid = transport_g2h->get_local_cid();
2400 		else if (transport_h2g)
2401 			cid = transport_h2g->get_local_cid();
2402 
2403 		if (put_user(cid, p) != 0)
2404 			retval = -EFAULT;
2405 		break;
2406 
2407 	default:
2408 		retval = -ENOIOCTLCMD;
2409 	}
2410 
2411 	return retval;
2412 }
2413 
vsock_dev_ioctl(struct file * filp,unsigned int cmd,unsigned long arg)2414 static long vsock_dev_ioctl(struct file *filp,
2415 			    unsigned int cmd, unsigned long arg)
2416 {
2417 	return vsock_dev_do_ioctl(filp, cmd, (void __user *)arg);
2418 }
2419 
2420 #ifdef CONFIG_COMPAT
vsock_dev_compat_ioctl(struct file * filp,unsigned int cmd,unsigned long arg)2421 static long vsock_dev_compat_ioctl(struct file *filp,
2422 				   unsigned int cmd, unsigned long arg)
2423 {
2424 	return vsock_dev_do_ioctl(filp, cmd, compat_ptr(arg));
2425 }
2426 #endif
2427 
2428 static const struct file_operations vsock_device_ops = {
2429 	.owner		= THIS_MODULE,
2430 	.unlocked_ioctl	= vsock_dev_ioctl,
2431 #ifdef CONFIG_COMPAT
2432 	.compat_ioctl	= vsock_dev_compat_ioctl,
2433 #endif
2434 	.open		= nonseekable_open,
2435 };
2436 
2437 static struct miscdevice vsock_device = {
2438 	.name		= "vsock",
2439 	.fops		= &vsock_device_ops,
2440 };
2441 
vsock_init(void)2442 static int __init vsock_init(void)
2443 {
2444 	int err = 0;
2445 
2446 	vsock_init_tables();
2447 
2448 	vsock_proto.owner = THIS_MODULE;
2449 	vsock_device.minor = MISC_DYNAMIC_MINOR;
2450 	err = misc_register(&vsock_device);
2451 	if (err) {
2452 		pr_err("Failed to register misc device\n");
2453 		goto err_reset_transport;
2454 	}
2455 
2456 	err = proto_register(&vsock_proto, 1);	/* we want our slab */
2457 	if (err) {
2458 		pr_err("Cannot register vsock protocol\n");
2459 		goto err_deregister_misc;
2460 	}
2461 
2462 	err = sock_register(&vsock_family_ops);
2463 	if (err) {
2464 		pr_err("could not register af_vsock (%d) address family: %d\n",
2465 		       AF_VSOCK, err);
2466 		goto err_unregister_proto;
2467 	}
2468 
2469 	vsock_bpf_build_proto();
2470 
2471 	return 0;
2472 
2473 err_unregister_proto:
2474 	proto_unregister(&vsock_proto);
2475 err_deregister_misc:
2476 	misc_deregister(&vsock_device);
2477 err_reset_transport:
2478 	return err;
2479 }
2480 
vsock_exit(void)2481 static void __exit vsock_exit(void)
2482 {
2483 	misc_deregister(&vsock_device);
2484 	sock_unregister(AF_VSOCK);
2485 	proto_unregister(&vsock_proto);
2486 }
2487 
vsock_core_get_transport(struct vsock_sock * vsk)2488 const struct vsock_transport *vsock_core_get_transport(struct vsock_sock *vsk)
2489 {
2490 	return vsk->transport;
2491 }
2492 EXPORT_SYMBOL_GPL(vsock_core_get_transport);
2493 
vsock_core_register(const struct vsock_transport * t,int features)2494 int vsock_core_register(const struct vsock_transport *t, int features)
2495 {
2496 	const struct vsock_transport *t_h2g, *t_g2h, *t_dgram, *t_local;
2497 	int err = mutex_lock_interruptible(&vsock_register_mutex);
2498 
2499 	if (err)
2500 		return err;
2501 
2502 	t_h2g = transport_h2g;
2503 	t_g2h = transport_g2h;
2504 	t_dgram = transport_dgram;
2505 	t_local = transport_local;
2506 
2507 	if (features & VSOCK_TRANSPORT_F_H2G) {
2508 		if (t_h2g) {
2509 			err = -EBUSY;
2510 			goto err_busy;
2511 		}
2512 		t_h2g = t;
2513 	}
2514 
2515 	if (features & VSOCK_TRANSPORT_F_G2H) {
2516 		if (t_g2h) {
2517 			err = -EBUSY;
2518 			goto err_busy;
2519 		}
2520 		t_g2h = t;
2521 	}
2522 
2523 	if (features & VSOCK_TRANSPORT_F_DGRAM) {
2524 		if (t_dgram) {
2525 			err = -EBUSY;
2526 			goto err_busy;
2527 		}
2528 		t_dgram = t;
2529 	}
2530 
2531 	if (features & VSOCK_TRANSPORT_F_LOCAL) {
2532 		if (t_local) {
2533 			err = -EBUSY;
2534 			goto err_busy;
2535 		}
2536 		t_local = t;
2537 	}
2538 
2539 	transport_h2g = t_h2g;
2540 	transport_g2h = t_g2h;
2541 	transport_dgram = t_dgram;
2542 	transport_local = t_local;
2543 
2544 err_busy:
2545 	mutex_unlock(&vsock_register_mutex);
2546 	return err;
2547 }
2548 EXPORT_SYMBOL_GPL(vsock_core_register);
2549 
vsock_core_unregister(const struct vsock_transport * t)2550 void vsock_core_unregister(const struct vsock_transport *t)
2551 {
2552 	mutex_lock(&vsock_register_mutex);
2553 
2554 	if (transport_h2g == t)
2555 		transport_h2g = NULL;
2556 
2557 	if (transport_g2h == t)
2558 		transport_g2h = NULL;
2559 
2560 	if (transport_dgram == t)
2561 		transport_dgram = NULL;
2562 
2563 	if (transport_local == t)
2564 		transport_local = NULL;
2565 
2566 	mutex_unlock(&vsock_register_mutex);
2567 }
2568 EXPORT_SYMBOL_GPL(vsock_core_unregister);
2569 
2570 module_init(vsock_init);
2571 module_exit(vsock_exit);
2572 
2573 MODULE_AUTHOR("VMware, Inc.");
2574 MODULE_DESCRIPTION("VMware Virtual Socket Family");
2575 MODULE_VERSION("1.0.2.0-k");
2576 MODULE_LICENSE("GPL v2");
2577