xref: /openbmc/linux/net/vmw_vsock/af_vsock.c (revision 877013bc)
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/types.h>
89 #include <linux/bitops.h>
90 #include <linux/cred.h>
91 #include <linux/init.h>
92 #include <linux/io.h>
93 #include <linux/kernel.h>
94 #include <linux/sched/signal.h>
95 #include <linux/kmod.h>
96 #include <linux/list.h>
97 #include <linux/miscdevice.h>
98 #include <linux/module.h>
99 #include <linux/mutex.h>
100 #include <linux/net.h>
101 #include <linux/poll.h>
102 #include <linux/random.h>
103 #include <linux/skbuff.h>
104 #include <linux/smp.h>
105 #include <linux/socket.h>
106 #include <linux/stddef.h>
107 #include <linux/unistd.h>
108 #include <linux/wait.h>
109 #include <linux/workqueue.h>
110 #include <net/sock.h>
111 #include <net/af_vsock.h>
112 
113 static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr);
114 static void vsock_sk_destruct(struct sock *sk);
115 static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
116 
117 /* Protocol family. */
118 static struct proto vsock_proto = {
119 	.name = "AF_VSOCK",
120 	.owner = THIS_MODULE,
121 	.obj_size = sizeof(struct vsock_sock),
122 };
123 
124 /* The default peer timeout indicates how long we will wait for a peer response
125  * to a control message.
126  */
127 #define VSOCK_DEFAULT_CONNECT_TIMEOUT (2 * HZ)
128 
129 #define VSOCK_DEFAULT_BUFFER_SIZE     (1024 * 256)
130 #define VSOCK_DEFAULT_BUFFER_MAX_SIZE (1024 * 256)
131 #define VSOCK_DEFAULT_BUFFER_MIN_SIZE 128
132 
133 /* Transport used for host->guest communication */
134 static const struct vsock_transport *transport_h2g;
135 /* Transport used for guest->host communication */
136 static const struct vsock_transport *transport_g2h;
137 /* Transport used for DGRAM communication */
138 static const struct vsock_transport *transport_dgram;
139 /* Transport used for local communication */
140 static const struct vsock_transport *transport_local;
141 static DEFINE_MUTEX(vsock_register_mutex);
142 
143 /**** UTILS ****/
144 
145 /* Each bound VSocket is stored in the bind hash table and each connected
146  * VSocket is stored in the connected hash table.
147  *
148  * Unbound sockets are all put on the same list attached to the end of the hash
149  * table (vsock_unbound_sockets).  Bound sockets are added to the hash table in
150  * the bucket that their local address hashes to (vsock_bound_sockets(addr)
151  * represents the list that addr hashes to).
152  *
153  * Specifically, we initialize the vsock_bind_table array to a size of
154  * VSOCK_HASH_SIZE + 1 so that vsock_bind_table[0] through
155  * vsock_bind_table[VSOCK_HASH_SIZE - 1] are for bound sockets and
156  * vsock_bind_table[VSOCK_HASH_SIZE] is for unbound sockets.  The hash function
157  * mods with VSOCK_HASH_SIZE to ensure this.
158  */
159 #define MAX_PORT_RETRIES        24
160 
161 #define VSOCK_HASH(addr)        ((addr)->svm_port % VSOCK_HASH_SIZE)
162 #define vsock_bound_sockets(addr) (&vsock_bind_table[VSOCK_HASH(addr)])
163 #define vsock_unbound_sockets     (&vsock_bind_table[VSOCK_HASH_SIZE])
164 
165 /* XXX This can probably be implemented in a better way. */
166 #define VSOCK_CONN_HASH(src, dst)				\
167 	(((src)->svm_cid ^ (dst)->svm_port) % VSOCK_HASH_SIZE)
168 #define vsock_connected_sockets(src, dst)		\
169 	(&vsock_connected_table[VSOCK_CONN_HASH(src, dst)])
170 #define vsock_connected_sockets_vsk(vsk)				\
171 	vsock_connected_sockets(&(vsk)->remote_addr, &(vsk)->local_addr)
172 
173 struct list_head vsock_bind_table[VSOCK_HASH_SIZE + 1];
174 EXPORT_SYMBOL_GPL(vsock_bind_table);
175 struct list_head vsock_connected_table[VSOCK_HASH_SIZE];
176 EXPORT_SYMBOL_GPL(vsock_connected_table);
177 DEFINE_SPINLOCK(vsock_table_lock);
178 EXPORT_SYMBOL_GPL(vsock_table_lock);
179 
180 /* Autobind this socket to the local address if necessary. */
181 static int vsock_auto_bind(struct vsock_sock *vsk)
182 {
183 	struct sock *sk = sk_vsock(vsk);
184 	struct sockaddr_vm local_addr;
185 
186 	if (vsock_addr_bound(&vsk->local_addr))
187 		return 0;
188 	vsock_addr_init(&local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
189 	return __vsock_bind(sk, &local_addr);
190 }
191 
192 static void vsock_init_tables(void)
193 {
194 	int i;
195 
196 	for (i = 0; i < ARRAY_SIZE(vsock_bind_table); i++)
197 		INIT_LIST_HEAD(&vsock_bind_table[i]);
198 
199 	for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++)
200 		INIT_LIST_HEAD(&vsock_connected_table[i]);
201 }
202 
203 static void __vsock_insert_bound(struct list_head *list,
204 				 struct vsock_sock *vsk)
205 {
206 	sock_hold(&vsk->sk);
207 	list_add(&vsk->bound_table, list);
208 }
209 
210 static void __vsock_insert_connected(struct list_head *list,
211 				     struct vsock_sock *vsk)
212 {
213 	sock_hold(&vsk->sk);
214 	list_add(&vsk->connected_table, list);
215 }
216 
217 static void __vsock_remove_bound(struct vsock_sock *vsk)
218 {
219 	list_del_init(&vsk->bound_table);
220 	sock_put(&vsk->sk);
221 }
222 
223 static void __vsock_remove_connected(struct vsock_sock *vsk)
224 {
225 	list_del_init(&vsk->connected_table);
226 	sock_put(&vsk->sk);
227 }
228 
229 static struct sock *__vsock_find_bound_socket(struct sockaddr_vm *addr)
230 {
231 	struct vsock_sock *vsk;
232 
233 	list_for_each_entry(vsk, vsock_bound_sockets(addr), bound_table) {
234 		if (vsock_addr_equals_addr(addr, &vsk->local_addr))
235 			return sk_vsock(vsk);
236 
237 		if (addr->svm_port == vsk->local_addr.svm_port &&
238 		    (vsk->local_addr.svm_cid == VMADDR_CID_ANY ||
239 		     addr->svm_cid == VMADDR_CID_ANY))
240 			return sk_vsock(vsk);
241 	}
242 
243 	return NULL;
244 }
245 
246 static struct sock *__vsock_find_connected_socket(struct sockaddr_vm *src,
247 						  struct sockaddr_vm *dst)
248 {
249 	struct vsock_sock *vsk;
250 
251 	list_for_each_entry(vsk, vsock_connected_sockets(src, dst),
252 			    connected_table) {
253 		if (vsock_addr_equals_addr(src, &vsk->remote_addr) &&
254 		    dst->svm_port == vsk->local_addr.svm_port) {
255 			return sk_vsock(vsk);
256 		}
257 	}
258 
259 	return NULL;
260 }
261 
262 static void vsock_insert_unbound(struct vsock_sock *vsk)
263 {
264 	spin_lock_bh(&vsock_table_lock);
265 	__vsock_insert_bound(vsock_unbound_sockets, vsk);
266 	spin_unlock_bh(&vsock_table_lock);
267 }
268 
269 void vsock_insert_connected(struct vsock_sock *vsk)
270 {
271 	struct list_head *list = vsock_connected_sockets(
272 		&vsk->remote_addr, &vsk->local_addr);
273 
274 	spin_lock_bh(&vsock_table_lock);
275 	__vsock_insert_connected(list, vsk);
276 	spin_unlock_bh(&vsock_table_lock);
277 }
278 EXPORT_SYMBOL_GPL(vsock_insert_connected);
279 
280 void vsock_remove_bound(struct vsock_sock *vsk)
281 {
282 	spin_lock_bh(&vsock_table_lock);
283 	if (__vsock_in_bound_table(vsk))
284 		__vsock_remove_bound(vsk);
285 	spin_unlock_bh(&vsock_table_lock);
286 }
287 EXPORT_SYMBOL_GPL(vsock_remove_bound);
288 
289 void vsock_remove_connected(struct vsock_sock *vsk)
290 {
291 	spin_lock_bh(&vsock_table_lock);
292 	if (__vsock_in_connected_table(vsk))
293 		__vsock_remove_connected(vsk);
294 	spin_unlock_bh(&vsock_table_lock);
295 }
296 EXPORT_SYMBOL_GPL(vsock_remove_connected);
297 
298 struct sock *vsock_find_bound_socket(struct sockaddr_vm *addr)
299 {
300 	struct sock *sk;
301 
302 	spin_lock_bh(&vsock_table_lock);
303 	sk = __vsock_find_bound_socket(addr);
304 	if (sk)
305 		sock_hold(sk);
306 
307 	spin_unlock_bh(&vsock_table_lock);
308 
309 	return sk;
310 }
311 EXPORT_SYMBOL_GPL(vsock_find_bound_socket);
312 
313 struct sock *vsock_find_connected_socket(struct sockaddr_vm *src,
314 					 struct sockaddr_vm *dst)
315 {
316 	struct sock *sk;
317 
318 	spin_lock_bh(&vsock_table_lock);
319 	sk = __vsock_find_connected_socket(src, dst);
320 	if (sk)
321 		sock_hold(sk);
322 
323 	spin_unlock_bh(&vsock_table_lock);
324 
325 	return sk;
326 }
327 EXPORT_SYMBOL_GPL(vsock_find_connected_socket);
328 
329 void vsock_remove_sock(struct vsock_sock *vsk)
330 {
331 	vsock_remove_bound(vsk);
332 	vsock_remove_connected(vsk);
333 }
334 EXPORT_SYMBOL_GPL(vsock_remove_sock);
335 
336 void vsock_for_each_connected_socket(void (*fn)(struct sock *sk))
337 {
338 	int i;
339 
340 	spin_lock_bh(&vsock_table_lock);
341 
342 	for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) {
343 		struct vsock_sock *vsk;
344 		list_for_each_entry(vsk, &vsock_connected_table[i],
345 				    connected_table)
346 			fn(sk_vsock(vsk));
347 	}
348 
349 	spin_unlock_bh(&vsock_table_lock);
350 }
351 EXPORT_SYMBOL_GPL(vsock_for_each_connected_socket);
352 
353 void vsock_add_pending(struct sock *listener, struct sock *pending)
354 {
355 	struct vsock_sock *vlistener;
356 	struct vsock_sock *vpending;
357 
358 	vlistener = vsock_sk(listener);
359 	vpending = vsock_sk(pending);
360 
361 	sock_hold(pending);
362 	sock_hold(listener);
363 	list_add_tail(&vpending->pending_links, &vlistener->pending_links);
364 }
365 EXPORT_SYMBOL_GPL(vsock_add_pending);
366 
367 void vsock_remove_pending(struct sock *listener, struct sock *pending)
368 {
369 	struct vsock_sock *vpending = vsock_sk(pending);
370 
371 	list_del_init(&vpending->pending_links);
372 	sock_put(listener);
373 	sock_put(pending);
374 }
375 EXPORT_SYMBOL_GPL(vsock_remove_pending);
376 
377 void vsock_enqueue_accept(struct sock *listener, struct sock *connected)
378 {
379 	struct vsock_sock *vlistener;
380 	struct vsock_sock *vconnected;
381 
382 	vlistener = vsock_sk(listener);
383 	vconnected = vsock_sk(connected);
384 
385 	sock_hold(connected);
386 	sock_hold(listener);
387 	list_add_tail(&vconnected->accept_queue, &vlistener->accept_queue);
388 }
389 EXPORT_SYMBOL_GPL(vsock_enqueue_accept);
390 
391 static bool vsock_use_local_transport(unsigned int remote_cid)
392 {
393 	if (!transport_local)
394 		return false;
395 
396 	if (remote_cid == VMADDR_CID_LOCAL)
397 		return true;
398 
399 	if (transport_g2h) {
400 		return remote_cid == transport_g2h->get_local_cid();
401 	} else {
402 		return remote_cid == VMADDR_CID_HOST;
403 	}
404 }
405 
406 static void vsock_deassign_transport(struct vsock_sock *vsk)
407 {
408 	if (!vsk->transport)
409 		return;
410 
411 	vsk->transport->destruct(vsk);
412 	module_put(vsk->transport->module);
413 	vsk->transport = NULL;
414 }
415 
416 /* Assign a transport to a socket and call the .init transport callback.
417  *
418  * Note: for stream socket this must be called when vsk->remote_addr is set
419  * (e.g. during the connect() or when a connection request on a listener
420  * socket is received).
421  * The vsk->remote_addr is used to decide which transport to use:
422  *  - remote CID == VMADDR_CID_LOCAL or g2h->local_cid or VMADDR_CID_HOST if
423  *    g2h is not loaded, will use local transport;
424  *  - remote CID <= VMADDR_CID_HOST or h2g is not loaded or remote flags field
425  *    includes VMADDR_FLAG_TO_HOST flag value, will use guest->host transport;
426  *  - remote CID > VMADDR_CID_HOST will use host->guest transport;
427  */
428 int vsock_assign_transport(struct vsock_sock *vsk, struct vsock_sock *psk)
429 {
430 	const struct vsock_transport *new_transport;
431 	struct sock *sk = sk_vsock(vsk);
432 	unsigned int remote_cid = vsk->remote_addr.svm_cid;
433 	__u8 remote_flags;
434 	int ret;
435 
436 	/* If the packet is coming with the source and destination CIDs higher
437 	 * than VMADDR_CID_HOST, then a vsock channel where all the packets are
438 	 * forwarded to the host should be established. Then the host will
439 	 * need to forward the packets to the guest.
440 	 *
441 	 * The flag is set on the (listen) receive path (psk is not NULL). On
442 	 * the connect path the flag can be set by the user space application.
443 	 */
444 	if (psk && vsk->local_addr.svm_cid > VMADDR_CID_HOST &&
445 	    vsk->remote_addr.svm_cid > VMADDR_CID_HOST)
446 		vsk->remote_addr.svm_flags |= VMADDR_FLAG_TO_HOST;
447 
448 	remote_flags = vsk->remote_addr.svm_flags;
449 
450 	switch (sk->sk_type) {
451 	case SOCK_DGRAM:
452 		new_transport = transport_dgram;
453 		break;
454 	case SOCK_STREAM:
455 		if (vsock_use_local_transport(remote_cid))
456 			new_transport = transport_local;
457 		else if (remote_cid <= VMADDR_CID_HOST || !transport_h2g ||
458 			 (remote_flags & VMADDR_FLAG_TO_HOST))
459 			new_transport = transport_g2h;
460 		else
461 			new_transport = transport_h2g;
462 		break;
463 	default:
464 		return -ESOCKTNOSUPPORT;
465 	}
466 
467 	if (vsk->transport) {
468 		if (vsk->transport == new_transport)
469 			return 0;
470 
471 		/* transport->release() must be called with sock lock acquired.
472 		 * This path can only be taken during vsock_stream_connect(),
473 		 * where we have already held the sock lock.
474 		 * In the other cases, this function is called on a new socket
475 		 * which is not assigned to any transport.
476 		 */
477 		vsk->transport->release(vsk);
478 		vsock_deassign_transport(vsk);
479 	}
480 
481 	/* We increase the module refcnt to prevent the transport unloading
482 	 * while there are open sockets assigned to it.
483 	 */
484 	if (!new_transport || !try_module_get(new_transport->module))
485 		return -ENODEV;
486 
487 	ret = new_transport->init(vsk, psk);
488 	if (ret) {
489 		module_put(new_transport->module);
490 		return ret;
491 	}
492 
493 	vsk->transport = new_transport;
494 
495 	return 0;
496 }
497 EXPORT_SYMBOL_GPL(vsock_assign_transport);
498 
499 bool vsock_find_cid(unsigned int cid)
500 {
501 	if (transport_g2h && cid == transport_g2h->get_local_cid())
502 		return true;
503 
504 	if (transport_h2g && cid == VMADDR_CID_HOST)
505 		return true;
506 
507 	if (transport_local && cid == VMADDR_CID_LOCAL)
508 		return true;
509 
510 	return false;
511 }
512 EXPORT_SYMBOL_GPL(vsock_find_cid);
513 
514 static struct sock *vsock_dequeue_accept(struct sock *listener)
515 {
516 	struct vsock_sock *vlistener;
517 	struct vsock_sock *vconnected;
518 
519 	vlistener = vsock_sk(listener);
520 
521 	if (list_empty(&vlistener->accept_queue))
522 		return NULL;
523 
524 	vconnected = list_entry(vlistener->accept_queue.next,
525 				struct vsock_sock, accept_queue);
526 
527 	list_del_init(&vconnected->accept_queue);
528 	sock_put(listener);
529 	/* The caller will need a reference on the connected socket so we let
530 	 * it call sock_put().
531 	 */
532 
533 	return sk_vsock(vconnected);
534 }
535 
536 static bool vsock_is_accept_queue_empty(struct sock *sk)
537 {
538 	struct vsock_sock *vsk = vsock_sk(sk);
539 	return list_empty(&vsk->accept_queue);
540 }
541 
542 static bool vsock_is_pending(struct sock *sk)
543 {
544 	struct vsock_sock *vsk = vsock_sk(sk);
545 	return !list_empty(&vsk->pending_links);
546 }
547 
548 static int vsock_send_shutdown(struct sock *sk, int mode)
549 {
550 	struct vsock_sock *vsk = vsock_sk(sk);
551 
552 	if (!vsk->transport)
553 		return -ENODEV;
554 
555 	return vsk->transport->shutdown(vsk, mode);
556 }
557 
558 static void vsock_pending_work(struct work_struct *work)
559 {
560 	struct sock *sk;
561 	struct sock *listener;
562 	struct vsock_sock *vsk;
563 	bool cleanup;
564 
565 	vsk = container_of(work, struct vsock_sock, pending_work.work);
566 	sk = sk_vsock(vsk);
567 	listener = vsk->listener;
568 	cleanup = true;
569 
570 	lock_sock(listener);
571 	lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
572 
573 	if (vsock_is_pending(sk)) {
574 		vsock_remove_pending(listener, sk);
575 
576 		sk_acceptq_removed(listener);
577 	} else if (!vsk->rejected) {
578 		/* We are not on the pending list and accept() did not reject
579 		 * us, so we must have been accepted by our user process.  We
580 		 * just need to drop our references to the sockets and be on
581 		 * our way.
582 		 */
583 		cleanup = false;
584 		goto out;
585 	}
586 
587 	/* We need to remove ourself from the global connected sockets list so
588 	 * incoming packets can't find this socket, and to reduce the reference
589 	 * count.
590 	 */
591 	vsock_remove_connected(vsk);
592 
593 	sk->sk_state = TCP_CLOSE;
594 
595 out:
596 	release_sock(sk);
597 	release_sock(listener);
598 	if (cleanup)
599 		sock_put(sk);
600 
601 	sock_put(sk);
602 	sock_put(listener);
603 }
604 
605 /**** SOCKET OPERATIONS ****/
606 
607 static int __vsock_bind_stream(struct vsock_sock *vsk,
608 			       struct sockaddr_vm *addr)
609 {
610 	static u32 port;
611 	struct sockaddr_vm new_addr;
612 
613 	if (!port)
614 		port = LAST_RESERVED_PORT + 1 +
615 			prandom_u32_max(U32_MAX - LAST_RESERVED_PORT);
616 
617 	vsock_addr_init(&new_addr, addr->svm_cid, addr->svm_port);
618 
619 	if (addr->svm_port == VMADDR_PORT_ANY) {
620 		bool found = false;
621 		unsigned int i;
622 
623 		for (i = 0; i < MAX_PORT_RETRIES; i++) {
624 			if (port <= LAST_RESERVED_PORT)
625 				port = LAST_RESERVED_PORT + 1;
626 
627 			new_addr.svm_port = port++;
628 
629 			if (!__vsock_find_bound_socket(&new_addr)) {
630 				found = true;
631 				break;
632 			}
633 		}
634 
635 		if (!found)
636 			return -EADDRNOTAVAIL;
637 	} else {
638 		/* If port is in reserved range, ensure caller
639 		 * has necessary privileges.
640 		 */
641 		if (addr->svm_port <= LAST_RESERVED_PORT &&
642 		    !capable(CAP_NET_BIND_SERVICE)) {
643 			return -EACCES;
644 		}
645 
646 		if (__vsock_find_bound_socket(&new_addr))
647 			return -EADDRINUSE;
648 	}
649 
650 	vsock_addr_init(&vsk->local_addr, new_addr.svm_cid, new_addr.svm_port);
651 
652 	/* Remove stream sockets from the unbound list and add them to the hash
653 	 * table for easy lookup by its address.  The unbound list is simply an
654 	 * extra entry at the end of the hash table, a trick used by AF_UNIX.
655 	 */
656 	__vsock_remove_bound(vsk);
657 	__vsock_insert_bound(vsock_bound_sockets(&vsk->local_addr), vsk);
658 
659 	return 0;
660 }
661 
662 static int __vsock_bind_dgram(struct vsock_sock *vsk,
663 			      struct sockaddr_vm *addr)
664 {
665 	return vsk->transport->dgram_bind(vsk, addr);
666 }
667 
668 static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr)
669 {
670 	struct vsock_sock *vsk = vsock_sk(sk);
671 	int retval;
672 
673 	/* First ensure this socket isn't already bound. */
674 	if (vsock_addr_bound(&vsk->local_addr))
675 		return -EINVAL;
676 
677 	/* Now bind to the provided address or select appropriate values if
678 	 * none are provided (VMADDR_CID_ANY and VMADDR_PORT_ANY).  Note that
679 	 * like AF_INET prevents binding to a non-local IP address (in most
680 	 * cases), we only allow binding to a local CID.
681 	 */
682 	if (addr->svm_cid != VMADDR_CID_ANY && !vsock_find_cid(addr->svm_cid))
683 		return -EADDRNOTAVAIL;
684 
685 	switch (sk->sk_socket->type) {
686 	case SOCK_STREAM:
687 		spin_lock_bh(&vsock_table_lock);
688 		retval = __vsock_bind_stream(vsk, addr);
689 		spin_unlock_bh(&vsock_table_lock);
690 		break;
691 
692 	case SOCK_DGRAM:
693 		retval = __vsock_bind_dgram(vsk, addr);
694 		break;
695 
696 	default:
697 		retval = -EINVAL;
698 		break;
699 	}
700 
701 	return retval;
702 }
703 
704 static void vsock_connect_timeout(struct work_struct *work);
705 
706 static struct sock *__vsock_create(struct net *net,
707 				   struct socket *sock,
708 				   struct sock *parent,
709 				   gfp_t priority,
710 				   unsigned short type,
711 				   int kern)
712 {
713 	struct sock *sk;
714 	struct vsock_sock *psk;
715 	struct vsock_sock *vsk;
716 
717 	sk = sk_alloc(net, AF_VSOCK, priority, &vsock_proto, kern);
718 	if (!sk)
719 		return NULL;
720 
721 	sock_init_data(sock, sk);
722 
723 	/* sk->sk_type is normally set in sock_init_data, but only if sock is
724 	 * non-NULL. We make sure that our sockets always have a type by
725 	 * setting it here if needed.
726 	 */
727 	if (!sock)
728 		sk->sk_type = type;
729 
730 	vsk = vsock_sk(sk);
731 	vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
732 	vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
733 
734 	sk->sk_destruct = vsock_sk_destruct;
735 	sk->sk_backlog_rcv = vsock_queue_rcv_skb;
736 	sock_reset_flag(sk, SOCK_DONE);
737 
738 	INIT_LIST_HEAD(&vsk->bound_table);
739 	INIT_LIST_HEAD(&vsk->connected_table);
740 	vsk->listener = NULL;
741 	INIT_LIST_HEAD(&vsk->pending_links);
742 	INIT_LIST_HEAD(&vsk->accept_queue);
743 	vsk->rejected = false;
744 	vsk->sent_request = false;
745 	vsk->ignore_connecting_rst = false;
746 	vsk->peer_shutdown = 0;
747 	INIT_DELAYED_WORK(&vsk->connect_work, vsock_connect_timeout);
748 	INIT_DELAYED_WORK(&vsk->pending_work, vsock_pending_work);
749 
750 	psk = parent ? vsock_sk(parent) : NULL;
751 	if (parent) {
752 		vsk->trusted = psk->trusted;
753 		vsk->owner = get_cred(psk->owner);
754 		vsk->connect_timeout = psk->connect_timeout;
755 		vsk->buffer_size = psk->buffer_size;
756 		vsk->buffer_min_size = psk->buffer_min_size;
757 		vsk->buffer_max_size = psk->buffer_max_size;
758 		security_sk_clone(parent, sk);
759 	} else {
760 		vsk->trusted = ns_capable_noaudit(&init_user_ns, CAP_NET_ADMIN);
761 		vsk->owner = get_current_cred();
762 		vsk->connect_timeout = VSOCK_DEFAULT_CONNECT_TIMEOUT;
763 		vsk->buffer_size = VSOCK_DEFAULT_BUFFER_SIZE;
764 		vsk->buffer_min_size = VSOCK_DEFAULT_BUFFER_MIN_SIZE;
765 		vsk->buffer_max_size = VSOCK_DEFAULT_BUFFER_MAX_SIZE;
766 	}
767 
768 	return sk;
769 }
770 
771 static void __vsock_release(struct sock *sk, int level)
772 {
773 	if (sk) {
774 		struct sock *pending;
775 		struct vsock_sock *vsk;
776 
777 		vsk = vsock_sk(sk);
778 		pending = NULL;	/* Compiler warning. */
779 
780 		/* When "level" is SINGLE_DEPTH_NESTING, use the nested
781 		 * version to avoid the warning "possible recursive locking
782 		 * detected". When "level" is 0, lock_sock_nested(sk, level)
783 		 * is the same as lock_sock(sk).
784 		 */
785 		lock_sock_nested(sk, level);
786 
787 		if (vsk->transport)
788 			vsk->transport->release(vsk);
789 		else if (sk->sk_type == SOCK_STREAM)
790 			vsock_remove_sock(vsk);
791 
792 		sock_orphan(sk);
793 		sk->sk_shutdown = SHUTDOWN_MASK;
794 
795 		skb_queue_purge(&sk->sk_receive_queue);
796 
797 		/* Clean up any sockets that never were accepted. */
798 		while ((pending = vsock_dequeue_accept(sk)) != NULL) {
799 			__vsock_release(pending, SINGLE_DEPTH_NESTING);
800 			sock_put(pending);
801 		}
802 
803 		release_sock(sk);
804 		sock_put(sk);
805 	}
806 }
807 
808 static void vsock_sk_destruct(struct sock *sk)
809 {
810 	struct vsock_sock *vsk = vsock_sk(sk);
811 
812 	vsock_deassign_transport(vsk);
813 
814 	/* When clearing these addresses, there's no need to set the family and
815 	 * possibly register the address family with the kernel.
816 	 */
817 	vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
818 	vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
819 
820 	put_cred(vsk->owner);
821 }
822 
823 static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
824 {
825 	int err;
826 
827 	err = sock_queue_rcv_skb(sk, skb);
828 	if (err)
829 		kfree_skb(skb);
830 
831 	return err;
832 }
833 
834 struct sock *vsock_create_connected(struct sock *parent)
835 {
836 	return __vsock_create(sock_net(parent), NULL, parent, GFP_KERNEL,
837 			      parent->sk_type, 0);
838 }
839 EXPORT_SYMBOL_GPL(vsock_create_connected);
840 
841 s64 vsock_stream_has_data(struct vsock_sock *vsk)
842 {
843 	return vsk->transport->stream_has_data(vsk);
844 }
845 EXPORT_SYMBOL_GPL(vsock_stream_has_data);
846 
847 s64 vsock_stream_has_space(struct vsock_sock *vsk)
848 {
849 	return vsk->transport->stream_has_space(vsk);
850 }
851 EXPORT_SYMBOL_GPL(vsock_stream_has_space);
852 
853 static int vsock_release(struct socket *sock)
854 {
855 	__vsock_release(sock->sk, 0);
856 	sock->sk = NULL;
857 	sock->state = SS_FREE;
858 
859 	return 0;
860 }
861 
862 static int
863 vsock_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
864 {
865 	int err;
866 	struct sock *sk;
867 	struct sockaddr_vm *vm_addr;
868 
869 	sk = sock->sk;
870 
871 	if (vsock_addr_cast(addr, addr_len, &vm_addr) != 0)
872 		return -EINVAL;
873 
874 	lock_sock(sk);
875 	err = __vsock_bind(sk, vm_addr);
876 	release_sock(sk);
877 
878 	return err;
879 }
880 
881 static int vsock_getname(struct socket *sock,
882 			 struct sockaddr *addr, int peer)
883 {
884 	int err;
885 	struct sock *sk;
886 	struct vsock_sock *vsk;
887 	struct sockaddr_vm *vm_addr;
888 
889 	sk = sock->sk;
890 	vsk = vsock_sk(sk);
891 	err = 0;
892 
893 	lock_sock(sk);
894 
895 	if (peer) {
896 		if (sock->state != SS_CONNECTED) {
897 			err = -ENOTCONN;
898 			goto out;
899 		}
900 		vm_addr = &vsk->remote_addr;
901 	} else {
902 		vm_addr = &vsk->local_addr;
903 	}
904 
905 	if (!vm_addr) {
906 		err = -EINVAL;
907 		goto out;
908 	}
909 
910 	/* sys_getsockname() and sys_getpeername() pass us a
911 	 * MAX_SOCK_ADDR-sized buffer and don't set addr_len.  Unfortunately
912 	 * that macro is defined in socket.c instead of .h, so we hardcode its
913 	 * value here.
914 	 */
915 	BUILD_BUG_ON(sizeof(*vm_addr) > 128);
916 	memcpy(addr, vm_addr, sizeof(*vm_addr));
917 	err = sizeof(*vm_addr);
918 
919 out:
920 	release_sock(sk);
921 	return err;
922 }
923 
924 static int vsock_shutdown(struct socket *sock, int mode)
925 {
926 	int err;
927 	struct sock *sk;
928 
929 	/* User level uses SHUT_RD (0) and SHUT_WR (1), but the kernel uses
930 	 * RCV_SHUTDOWN (1) and SEND_SHUTDOWN (2), so we must increment mode
931 	 * here like the other address families do.  Note also that the
932 	 * increment makes SHUT_RDWR (2) into RCV_SHUTDOWN | SEND_SHUTDOWN (3),
933 	 * which is what we want.
934 	 */
935 	mode++;
936 
937 	if ((mode & ~SHUTDOWN_MASK) || !mode)
938 		return -EINVAL;
939 
940 	/* If this is a STREAM socket and it is not connected then bail out
941 	 * immediately.  If it is a DGRAM socket then we must first kick the
942 	 * socket so that it wakes up from any sleeping calls, for example
943 	 * recv(), and then afterwards return the error.
944 	 */
945 
946 	sk = sock->sk;
947 
948 	lock_sock(sk);
949 	if (sock->state == SS_UNCONNECTED) {
950 		err = -ENOTCONN;
951 		if (sk->sk_type == SOCK_STREAM)
952 			goto out;
953 	} else {
954 		sock->state = SS_DISCONNECTING;
955 		err = 0;
956 	}
957 
958 	/* Receive and send shutdowns are treated alike. */
959 	mode = mode & (RCV_SHUTDOWN | SEND_SHUTDOWN);
960 	if (mode) {
961 		sk->sk_shutdown |= mode;
962 		sk->sk_state_change(sk);
963 
964 		if (sk->sk_type == SOCK_STREAM) {
965 			sock_reset_flag(sk, SOCK_DONE);
966 			vsock_send_shutdown(sk, mode);
967 		}
968 	}
969 
970 out:
971 	release_sock(sk);
972 	return err;
973 }
974 
975 static __poll_t vsock_poll(struct file *file, struct socket *sock,
976 			       poll_table *wait)
977 {
978 	struct sock *sk;
979 	__poll_t mask;
980 	struct vsock_sock *vsk;
981 
982 	sk = sock->sk;
983 	vsk = vsock_sk(sk);
984 
985 	poll_wait(file, sk_sleep(sk), wait);
986 	mask = 0;
987 
988 	if (sk->sk_err)
989 		/* Signify that there has been an error on this socket. */
990 		mask |= EPOLLERR;
991 
992 	/* INET sockets treat local write shutdown and peer write shutdown as a
993 	 * case of EPOLLHUP set.
994 	 */
995 	if ((sk->sk_shutdown == SHUTDOWN_MASK) ||
996 	    ((sk->sk_shutdown & SEND_SHUTDOWN) &&
997 	     (vsk->peer_shutdown & SEND_SHUTDOWN))) {
998 		mask |= EPOLLHUP;
999 	}
1000 
1001 	if (sk->sk_shutdown & RCV_SHUTDOWN ||
1002 	    vsk->peer_shutdown & SEND_SHUTDOWN) {
1003 		mask |= EPOLLRDHUP;
1004 	}
1005 
1006 	if (sock->type == SOCK_DGRAM) {
1007 		/* For datagram sockets we can read if there is something in
1008 		 * the queue and write as long as the socket isn't shutdown for
1009 		 * sending.
1010 		 */
1011 		if (!skb_queue_empty_lockless(&sk->sk_receive_queue) ||
1012 		    (sk->sk_shutdown & RCV_SHUTDOWN)) {
1013 			mask |= EPOLLIN | EPOLLRDNORM;
1014 		}
1015 
1016 		if (!(sk->sk_shutdown & SEND_SHUTDOWN))
1017 			mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND;
1018 
1019 	} else if (sock->type == SOCK_STREAM) {
1020 		const struct vsock_transport *transport;
1021 
1022 		lock_sock(sk);
1023 
1024 		transport = vsk->transport;
1025 
1026 		/* Listening sockets that have connections in their accept
1027 		 * queue can be read.
1028 		 */
1029 		if (sk->sk_state == TCP_LISTEN
1030 		    && !vsock_is_accept_queue_empty(sk))
1031 			mask |= EPOLLIN | EPOLLRDNORM;
1032 
1033 		/* If there is something in the queue then we can read. */
1034 		if (transport && transport->stream_is_active(vsk) &&
1035 		    !(sk->sk_shutdown & RCV_SHUTDOWN)) {
1036 			bool data_ready_now = false;
1037 			int ret = transport->notify_poll_in(
1038 					vsk, 1, &data_ready_now);
1039 			if (ret < 0) {
1040 				mask |= EPOLLERR;
1041 			} else {
1042 				if (data_ready_now)
1043 					mask |= EPOLLIN | EPOLLRDNORM;
1044 
1045 			}
1046 		}
1047 
1048 		/* Sockets whose connections have been closed, reset, or
1049 		 * terminated should also be considered read, and we check the
1050 		 * shutdown flag for that.
1051 		 */
1052 		if (sk->sk_shutdown & RCV_SHUTDOWN ||
1053 		    vsk->peer_shutdown & SEND_SHUTDOWN) {
1054 			mask |= EPOLLIN | EPOLLRDNORM;
1055 		}
1056 
1057 		/* Connected sockets that can produce data can be written. */
1058 		if (transport && sk->sk_state == TCP_ESTABLISHED) {
1059 			if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1060 				bool space_avail_now = false;
1061 				int ret = transport->notify_poll_out(
1062 						vsk, 1, &space_avail_now);
1063 				if (ret < 0) {
1064 					mask |= EPOLLERR;
1065 				} else {
1066 					if (space_avail_now)
1067 						/* Remove EPOLLWRBAND since INET
1068 						 * sockets are not setting it.
1069 						 */
1070 						mask |= EPOLLOUT | EPOLLWRNORM;
1071 
1072 				}
1073 			}
1074 		}
1075 
1076 		/* Simulate INET socket poll behaviors, which sets
1077 		 * EPOLLOUT|EPOLLWRNORM when peer is closed and nothing to read,
1078 		 * but local send is not shutdown.
1079 		 */
1080 		if (sk->sk_state == TCP_CLOSE || sk->sk_state == TCP_CLOSING) {
1081 			if (!(sk->sk_shutdown & SEND_SHUTDOWN))
1082 				mask |= EPOLLOUT | EPOLLWRNORM;
1083 
1084 		}
1085 
1086 		release_sock(sk);
1087 	}
1088 
1089 	return mask;
1090 }
1091 
1092 static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg,
1093 			       size_t len)
1094 {
1095 	int err;
1096 	struct sock *sk;
1097 	struct vsock_sock *vsk;
1098 	struct sockaddr_vm *remote_addr;
1099 	const struct vsock_transport *transport;
1100 
1101 	if (msg->msg_flags & MSG_OOB)
1102 		return -EOPNOTSUPP;
1103 
1104 	/* For now, MSG_DONTWAIT is always assumed... */
1105 	err = 0;
1106 	sk = sock->sk;
1107 	vsk = vsock_sk(sk);
1108 
1109 	lock_sock(sk);
1110 
1111 	transport = vsk->transport;
1112 
1113 	err = vsock_auto_bind(vsk);
1114 	if (err)
1115 		goto out;
1116 
1117 
1118 	/* If the provided message contains an address, use that.  Otherwise
1119 	 * fall back on the socket's remote handle (if it has been connected).
1120 	 */
1121 	if (msg->msg_name &&
1122 	    vsock_addr_cast(msg->msg_name, msg->msg_namelen,
1123 			    &remote_addr) == 0) {
1124 		/* Ensure this address is of the right type and is a valid
1125 		 * destination.
1126 		 */
1127 
1128 		if (remote_addr->svm_cid == VMADDR_CID_ANY)
1129 			remote_addr->svm_cid = transport->get_local_cid();
1130 
1131 		if (!vsock_addr_bound(remote_addr)) {
1132 			err = -EINVAL;
1133 			goto out;
1134 		}
1135 	} else if (sock->state == SS_CONNECTED) {
1136 		remote_addr = &vsk->remote_addr;
1137 
1138 		if (remote_addr->svm_cid == VMADDR_CID_ANY)
1139 			remote_addr->svm_cid = transport->get_local_cid();
1140 
1141 		/* XXX Should connect() or this function ensure remote_addr is
1142 		 * bound?
1143 		 */
1144 		if (!vsock_addr_bound(&vsk->remote_addr)) {
1145 			err = -EINVAL;
1146 			goto out;
1147 		}
1148 	} else {
1149 		err = -EINVAL;
1150 		goto out;
1151 	}
1152 
1153 	if (!transport->dgram_allow(remote_addr->svm_cid,
1154 				    remote_addr->svm_port)) {
1155 		err = -EINVAL;
1156 		goto out;
1157 	}
1158 
1159 	err = transport->dgram_enqueue(vsk, remote_addr, msg, len);
1160 
1161 out:
1162 	release_sock(sk);
1163 	return err;
1164 }
1165 
1166 static int vsock_dgram_connect(struct socket *sock,
1167 			       struct sockaddr *addr, int addr_len, int flags)
1168 {
1169 	int err;
1170 	struct sock *sk;
1171 	struct vsock_sock *vsk;
1172 	struct sockaddr_vm *remote_addr;
1173 
1174 	sk = sock->sk;
1175 	vsk = vsock_sk(sk);
1176 
1177 	err = vsock_addr_cast(addr, addr_len, &remote_addr);
1178 	if (err == -EAFNOSUPPORT && remote_addr->svm_family == AF_UNSPEC) {
1179 		lock_sock(sk);
1180 		vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY,
1181 				VMADDR_PORT_ANY);
1182 		sock->state = SS_UNCONNECTED;
1183 		release_sock(sk);
1184 		return 0;
1185 	} else if (err != 0)
1186 		return -EINVAL;
1187 
1188 	lock_sock(sk);
1189 
1190 	err = vsock_auto_bind(vsk);
1191 	if (err)
1192 		goto out;
1193 
1194 	if (!vsk->transport->dgram_allow(remote_addr->svm_cid,
1195 					 remote_addr->svm_port)) {
1196 		err = -EINVAL;
1197 		goto out;
1198 	}
1199 
1200 	memcpy(&vsk->remote_addr, remote_addr, sizeof(vsk->remote_addr));
1201 	sock->state = SS_CONNECTED;
1202 
1203 out:
1204 	release_sock(sk);
1205 	return err;
1206 }
1207 
1208 static int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg,
1209 			       size_t len, int flags)
1210 {
1211 	struct vsock_sock *vsk = vsock_sk(sock->sk);
1212 
1213 	return vsk->transport->dgram_dequeue(vsk, msg, len, flags);
1214 }
1215 
1216 static const struct proto_ops vsock_dgram_ops = {
1217 	.family = PF_VSOCK,
1218 	.owner = THIS_MODULE,
1219 	.release = vsock_release,
1220 	.bind = vsock_bind,
1221 	.connect = vsock_dgram_connect,
1222 	.socketpair = sock_no_socketpair,
1223 	.accept = sock_no_accept,
1224 	.getname = vsock_getname,
1225 	.poll = vsock_poll,
1226 	.ioctl = sock_no_ioctl,
1227 	.listen = sock_no_listen,
1228 	.shutdown = vsock_shutdown,
1229 	.sendmsg = vsock_dgram_sendmsg,
1230 	.recvmsg = vsock_dgram_recvmsg,
1231 	.mmap = sock_no_mmap,
1232 	.sendpage = sock_no_sendpage,
1233 };
1234 
1235 static int vsock_transport_cancel_pkt(struct vsock_sock *vsk)
1236 {
1237 	const struct vsock_transport *transport = vsk->transport;
1238 
1239 	if (!transport || !transport->cancel_pkt)
1240 		return -EOPNOTSUPP;
1241 
1242 	return transport->cancel_pkt(vsk);
1243 }
1244 
1245 static void vsock_connect_timeout(struct work_struct *work)
1246 {
1247 	struct sock *sk;
1248 	struct vsock_sock *vsk;
1249 
1250 	vsk = container_of(work, struct vsock_sock, connect_work.work);
1251 	sk = sk_vsock(vsk);
1252 
1253 	lock_sock(sk);
1254 	if (sk->sk_state == TCP_SYN_SENT &&
1255 	    (sk->sk_shutdown != SHUTDOWN_MASK)) {
1256 		sk->sk_state = TCP_CLOSE;
1257 		sk->sk_err = ETIMEDOUT;
1258 		sk->sk_error_report(sk);
1259 		vsock_transport_cancel_pkt(vsk);
1260 	}
1261 	release_sock(sk);
1262 
1263 	sock_put(sk);
1264 }
1265 
1266 static int vsock_stream_connect(struct socket *sock, struct sockaddr *addr,
1267 				int addr_len, int flags)
1268 {
1269 	int err;
1270 	struct sock *sk;
1271 	struct vsock_sock *vsk;
1272 	const struct vsock_transport *transport;
1273 	struct sockaddr_vm *remote_addr;
1274 	long timeout;
1275 	DEFINE_WAIT(wait);
1276 
1277 	err = 0;
1278 	sk = sock->sk;
1279 	vsk = vsock_sk(sk);
1280 
1281 	lock_sock(sk);
1282 
1283 	/* XXX AF_UNSPEC should make us disconnect like AF_INET. */
1284 	switch (sock->state) {
1285 	case SS_CONNECTED:
1286 		err = -EISCONN;
1287 		goto out;
1288 	case SS_DISCONNECTING:
1289 		err = -EINVAL;
1290 		goto out;
1291 	case SS_CONNECTING:
1292 		/* This continues on so we can move sock into the SS_CONNECTED
1293 		 * state once the connection has completed (at which point err
1294 		 * will be set to zero also).  Otherwise, we will either wait
1295 		 * for the connection or return -EALREADY should this be a
1296 		 * non-blocking call.
1297 		 */
1298 		err = -EALREADY;
1299 		break;
1300 	default:
1301 		if ((sk->sk_state == TCP_LISTEN) ||
1302 		    vsock_addr_cast(addr, addr_len, &remote_addr) != 0) {
1303 			err = -EINVAL;
1304 			goto out;
1305 		}
1306 
1307 		/* Set the remote address that we are connecting to. */
1308 		memcpy(&vsk->remote_addr, remote_addr,
1309 		       sizeof(vsk->remote_addr));
1310 
1311 		err = vsock_assign_transport(vsk, NULL);
1312 		if (err)
1313 			goto out;
1314 
1315 		transport = vsk->transport;
1316 
1317 		/* The hypervisor and well-known contexts do not have socket
1318 		 * endpoints.
1319 		 */
1320 		if (!transport ||
1321 		    !transport->stream_allow(remote_addr->svm_cid,
1322 					     remote_addr->svm_port)) {
1323 			err = -ENETUNREACH;
1324 			goto out;
1325 		}
1326 
1327 		err = vsock_auto_bind(vsk);
1328 		if (err)
1329 			goto out;
1330 
1331 		sk->sk_state = TCP_SYN_SENT;
1332 
1333 		err = transport->connect(vsk);
1334 		if (err < 0)
1335 			goto out;
1336 
1337 		/* Mark sock as connecting and set the error code to in
1338 		 * progress in case this is a non-blocking connect.
1339 		 */
1340 		sock->state = SS_CONNECTING;
1341 		err = -EINPROGRESS;
1342 	}
1343 
1344 	/* The receive path will handle all communication until we are able to
1345 	 * enter the connected state.  Here we wait for the connection to be
1346 	 * completed or a notification of an error.
1347 	 */
1348 	timeout = vsk->connect_timeout;
1349 	prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1350 
1351 	while (sk->sk_state != TCP_ESTABLISHED && sk->sk_err == 0) {
1352 		if (flags & O_NONBLOCK) {
1353 			/* If we're not going to block, we schedule a timeout
1354 			 * function to generate a timeout on the connection
1355 			 * attempt, in case the peer doesn't respond in a
1356 			 * timely manner. We hold on to the socket until the
1357 			 * timeout fires.
1358 			 */
1359 			sock_hold(sk);
1360 			schedule_delayed_work(&vsk->connect_work, timeout);
1361 
1362 			/* Skip ahead to preserve error code set above. */
1363 			goto out_wait;
1364 		}
1365 
1366 		release_sock(sk);
1367 		timeout = schedule_timeout(timeout);
1368 		lock_sock(sk);
1369 
1370 		if (signal_pending(current)) {
1371 			err = sock_intr_errno(timeout);
1372 			sk->sk_state = TCP_CLOSE;
1373 			sock->state = SS_UNCONNECTED;
1374 			vsock_transport_cancel_pkt(vsk);
1375 			goto out_wait;
1376 		} else if (timeout == 0) {
1377 			err = -ETIMEDOUT;
1378 			sk->sk_state = TCP_CLOSE;
1379 			sock->state = SS_UNCONNECTED;
1380 			vsock_transport_cancel_pkt(vsk);
1381 			goto out_wait;
1382 		}
1383 
1384 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1385 	}
1386 
1387 	if (sk->sk_err) {
1388 		err = -sk->sk_err;
1389 		sk->sk_state = TCP_CLOSE;
1390 		sock->state = SS_UNCONNECTED;
1391 	} else {
1392 		err = 0;
1393 	}
1394 
1395 out_wait:
1396 	finish_wait(sk_sleep(sk), &wait);
1397 out:
1398 	release_sock(sk);
1399 	return err;
1400 }
1401 
1402 static int vsock_accept(struct socket *sock, struct socket *newsock, int flags,
1403 			bool kern)
1404 {
1405 	struct sock *listener;
1406 	int err;
1407 	struct sock *connected;
1408 	struct vsock_sock *vconnected;
1409 	long timeout;
1410 	DEFINE_WAIT(wait);
1411 
1412 	err = 0;
1413 	listener = sock->sk;
1414 
1415 	lock_sock(listener);
1416 
1417 	if (sock->type != SOCK_STREAM) {
1418 		err = -EOPNOTSUPP;
1419 		goto out;
1420 	}
1421 
1422 	if (listener->sk_state != TCP_LISTEN) {
1423 		err = -EINVAL;
1424 		goto out;
1425 	}
1426 
1427 	/* Wait for children sockets to appear; these are the new sockets
1428 	 * created upon connection establishment.
1429 	 */
1430 	timeout = sock_rcvtimeo(listener, flags & O_NONBLOCK);
1431 	prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1432 
1433 	while ((connected = vsock_dequeue_accept(listener)) == NULL &&
1434 	       listener->sk_err == 0) {
1435 		release_sock(listener);
1436 		timeout = schedule_timeout(timeout);
1437 		finish_wait(sk_sleep(listener), &wait);
1438 		lock_sock(listener);
1439 
1440 		if (signal_pending(current)) {
1441 			err = sock_intr_errno(timeout);
1442 			goto out;
1443 		} else if (timeout == 0) {
1444 			err = -EAGAIN;
1445 			goto out;
1446 		}
1447 
1448 		prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1449 	}
1450 	finish_wait(sk_sleep(listener), &wait);
1451 
1452 	if (listener->sk_err)
1453 		err = -listener->sk_err;
1454 
1455 	if (connected) {
1456 		sk_acceptq_removed(listener);
1457 
1458 		lock_sock_nested(connected, SINGLE_DEPTH_NESTING);
1459 		vconnected = vsock_sk(connected);
1460 
1461 		/* If the listener socket has received an error, then we should
1462 		 * reject this socket and return.  Note that we simply mark the
1463 		 * socket rejected, drop our reference, and let the cleanup
1464 		 * function handle the cleanup; the fact that we found it in
1465 		 * the listener's accept queue guarantees that the cleanup
1466 		 * function hasn't run yet.
1467 		 */
1468 		if (err) {
1469 			vconnected->rejected = true;
1470 		} else {
1471 			newsock->state = SS_CONNECTED;
1472 			sock_graft(connected, newsock);
1473 		}
1474 
1475 		release_sock(connected);
1476 		sock_put(connected);
1477 	}
1478 
1479 out:
1480 	release_sock(listener);
1481 	return err;
1482 }
1483 
1484 static int vsock_listen(struct socket *sock, int backlog)
1485 {
1486 	int err;
1487 	struct sock *sk;
1488 	struct vsock_sock *vsk;
1489 
1490 	sk = sock->sk;
1491 
1492 	lock_sock(sk);
1493 
1494 	if (sock->type != SOCK_STREAM) {
1495 		err = -EOPNOTSUPP;
1496 		goto out;
1497 	}
1498 
1499 	if (sock->state != SS_UNCONNECTED) {
1500 		err = -EINVAL;
1501 		goto out;
1502 	}
1503 
1504 	vsk = vsock_sk(sk);
1505 
1506 	if (!vsock_addr_bound(&vsk->local_addr)) {
1507 		err = -EINVAL;
1508 		goto out;
1509 	}
1510 
1511 	sk->sk_max_ack_backlog = backlog;
1512 	sk->sk_state = TCP_LISTEN;
1513 
1514 	err = 0;
1515 
1516 out:
1517 	release_sock(sk);
1518 	return err;
1519 }
1520 
1521 static void vsock_update_buffer_size(struct vsock_sock *vsk,
1522 				     const struct vsock_transport *transport,
1523 				     u64 val)
1524 {
1525 	if (val > vsk->buffer_max_size)
1526 		val = vsk->buffer_max_size;
1527 
1528 	if (val < vsk->buffer_min_size)
1529 		val = vsk->buffer_min_size;
1530 
1531 	if (val != vsk->buffer_size &&
1532 	    transport && transport->notify_buffer_size)
1533 		transport->notify_buffer_size(vsk, &val);
1534 
1535 	vsk->buffer_size = val;
1536 }
1537 
1538 static int vsock_stream_setsockopt(struct socket *sock,
1539 				   int level,
1540 				   int optname,
1541 				   sockptr_t optval,
1542 				   unsigned int optlen)
1543 {
1544 	int err;
1545 	struct sock *sk;
1546 	struct vsock_sock *vsk;
1547 	const struct vsock_transport *transport;
1548 	u64 val;
1549 
1550 	if (level != AF_VSOCK)
1551 		return -ENOPROTOOPT;
1552 
1553 #define COPY_IN(_v)                                       \
1554 	do {						  \
1555 		if (optlen < sizeof(_v)) {		  \
1556 			err = -EINVAL;			  \
1557 			goto exit;			  \
1558 		}					  \
1559 		if (copy_from_sockptr(&_v, optval, sizeof(_v)) != 0) {	\
1560 			err = -EFAULT;					\
1561 			goto exit;					\
1562 		}							\
1563 	} while (0)
1564 
1565 	err = 0;
1566 	sk = sock->sk;
1567 	vsk = vsock_sk(sk);
1568 
1569 	lock_sock(sk);
1570 
1571 	transport = vsk->transport;
1572 
1573 	switch (optname) {
1574 	case SO_VM_SOCKETS_BUFFER_SIZE:
1575 		COPY_IN(val);
1576 		vsock_update_buffer_size(vsk, transport, val);
1577 		break;
1578 
1579 	case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1580 		COPY_IN(val);
1581 		vsk->buffer_max_size = val;
1582 		vsock_update_buffer_size(vsk, transport, vsk->buffer_size);
1583 		break;
1584 
1585 	case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1586 		COPY_IN(val);
1587 		vsk->buffer_min_size = val;
1588 		vsock_update_buffer_size(vsk, transport, vsk->buffer_size);
1589 		break;
1590 
1591 	case SO_VM_SOCKETS_CONNECT_TIMEOUT: {
1592 		struct __kernel_old_timeval tv;
1593 		COPY_IN(tv);
1594 		if (tv.tv_sec >= 0 && tv.tv_usec < USEC_PER_SEC &&
1595 		    tv.tv_sec < (MAX_SCHEDULE_TIMEOUT / HZ - 1)) {
1596 			vsk->connect_timeout = tv.tv_sec * HZ +
1597 			    DIV_ROUND_UP(tv.tv_usec, (1000000 / HZ));
1598 			if (vsk->connect_timeout == 0)
1599 				vsk->connect_timeout =
1600 				    VSOCK_DEFAULT_CONNECT_TIMEOUT;
1601 
1602 		} else {
1603 			err = -ERANGE;
1604 		}
1605 		break;
1606 	}
1607 
1608 	default:
1609 		err = -ENOPROTOOPT;
1610 		break;
1611 	}
1612 
1613 #undef COPY_IN
1614 
1615 exit:
1616 	release_sock(sk);
1617 	return err;
1618 }
1619 
1620 static int vsock_stream_getsockopt(struct socket *sock,
1621 				   int level, int optname,
1622 				   char __user *optval,
1623 				   int __user *optlen)
1624 {
1625 	int err;
1626 	int len;
1627 	struct sock *sk;
1628 	struct vsock_sock *vsk;
1629 	u64 val;
1630 
1631 	if (level != AF_VSOCK)
1632 		return -ENOPROTOOPT;
1633 
1634 	err = get_user(len, optlen);
1635 	if (err != 0)
1636 		return err;
1637 
1638 #define COPY_OUT(_v)                            \
1639 	do {					\
1640 		if (len < sizeof(_v))		\
1641 			return -EINVAL;		\
1642 						\
1643 		len = sizeof(_v);		\
1644 		if (copy_to_user(optval, &_v, len) != 0)	\
1645 			return -EFAULT;				\
1646 								\
1647 	} while (0)
1648 
1649 	err = 0;
1650 	sk = sock->sk;
1651 	vsk = vsock_sk(sk);
1652 
1653 	switch (optname) {
1654 	case SO_VM_SOCKETS_BUFFER_SIZE:
1655 		val = vsk->buffer_size;
1656 		COPY_OUT(val);
1657 		break;
1658 
1659 	case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1660 		val = vsk->buffer_max_size;
1661 		COPY_OUT(val);
1662 		break;
1663 
1664 	case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1665 		val = vsk->buffer_min_size;
1666 		COPY_OUT(val);
1667 		break;
1668 
1669 	case SO_VM_SOCKETS_CONNECT_TIMEOUT: {
1670 		struct __kernel_old_timeval tv;
1671 		tv.tv_sec = vsk->connect_timeout / HZ;
1672 		tv.tv_usec =
1673 		    (vsk->connect_timeout -
1674 		     tv.tv_sec * HZ) * (1000000 / HZ);
1675 		COPY_OUT(tv);
1676 		break;
1677 	}
1678 	default:
1679 		return -ENOPROTOOPT;
1680 	}
1681 
1682 	err = put_user(len, optlen);
1683 	if (err != 0)
1684 		return -EFAULT;
1685 
1686 #undef COPY_OUT
1687 
1688 	return 0;
1689 }
1690 
1691 static int vsock_stream_sendmsg(struct socket *sock, struct msghdr *msg,
1692 				size_t len)
1693 {
1694 	struct sock *sk;
1695 	struct vsock_sock *vsk;
1696 	const struct vsock_transport *transport;
1697 	ssize_t total_written;
1698 	long timeout;
1699 	int err;
1700 	struct vsock_transport_send_notify_data send_data;
1701 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
1702 
1703 	sk = sock->sk;
1704 	vsk = vsock_sk(sk);
1705 	total_written = 0;
1706 	err = 0;
1707 
1708 	if (msg->msg_flags & MSG_OOB)
1709 		return -EOPNOTSUPP;
1710 
1711 	lock_sock(sk);
1712 
1713 	transport = vsk->transport;
1714 
1715 	/* Callers should not provide a destination with stream sockets. */
1716 	if (msg->msg_namelen) {
1717 		err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP;
1718 		goto out;
1719 	}
1720 
1721 	/* Send data only if both sides are not shutdown in the direction. */
1722 	if (sk->sk_shutdown & SEND_SHUTDOWN ||
1723 	    vsk->peer_shutdown & RCV_SHUTDOWN) {
1724 		err = -EPIPE;
1725 		goto out;
1726 	}
1727 
1728 	if (!transport || sk->sk_state != TCP_ESTABLISHED ||
1729 	    !vsock_addr_bound(&vsk->local_addr)) {
1730 		err = -ENOTCONN;
1731 		goto out;
1732 	}
1733 
1734 	if (!vsock_addr_bound(&vsk->remote_addr)) {
1735 		err = -EDESTADDRREQ;
1736 		goto out;
1737 	}
1738 
1739 	/* Wait for room in the produce queue to enqueue our user's data. */
1740 	timeout = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1741 
1742 	err = transport->notify_send_init(vsk, &send_data);
1743 	if (err < 0)
1744 		goto out;
1745 
1746 	while (total_written < len) {
1747 		ssize_t written;
1748 
1749 		add_wait_queue(sk_sleep(sk), &wait);
1750 		while (vsock_stream_has_space(vsk) == 0 &&
1751 		       sk->sk_err == 0 &&
1752 		       !(sk->sk_shutdown & SEND_SHUTDOWN) &&
1753 		       !(vsk->peer_shutdown & RCV_SHUTDOWN)) {
1754 
1755 			/* Don't wait for non-blocking sockets. */
1756 			if (timeout == 0) {
1757 				err = -EAGAIN;
1758 				remove_wait_queue(sk_sleep(sk), &wait);
1759 				goto out_err;
1760 			}
1761 
1762 			err = transport->notify_send_pre_block(vsk, &send_data);
1763 			if (err < 0) {
1764 				remove_wait_queue(sk_sleep(sk), &wait);
1765 				goto out_err;
1766 			}
1767 
1768 			release_sock(sk);
1769 			timeout = wait_woken(&wait, TASK_INTERRUPTIBLE, timeout);
1770 			lock_sock(sk);
1771 			if (signal_pending(current)) {
1772 				err = sock_intr_errno(timeout);
1773 				remove_wait_queue(sk_sleep(sk), &wait);
1774 				goto out_err;
1775 			} else if (timeout == 0) {
1776 				err = -EAGAIN;
1777 				remove_wait_queue(sk_sleep(sk), &wait);
1778 				goto out_err;
1779 			}
1780 		}
1781 		remove_wait_queue(sk_sleep(sk), &wait);
1782 
1783 		/* These checks occur both as part of and after the loop
1784 		 * conditional since we need to check before and after
1785 		 * sleeping.
1786 		 */
1787 		if (sk->sk_err) {
1788 			err = -sk->sk_err;
1789 			goto out_err;
1790 		} else if ((sk->sk_shutdown & SEND_SHUTDOWN) ||
1791 			   (vsk->peer_shutdown & RCV_SHUTDOWN)) {
1792 			err = -EPIPE;
1793 			goto out_err;
1794 		}
1795 
1796 		err = transport->notify_send_pre_enqueue(vsk, &send_data);
1797 		if (err < 0)
1798 			goto out_err;
1799 
1800 		/* Note that enqueue will only write as many bytes as are free
1801 		 * in the produce queue, so we don't need to ensure len is
1802 		 * smaller than the queue size.  It is the caller's
1803 		 * responsibility to check how many bytes we were able to send.
1804 		 */
1805 
1806 		written = transport->stream_enqueue(
1807 				vsk, msg,
1808 				len - total_written);
1809 		if (written < 0) {
1810 			err = -ENOMEM;
1811 			goto out_err;
1812 		}
1813 
1814 		total_written += written;
1815 
1816 		err = transport->notify_send_post_enqueue(
1817 				vsk, written, &send_data);
1818 		if (err < 0)
1819 			goto out_err;
1820 
1821 	}
1822 
1823 out_err:
1824 	if (total_written > 0)
1825 		err = total_written;
1826 out:
1827 	release_sock(sk);
1828 	return err;
1829 }
1830 
1831 
1832 static int
1833 vsock_stream_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
1834 		     int flags)
1835 {
1836 	struct sock *sk;
1837 	struct vsock_sock *vsk;
1838 	const struct vsock_transport *transport;
1839 	int err;
1840 	size_t target;
1841 	ssize_t copied;
1842 	long timeout;
1843 	struct vsock_transport_recv_notify_data recv_data;
1844 
1845 	DEFINE_WAIT(wait);
1846 
1847 	sk = sock->sk;
1848 	vsk = vsock_sk(sk);
1849 	err = 0;
1850 
1851 	lock_sock(sk);
1852 
1853 	transport = vsk->transport;
1854 
1855 	if (!transport || sk->sk_state != TCP_ESTABLISHED) {
1856 		/* Recvmsg is supposed to return 0 if a peer performs an
1857 		 * orderly shutdown. Differentiate between that case and when a
1858 		 * peer has not connected or a local shutdown occurred with the
1859 		 * SOCK_DONE flag.
1860 		 */
1861 		if (sock_flag(sk, SOCK_DONE))
1862 			err = 0;
1863 		else
1864 			err = -ENOTCONN;
1865 
1866 		goto out;
1867 	}
1868 
1869 	if (flags & MSG_OOB) {
1870 		err = -EOPNOTSUPP;
1871 		goto out;
1872 	}
1873 
1874 	/* We don't check peer_shutdown flag here since peer may actually shut
1875 	 * down, but there can be data in the queue that a local socket can
1876 	 * receive.
1877 	 */
1878 	if (sk->sk_shutdown & RCV_SHUTDOWN) {
1879 		err = 0;
1880 		goto out;
1881 	}
1882 
1883 	/* It is valid on Linux to pass in a zero-length receive buffer.  This
1884 	 * is not an error.  We may as well bail out now.
1885 	 */
1886 	if (!len) {
1887 		err = 0;
1888 		goto out;
1889 	}
1890 
1891 	/* We must not copy less than target bytes into the user's buffer
1892 	 * before returning successfully, so we wait for the consume queue to
1893 	 * have that much data to consume before dequeueing.  Note that this
1894 	 * makes it impossible to handle cases where target is greater than the
1895 	 * queue size.
1896 	 */
1897 	target = sock_rcvlowat(sk, flags & MSG_WAITALL, len);
1898 	if (target >= transport->stream_rcvhiwat(vsk)) {
1899 		err = -ENOMEM;
1900 		goto out;
1901 	}
1902 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1903 	copied = 0;
1904 
1905 	err = transport->notify_recv_init(vsk, target, &recv_data);
1906 	if (err < 0)
1907 		goto out;
1908 
1909 
1910 	while (1) {
1911 		s64 ready;
1912 
1913 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1914 		ready = vsock_stream_has_data(vsk);
1915 
1916 		if (ready == 0) {
1917 			if (sk->sk_err != 0 ||
1918 			    (sk->sk_shutdown & RCV_SHUTDOWN) ||
1919 			    (vsk->peer_shutdown & SEND_SHUTDOWN)) {
1920 				finish_wait(sk_sleep(sk), &wait);
1921 				break;
1922 			}
1923 			/* Don't wait for non-blocking sockets. */
1924 			if (timeout == 0) {
1925 				err = -EAGAIN;
1926 				finish_wait(sk_sleep(sk), &wait);
1927 				break;
1928 			}
1929 
1930 			err = transport->notify_recv_pre_block(
1931 					vsk, target, &recv_data);
1932 			if (err < 0) {
1933 				finish_wait(sk_sleep(sk), &wait);
1934 				break;
1935 			}
1936 			release_sock(sk);
1937 			timeout = schedule_timeout(timeout);
1938 			lock_sock(sk);
1939 
1940 			if (signal_pending(current)) {
1941 				err = sock_intr_errno(timeout);
1942 				finish_wait(sk_sleep(sk), &wait);
1943 				break;
1944 			} else if (timeout == 0) {
1945 				err = -EAGAIN;
1946 				finish_wait(sk_sleep(sk), &wait);
1947 				break;
1948 			}
1949 		} else {
1950 			ssize_t read;
1951 
1952 			finish_wait(sk_sleep(sk), &wait);
1953 
1954 			if (ready < 0) {
1955 				/* Invalid queue pair content. XXX This should
1956 				* be changed to a connection reset in a later
1957 				* change.
1958 				*/
1959 
1960 				err = -ENOMEM;
1961 				goto out;
1962 			}
1963 
1964 			err = transport->notify_recv_pre_dequeue(
1965 					vsk, target, &recv_data);
1966 			if (err < 0)
1967 				break;
1968 
1969 			read = transport->stream_dequeue(
1970 					vsk, msg,
1971 					len - copied, flags);
1972 			if (read < 0) {
1973 				err = -ENOMEM;
1974 				break;
1975 			}
1976 
1977 			copied += read;
1978 
1979 			err = transport->notify_recv_post_dequeue(
1980 					vsk, target, read,
1981 					!(flags & MSG_PEEK), &recv_data);
1982 			if (err < 0)
1983 				goto out;
1984 
1985 			if (read >= target || flags & MSG_PEEK)
1986 				break;
1987 
1988 			target -= read;
1989 		}
1990 	}
1991 
1992 	if (sk->sk_err)
1993 		err = -sk->sk_err;
1994 	else if (sk->sk_shutdown & RCV_SHUTDOWN)
1995 		err = 0;
1996 
1997 	if (copied > 0)
1998 		err = copied;
1999 
2000 out:
2001 	release_sock(sk);
2002 	return err;
2003 }
2004 
2005 static const struct proto_ops vsock_stream_ops = {
2006 	.family = PF_VSOCK,
2007 	.owner = THIS_MODULE,
2008 	.release = vsock_release,
2009 	.bind = vsock_bind,
2010 	.connect = vsock_stream_connect,
2011 	.socketpair = sock_no_socketpair,
2012 	.accept = vsock_accept,
2013 	.getname = vsock_getname,
2014 	.poll = vsock_poll,
2015 	.ioctl = sock_no_ioctl,
2016 	.listen = vsock_listen,
2017 	.shutdown = vsock_shutdown,
2018 	.setsockopt = vsock_stream_setsockopt,
2019 	.getsockopt = vsock_stream_getsockopt,
2020 	.sendmsg = vsock_stream_sendmsg,
2021 	.recvmsg = vsock_stream_recvmsg,
2022 	.mmap = sock_no_mmap,
2023 	.sendpage = sock_no_sendpage,
2024 };
2025 
2026 static int vsock_create(struct net *net, struct socket *sock,
2027 			int protocol, int kern)
2028 {
2029 	struct vsock_sock *vsk;
2030 	struct sock *sk;
2031 	int ret;
2032 
2033 	if (!sock)
2034 		return -EINVAL;
2035 
2036 	if (protocol && protocol != PF_VSOCK)
2037 		return -EPROTONOSUPPORT;
2038 
2039 	switch (sock->type) {
2040 	case SOCK_DGRAM:
2041 		sock->ops = &vsock_dgram_ops;
2042 		break;
2043 	case SOCK_STREAM:
2044 		sock->ops = &vsock_stream_ops;
2045 		break;
2046 	default:
2047 		return -ESOCKTNOSUPPORT;
2048 	}
2049 
2050 	sock->state = SS_UNCONNECTED;
2051 
2052 	sk = __vsock_create(net, sock, NULL, GFP_KERNEL, 0, kern);
2053 	if (!sk)
2054 		return -ENOMEM;
2055 
2056 	vsk = vsock_sk(sk);
2057 
2058 	if (sock->type == SOCK_DGRAM) {
2059 		ret = vsock_assign_transport(vsk, NULL);
2060 		if (ret < 0) {
2061 			sock_put(sk);
2062 			return ret;
2063 		}
2064 	}
2065 
2066 	vsock_insert_unbound(vsk);
2067 
2068 	return 0;
2069 }
2070 
2071 static const struct net_proto_family vsock_family_ops = {
2072 	.family = AF_VSOCK,
2073 	.create = vsock_create,
2074 	.owner = THIS_MODULE,
2075 };
2076 
2077 static long vsock_dev_do_ioctl(struct file *filp,
2078 			       unsigned int cmd, void __user *ptr)
2079 {
2080 	u32 __user *p = ptr;
2081 	u32 cid = VMADDR_CID_ANY;
2082 	int retval = 0;
2083 
2084 	switch (cmd) {
2085 	case IOCTL_VM_SOCKETS_GET_LOCAL_CID:
2086 		/* To be compatible with the VMCI behavior, we prioritize the
2087 		 * guest CID instead of well-know host CID (VMADDR_CID_HOST).
2088 		 */
2089 		if (transport_g2h)
2090 			cid = transport_g2h->get_local_cid();
2091 		else if (transport_h2g)
2092 			cid = transport_h2g->get_local_cid();
2093 
2094 		if (put_user(cid, p) != 0)
2095 			retval = -EFAULT;
2096 		break;
2097 
2098 	default:
2099 		retval = -ENOIOCTLCMD;
2100 	}
2101 
2102 	return retval;
2103 }
2104 
2105 static long vsock_dev_ioctl(struct file *filp,
2106 			    unsigned int cmd, unsigned long arg)
2107 {
2108 	return vsock_dev_do_ioctl(filp, cmd, (void __user *)arg);
2109 }
2110 
2111 #ifdef CONFIG_COMPAT
2112 static long vsock_dev_compat_ioctl(struct file *filp,
2113 				   unsigned int cmd, unsigned long arg)
2114 {
2115 	return vsock_dev_do_ioctl(filp, cmd, compat_ptr(arg));
2116 }
2117 #endif
2118 
2119 static const struct file_operations vsock_device_ops = {
2120 	.owner		= THIS_MODULE,
2121 	.unlocked_ioctl	= vsock_dev_ioctl,
2122 #ifdef CONFIG_COMPAT
2123 	.compat_ioctl	= vsock_dev_compat_ioctl,
2124 #endif
2125 	.open		= nonseekable_open,
2126 };
2127 
2128 static struct miscdevice vsock_device = {
2129 	.name		= "vsock",
2130 	.fops		= &vsock_device_ops,
2131 };
2132 
2133 static int __init vsock_init(void)
2134 {
2135 	int err = 0;
2136 
2137 	vsock_init_tables();
2138 
2139 	vsock_proto.owner = THIS_MODULE;
2140 	vsock_device.minor = MISC_DYNAMIC_MINOR;
2141 	err = misc_register(&vsock_device);
2142 	if (err) {
2143 		pr_err("Failed to register misc device\n");
2144 		goto err_reset_transport;
2145 	}
2146 
2147 	err = proto_register(&vsock_proto, 1);	/* we want our slab */
2148 	if (err) {
2149 		pr_err("Cannot register vsock protocol\n");
2150 		goto err_deregister_misc;
2151 	}
2152 
2153 	err = sock_register(&vsock_family_ops);
2154 	if (err) {
2155 		pr_err("could not register af_vsock (%d) address family: %d\n",
2156 		       AF_VSOCK, err);
2157 		goto err_unregister_proto;
2158 	}
2159 
2160 	return 0;
2161 
2162 err_unregister_proto:
2163 	proto_unregister(&vsock_proto);
2164 err_deregister_misc:
2165 	misc_deregister(&vsock_device);
2166 err_reset_transport:
2167 	return err;
2168 }
2169 
2170 static void __exit vsock_exit(void)
2171 {
2172 	misc_deregister(&vsock_device);
2173 	sock_unregister(AF_VSOCK);
2174 	proto_unregister(&vsock_proto);
2175 }
2176 
2177 const struct vsock_transport *vsock_core_get_transport(struct vsock_sock *vsk)
2178 {
2179 	return vsk->transport;
2180 }
2181 EXPORT_SYMBOL_GPL(vsock_core_get_transport);
2182 
2183 int vsock_core_register(const struct vsock_transport *t, int features)
2184 {
2185 	const struct vsock_transport *t_h2g, *t_g2h, *t_dgram, *t_local;
2186 	int err = mutex_lock_interruptible(&vsock_register_mutex);
2187 
2188 	if (err)
2189 		return err;
2190 
2191 	t_h2g = transport_h2g;
2192 	t_g2h = transport_g2h;
2193 	t_dgram = transport_dgram;
2194 	t_local = transport_local;
2195 
2196 	if (features & VSOCK_TRANSPORT_F_H2G) {
2197 		if (t_h2g) {
2198 			err = -EBUSY;
2199 			goto err_busy;
2200 		}
2201 		t_h2g = t;
2202 	}
2203 
2204 	if (features & VSOCK_TRANSPORT_F_G2H) {
2205 		if (t_g2h) {
2206 			err = -EBUSY;
2207 			goto err_busy;
2208 		}
2209 		t_g2h = t;
2210 	}
2211 
2212 	if (features & VSOCK_TRANSPORT_F_DGRAM) {
2213 		if (t_dgram) {
2214 			err = -EBUSY;
2215 			goto err_busy;
2216 		}
2217 		t_dgram = t;
2218 	}
2219 
2220 	if (features & VSOCK_TRANSPORT_F_LOCAL) {
2221 		if (t_local) {
2222 			err = -EBUSY;
2223 			goto err_busy;
2224 		}
2225 		t_local = t;
2226 	}
2227 
2228 	transport_h2g = t_h2g;
2229 	transport_g2h = t_g2h;
2230 	transport_dgram = t_dgram;
2231 	transport_local = t_local;
2232 
2233 err_busy:
2234 	mutex_unlock(&vsock_register_mutex);
2235 	return err;
2236 }
2237 EXPORT_SYMBOL_GPL(vsock_core_register);
2238 
2239 void vsock_core_unregister(const struct vsock_transport *t)
2240 {
2241 	mutex_lock(&vsock_register_mutex);
2242 
2243 	if (transport_h2g == t)
2244 		transport_h2g = NULL;
2245 
2246 	if (transport_g2h == t)
2247 		transport_g2h = NULL;
2248 
2249 	if (transport_dgram == t)
2250 		transport_dgram = NULL;
2251 
2252 	if (transport_local == t)
2253 		transport_local = NULL;
2254 
2255 	mutex_unlock(&vsock_register_mutex);
2256 }
2257 EXPORT_SYMBOL_GPL(vsock_core_unregister);
2258 
2259 module_init(vsock_init);
2260 module_exit(vsock_exit);
2261 
2262 MODULE_AUTHOR("VMware, Inc.");
2263 MODULE_DESCRIPTION("VMware Virtual Socket Family");
2264 MODULE_VERSION("1.0.2.0-k");
2265 MODULE_LICENSE("GPL v2");
2266