xref: /openbmc/linux/net/vmw_vsock/af_vsock.c (revision 1b39eacd)
1 /*
2  * VMware vSockets Driver
3  *
4  * Copyright (C) 2007-2013 VMware, Inc. All rights reserved.
5  *
6  * This program is free software; you can redistribute it and/or modify it
7  * under the terms of the GNU General Public License as published by the Free
8  * Software Foundation version 2 and no later version.
9  *
10  * This program is distributed in the hope that it will be useful, but WITHOUT
11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13  * more details.
14  */
15 
16 /* Implementation notes:
17  *
18  * - There are two kinds of sockets: those created by user action (such as
19  * calling socket(2)) and those created by incoming connection request packets.
20  *
21  * - There are two "global" tables, one for bound sockets (sockets that have
22  * specified an address that they are responsible for) and one for connected
23  * sockets (sockets that have established a connection with another socket).
24  * These tables are "global" in that all sockets on the system are placed
25  * within them. - Note, though, that the bound table contains an extra entry
26  * for a list of unbound sockets and SOCK_DGRAM sockets will always remain in
27  * that list. The bound table is used solely for lookup of sockets when packets
28  * are received and that's not necessary for SOCK_DGRAM sockets since we create
29  * a datagram handle for each and need not perform a lookup.  Keeping SOCK_DGRAM
30  * sockets out of the bound hash buckets will reduce the chance of collisions
31  * when looking for SOCK_STREAM sockets and prevents us from having to check the
32  * socket type in the hash table lookups.
33  *
34  * - Sockets created by user action will either be "client" sockets that
35  * initiate a connection or "server" sockets that listen for connections; we do
36  * not support simultaneous connects (two "client" sockets connecting).
37  *
38  * - "Server" sockets are referred to as listener sockets throughout this
39  * implementation because they are in the TCP_LISTEN state.  When a
40  * connection request is received (the second kind of socket mentioned above),
41  * we create a new socket and refer to it as a pending socket.  These pending
42  * sockets are placed on the pending connection list of the listener socket.
43  * When future packets are received for the address the listener socket is
44  * bound to, we check if the source of the packet is from one that has an
45  * existing pending connection.  If it does, we process the packet for the
46  * pending socket.  When that socket reaches the connected state, it is removed
47  * from the listener socket's pending list and enqueued in the listener
48  * socket's accept queue.  Callers of accept(2) will accept connected sockets
49  * from the listener socket's accept queue.  If the socket cannot be accepted
50  * for some reason then it is marked rejected.  Once the connection is
51  * accepted, it is owned by the user process and the responsibility for cleanup
52  * falls with that user process.
53  *
54  * - It is possible that these pending sockets will never reach the connected
55  * state; in fact, we may never receive another packet after the connection
56  * request.  Because of this, we must schedule a cleanup function to run in the
57  * future, after some amount of time passes where a connection should have been
58  * established.  This function ensures that the socket is off all lists so it
59  * cannot be retrieved, then drops all references to the socket so it is cleaned
60  * up (sock_put() -> sk_free() -> our sk_destruct implementation).  Note this
61  * function will also cleanup rejected sockets, those that reach the connected
62  * state but leave it before they have been accepted.
63  *
64  * - Lock ordering for pending or accept queue sockets is:
65  *
66  *     lock_sock(listener);
67  *     lock_sock_nested(pending, SINGLE_DEPTH_NESTING);
68  *
69  * Using explicit nested locking keeps lockdep happy since normally only one
70  * lock of a given class may be taken at a time.
71  *
72  * - Sockets created by user action will be cleaned up when the user process
73  * calls close(2), causing our release implementation to be called. Our release
74  * implementation will perform some cleanup then drop the last reference so our
75  * sk_destruct implementation is invoked.  Our sk_destruct implementation will
76  * perform additional cleanup that's common for both types of sockets.
77  *
78  * - A socket's reference count is what ensures that the structure won't be
79  * freed.  Each entry in a list (such as the "global" bound and connected tables
80  * and the listener socket's pending list and connected queue) ensures a
81  * reference.  When we defer work until process context and pass a socket as our
82  * argument, we must ensure the reference count is increased to ensure the
83  * socket isn't freed before the function is run; the deferred function will
84  * then drop the reference.
85  *
86  * - sk->sk_state uses the TCP state constants because they are widely used by
87  * other address families and exposed to userspace tools like ss(8):
88  *
89  *   TCP_CLOSE - unconnected
90  *   TCP_SYN_SENT - connecting
91  *   TCP_ESTABLISHED - connected
92  *   TCP_CLOSING - disconnecting
93  *   TCP_LISTEN - listening
94  */
95 
96 #include <linux/types.h>
97 #include <linux/bitops.h>
98 #include <linux/cred.h>
99 #include <linux/init.h>
100 #include <linux/io.h>
101 #include <linux/kernel.h>
102 #include <linux/sched/signal.h>
103 #include <linux/kmod.h>
104 #include <linux/list.h>
105 #include <linux/miscdevice.h>
106 #include <linux/module.h>
107 #include <linux/mutex.h>
108 #include <linux/net.h>
109 #include <linux/poll.h>
110 #include <linux/skbuff.h>
111 #include <linux/smp.h>
112 #include <linux/socket.h>
113 #include <linux/stddef.h>
114 #include <linux/unistd.h>
115 #include <linux/wait.h>
116 #include <linux/workqueue.h>
117 #include <net/sock.h>
118 #include <net/af_vsock.h>
119 
120 static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr);
121 static void vsock_sk_destruct(struct sock *sk);
122 static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
123 
124 /* Protocol family. */
125 static struct proto vsock_proto = {
126 	.name = "AF_VSOCK",
127 	.owner = THIS_MODULE,
128 	.obj_size = sizeof(struct vsock_sock),
129 };
130 
131 /* The default peer timeout indicates how long we will wait for a peer response
132  * to a control message.
133  */
134 #define VSOCK_DEFAULT_CONNECT_TIMEOUT (2 * HZ)
135 
136 static const struct vsock_transport *transport;
137 static DEFINE_MUTEX(vsock_register_mutex);
138 
139 /**** EXPORTS ****/
140 
141 /* Get the ID of the local context.  This is transport dependent. */
142 
143 int vm_sockets_get_local_cid(void)
144 {
145 	return transport->get_local_cid();
146 }
147 EXPORT_SYMBOL_GPL(vm_sockets_get_local_cid);
148 
149 /**** UTILS ****/
150 
151 /* Each bound VSocket is stored in the bind hash table and each connected
152  * VSocket is stored in the connected hash table.
153  *
154  * Unbound sockets are all put on the same list attached to the end of the hash
155  * table (vsock_unbound_sockets).  Bound sockets are added to the hash table in
156  * the bucket that their local address hashes to (vsock_bound_sockets(addr)
157  * represents the list that addr hashes to).
158  *
159  * Specifically, we initialize the vsock_bind_table array to a size of
160  * VSOCK_HASH_SIZE + 1 so that vsock_bind_table[0] through
161  * vsock_bind_table[VSOCK_HASH_SIZE - 1] are for bound sockets and
162  * vsock_bind_table[VSOCK_HASH_SIZE] is for unbound sockets.  The hash function
163  * mods with VSOCK_HASH_SIZE to ensure this.
164  */
165 #define MAX_PORT_RETRIES        24
166 
167 #define VSOCK_HASH(addr)        ((addr)->svm_port % VSOCK_HASH_SIZE)
168 #define vsock_bound_sockets(addr) (&vsock_bind_table[VSOCK_HASH(addr)])
169 #define vsock_unbound_sockets     (&vsock_bind_table[VSOCK_HASH_SIZE])
170 
171 /* XXX This can probably be implemented in a better way. */
172 #define VSOCK_CONN_HASH(src, dst)				\
173 	(((src)->svm_cid ^ (dst)->svm_port) % VSOCK_HASH_SIZE)
174 #define vsock_connected_sockets(src, dst)		\
175 	(&vsock_connected_table[VSOCK_CONN_HASH(src, dst)])
176 #define vsock_connected_sockets_vsk(vsk)				\
177 	vsock_connected_sockets(&(vsk)->remote_addr, &(vsk)->local_addr)
178 
179 struct list_head vsock_bind_table[VSOCK_HASH_SIZE + 1];
180 EXPORT_SYMBOL_GPL(vsock_bind_table);
181 struct list_head vsock_connected_table[VSOCK_HASH_SIZE];
182 EXPORT_SYMBOL_GPL(vsock_connected_table);
183 DEFINE_SPINLOCK(vsock_table_lock);
184 EXPORT_SYMBOL_GPL(vsock_table_lock);
185 
186 /* Autobind this socket to the local address if necessary. */
187 static int vsock_auto_bind(struct vsock_sock *vsk)
188 {
189 	struct sock *sk = sk_vsock(vsk);
190 	struct sockaddr_vm local_addr;
191 
192 	if (vsock_addr_bound(&vsk->local_addr))
193 		return 0;
194 	vsock_addr_init(&local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
195 	return __vsock_bind(sk, &local_addr);
196 }
197 
198 static int __init vsock_init_tables(void)
199 {
200 	int i;
201 
202 	for (i = 0; i < ARRAY_SIZE(vsock_bind_table); i++)
203 		INIT_LIST_HEAD(&vsock_bind_table[i]);
204 
205 	for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++)
206 		INIT_LIST_HEAD(&vsock_connected_table[i]);
207 	return 0;
208 }
209 
210 static void __vsock_insert_bound(struct list_head *list,
211 				 struct vsock_sock *vsk)
212 {
213 	sock_hold(&vsk->sk);
214 	list_add(&vsk->bound_table, list);
215 }
216 
217 static void __vsock_insert_connected(struct list_head *list,
218 				     struct vsock_sock *vsk)
219 {
220 	sock_hold(&vsk->sk);
221 	list_add(&vsk->connected_table, list);
222 }
223 
224 static void __vsock_remove_bound(struct vsock_sock *vsk)
225 {
226 	list_del_init(&vsk->bound_table);
227 	sock_put(&vsk->sk);
228 }
229 
230 static void __vsock_remove_connected(struct vsock_sock *vsk)
231 {
232 	list_del_init(&vsk->connected_table);
233 	sock_put(&vsk->sk);
234 }
235 
236 static struct sock *__vsock_find_bound_socket(struct sockaddr_vm *addr)
237 {
238 	struct vsock_sock *vsk;
239 
240 	list_for_each_entry(vsk, vsock_bound_sockets(addr), bound_table)
241 		if (addr->svm_port == vsk->local_addr.svm_port)
242 			return sk_vsock(vsk);
243 
244 	return NULL;
245 }
246 
247 static struct sock *__vsock_find_connected_socket(struct sockaddr_vm *src,
248 						  struct sockaddr_vm *dst)
249 {
250 	struct vsock_sock *vsk;
251 
252 	list_for_each_entry(vsk, vsock_connected_sockets(src, dst),
253 			    connected_table) {
254 		if (vsock_addr_equals_addr(src, &vsk->remote_addr) &&
255 		    dst->svm_port == vsk->local_addr.svm_port) {
256 			return sk_vsock(vsk);
257 		}
258 	}
259 
260 	return NULL;
261 }
262 
263 static void vsock_insert_unbound(struct vsock_sock *vsk)
264 {
265 	spin_lock_bh(&vsock_table_lock);
266 	__vsock_insert_bound(vsock_unbound_sockets, vsk);
267 	spin_unlock_bh(&vsock_table_lock);
268 }
269 
270 void vsock_insert_connected(struct vsock_sock *vsk)
271 {
272 	struct list_head *list = vsock_connected_sockets(
273 		&vsk->remote_addr, &vsk->local_addr);
274 
275 	spin_lock_bh(&vsock_table_lock);
276 	__vsock_insert_connected(list, vsk);
277 	spin_unlock_bh(&vsock_table_lock);
278 }
279 EXPORT_SYMBOL_GPL(vsock_insert_connected);
280 
281 void vsock_remove_bound(struct vsock_sock *vsk)
282 {
283 	spin_lock_bh(&vsock_table_lock);
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 	__vsock_remove_connected(vsk);
293 	spin_unlock_bh(&vsock_table_lock);
294 }
295 EXPORT_SYMBOL_GPL(vsock_remove_connected);
296 
297 struct sock *vsock_find_bound_socket(struct sockaddr_vm *addr)
298 {
299 	struct sock *sk;
300 
301 	spin_lock_bh(&vsock_table_lock);
302 	sk = __vsock_find_bound_socket(addr);
303 	if (sk)
304 		sock_hold(sk);
305 
306 	spin_unlock_bh(&vsock_table_lock);
307 
308 	return sk;
309 }
310 EXPORT_SYMBOL_GPL(vsock_find_bound_socket);
311 
312 struct sock *vsock_find_connected_socket(struct sockaddr_vm *src,
313 					 struct sockaddr_vm *dst)
314 {
315 	struct sock *sk;
316 
317 	spin_lock_bh(&vsock_table_lock);
318 	sk = __vsock_find_connected_socket(src, dst);
319 	if (sk)
320 		sock_hold(sk);
321 
322 	spin_unlock_bh(&vsock_table_lock);
323 
324 	return sk;
325 }
326 EXPORT_SYMBOL_GPL(vsock_find_connected_socket);
327 
328 static bool vsock_in_bound_table(struct vsock_sock *vsk)
329 {
330 	bool ret;
331 
332 	spin_lock_bh(&vsock_table_lock);
333 	ret = __vsock_in_bound_table(vsk);
334 	spin_unlock_bh(&vsock_table_lock);
335 
336 	return ret;
337 }
338 
339 static bool vsock_in_connected_table(struct vsock_sock *vsk)
340 {
341 	bool ret;
342 
343 	spin_lock_bh(&vsock_table_lock);
344 	ret = __vsock_in_connected_table(vsk);
345 	spin_unlock_bh(&vsock_table_lock);
346 
347 	return ret;
348 }
349 
350 void vsock_remove_sock(struct vsock_sock *vsk)
351 {
352 	if (vsock_in_bound_table(vsk))
353 		vsock_remove_bound(vsk);
354 
355 	if (vsock_in_connected_table(vsk))
356 		vsock_remove_connected(vsk);
357 }
358 EXPORT_SYMBOL_GPL(vsock_remove_sock);
359 
360 void vsock_for_each_connected_socket(void (*fn)(struct sock *sk))
361 {
362 	int i;
363 
364 	spin_lock_bh(&vsock_table_lock);
365 
366 	for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) {
367 		struct vsock_sock *vsk;
368 		list_for_each_entry(vsk, &vsock_connected_table[i],
369 				    connected_table)
370 			fn(sk_vsock(vsk));
371 	}
372 
373 	spin_unlock_bh(&vsock_table_lock);
374 }
375 EXPORT_SYMBOL_GPL(vsock_for_each_connected_socket);
376 
377 void vsock_add_pending(struct sock *listener, struct sock *pending)
378 {
379 	struct vsock_sock *vlistener;
380 	struct vsock_sock *vpending;
381 
382 	vlistener = vsock_sk(listener);
383 	vpending = vsock_sk(pending);
384 
385 	sock_hold(pending);
386 	sock_hold(listener);
387 	list_add_tail(&vpending->pending_links, &vlistener->pending_links);
388 }
389 EXPORT_SYMBOL_GPL(vsock_add_pending);
390 
391 void vsock_remove_pending(struct sock *listener, struct sock *pending)
392 {
393 	struct vsock_sock *vpending = vsock_sk(pending);
394 
395 	list_del_init(&vpending->pending_links);
396 	sock_put(listener);
397 	sock_put(pending);
398 }
399 EXPORT_SYMBOL_GPL(vsock_remove_pending);
400 
401 void vsock_enqueue_accept(struct sock *listener, struct sock *connected)
402 {
403 	struct vsock_sock *vlistener;
404 	struct vsock_sock *vconnected;
405 
406 	vlistener = vsock_sk(listener);
407 	vconnected = vsock_sk(connected);
408 
409 	sock_hold(connected);
410 	sock_hold(listener);
411 	list_add_tail(&vconnected->accept_queue, &vlistener->accept_queue);
412 }
413 EXPORT_SYMBOL_GPL(vsock_enqueue_accept);
414 
415 static struct sock *vsock_dequeue_accept(struct sock *listener)
416 {
417 	struct vsock_sock *vlistener;
418 	struct vsock_sock *vconnected;
419 
420 	vlistener = vsock_sk(listener);
421 
422 	if (list_empty(&vlistener->accept_queue))
423 		return NULL;
424 
425 	vconnected = list_entry(vlistener->accept_queue.next,
426 				struct vsock_sock, accept_queue);
427 
428 	list_del_init(&vconnected->accept_queue);
429 	sock_put(listener);
430 	/* The caller will need a reference on the connected socket so we let
431 	 * it call sock_put().
432 	 */
433 
434 	return sk_vsock(vconnected);
435 }
436 
437 static bool vsock_is_accept_queue_empty(struct sock *sk)
438 {
439 	struct vsock_sock *vsk = vsock_sk(sk);
440 	return list_empty(&vsk->accept_queue);
441 }
442 
443 static bool vsock_is_pending(struct sock *sk)
444 {
445 	struct vsock_sock *vsk = vsock_sk(sk);
446 	return !list_empty(&vsk->pending_links);
447 }
448 
449 static int vsock_send_shutdown(struct sock *sk, int mode)
450 {
451 	return transport->shutdown(vsock_sk(sk), mode);
452 }
453 
454 void vsock_pending_work(struct work_struct *work)
455 {
456 	struct sock *sk;
457 	struct sock *listener;
458 	struct vsock_sock *vsk;
459 	bool cleanup;
460 
461 	vsk = container_of(work, struct vsock_sock, dwork.work);
462 	sk = sk_vsock(vsk);
463 	listener = vsk->listener;
464 	cleanup = true;
465 
466 	lock_sock(listener);
467 	lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
468 
469 	if (vsock_is_pending(sk)) {
470 		vsock_remove_pending(listener, sk);
471 
472 		listener->sk_ack_backlog--;
473 	} else if (!vsk->rejected) {
474 		/* We are not on the pending list and accept() did not reject
475 		 * us, so we must have been accepted by our user process.  We
476 		 * just need to drop our references to the sockets and be on
477 		 * our way.
478 		 */
479 		cleanup = false;
480 		goto out;
481 	}
482 
483 	/* We need to remove ourself from the global connected sockets list so
484 	 * incoming packets can't find this socket, and to reduce the reference
485 	 * count.
486 	 */
487 	if (vsock_in_connected_table(vsk))
488 		vsock_remove_connected(vsk);
489 
490 	sk->sk_state = TCP_CLOSE;
491 
492 out:
493 	release_sock(sk);
494 	release_sock(listener);
495 	if (cleanup)
496 		sock_put(sk);
497 
498 	sock_put(sk);
499 	sock_put(listener);
500 }
501 EXPORT_SYMBOL_GPL(vsock_pending_work);
502 
503 /**** SOCKET OPERATIONS ****/
504 
505 static int __vsock_bind_stream(struct vsock_sock *vsk,
506 			       struct sockaddr_vm *addr)
507 {
508 	static u32 port = LAST_RESERVED_PORT + 1;
509 	struct sockaddr_vm new_addr;
510 
511 	vsock_addr_init(&new_addr, addr->svm_cid, addr->svm_port);
512 
513 	if (addr->svm_port == VMADDR_PORT_ANY) {
514 		bool found = false;
515 		unsigned int i;
516 
517 		for (i = 0; i < MAX_PORT_RETRIES; i++) {
518 			if (port <= LAST_RESERVED_PORT)
519 				port = LAST_RESERVED_PORT + 1;
520 
521 			new_addr.svm_port = port++;
522 
523 			if (!__vsock_find_bound_socket(&new_addr)) {
524 				found = true;
525 				break;
526 			}
527 		}
528 
529 		if (!found)
530 			return -EADDRNOTAVAIL;
531 	} else {
532 		/* If port is in reserved range, ensure caller
533 		 * has necessary privileges.
534 		 */
535 		if (addr->svm_port <= LAST_RESERVED_PORT &&
536 		    !capable(CAP_NET_BIND_SERVICE)) {
537 			return -EACCES;
538 		}
539 
540 		if (__vsock_find_bound_socket(&new_addr))
541 			return -EADDRINUSE;
542 	}
543 
544 	vsock_addr_init(&vsk->local_addr, new_addr.svm_cid, new_addr.svm_port);
545 
546 	/* Remove stream sockets from the unbound list and add them to the hash
547 	 * table for easy lookup by its address.  The unbound list is simply an
548 	 * extra entry at the end of the hash table, a trick used by AF_UNIX.
549 	 */
550 	__vsock_remove_bound(vsk);
551 	__vsock_insert_bound(vsock_bound_sockets(&vsk->local_addr), vsk);
552 
553 	return 0;
554 }
555 
556 static int __vsock_bind_dgram(struct vsock_sock *vsk,
557 			      struct sockaddr_vm *addr)
558 {
559 	return transport->dgram_bind(vsk, addr);
560 }
561 
562 static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr)
563 {
564 	struct vsock_sock *vsk = vsock_sk(sk);
565 	u32 cid;
566 	int retval;
567 
568 	/* First ensure this socket isn't already bound. */
569 	if (vsock_addr_bound(&vsk->local_addr))
570 		return -EINVAL;
571 
572 	/* Now bind to the provided address or select appropriate values if
573 	 * none are provided (VMADDR_CID_ANY and VMADDR_PORT_ANY).  Note that
574 	 * like AF_INET prevents binding to a non-local IP address (in most
575 	 * cases), we only allow binding to the local CID.
576 	 */
577 	cid = transport->get_local_cid();
578 	if (addr->svm_cid != cid && addr->svm_cid != VMADDR_CID_ANY)
579 		return -EADDRNOTAVAIL;
580 
581 	switch (sk->sk_socket->type) {
582 	case SOCK_STREAM:
583 		spin_lock_bh(&vsock_table_lock);
584 		retval = __vsock_bind_stream(vsk, addr);
585 		spin_unlock_bh(&vsock_table_lock);
586 		break;
587 
588 	case SOCK_DGRAM:
589 		retval = __vsock_bind_dgram(vsk, addr);
590 		break;
591 
592 	default:
593 		retval = -EINVAL;
594 		break;
595 	}
596 
597 	return retval;
598 }
599 
600 struct sock *__vsock_create(struct net *net,
601 			    struct socket *sock,
602 			    struct sock *parent,
603 			    gfp_t priority,
604 			    unsigned short type,
605 			    int kern)
606 {
607 	struct sock *sk;
608 	struct vsock_sock *psk;
609 	struct vsock_sock *vsk;
610 
611 	sk = sk_alloc(net, AF_VSOCK, priority, &vsock_proto, kern);
612 	if (!sk)
613 		return NULL;
614 
615 	sock_init_data(sock, sk);
616 
617 	/* sk->sk_type is normally set in sock_init_data, but only if sock is
618 	 * non-NULL. We make sure that our sockets always have a type by
619 	 * setting it here if needed.
620 	 */
621 	if (!sock)
622 		sk->sk_type = type;
623 
624 	vsk = vsock_sk(sk);
625 	vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
626 	vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
627 
628 	sk->sk_destruct = vsock_sk_destruct;
629 	sk->sk_backlog_rcv = vsock_queue_rcv_skb;
630 	sock_reset_flag(sk, SOCK_DONE);
631 
632 	INIT_LIST_HEAD(&vsk->bound_table);
633 	INIT_LIST_HEAD(&vsk->connected_table);
634 	vsk->listener = NULL;
635 	INIT_LIST_HEAD(&vsk->pending_links);
636 	INIT_LIST_HEAD(&vsk->accept_queue);
637 	vsk->rejected = false;
638 	vsk->sent_request = false;
639 	vsk->ignore_connecting_rst = false;
640 	vsk->peer_shutdown = 0;
641 
642 	psk = parent ? vsock_sk(parent) : NULL;
643 	if (parent) {
644 		vsk->trusted = psk->trusted;
645 		vsk->owner = get_cred(psk->owner);
646 		vsk->connect_timeout = psk->connect_timeout;
647 	} else {
648 		vsk->trusted = capable(CAP_NET_ADMIN);
649 		vsk->owner = get_current_cred();
650 		vsk->connect_timeout = VSOCK_DEFAULT_CONNECT_TIMEOUT;
651 	}
652 
653 	if (transport->init(vsk, psk) < 0) {
654 		sk_free(sk);
655 		return NULL;
656 	}
657 
658 	if (sock)
659 		vsock_insert_unbound(vsk);
660 
661 	return sk;
662 }
663 EXPORT_SYMBOL_GPL(__vsock_create);
664 
665 static void __vsock_release(struct sock *sk)
666 {
667 	if (sk) {
668 		struct sk_buff *skb;
669 		struct sock *pending;
670 		struct vsock_sock *vsk;
671 
672 		vsk = vsock_sk(sk);
673 		pending = NULL;	/* Compiler warning. */
674 
675 		transport->release(vsk);
676 
677 		lock_sock(sk);
678 		sock_orphan(sk);
679 		sk->sk_shutdown = SHUTDOWN_MASK;
680 
681 		while ((skb = skb_dequeue(&sk->sk_receive_queue)))
682 			kfree_skb(skb);
683 
684 		/* Clean up any sockets that never were accepted. */
685 		while ((pending = vsock_dequeue_accept(sk)) != NULL) {
686 			__vsock_release(pending);
687 			sock_put(pending);
688 		}
689 
690 		release_sock(sk);
691 		sock_put(sk);
692 	}
693 }
694 
695 static void vsock_sk_destruct(struct sock *sk)
696 {
697 	struct vsock_sock *vsk = vsock_sk(sk);
698 
699 	transport->destruct(vsk);
700 
701 	/* When clearing these addresses, there's no need to set the family and
702 	 * possibly register the address family with the kernel.
703 	 */
704 	vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
705 	vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
706 
707 	put_cred(vsk->owner);
708 }
709 
710 static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
711 {
712 	int err;
713 
714 	err = sock_queue_rcv_skb(sk, skb);
715 	if (err)
716 		kfree_skb(skb);
717 
718 	return err;
719 }
720 
721 s64 vsock_stream_has_data(struct vsock_sock *vsk)
722 {
723 	return transport->stream_has_data(vsk);
724 }
725 EXPORT_SYMBOL_GPL(vsock_stream_has_data);
726 
727 s64 vsock_stream_has_space(struct vsock_sock *vsk)
728 {
729 	return transport->stream_has_space(vsk);
730 }
731 EXPORT_SYMBOL_GPL(vsock_stream_has_space);
732 
733 static int vsock_release(struct socket *sock)
734 {
735 	__vsock_release(sock->sk);
736 	sock->sk = NULL;
737 	sock->state = SS_FREE;
738 
739 	return 0;
740 }
741 
742 static int
743 vsock_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
744 {
745 	int err;
746 	struct sock *sk;
747 	struct sockaddr_vm *vm_addr;
748 
749 	sk = sock->sk;
750 
751 	if (vsock_addr_cast(addr, addr_len, &vm_addr) != 0)
752 		return -EINVAL;
753 
754 	lock_sock(sk);
755 	err = __vsock_bind(sk, vm_addr);
756 	release_sock(sk);
757 
758 	return err;
759 }
760 
761 static int vsock_getname(struct socket *sock,
762 			 struct sockaddr *addr, int *addr_len, int peer)
763 {
764 	int err;
765 	struct sock *sk;
766 	struct vsock_sock *vsk;
767 	struct sockaddr_vm *vm_addr;
768 
769 	sk = sock->sk;
770 	vsk = vsock_sk(sk);
771 	err = 0;
772 
773 	lock_sock(sk);
774 
775 	if (peer) {
776 		if (sock->state != SS_CONNECTED) {
777 			err = -ENOTCONN;
778 			goto out;
779 		}
780 		vm_addr = &vsk->remote_addr;
781 	} else {
782 		vm_addr = &vsk->local_addr;
783 	}
784 
785 	if (!vm_addr) {
786 		err = -EINVAL;
787 		goto out;
788 	}
789 
790 	/* sys_getsockname() and sys_getpeername() pass us a
791 	 * MAX_SOCK_ADDR-sized buffer and don't set addr_len.  Unfortunately
792 	 * that macro is defined in socket.c instead of .h, so we hardcode its
793 	 * value here.
794 	 */
795 	BUILD_BUG_ON(sizeof(*vm_addr) > 128);
796 	memcpy(addr, vm_addr, sizeof(*vm_addr));
797 	*addr_len = sizeof(*vm_addr);
798 
799 out:
800 	release_sock(sk);
801 	return err;
802 }
803 
804 static int vsock_shutdown(struct socket *sock, int mode)
805 {
806 	int err;
807 	struct sock *sk;
808 
809 	/* User level uses SHUT_RD (0) and SHUT_WR (1), but the kernel uses
810 	 * RCV_SHUTDOWN (1) and SEND_SHUTDOWN (2), so we must increment mode
811 	 * here like the other address families do.  Note also that the
812 	 * increment makes SHUT_RDWR (2) into RCV_SHUTDOWN | SEND_SHUTDOWN (3),
813 	 * which is what we want.
814 	 */
815 	mode++;
816 
817 	if ((mode & ~SHUTDOWN_MASK) || !mode)
818 		return -EINVAL;
819 
820 	/* If this is a STREAM socket and it is not connected then bail out
821 	 * immediately.  If it is a DGRAM socket then we must first kick the
822 	 * socket so that it wakes up from any sleeping calls, for example
823 	 * recv(), and then afterwards return the error.
824 	 */
825 
826 	sk = sock->sk;
827 	if (sock->state == SS_UNCONNECTED) {
828 		err = -ENOTCONN;
829 		if (sk->sk_type == SOCK_STREAM)
830 			return err;
831 	} else {
832 		sock->state = SS_DISCONNECTING;
833 		err = 0;
834 	}
835 
836 	/* Receive and send shutdowns are treated alike. */
837 	mode = mode & (RCV_SHUTDOWN | SEND_SHUTDOWN);
838 	if (mode) {
839 		lock_sock(sk);
840 		sk->sk_shutdown |= mode;
841 		sk->sk_state_change(sk);
842 		release_sock(sk);
843 
844 		if (sk->sk_type == SOCK_STREAM) {
845 			sock_reset_flag(sk, SOCK_DONE);
846 			vsock_send_shutdown(sk, mode);
847 		}
848 	}
849 
850 	return err;
851 }
852 
853 static unsigned int vsock_poll(struct file *file, struct socket *sock,
854 			       poll_table *wait)
855 {
856 	struct sock *sk;
857 	unsigned int mask;
858 	struct vsock_sock *vsk;
859 
860 	sk = sock->sk;
861 	vsk = vsock_sk(sk);
862 
863 	poll_wait(file, sk_sleep(sk), wait);
864 	mask = 0;
865 
866 	if (sk->sk_err)
867 		/* Signify that there has been an error on this socket. */
868 		mask |= POLLERR;
869 
870 	/* INET sockets treat local write shutdown and peer write shutdown as a
871 	 * case of POLLHUP set.
872 	 */
873 	if ((sk->sk_shutdown == SHUTDOWN_MASK) ||
874 	    ((sk->sk_shutdown & SEND_SHUTDOWN) &&
875 	     (vsk->peer_shutdown & SEND_SHUTDOWN))) {
876 		mask |= POLLHUP;
877 	}
878 
879 	if (sk->sk_shutdown & RCV_SHUTDOWN ||
880 	    vsk->peer_shutdown & SEND_SHUTDOWN) {
881 		mask |= POLLRDHUP;
882 	}
883 
884 	if (sock->type == SOCK_DGRAM) {
885 		/* For datagram sockets we can read if there is something in
886 		 * the queue and write as long as the socket isn't shutdown for
887 		 * sending.
888 		 */
889 		if (!skb_queue_empty(&sk->sk_receive_queue) ||
890 		    (sk->sk_shutdown & RCV_SHUTDOWN)) {
891 			mask |= POLLIN | POLLRDNORM;
892 		}
893 
894 		if (!(sk->sk_shutdown & SEND_SHUTDOWN))
895 			mask |= POLLOUT | POLLWRNORM | POLLWRBAND;
896 
897 	} else if (sock->type == SOCK_STREAM) {
898 		lock_sock(sk);
899 
900 		/* Listening sockets that have connections in their accept
901 		 * queue can be read.
902 		 */
903 		if (sk->sk_state == TCP_LISTEN
904 		    && !vsock_is_accept_queue_empty(sk))
905 			mask |= POLLIN | POLLRDNORM;
906 
907 		/* If there is something in the queue then we can read. */
908 		if (transport->stream_is_active(vsk) &&
909 		    !(sk->sk_shutdown & RCV_SHUTDOWN)) {
910 			bool data_ready_now = false;
911 			int ret = transport->notify_poll_in(
912 					vsk, 1, &data_ready_now);
913 			if (ret < 0) {
914 				mask |= POLLERR;
915 			} else {
916 				if (data_ready_now)
917 					mask |= POLLIN | POLLRDNORM;
918 
919 			}
920 		}
921 
922 		/* Sockets whose connections have been closed, reset, or
923 		 * terminated should also be considered read, and we check the
924 		 * shutdown flag for that.
925 		 */
926 		if (sk->sk_shutdown & RCV_SHUTDOWN ||
927 		    vsk->peer_shutdown & SEND_SHUTDOWN) {
928 			mask |= POLLIN | POLLRDNORM;
929 		}
930 
931 		/* Connected sockets that can produce data can be written. */
932 		if (sk->sk_state == TCP_ESTABLISHED) {
933 			if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
934 				bool space_avail_now = false;
935 				int ret = transport->notify_poll_out(
936 						vsk, 1, &space_avail_now);
937 				if (ret < 0) {
938 					mask |= POLLERR;
939 				} else {
940 					if (space_avail_now)
941 						/* Remove POLLWRBAND since INET
942 						 * sockets are not setting it.
943 						 */
944 						mask |= POLLOUT | POLLWRNORM;
945 
946 				}
947 			}
948 		}
949 
950 		/* Simulate INET socket poll behaviors, which sets
951 		 * POLLOUT|POLLWRNORM when peer is closed and nothing to read,
952 		 * but local send is not shutdown.
953 		 */
954 		if (sk->sk_state == TCP_CLOSE) {
955 			if (!(sk->sk_shutdown & SEND_SHUTDOWN))
956 				mask |= POLLOUT | POLLWRNORM;
957 
958 		}
959 
960 		release_sock(sk);
961 	}
962 
963 	return mask;
964 }
965 
966 static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg,
967 			       size_t len)
968 {
969 	int err;
970 	struct sock *sk;
971 	struct vsock_sock *vsk;
972 	struct sockaddr_vm *remote_addr;
973 
974 	if (msg->msg_flags & MSG_OOB)
975 		return -EOPNOTSUPP;
976 
977 	/* For now, MSG_DONTWAIT is always assumed... */
978 	err = 0;
979 	sk = sock->sk;
980 	vsk = vsock_sk(sk);
981 
982 	lock_sock(sk);
983 
984 	err = vsock_auto_bind(vsk);
985 	if (err)
986 		goto out;
987 
988 
989 	/* If the provided message contains an address, use that.  Otherwise
990 	 * fall back on the socket's remote handle (if it has been connected).
991 	 */
992 	if (msg->msg_name &&
993 	    vsock_addr_cast(msg->msg_name, msg->msg_namelen,
994 			    &remote_addr) == 0) {
995 		/* Ensure this address is of the right type and is a valid
996 		 * destination.
997 		 */
998 
999 		if (remote_addr->svm_cid == VMADDR_CID_ANY)
1000 			remote_addr->svm_cid = transport->get_local_cid();
1001 
1002 		if (!vsock_addr_bound(remote_addr)) {
1003 			err = -EINVAL;
1004 			goto out;
1005 		}
1006 	} else if (sock->state == SS_CONNECTED) {
1007 		remote_addr = &vsk->remote_addr;
1008 
1009 		if (remote_addr->svm_cid == VMADDR_CID_ANY)
1010 			remote_addr->svm_cid = transport->get_local_cid();
1011 
1012 		/* XXX Should connect() or this function ensure remote_addr is
1013 		 * bound?
1014 		 */
1015 		if (!vsock_addr_bound(&vsk->remote_addr)) {
1016 			err = -EINVAL;
1017 			goto out;
1018 		}
1019 	} else {
1020 		err = -EINVAL;
1021 		goto out;
1022 	}
1023 
1024 	if (!transport->dgram_allow(remote_addr->svm_cid,
1025 				    remote_addr->svm_port)) {
1026 		err = -EINVAL;
1027 		goto out;
1028 	}
1029 
1030 	err = transport->dgram_enqueue(vsk, remote_addr, msg, len);
1031 
1032 out:
1033 	release_sock(sk);
1034 	return err;
1035 }
1036 
1037 static int vsock_dgram_connect(struct socket *sock,
1038 			       struct sockaddr *addr, int addr_len, int flags)
1039 {
1040 	int err;
1041 	struct sock *sk;
1042 	struct vsock_sock *vsk;
1043 	struct sockaddr_vm *remote_addr;
1044 
1045 	sk = sock->sk;
1046 	vsk = vsock_sk(sk);
1047 
1048 	err = vsock_addr_cast(addr, addr_len, &remote_addr);
1049 	if (err == -EAFNOSUPPORT && remote_addr->svm_family == AF_UNSPEC) {
1050 		lock_sock(sk);
1051 		vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY,
1052 				VMADDR_PORT_ANY);
1053 		sock->state = SS_UNCONNECTED;
1054 		release_sock(sk);
1055 		return 0;
1056 	} else if (err != 0)
1057 		return -EINVAL;
1058 
1059 	lock_sock(sk);
1060 
1061 	err = vsock_auto_bind(vsk);
1062 	if (err)
1063 		goto out;
1064 
1065 	if (!transport->dgram_allow(remote_addr->svm_cid,
1066 				    remote_addr->svm_port)) {
1067 		err = -EINVAL;
1068 		goto out;
1069 	}
1070 
1071 	memcpy(&vsk->remote_addr, remote_addr, sizeof(vsk->remote_addr));
1072 	sock->state = SS_CONNECTED;
1073 
1074 out:
1075 	release_sock(sk);
1076 	return err;
1077 }
1078 
1079 static int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg,
1080 			       size_t len, int flags)
1081 {
1082 	return transport->dgram_dequeue(vsock_sk(sock->sk), msg, len, flags);
1083 }
1084 
1085 static const struct proto_ops vsock_dgram_ops = {
1086 	.family = PF_VSOCK,
1087 	.owner = THIS_MODULE,
1088 	.release = vsock_release,
1089 	.bind = vsock_bind,
1090 	.connect = vsock_dgram_connect,
1091 	.socketpair = sock_no_socketpair,
1092 	.accept = sock_no_accept,
1093 	.getname = vsock_getname,
1094 	.poll = vsock_poll,
1095 	.ioctl = sock_no_ioctl,
1096 	.listen = sock_no_listen,
1097 	.shutdown = vsock_shutdown,
1098 	.setsockopt = sock_no_setsockopt,
1099 	.getsockopt = sock_no_getsockopt,
1100 	.sendmsg = vsock_dgram_sendmsg,
1101 	.recvmsg = vsock_dgram_recvmsg,
1102 	.mmap = sock_no_mmap,
1103 	.sendpage = sock_no_sendpage,
1104 };
1105 
1106 static int vsock_transport_cancel_pkt(struct vsock_sock *vsk)
1107 {
1108 	if (!transport->cancel_pkt)
1109 		return -EOPNOTSUPP;
1110 
1111 	return transport->cancel_pkt(vsk);
1112 }
1113 
1114 static void vsock_connect_timeout(struct work_struct *work)
1115 {
1116 	struct sock *sk;
1117 	struct vsock_sock *vsk;
1118 	int cancel = 0;
1119 
1120 	vsk = container_of(work, struct vsock_sock, dwork.work);
1121 	sk = sk_vsock(vsk);
1122 
1123 	lock_sock(sk);
1124 	if (sk->sk_state == TCP_SYN_SENT &&
1125 	    (sk->sk_shutdown != SHUTDOWN_MASK)) {
1126 		sk->sk_state = TCP_CLOSE;
1127 		sk->sk_err = ETIMEDOUT;
1128 		sk->sk_error_report(sk);
1129 		cancel = 1;
1130 	}
1131 	release_sock(sk);
1132 	if (cancel)
1133 		vsock_transport_cancel_pkt(vsk);
1134 
1135 	sock_put(sk);
1136 }
1137 
1138 static int vsock_stream_connect(struct socket *sock, struct sockaddr *addr,
1139 				int addr_len, int flags)
1140 {
1141 	int err;
1142 	struct sock *sk;
1143 	struct vsock_sock *vsk;
1144 	struct sockaddr_vm *remote_addr;
1145 	long timeout;
1146 	DEFINE_WAIT(wait);
1147 
1148 	err = 0;
1149 	sk = sock->sk;
1150 	vsk = vsock_sk(sk);
1151 
1152 	lock_sock(sk);
1153 
1154 	/* XXX AF_UNSPEC should make us disconnect like AF_INET. */
1155 	switch (sock->state) {
1156 	case SS_CONNECTED:
1157 		err = -EISCONN;
1158 		goto out;
1159 	case SS_DISCONNECTING:
1160 		err = -EINVAL;
1161 		goto out;
1162 	case SS_CONNECTING:
1163 		/* This continues on so we can move sock into the SS_CONNECTED
1164 		 * state once the connection has completed (at which point err
1165 		 * will be set to zero also).  Otherwise, we will either wait
1166 		 * for the connection or return -EALREADY should this be a
1167 		 * non-blocking call.
1168 		 */
1169 		err = -EALREADY;
1170 		break;
1171 	default:
1172 		if ((sk->sk_state == TCP_LISTEN) ||
1173 		    vsock_addr_cast(addr, addr_len, &remote_addr) != 0) {
1174 			err = -EINVAL;
1175 			goto out;
1176 		}
1177 
1178 		/* The hypervisor and well-known contexts do not have socket
1179 		 * endpoints.
1180 		 */
1181 		if (!transport->stream_allow(remote_addr->svm_cid,
1182 					     remote_addr->svm_port)) {
1183 			err = -ENETUNREACH;
1184 			goto out;
1185 		}
1186 
1187 		/* Set the remote address that we are connecting to. */
1188 		memcpy(&vsk->remote_addr, remote_addr,
1189 		       sizeof(vsk->remote_addr));
1190 
1191 		err = vsock_auto_bind(vsk);
1192 		if (err)
1193 			goto out;
1194 
1195 		sk->sk_state = TCP_SYN_SENT;
1196 
1197 		err = transport->connect(vsk);
1198 		if (err < 0)
1199 			goto out;
1200 
1201 		/* Mark sock as connecting and set the error code to in
1202 		 * progress in case this is a non-blocking connect.
1203 		 */
1204 		sock->state = SS_CONNECTING;
1205 		err = -EINPROGRESS;
1206 	}
1207 
1208 	/* The receive path will handle all communication until we are able to
1209 	 * enter the connected state.  Here we wait for the connection to be
1210 	 * completed or a notification of an error.
1211 	 */
1212 	timeout = vsk->connect_timeout;
1213 	prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1214 
1215 	while (sk->sk_state != TCP_ESTABLISHED && sk->sk_err == 0) {
1216 		if (flags & O_NONBLOCK) {
1217 			/* If we're not going to block, we schedule a timeout
1218 			 * function to generate a timeout on the connection
1219 			 * attempt, in case the peer doesn't respond in a
1220 			 * timely manner. We hold on to the socket until the
1221 			 * timeout fires.
1222 			 */
1223 			sock_hold(sk);
1224 			INIT_DELAYED_WORK(&vsk->dwork,
1225 					  vsock_connect_timeout);
1226 			schedule_delayed_work(&vsk->dwork, timeout);
1227 
1228 			/* Skip ahead to preserve error code set above. */
1229 			goto out_wait;
1230 		}
1231 
1232 		release_sock(sk);
1233 		timeout = schedule_timeout(timeout);
1234 		lock_sock(sk);
1235 
1236 		if (signal_pending(current)) {
1237 			err = sock_intr_errno(timeout);
1238 			sk->sk_state = TCP_CLOSE;
1239 			sock->state = SS_UNCONNECTED;
1240 			vsock_transport_cancel_pkt(vsk);
1241 			goto out_wait;
1242 		} else if (timeout == 0) {
1243 			err = -ETIMEDOUT;
1244 			sk->sk_state = TCP_CLOSE;
1245 			sock->state = SS_UNCONNECTED;
1246 			vsock_transport_cancel_pkt(vsk);
1247 			goto out_wait;
1248 		}
1249 
1250 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1251 	}
1252 
1253 	if (sk->sk_err) {
1254 		err = -sk->sk_err;
1255 		sk->sk_state = TCP_CLOSE;
1256 		sock->state = SS_UNCONNECTED;
1257 	} else {
1258 		err = 0;
1259 	}
1260 
1261 out_wait:
1262 	finish_wait(sk_sleep(sk), &wait);
1263 out:
1264 	release_sock(sk);
1265 	return err;
1266 }
1267 
1268 static int vsock_accept(struct socket *sock, struct socket *newsock, int flags,
1269 			bool kern)
1270 {
1271 	struct sock *listener;
1272 	int err;
1273 	struct sock *connected;
1274 	struct vsock_sock *vconnected;
1275 	long timeout;
1276 	DEFINE_WAIT(wait);
1277 
1278 	err = 0;
1279 	listener = sock->sk;
1280 
1281 	lock_sock(listener);
1282 
1283 	if (sock->type != SOCK_STREAM) {
1284 		err = -EOPNOTSUPP;
1285 		goto out;
1286 	}
1287 
1288 	if (listener->sk_state != TCP_LISTEN) {
1289 		err = -EINVAL;
1290 		goto out;
1291 	}
1292 
1293 	/* Wait for children sockets to appear; these are the new sockets
1294 	 * created upon connection establishment.
1295 	 */
1296 	timeout = sock_sndtimeo(listener, flags & O_NONBLOCK);
1297 	prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1298 
1299 	while ((connected = vsock_dequeue_accept(listener)) == NULL &&
1300 	       listener->sk_err == 0) {
1301 		release_sock(listener);
1302 		timeout = schedule_timeout(timeout);
1303 		finish_wait(sk_sleep(listener), &wait);
1304 		lock_sock(listener);
1305 
1306 		if (signal_pending(current)) {
1307 			err = sock_intr_errno(timeout);
1308 			goto out;
1309 		} else if (timeout == 0) {
1310 			err = -EAGAIN;
1311 			goto out;
1312 		}
1313 
1314 		prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1315 	}
1316 	finish_wait(sk_sleep(listener), &wait);
1317 
1318 	if (listener->sk_err)
1319 		err = -listener->sk_err;
1320 
1321 	if (connected) {
1322 		listener->sk_ack_backlog--;
1323 
1324 		lock_sock_nested(connected, SINGLE_DEPTH_NESTING);
1325 		vconnected = vsock_sk(connected);
1326 
1327 		/* If the listener socket has received an error, then we should
1328 		 * reject this socket and return.  Note that we simply mark the
1329 		 * socket rejected, drop our reference, and let the cleanup
1330 		 * function handle the cleanup; the fact that we found it in
1331 		 * the listener's accept queue guarantees that the cleanup
1332 		 * function hasn't run yet.
1333 		 */
1334 		if (err) {
1335 			vconnected->rejected = true;
1336 		} else {
1337 			newsock->state = SS_CONNECTED;
1338 			sock_graft(connected, newsock);
1339 		}
1340 
1341 		release_sock(connected);
1342 		sock_put(connected);
1343 	}
1344 
1345 out:
1346 	release_sock(listener);
1347 	return err;
1348 }
1349 
1350 static int vsock_listen(struct socket *sock, int backlog)
1351 {
1352 	int err;
1353 	struct sock *sk;
1354 	struct vsock_sock *vsk;
1355 
1356 	sk = sock->sk;
1357 
1358 	lock_sock(sk);
1359 
1360 	if (sock->type != SOCK_STREAM) {
1361 		err = -EOPNOTSUPP;
1362 		goto out;
1363 	}
1364 
1365 	if (sock->state != SS_UNCONNECTED) {
1366 		err = -EINVAL;
1367 		goto out;
1368 	}
1369 
1370 	vsk = vsock_sk(sk);
1371 
1372 	if (!vsock_addr_bound(&vsk->local_addr)) {
1373 		err = -EINVAL;
1374 		goto out;
1375 	}
1376 
1377 	sk->sk_max_ack_backlog = backlog;
1378 	sk->sk_state = TCP_LISTEN;
1379 
1380 	err = 0;
1381 
1382 out:
1383 	release_sock(sk);
1384 	return err;
1385 }
1386 
1387 static int vsock_stream_setsockopt(struct socket *sock,
1388 				   int level,
1389 				   int optname,
1390 				   char __user *optval,
1391 				   unsigned int optlen)
1392 {
1393 	int err;
1394 	struct sock *sk;
1395 	struct vsock_sock *vsk;
1396 	u64 val;
1397 
1398 	if (level != AF_VSOCK)
1399 		return -ENOPROTOOPT;
1400 
1401 #define COPY_IN(_v)                                       \
1402 	do {						  \
1403 		if (optlen < sizeof(_v)) {		  \
1404 			err = -EINVAL;			  \
1405 			goto exit;			  \
1406 		}					  \
1407 		if (copy_from_user(&_v, optval, sizeof(_v)) != 0) {	\
1408 			err = -EFAULT;					\
1409 			goto exit;					\
1410 		}							\
1411 	} while (0)
1412 
1413 	err = 0;
1414 	sk = sock->sk;
1415 	vsk = vsock_sk(sk);
1416 
1417 	lock_sock(sk);
1418 
1419 	switch (optname) {
1420 	case SO_VM_SOCKETS_BUFFER_SIZE:
1421 		COPY_IN(val);
1422 		transport->set_buffer_size(vsk, val);
1423 		break;
1424 
1425 	case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1426 		COPY_IN(val);
1427 		transport->set_max_buffer_size(vsk, val);
1428 		break;
1429 
1430 	case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1431 		COPY_IN(val);
1432 		transport->set_min_buffer_size(vsk, val);
1433 		break;
1434 
1435 	case SO_VM_SOCKETS_CONNECT_TIMEOUT: {
1436 		struct timeval tv;
1437 		COPY_IN(tv);
1438 		if (tv.tv_sec >= 0 && tv.tv_usec < USEC_PER_SEC &&
1439 		    tv.tv_sec < (MAX_SCHEDULE_TIMEOUT / HZ - 1)) {
1440 			vsk->connect_timeout = tv.tv_sec * HZ +
1441 			    DIV_ROUND_UP(tv.tv_usec, (1000000 / HZ));
1442 			if (vsk->connect_timeout == 0)
1443 				vsk->connect_timeout =
1444 				    VSOCK_DEFAULT_CONNECT_TIMEOUT;
1445 
1446 		} else {
1447 			err = -ERANGE;
1448 		}
1449 		break;
1450 	}
1451 
1452 	default:
1453 		err = -ENOPROTOOPT;
1454 		break;
1455 	}
1456 
1457 #undef COPY_IN
1458 
1459 exit:
1460 	release_sock(sk);
1461 	return err;
1462 }
1463 
1464 static int vsock_stream_getsockopt(struct socket *sock,
1465 				   int level, int optname,
1466 				   char __user *optval,
1467 				   int __user *optlen)
1468 {
1469 	int err;
1470 	int len;
1471 	struct sock *sk;
1472 	struct vsock_sock *vsk;
1473 	u64 val;
1474 
1475 	if (level != AF_VSOCK)
1476 		return -ENOPROTOOPT;
1477 
1478 	err = get_user(len, optlen);
1479 	if (err != 0)
1480 		return err;
1481 
1482 #define COPY_OUT(_v)                            \
1483 	do {					\
1484 		if (len < sizeof(_v))		\
1485 			return -EINVAL;		\
1486 						\
1487 		len = sizeof(_v);		\
1488 		if (copy_to_user(optval, &_v, len) != 0)	\
1489 			return -EFAULT;				\
1490 								\
1491 	} while (0)
1492 
1493 	err = 0;
1494 	sk = sock->sk;
1495 	vsk = vsock_sk(sk);
1496 
1497 	switch (optname) {
1498 	case SO_VM_SOCKETS_BUFFER_SIZE:
1499 		val = transport->get_buffer_size(vsk);
1500 		COPY_OUT(val);
1501 		break;
1502 
1503 	case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1504 		val = transport->get_max_buffer_size(vsk);
1505 		COPY_OUT(val);
1506 		break;
1507 
1508 	case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1509 		val = transport->get_min_buffer_size(vsk);
1510 		COPY_OUT(val);
1511 		break;
1512 
1513 	case SO_VM_SOCKETS_CONNECT_TIMEOUT: {
1514 		struct timeval tv;
1515 		tv.tv_sec = vsk->connect_timeout / HZ;
1516 		tv.tv_usec =
1517 		    (vsk->connect_timeout -
1518 		     tv.tv_sec * HZ) * (1000000 / HZ);
1519 		COPY_OUT(tv);
1520 		break;
1521 	}
1522 	default:
1523 		return -ENOPROTOOPT;
1524 	}
1525 
1526 	err = put_user(len, optlen);
1527 	if (err != 0)
1528 		return -EFAULT;
1529 
1530 #undef COPY_OUT
1531 
1532 	return 0;
1533 }
1534 
1535 static int vsock_stream_sendmsg(struct socket *sock, struct msghdr *msg,
1536 				size_t len)
1537 {
1538 	struct sock *sk;
1539 	struct vsock_sock *vsk;
1540 	ssize_t total_written;
1541 	long timeout;
1542 	int err;
1543 	struct vsock_transport_send_notify_data send_data;
1544 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
1545 
1546 	sk = sock->sk;
1547 	vsk = vsock_sk(sk);
1548 	total_written = 0;
1549 	err = 0;
1550 
1551 	if (msg->msg_flags & MSG_OOB)
1552 		return -EOPNOTSUPP;
1553 
1554 	lock_sock(sk);
1555 
1556 	/* Callers should not provide a destination with stream sockets. */
1557 	if (msg->msg_namelen) {
1558 		err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP;
1559 		goto out;
1560 	}
1561 
1562 	/* Send data only if both sides are not shutdown in the direction. */
1563 	if (sk->sk_shutdown & SEND_SHUTDOWN ||
1564 	    vsk->peer_shutdown & RCV_SHUTDOWN) {
1565 		err = -EPIPE;
1566 		goto out;
1567 	}
1568 
1569 	if (sk->sk_state != TCP_ESTABLISHED ||
1570 	    !vsock_addr_bound(&vsk->local_addr)) {
1571 		err = -ENOTCONN;
1572 		goto out;
1573 	}
1574 
1575 	if (!vsock_addr_bound(&vsk->remote_addr)) {
1576 		err = -EDESTADDRREQ;
1577 		goto out;
1578 	}
1579 
1580 	/* Wait for room in the produce queue to enqueue our user's data. */
1581 	timeout = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1582 
1583 	err = transport->notify_send_init(vsk, &send_data);
1584 	if (err < 0)
1585 		goto out;
1586 
1587 	while (total_written < len) {
1588 		ssize_t written;
1589 
1590 		add_wait_queue(sk_sleep(sk), &wait);
1591 		while (vsock_stream_has_space(vsk) == 0 &&
1592 		       sk->sk_err == 0 &&
1593 		       !(sk->sk_shutdown & SEND_SHUTDOWN) &&
1594 		       !(vsk->peer_shutdown & RCV_SHUTDOWN)) {
1595 
1596 			/* Don't wait for non-blocking sockets. */
1597 			if (timeout == 0) {
1598 				err = -EAGAIN;
1599 				remove_wait_queue(sk_sleep(sk), &wait);
1600 				goto out_err;
1601 			}
1602 
1603 			err = transport->notify_send_pre_block(vsk, &send_data);
1604 			if (err < 0) {
1605 				remove_wait_queue(sk_sleep(sk), &wait);
1606 				goto out_err;
1607 			}
1608 
1609 			release_sock(sk);
1610 			timeout = wait_woken(&wait, TASK_INTERRUPTIBLE, timeout);
1611 			lock_sock(sk);
1612 			if (signal_pending(current)) {
1613 				err = sock_intr_errno(timeout);
1614 				remove_wait_queue(sk_sleep(sk), &wait);
1615 				goto out_err;
1616 			} else if (timeout == 0) {
1617 				err = -EAGAIN;
1618 				remove_wait_queue(sk_sleep(sk), &wait);
1619 				goto out_err;
1620 			}
1621 		}
1622 		remove_wait_queue(sk_sleep(sk), &wait);
1623 
1624 		/* These checks occur both as part of and after the loop
1625 		 * conditional since we need to check before and after
1626 		 * sleeping.
1627 		 */
1628 		if (sk->sk_err) {
1629 			err = -sk->sk_err;
1630 			goto out_err;
1631 		} else if ((sk->sk_shutdown & SEND_SHUTDOWN) ||
1632 			   (vsk->peer_shutdown & RCV_SHUTDOWN)) {
1633 			err = -EPIPE;
1634 			goto out_err;
1635 		}
1636 
1637 		err = transport->notify_send_pre_enqueue(vsk, &send_data);
1638 		if (err < 0)
1639 			goto out_err;
1640 
1641 		/* Note that enqueue will only write as many bytes as are free
1642 		 * in the produce queue, so we don't need to ensure len is
1643 		 * smaller than the queue size.  It is the caller's
1644 		 * responsibility to check how many bytes we were able to send.
1645 		 */
1646 
1647 		written = transport->stream_enqueue(
1648 				vsk, msg,
1649 				len - total_written);
1650 		if (written < 0) {
1651 			err = -ENOMEM;
1652 			goto out_err;
1653 		}
1654 
1655 		total_written += written;
1656 
1657 		err = transport->notify_send_post_enqueue(
1658 				vsk, written, &send_data);
1659 		if (err < 0)
1660 			goto out_err;
1661 
1662 	}
1663 
1664 out_err:
1665 	if (total_written > 0)
1666 		err = total_written;
1667 out:
1668 	release_sock(sk);
1669 	return err;
1670 }
1671 
1672 
1673 static int
1674 vsock_stream_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
1675 		     int flags)
1676 {
1677 	struct sock *sk;
1678 	struct vsock_sock *vsk;
1679 	int err;
1680 	size_t target;
1681 	ssize_t copied;
1682 	long timeout;
1683 	struct vsock_transport_recv_notify_data recv_data;
1684 
1685 	DEFINE_WAIT(wait);
1686 
1687 	sk = sock->sk;
1688 	vsk = vsock_sk(sk);
1689 	err = 0;
1690 
1691 	lock_sock(sk);
1692 
1693 	if (sk->sk_state != TCP_ESTABLISHED) {
1694 		/* Recvmsg is supposed to return 0 if a peer performs an
1695 		 * orderly shutdown. Differentiate between that case and when a
1696 		 * peer has not connected or a local shutdown occured with the
1697 		 * SOCK_DONE flag.
1698 		 */
1699 		if (sock_flag(sk, SOCK_DONE))
1700 			err = 0;
1701 		else
1702 			err = -ENOTCONN;
1703 
1704 		goto out;
1705 	}
1706 
1707 	if (flags & MSG_OOB) {
1708 		err = -EOPNOTSUPP;
1709 		goto out;
1710 	}
1711 
1712 	/* We don't check peer_shutdown flag here since peer may actually shut
1713 	 * down, but there can be data in the queue that a local socket can
1714 	 * receive.
1715 	 */
1716 	if (sk->sk_shutdown & RCV_SHUTDOWN) {
1717 		err = 0;
1718 		goto out;
1719 	}
1720 
1721 	/* It is valid on Linux to pass in a zero-length receive buffer.  This
1722 	 * is not an error.  We may as well bail out now.
1723 	 */
1724 	if (!len) {
1725 		err = 0;
1726 		goto out;
1727 	}
1728 
1729 	/* We must not copy less than target bytes into the user's buffer
1730 	 * before returning successfully, so we wait for the consume queue to
1731 	 * have that much data to consume before dequeueing.  Note that this
1732 	 * makes it impossible to handle cases where target is greater than the
1733 	 * queue size.
1734 	 */
1735 	target = sock_rcvlowat(sk, flags & MSG_WAITALL, len);
1736 	if (target >= transport->stream_rcvhiwat(vsk)) {
1737 		err = -ENOMEM;
1738 		goto out;
1739 	}
1740 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1741 	copied = 0;
1742 
1743 	err = transport->notify_recv_init(vsk, target, &recv_data);
1744 	if (err < 0)
1745 		goto out;
1746 
1747 
1748 	while (1) {
1749 		s64 ready;
1750 
1751 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1752 		ready = vsock_stream_has_data(vsk);
1753 
1754 		if (ready == 0) {
1755 			if (sk->sk_err != 0 ||
1756 			    (sk->sk_shutdown & RCV_SHUTDOWN) ||
1757 			    (vsk->peer_shutdown & SEND_SHUTDOWN)) {
1758 				finish_wait(sk_sleep(sk), &wait);
1759 				break;
1760 			}
1761 			/* Don't wait for non-blocking sockets. */
1762 			if (timeout == 0) {
1763 				err = -EAGAIN;
1764 				finish_wait(sk_sleep(sk), &wait);
1765 				break;
1766 			}
1767 
1768 			err = transport->notify_recv_pre_block(
1769 					vsk, target, &recv_data);
1770 			if (err < 0) {
1771 				finish_wait(sk_sleep(sk), &wait);
1772 				break;
1773 			}
1774 			release_sock(sk);
1775 			timeout = schedule_timeout(timeout);
1776 			lock_sock(sk);
1777 
1778 			if (signal_pending(current)) {
1779 				err = sock_intr_errno(timeout);
1780 				finish_wait(sk_sleep(sk), &wait);
1781 				break;
1782 			} else if (timeout == 0) {
1783 				err = -EAGAIN;
1784 				finish_wait(sk_sleep(sk), &wait);
1785 				break;
1786 			}
1787 		} else {
1788 			ssize_t read;
1789 
1790 			finish_wait(sk_sleep(sk), &wait);
1791 
1792 			if (ready < 0) {
1793 				/* Invalid queue pair content. XXX This should
1794 				* be changed to a connection reset in a later
1795 				* change.
1796 				*/
1797 
1798 				err = -ENOMEM;
1799 				goto out;
1800 			}
1801 
1802 			err = transport->notify_recv_pre_dequeue(
1803 					vsk, target, &recv_data);
1804 			if (err < 0)
1805 				break;
1806 
1807 			read = transport->stream_dequeue(
1808 					vsk, msg,
1809 					len - copied, flags);
1810 			if (read < 0) {
1811 				err = -ENOMEM;
1812 				break;
1813 			}
1814 
1815 			copied += read;
1816 
1817 			err = transport->notify_recv_post_dequeue(
1818 					vsk, target, read,
1819 					!(flags & MSG_PEEK), &recv_data);
1820 			if (err < 0)
1821 				goto out;
1822 
1823 			if (read >= target || flags & MSG_PEEK)
1824 				break;
1825 
1826 			target -= read;
1827 		}
1828 	}
1829 
1830 	if (sk->sk_err)
1831 		err = -sk->sk_err;
1832 	else if (sk->sk_shutdown & RCV_SHUTDOWN)
1833 		err = 0;
1834 
1835 	if (copied > 0)
1836 		err = copied;
1837 
1838 out:
1839 	release_sock(sk);
1840 	return err;
1841 }
1842 
1843 static const struct proto_ops vsock_stream_ops = {
1844 	.family = PF_VSOCK,
1845 	.owner = THIS_MODULE,
1846 	.release = vsock_release,
1847 	.bind = vsock_bind,
1848 	.connect = vsock_stream_connect,
1849 	.socketpair = sock_no_socketpair,
1850 	.accept = vsock_accept,
1851 	.getname = vsock_getname,
1852 	.poll = vsock_poll,
1853 	.ioctl = sock_no_ioctl,
1854 	.listen = vsock_listen,
1855 	.shutdown = vsock_shutdown,
1856 	.setsockopt = vsock_stream_setsockopt,
1857 	.getsockopt = vsock_stream_getsockopt,
1858 	.sendmsg = vsock_stream_sendmsg,
1859 	.recvmsg = vsock_stream_recvmsg,
1860 	.mmap = sock_no_mmap,
1861 	.sendpage = sock_no_sendpage,
1862 };
1863 
1864 static int vsock_create(struct net *net, struct socket *sock,
1865 			int protocol, int kern)
1866 {
1867 	if (!sock)
1868 		return -EINVAL;
1869 
1870 	if (protocol && protocol != PF_VSOCK)
1871 		return -EPROTONOSUPPORT;
1872 
1873 	switch (sock->type) {
1874 	case SOCK_DGRAM:
1875 		sock->ops = &vsock_dgram_ops;
1876 		break;
1877 	case SOCK_STREAM:
1878 		sock->ops = &vsock_stream_ops;
1879 		break;
1880 	default:
1881 		return -ESOCKTNOSUPPORT;
1882 	}
1883 
1884 	sock->state = SS_UNCONNECTED;
1885 
1886 	return __vsock_create(net, sock, NULL, GFP_KERNEL, 0, kern) ? 0 : -ENOMEM;
1887 }
1888 
1889 static const struct net_proto_family vsock_family_ops = {
1890 	.family = AF_VSOCK,
1891 	.create = vsock_create,
1892 	.owner = THIS_MODULE,
1893 };
1894 
1895 static long vsock_dev_do_ioctl(struct file *filp,
1896 			       unsigned int cmd, void __user *ptr)
1897 {
1898 	u32 __user *p = ptr;
1899 	int retval = 0;
1900 
1901 	switch (cmd) {
1902 	case IOCTL_VM_SOCKETS_GET_LOCAL_CID:
1903 		if (put_user(transport->get_local_cid(), p) != 0)
1904 			retval = -EFAULT;
1905 		break;
1906 
1907 	default:
1908 		pr_err("Unknown ioctl %d\n", cmd);
1909 		retval = -EINVAL;
1910 	}
1911 
1912 	return retval;
1913 }
1914 
1915 static long vsock_dev_ioctl(struct file *filp,
1916 			    unsigned int cmd, unsigned long arg)
1917 {
1918 	return vsock_dev_do_ioctl(filp, cmd, (void __user *)arg);
1919 }
1920 
1921 #ifdef CONFIG_COMPAT
1922 static long vsock_dev_compat_ioctl(struct file *filp,
1923 				   unsigned int cmd, unsigned long arg)
1924 {
1925 	return vsock_dev_do_ioctl(filp, cmd, compat_ptr(arg));
1926 }
1927 #endif
1928 
1929 static const struct file_operations vsock_device_ops = {
1930 	.owner		= THIS_MODULE,
1931 	.unlocked_ioctl	= vsock_dev_ioctl,
1932 #ifdef CONFIG_COMPAT
1933 	.compat_ioctl	= vsock_dev_compat_ioctl,
1934 #endif
1935 	.open		= nonseekable_open,
1936 };
1937 
1938 static struct miscdevice vsock_device = {
1939 	.name		= "vsock",
1940 	.fops		= &vsock_device_ops,
1941 };
1942 
1943 int __vsock_core_init(const struct vsock_transport *t, struct module *owner)
1944 {
1945 	int err = mutex_lock_interruptible(&vsock_register_mutex);
1946 
1947 	if (err)
1948 		return err;
1949 
1950 	if (transport) {
1951 		err = -EBUSY;
1952 		goto err_busy;
1953 	}
1954 
1955 	/* Transport must be the owner of the protocol so that it can't
1956 	 * unload while there are open sockets.
1957 	 */
1958 	vsock_proto.owner = owner;
1959 	transport = t;
1960 
1961 	vsock_device.minor = MISC_DYNAMIC_MINOR;
1962 	err = misc_register(&vsock_device);
1963 	if (err) {
1964 		pr_err("Failed to register misc device\n");
1965 		goto err_reset_transport;
1966 	}
1967 
1968 	err = proto_register(&vsock_proto, 1);	/* we want our slab */
1969 	if (err) {
1970 		pr_err("Cannot register vsock protocol\n");
1971 		goto err_deregister_misc;
1972 	}
1973 
1974 	err = sock_register(&vsock_family_ops);
1975 	if (err) {
1976 		pr_err("could not register af_vsock (%d) address family: %d\n",
1977 		       AF_VSOCK, err);
1978 		goto err_unregister_proto;
1979 	}
1980 
1981 	mutex_unlock(&vsock_register_mutex);
1982 	return 0;
1983 
1984 err_unregister_proto:
1985 	proto_unregister(&vsock_proto);
1986 err_deregister_misc:
1987 	misc_deregister(&vsock_device);
1988 err_reset_transport:
1989 	transport = NULL;
1990 err_busy:
1991 	mutex_unlock(&vsock_register_mutex);
1992 	return err;
1993 }
1994 EXPORT_SYMBOL_GPL(__vsock_core_init);
1995 
1996 void vsock_core_exit(void)
1997 {
1998 	mutex_lock(&vsock_register_mutex);
1999 
2000 	misc_deregister(&vsock_device);
2001 	sock_unregister(AF_VSOCK);
2002 	proto_unregister(&vsock_proto);
2003 
2004 	/* We do not want the assignment below re-ordered. */
2005 	mb();
2006 	transport = NULL;
2007 
2008 	mutex_unlock(&vsock_register_mutex);
2009 }
2010 EXPORT_SYMBOL_GPL(vsock_core_exit);
2011 
2012 const struct vsock_transport *vsock_core_get_transport(void)
2013 {
2014 	/* vsock_register_mutex not taken since only the transport uses this
2015 	 * function and only while registered.
2016 	 */
2017 	return transport;
2018 }
2019 EXPORT_SYMBOL_GPL(vsock_core_get_transport);
2020 
2021 module_init(vsock_init_tables);
2022 
2023 MODULE_AUTHOR("VMware, Inc.");
2024 MODULE_DESCRIPTION("VMware Virtual Socket Family");
2025 MODULE_VERSION("1.0.2.0-k");
2026 MODULE_LICENSE("GPL v2");
2027