xref: /openbmc/linux/net/vmw_vsock/af_vsock.c (revision 76ce0265)
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 will use guest->host transport;
425  *  - remote CID > VMADDR_CID_HOST will use host->guest transport;
426  */
427 int vsock_assign_transport(struct vsock_sock *vsk, struct vsock_sock *psk)
428 {
429 	const struct vsock_transport *new_transport;
430 	struct sock *sk = sk_vsock(vsk);
431 	unsigned int remote_cid = vsk->remote_addr.svm_cid;
432 	int ret;
433 
434 	switch (sk->sk_type) {
435 	case SOCK_DGRAM:
436 		new_transport = transport_dgram;
437 		break;
438 	case SOCK_STREAM:
439 		if (vsock_use_local_transport(remote_cid))
440 			new_transport = transport_local;
441 		else if (remote_cid <= VMADDR_CID_HOST)
442 			new_transport = transport_g2h;
443 		else
444 			new_transport = transport_h2g;
445 		break;
446 	default:
447 		return -ESOCKTNOSUPPORT;
448 	}
449 
450 	if (vsk->transport) {
451 		if (vsk->transport == new_transport)
452 			return 0;
453 
454 		/* transport->release() must be called with sock lock acquired.
455 		 * This path can only be taken during vsock_stream_connect(),
456 		 * where we have already held the sock lock.
457 		 * In the other cases, this function is called on a new socket
458 		 * which is not assigned to any transport.
459 		 */
460 		vsk->transport->release(vsk);
461 		vsock_deassign_transport(vsk);
462 	}
463 
464 	/* We increase the module refcnt to prevent the transport unloading
465 	 * while there are open sockets assigned to it.
466 	 */
467 	if (!new_transport || !try_module_get(new_transport->module))
468 		return -ENODEV;
469 
470 	ret = new_transport->init(vsk, psk);
471 	if (ret) {
472 		module_put(new_transport->module);
473 		return ret;
474 	}
475 
476 	vsk->transport = new_transport;
477 
478 	return 0;
479 }
480 EXPORT_SYMBOL_GPL(vsock_assign_transport);
481 
482 bool vsock_find_cid(unsigned int cid)
483 {
484 	if (transport_g2h && cid == transport_g2h->get_local_cid())
485 		return true;
486 
487 	if (transport_h2g && cid == VMADDR_CID_HOST)
488 		return true;
489 
490 	if (transport_local && cid == VMADDR_CID_LOCAL)
491 		return true;
492 
493 	return false;
494 }
495 EXPORT_SYMBOL_GPL(vsock_find_cid);
496 
497 static struct sock *vsock_dequeue_accept(struct sock *listener)
498 {
499 	struct vsock_sock *vlistener;
500 	struct vsock_sock *vconnected;
501 
502 	vlistener = vsock_sk(listener);
503 
504 	if (list_empty(&vlistener->accept_queue))
505 		return NULL;
506 
507 	vconnected = list_entry(vlistener->accept_queue.next,
508 				struct vsock_sock, accept_queue);
509 
510 	list_del_init(&vconnected->accept_queue);
511 	sock_put(listener);
512 	/* The caller will need a reference on the connected socket so we let
513 	 * it call sock_put().
514 	 */
515 
516 	return sk_vsock(vconnected);
517 }
518 
519 static bool vsock_is_accept_queue_empty(struct sock *sk)
520 {
521 	struct vsock_sock *vsk = vsock_sk(sk);
522 	return list_empty(&vsk->accept_queue);
523 }
524 
525 static bool vsock_is_pending(struct sock *sk)
526 {
527 	struct vsock_sock *vsk = vsock_sk(sk);
528 	return !list_empty(&vsk->pending_links);
529 }
530 
531 static int vsock_send_shutdown(struct sock *sk, int mode)
532 {
533 	struct vsock_sock *vsk = vsock_sk(sk);
534 
535 	if (!vsk->transport)
536 		return -ENODEV;
537 
538 	return vsk->transport->shutdown(vsk, mode);
539 }
540 
541 static void vsock_pending_work(struct work_struct *work)
542 {
543 	struct sock *sk;
544 	struct sock *listener;
545 	struct vsock_sock *vsk;
546 	bool cleanup;
547 
548 	vsk = container_of(work, struct vsock_sock, pending_work.work);
549 	sk = sk_vsock(vsk);
550 	listener = vsk->listener;
551 	cleanup = true;
552 
553 	lock_sock(listener);
554 	lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
555 
556 	if (vsock_is_pending(sk)) {
557 		vsock_remove_pending(listener, sk);
558 
559 		sk_acceptq_removed(listener);
560 	} else if (!vsk->rejected) {
561 		/* We are not on the pending list and accept() did not reject
562 		 * us, so we must have been accepted by our user process.  We
563 		 * just need to drop our references to the sockets and be on
564 		 * our way.
565 		 */
566 		cleanup = false;
567 		goto out;
568 	}
569 
570 	/* We need to remove ourself from the global connected sockets list so
571 	 * incoming packets can't find this socket, and to reduce the reference
572 	 * count.
573 	 */
574 	vsock_remove_connected(vsk);
575 
576 	sk->sk_state = TCP_CLOSE;
577 
578 out:
579 	release_sock(sk);
580 	release_sock(listener);
581 	if (cleanup)
582 		sock_put(sk);
583 
584 	sock_put(sk);
585 	sock_put(listener);
586 }
587 
588 /**** SOCKET OPERATIONS ****/
589 
590 static int __vsock_bind_stream(struct vsock_sock *vsk,
591 			       struct sockaddr_vm *addr)
592 {
593 	static u32 port;
594 	struct sockaddr_vm new_addr;
595 
596 	if (!port)
597 		port = LAST_RESERVED_PORT + 1 +
598 			prandom_u32_max(U32_MAX - LAST_RESERVED_PORT);
599 
600 	vsock_addr_init(&new_addr, addr->svm_cid, addr->svm_port);
601 
602 	if (addr->svm_port == VMADDR_PORT_ANY) {
603 		bool found = false;
604 		unsigned int i;
605 
606 		for (i = 0; i < MAX_PORT_RETRIES; i++) {
607 			if (port <= LAST_RESERVED_PORT)
608 				port = LAST_RESERVED_PORT + 1;
609 
610 			new_addr.svm_port = port++;
611 
612 			if (!__vsock_find_bound_socket(&new_addr)) {
613 				found = true;
614 				break;
615 			}
616 		}
617 
618 		if (!found)
619 			return -EADDRNOTAVAIL;
620 	} else {
621 		/* If port is in reserved range, ensure caller
622 		 * has necessary privileges.
623 		 */
624 		if (addr->svm_port <= LAST_RESERVED_PORT &&
625 		    !capable(CAP_NET_BIND_SERVICE)) {
626 			return -EACCES;
627 		}
628 
629 		if (__vsock_find_bound_socket(&new_addr))
630 			return -EADDRINUSE;
631 	}
632 
633 	vsock_addr_init(&vsk->local_addr, new_addr.svm_cid, new_addr.svm_port);
634 
635 	/* Remove stream sockets from the unbound list and add them to the hash
636 	 * table for easy lookup by its address.  The unbound list is simply an
637 	 * extra entry at the end of the hash table, a trick used by AF_UNIX.
638 	 */
639 	__vsock_remove_bound(vsk);
640 	__vsock_insert_bound(vsock_bound_sockets(&vsk->local_addr), vsk);
641 
642 	return 0;
643 }
644 
645 static int __vsock_bind_dgram(struct vsock_sock *vsk,
646 			      struct sockaddr_vm *addr)
647 {
648 	return vsk->transport->dgram_bind(vsk, addr);
649 }
650 
651 static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr)
652 {
653 	struct vsock_sock *vsk = vsock_sk(sk);
654 	int retval;
655 
656 	/* First ensure this socket isn't already bound. */
657 	if (vsock_addr_bound(&vsk->local_addr))
658 		return -EINVAL;
659 
660 	/* Now bind to the provided address or select appropriate values if
661 	 * none are provided (VMADDR_CID_ANY and VMADDR_PORT_ANY).  Note that
662 	 * like AF_INET prevents binding to a non-local IP address (in most
663 	 * cases), we only allow binding to a local CID.
664 	 */
665 	if (addr->svm_cid != VMADDR_CID_ANY && !vsock_find_cid(addr->svm_cid))
666 		return -EADDRNOTAVAIL;
667 
668 	switch (sk->sk_socket->type) {
669 	case SOCK_STREAM:
670 		spin_lock_bh(&vsock_table_lock);
671 		retval = __vsock_bind_stream(vsk, addr);
672 		spin_unlock_bh(&vsock_table_lock);
673 		break;
674 
675 	case SOCK_DGRAM:
676 		retval = __vsock_bind_dgram(vsk, addr);
677 		break;
678 
679 	default:
680 		retval = -EINVAL;
681 		break;
682 	}
683 
684 	return retval;
685 }
686 
687 static void vsock_connect_timeout(struct work_struct *work);
688 
689 static struct sock *__vsock_create(struct net *net,
690 				   struct socket *sock,
691 				   struct sock *parent,
692 				   gfp_t priority,
693 				   unsigned short type,
694 				   int kern)
695 {
696 	struct sock *sk;
697 	struct vsock_sock *psk;
698 	struct vsock_sock *vsk;
699 
700 	sk = sk_alloc(net, AF_VSOCK, priority, &vsock_proto, kern);
701 	if (!sk)
702 		return NULL;
703 
704 	sock_init_data(sock, sk);
705 
706 	/* sk->sk_type is normally set in sock_init_data, but only if sock is
707 	 * non-NULL. We make sure that our sockets always have a type by
708 	 * setting it here if needed.
709 	 */
710 	if (!sock)
711 		sk->sk_type = type;
712 
713 	vsk = vsock_sk(sk);
714 	vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
715 	vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
716 
717 	sk->sk_destruct = vsock_sk_destruct;
718 	sk->sk_backlog_rcv = vsock_queue_rcv_skb;
719 	sock_reset_flag(sk, SOCK_DONE);
720 
721 	INIT_LIST_HEAD(&vsk->bound_table);
722 	INIT_LIST_HEAD(&vsk->connected_table);
723 	vsk->listener = NULL;
724 	INIT_LIST_HEAD(&vsk->pending_links);
725 	INIT_LIST_HEAD(&vsk->accept_queue);
726 	vsk->rejected = false;
727 	vsk->sent_request = false;
728 	vsk->ignore_connecting_rst = false;
729 	vsk->peer_shutdown = 0;
730 	INIT_DELAYED_WORK(&vsk->connect_work, vsock_connect_timeout);
731 	INIT_DELAYED_WORK(&vsk->pending_work, vsock_pending_work);
732 
733 	psk = parent ? vsock_sk(parent) : NULL;
734 	if (parent) {
735 		vsk->trusted = psk->trusted;
736 		vsk->owner = get_cred(psk->owner);
737 		vsk->connect_timeout = psk->connect_timeout;
738 		vsk->buffer_size = psk->buffer_size;
739 		vsk->buffer_min_size = psk->buffer_min_size;
740 		vsk->buffer_max_size = psk->buffer_max_size;
741 	} else {
742 		vsk->trusted = capable(CAP_NET_ADMIN);
743 		vsk->owner = get_current_cred();
744 		vsk->connect_timeout = VSOCK_DEFAULT_CONNECT_TIMEOUT;
745 		vsk->buffer_size = VSOCK_DEFAULT_BUFFER_SIZE;
746 		vsk->buffer_min_size = VSOCK_DEFAULT_BUFFER_MIN_SIZE;
747 		vsk->buffer_max_size = VSOCK_DEFAULT_BUFFER_MAX_SIZE;
748 	}
749 
750 	return sk;
751 }
752 
753 static void __vsock_release(struct sock *sk, int level)
754 {
755 	if (sk) {
756 		struct sock *pending;
757 		struct vsock_sock *vsk;
758 
759 		vsk = vsock_sk(sk);
760 		pending = NULL;	/* Compiler warning. */
761 
762 		/* When "level" is SINGLE_DEPTH_NESTING, use the nested
763 		 * version to avoid the warning "possible recursive locking
764 		 * detected". When "level" is 0, lock_sock_nested(sk, level)
765 		 * is the same as lock_sock(sk).
766 		 */
767 		lock_sock_nested(sk, level);
768 
769 		if (vsk->transport)
770 			vsk->transport->release(vsk);
771 		else if (sk->sk_type == SOCK_STREAM)
772 			vsock_remove_sock(vsk);
773 
774 		sock_orphan(sk);
775 		sk->sk_shutdown = SHUTDOWN_MASK;
776 
777 		skb_queue_purge(&sk->sk_receive_queue);
778 
779 		/* Clean up any sockets that never were accepted. */
780 		while ((pending = vsock_dequeue_accept(sk)) != NULL) {
781 			__vsock_release(pending, SINGLE_DEPTH_NESTING);
782 			sock_put(pending);
783 		}
784 
785 		release_sock(sk);
786 		sock_put(sk);
787 	}
788 }
789 
790 static void vsock_sk_destruct(struct sock *sk)
791 {
792 	struct vsock_sock *vsk = vsock_sk(sk);
793 
794 	vsock_deassign_transport(vsk);
795 
796 	/* When clearing these addresses, there's no need to set the family and
797 	 * possibly register the address family with the kernel.
798 	 */
799 	vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
800 	vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
801 
802 	put_cred(vsk->owner);
803 }
804 
805 static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
806 {
807 	int err;
808 
809 	err = sock_queue_rcv_skb(sk, skb);
810 	if (err)
811 		kfree_skb(skb);
812 
813 	return err;
814 }
815 
816 struct sock *vsock_create_connected(struct sock *parent)
817 {
818 	return __vsock_create(sock_net(parent), NULL, parent, GFP_KERNEL,
819 			      parent->sk_type, 0);
820 }
821 EXPORT_SYMBOL_GPL(vsock_create_connected);
822 
823 s64 vsock_stream_has_data(struct vsock_sock *vsk)
824 {
825 	return vsk->transport->stream_has_data(vsk);
826 }
827 EXPORT_SYMBOL_GPL(vsock_stream_has_data);
828 
829 s64 vsock_stream_has_space(struct vsock_sock *vsk)
830 {
831 	return vsk->transport->stream_has_space(vsk);
832 }
833 EXPORT_SYMBOL_GPL(vsock_stream_has_space);
834 
835 static int vsock_release(struct socket *sock)
836 {
837 	__vsock_release(sock->sk, 0);
838 	sock->sk = NULL;
839 	sock->state = SS_FREE;
840 
841 	return 0;
842 }
843 
844 static int
845 vsock_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
846 {
847 	int err;
848 	struct sock *sk;
849 	struct sockaddr_vm *vm_addr;
850 
851 	sk = sock->sk;
852 
853 	if (vsock_addr_cast(addr, addr_len, &vm_addr) != 0)
854 		return -EINVAL;
855 
856 	lock_sock(sk);
857 	err = __vsock_bind(sk, vm_addr);
858 	release_sock(sk);
859 
860 	return err;
861 }
862 
863 static int vsock_getname(struct socket *sock,
864 			 struct sockaddr *addr, int peer)
865 {
866 	int err;
867 	struct sock *sk;
868 	struct vsock_sock *vsk;
869 	struct sockaddr_vm *vm_addr;
870 
871 	sk = sock->sk;
872 	vsk = vsock_sk(sk);
873 	err = 0;
874 
875 	lock_sock(sk);
876 
877 	if (peer) {
878 		if (sock->state != SS_CONNECTED) {
879 			err = -ENOTCONN;
880 			goto out;
881 		}
882 		vm_addr = &vsk->remote_addr;
883 	} else {
884 		vm_addr = &vsk->local_addr;
885 	}
886 
887 	if (!vm_addr) {
888 		err = -EINVAL;
889 		goto out;
890 	}
891 
892 	/* sys_getsockname() and sys_getpeername() pass us a
893 	 * MAX_SOCK_ADDR-sized buffer and don't set addr_len.  Unfortunately
894 	 * that macro is defined in socket.c instead of .h, so we hardcode its
895 	 * value here.
896 	 */
897 	BUILD_BUG_ON(sizeof(*vm_addr) > 128);
898 	memcpy(addr, vm_addr, sizeof(*vm_addr));
899 	err = sizeof(*vm_addr);
900 
901 out:
902 	release_sock(sk);
903 	return err;
904 }
905 
906 static int vsock_shutdown(struct socket *sock, int mode)
907 {
908 	int err;
909 	struct sock *sk;
910 
911 	/* User level uses SHUT_RD (0) and SHUT_WR (1), but the kernel uses
912 	 * RCV_SHUTDOWN (1) and SEND_SHUTDOWN (2), so we must increment mode
913 	 * here like the other address families do.  Note also that the
914 	 * increment makes SHUT_RDWR (2) into RCV_SHUTDOWN | SEND_SHUTDOWN (3),
915 	 * which is what we want.
916 	 */
917 	mode++;
918 
919 	if ((mode & ~SHUTDOWN_MASK) || !mode)
920 		return -EINVAL;
921 
922 	/* If this is a STREAM socket and it is not connected then bail out
923 	 * immediately.  If it is a DGRAM socket then we must first kick the
924 	 * socket so that it wakes up from any sleeping calls, for example
925 	 * recv(), and then afterwards return the error.
926 	 */
927 
928 	sk = sock->sk;
929 	if (sock->state == SS_UNCONNECTED) {
930 		err = -ENOTCONN;
931 		if (sk->sk_type == SOCK_STREAM)
932 			return err;
933 	} else {
934 		sock->state = SS_DISCONNECTING;
935 		err = 0;
936 	}
937 
938 	/* Receive and send shutdowns are treated alike. */
939 	mode = mode & (RCV_SHUTDOWN | SEND_SHUTDOWN);
940 	if (mode) {
941 		lock_sock(sk);
942 		sk->sk_shutdown |= mode;
943 		sk->sk_state_change(sk);
944 		release_sock(sk);
945 
946 		if (sk->sk_type == SOCK_STREAM) {
947 			sock_reset_flag(sk, SOCK_DONE);
948 			vsock_send_shutdown(sk, mode);
949 		}
950 	}
951 
952 	return err;
953 }
954 
955 static __poll_t vsock_poll(struct file *file, struct socket *sock,
956 			       poll_table *wait)
957 {
958 	struct sock *sk;
959 	__poll_t mask;
960 	struct vsock_sock *vsk;
961 
962 	sk = sock->sk;
963 	vsk = vsock_sk(sk);
964 
965 	poll_wait(file, sk_sleep(sk), wait);
966 	mask = 0;
967 
968 	if (sk->sk_err)
969 		/* Signify that there has been an error on this socket. */
970 		mask |= EPOLLERR;
971 
972 	/* INET sockets treat local write shutdown and peer write shutdown as a
973 	 * case of EPOLLHUP set.
974 	 */
975 	if ((sk->sk_shutdown == SHUTDOWN_MASK) ||
976 	    ((sk->sk_shutdown & SEND_SHUTDOWN) &&
977 	     (vsk->peer_shutdown & SEND_SHUTDOWN))) {
978 		mask |= EPOLLHUP;
979 	}
980 
981 	if (sk->sk_shutdown & RCV_SHUTDOWN ||
982 	    vsk->peer_shutdown & SEND_SHUTDOWN) {
983 		mask |= EPOLLRDHUP;
984 	}
985 
986 	if (sock->type == SOCK_DGRAM) {
987 		/* For datagram sockets we can read if there is something in
988 		 * the queue and write as long as the socket isn't shutdown for
989 		 * sending.
990 		 */
991 		if (!skb_queue_empty_lockless(&sk->sk_receive_queue) ||
992 		    (sk->sk_shutdown & RCV_SHUTDOWN)) {
993 			mask |= EPOLLIN | EPOLLRDNORM;
994 		}
995 
996 		if (!(sk->sk_shutdown & SEND_SHUTDOWN))
997 			mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND;
998 
999 	} else if (sock->type == SOCK_STREAM) {
1000 		const struct vsock_transport *transport = vsk->transport;
1001 		lock_sock(sk);
1002 
1003 		/* Listening sockets that have connections in their accept
1004 		 * queue can be read.
1005 		 */
1006 		if (sk->sk_state == TCP_LISTEN
1007 		    && !vsock_is_accept_queue_empty(sk))
1008 			mask |= EPOLLIN | EPOLLRDNORM;
1009 
1010 		/* If there is something in the queue then we can read. */
1011 		if (transport && transport->stream_is_active(vsk) &&
1012 		    !(sk->sk_shutdown & RCV_SHUTDOWN)) {
1013 			bool data_ready_now = false;
1014 			int ret = transport->notify_poll_in(
1015 					vsk, 1, &data_ready_now);
1016 			if (ret < 0) {
1017 				mask |= EPOLLERR;
1018 			} else {
1019 				if (data_ready_now)
1020 					mask |= EPOLLIN | EPOLLRDNORM;
1021 
1022 			}
1023 		}
1024 
1025 		/* Sockets whose connections have been closed, reset, or
1026 		 * terminated should also be considered read, and we check the
1027 		 * shutdown flag for that.
1028 		 */
1029 		if (sk->sk_shutdown & RCV_SHUTDOWN ||
1030 		    vsk->peer_shutdown & SEND_SHUTDOWN) {
1031 			mask |= EPOLLIN | EPOLLRDNORM;
1032 		}
1033 
1034 		/* Connected sockets that can produce data can be written. */
1035 		if (sk->sk_state == TCP_ESTABLISHED) {
1036 			if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1037 				bool space_avail_now = false;
1038 				int ret = transport->notify_poll_out(
1039 						vsk, 1, &space_avail_now);
1040 				if (ret < 0) {
1041 					mask |= EPOLLERR;
1042 				} else {
1043 					if (space_avail_now)
1044 						/* Remove EPOLLWRBAND since INET
1045 						 * sockets are not setting it.
1046 						 */
1047 						mask |= EPOLLOUT | EPOLLWRNORM;
1048 
1049 				}
1050 			}
1051 		}
1052 
1053 		/* Simulate INET socket poll behaviors, which sets
1054 		 * EPOLLOUT|EPOLLWRNORM when peer is closed and nothing to read,
1055 		 * but local send is not shutdown.
1056 		 */
1057 		if (sk->sk_state == TCP_CLOSE || sk->sk_state == TCP_CLOSING) {
1058 			if (!(sk->sk_shutdown & SEND_SHUTDOWN))
1059 				mask |= EPOLLOUT | EPOLLWRNORM;
1060 
1061 		}
1062 
1063 		release_sock(sk);
1064 	}
1065 
1066 	return mask;
1067 }
1068 
1069 static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg,
1070 			       size_t len)
1071 {
1072 	int err;
1073 	struct sock *sk;
1074 	struct vsock_sock *vsk;
1075 	struct sockaddr_vm *remote_addr;
1076 	const struct vsock_transport *transport;
1077 
1078 	if (msg->msg_flags & MSG_OOB)
1079 		return -EOPNOTSUPP;
1080 
1081 	/* For now, MSG_DONTWAIT is always assumed... */
1082 	err = 0;
1083 	sk = sock->sk;
1084 	vsk = vsock_sk(sk);
1085 	transport = vsk->transport;
1086 
1087 	lock_sock(sk);
1088 
1089 	err = vsock_auto_bind(vsk);
1090 	if (err)
1091 		goto out;
1092 
1093 
1094 	/* If the provided message contains an address, use that.  Otherwise
1095 	 * fall back on the socket's remote handle (if it has been connected).
1096 	 */
1097 	if (msg->msg_name &&
1098 	    vsock_addr_cast(msg->msg_name, msg->msg_namelen,
1099 			    &remote_addr) == 0) {
1100 		/* Ensure this address is of the right type and is a valid
1101 		 * destination.
1102 		 */
1103 
1104 		if (remote_addr->svm_cid == VMADDR_CID_ANY)
1105 			remote_addr->svm_cid = transport->get_local_cid();
1106 
1107 		if (!vsock_addr_bound(remote_addr)) {
1108 			err = -EINVAL;
1109 			goto out;
1110 		}
1111 	} else if (sock->state == SS_CONNECTED) {
1112 		remote_addr = &vsk->remote_addr;
1113 
1114 		if (remote_addr->svm_cid == VMADDR_CID_ANY)
1115 			remote_addr->svm_cid = transport->get_local_cid();
1116 
1117 		/* XXX Should connect() or this function ensure remote_addr is
1118 		 * bound?
1119 		 */
1120 		if (!vsock_addr_bound(&vsk->remote_addr)) {
1121 			err = -EINVAL;
1122 			goto out;
1123 		}
1124 	} else {
1125 		err = -EINVAL;
1126 		goto out;
1127 	}
1128 
1129 	if (!transport->dgram_allow(remote_addr->svm_cid,
1130 				    remote_addr->svm_port)) {
1131 		err = -EINVAL;
1132 		goto out;
1133 	}
1134 
1135 	err = transport->dgram_enqueue(vsk, remote_addr, msg, len);
1136 
1137 out:
1138 	release_sock(sk);
1139 	return err;
1140 }
1141 
1142 static int vsock_dgram_connect(struct socket *sock,
1143 			       struct sockaddr *addr, int addr_len, int flags)
1144 {
1145 	int err;
1146 	struct sock *sk;
1147 	struct vsock_sock *vsk;
1148 	struct sockaddr_vm *remote_addr;
1149 
1150 	sk = sock->sk;
1151 	vsk = vsock_sk(sk);
1152 
1153 	err = vsock_addr_cast(addr, addr_len, &remote_addr);
1154 	if (err == -EAFNOSUPPORT && remote_addr->svm_family == AF_UNSPEC) {
1155 		lock_sock(sk);
1156 		vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY,
1157 				VMADDR_PORT_ANY);
1158 		sock->state = SS_UNCONNECTED;
1159 		release_sock(sk);
1160 		return 0;
1161 	} else if (err != 0)
1162 		return -EINVAL;
1163 
1164 	lock_sock(sk);
1165 
1166 	err = vsock_auto_bind(vsk);
1167 	if (err)
1168 		goto out;
1169 
1170 	if (!vsk->transport->dgram_allow(remote_addr->svm_cid,
1171 					 remote_addr->svm_port)) {
1172 		err = -EINVAL;
1173 		goto out;
1174 	}
1175 
1176 	memcpy(&vsk->remote_addr, remote_addr, sizeof(vsk->remote_addr));
1177 	sock->state = SS_CONNECTED;
1178 
1179 out:
1180 	release_sock(sk);
1181 	return err;
1182 }
1183 
1184 static int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg,
1185 			       size_t len, int flags)
1186 {
1187 	struct vsock_sock *vsk = vsock_sk(sock->sk);
1188 
1189 	return vsk->transport->dgram_dequeue(vsk, msg, len, flags);
1190 }
1191 
1192 static const struct proto_ops vsock_dgram_ops = {
1193 	.family = PF_VSOCK,
1194 	.owner = THIS_MODULE,
1195 	.release = vsock_release,
1196 	.bind = vsock_bind,
1197 	.connect = vsock_dgram_connect,
1198 	.socketpair = sock_no_socketpair,
1199 	.accept = sock_no_accept,
1200 	.getname = vsock_getname,
1201 	.poll = vsock_poll,
1202 	.ioctl = sock_no_ioctl,
1203 	.listen = sock_no_listen,
1204 	.shutdown = vsock_shutdown,
1205 	.setsockopt = sock_no_setsockopt,
1206 	.getsockopt = sock_no_getsockopt,
1207 	.sendmsg = vsock_dgram_sendmsg,
1208 	.recvmsg = vsock_dgram_recvmsg,
1209 	.mmap = sock_no_mmap,
1210 	.sendpage = sock_no_sendpage,
1211 };
1212 
1213 static int vsock_transport_cancel_pkt(struct vsock_sock *vsk)
1214 {
1215 	const struct vsock_transport *transport = vsk->transport;
1216 
1217 	if (!transport->cancel_pkt)
1218 		return -EOPNOTSUPP;
1219 
1220 	return transport->cancel_pkt(vsk);
1221 }
1222 
1223 static void vsock_connect_timeout(struct work_struct *work)
1224 {
1225 	struct sock *sk;
1226 	struct vsock_sock *vsk;
1227 	int cancel = 0;
1228 
1229 	vsk = container_of(work, struct vsock_sock, connect_work.work);
1230 	sk = sk_vsock(vsk);
1231 
1232 	lock_sock(sk);
1233 	if (sk->sk_state == TCP_SYN_SENT &&
1234 	    (sk->sk_shutdown != SHUTDOWN_MASK)) {
1235 		sk->sk_state = TCP_CLOSE;
1236 		sk->sk_err = ETIMEDOUT;
1237 		sk->sk_error_report(sk);
1238 		cancel = 1;
1239 	}
1240 	release_sock(sk);
1241 	if (cancel)
1242 		vsock_transport_cancel_pkt(vsk);
1243 
1244 	sock_put(sk);
1245 }
1246 
1247 static int vsock_stream_connect(struct socket *sock, struct sockaddr *addr,
1248 				int addr_len, int flags)
1249 {
1250 	int err;
1251 	struct sock *sk;
1252 	struct vsock_sock *vsk;
1253 	const struct vsock_transport *transport;
1254 	struct sockaddr_vm *remote_addr;
1255 	long timeout;
1256 	DEFINE_WAIT(wait);
1257 
1258 	err = 0;
1259 	sk = sock->sk;
1260 	vsk = vsock_sk(sk);
1261 
1262 	lock_sock(sk);
1263 
1264 	/* XXX AF_UNSPEC should make us disconnect like AF_INET. */
1265 	switch (sock->state) {
1266 	case SS_CONNECTED:
1267 		err = -EISCONN;
1268 		goto out;
1269 	case SS_DISCONNECTING:
1270 		err = -EINVAL;
1271 		goto out;
1272 	case SS_CONNECTING:
1273 		/* This continues on so we can move sock into the SS_CONNECTED
1274 		 * state once the connection has completed (at which point err
1275 		 * will be set to zero also).  Otherwise, we will either wait
1276 		 * for the connection or return -EALREADY should this be a
1277 		 * non-blocking call.
1278 		 */
1279 		err = -EALREADY;
1280 		break;
1281 	default:
1282 		if ((sk->sk_state == TCP_LISTEN) ||
1283 		    vsock_addr_cast(addr, addr_len, &remote_addr) != 0) {
1284 			err = -EINVAL;
1285 			goto out;
1286 		}
1287 
1288 		/* Set the remote address that we are connecting to. */
1289 		memcpy(&vsk->remote_addr, remote_addr,
1290 		       sizeof(vsk->remote_addr));
1291 
1292 		err = vsock_assign_transport(vsk, NULL);
1293 		if (err)
1294 			goto out;
1295 
1296 		transport = vsk->transport;
1297 
1298 		/* The hypervisor and well-known contexts do not have socket
1299 		 * endpoints.
1300 		 */
1301 		if (!transport ||
1302 		    !transport->stream_allow(remote_addr->svm_cid,
1303 					     remote_addr->svm_port)) {
1304 			err = -ENETUNREACH;
1305 			goto out;
1306 		}
1307 
1308 		err = vsock_auto_bind(vsk);
1309 		if (err)
1310 			goto out;
1311 
1312 		sk->sk_state = TCP_SYN_SENT;
1313 
1314 		err = transport->connect(vsk);
1315 		if (err < 0)
1316 			goto out;
1317 
1318 		/* Mark sock as connecting and set the error code to in
1319 		 * progress in case this is a non-blocking connect.
1320 		 */
1321 		sock->state = SS_CONNECTING;
1322 		err = -EINPROGRESS;
1323 	}
1324 
1325 	/* The receive path will handle all communication until we are able to
1326 	 * enter the connected state.  Here we wait for the connection to be
1327 	 * completed or a notification of an error.
1328 	 */
1329 	timeout = vsk->connect_timeout;
1330 	prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1331 
1332 	while (sk->sk_state != TCP_ESTABLISHED && sk->sk_err == 0) {
1333 		if (flags & O_NONBLOCK) {
1334 			/* If we're not going to block, we schedule a timeout
1335 			 * function to generate a timeout on the connection
1336 			 * attempt, in case the peer doesn't respond in a
1337 			 * timely manner. We hold on to the socket until the
1338 			 * timeout fires.
1339 			 */
1340 			sock_hold(sk);
1341 			schedule_delayed_work(&vsk->connect_work, timeout);
1342 
1343 			/* Skip ahead to preserve error code set above. */
1344 			goto out_wait;
1345 		}
1346 
1347 		release_sock(sk);
1348 		timeout = schedule_timeout(timeout);
1349 		lock_sock(sk);
1350 
1351 		if (signal_pending(current)) {
1352 			err = sock_intr_errno(timeout);
1353 			sk->sk_state = TCP_CLOSE;
1354 			sock->state = SS_UNCONNECTED;
1355 			vsock_transport_cancel_pkt(vsk);
1356 			goto out_wait;
1357 		} else if (timeout == 0) {
1358 			err = -ETIMEDOUT;
1359 			sk->sk_state = TCP_CLOSE;
1360 			sock->state = SS_UNCONNECTED;
1361 			vsock_transport_cancel_pkt(vsk);
1362 			goto out_wait;
1363 		}
1364 
1365 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1366 	}
1367 
1368 	if (sk->sk_err) {
1369 		err = -sk->sk_err;
1370 		sk->sk_state = TCP_CLOSE;
1371 		sock->state = SS_UNCONNECTED;
1372 	} else {
1373 		err = 0;
1374 	}
1375 
1376 out_wait:
1377 	finish_wait(sk_sleep(sk), &wait);
1378 out:
1379 	release_sock(sk);
1380 	return err;
1381 }
1382 
1383 static int vsock_accept(struct socket *sock, struct socket *newsock, int flags,
1384 			bool kern)
1385 {
1386 	struct sock *listener;
1387 	int err;
1388 	struct sock *connected;
1389 	struct vsock_sock *vconnected;
1390 	long timeout;
1391 	DEFINE_WAIT(wait);
1392 
1393 	err = 0;
1394 	listener = sock->sk;
1395 
1396 	lock_sock(listener);
1397 
1398 	if (sock->type != SOCK_STREAM) {
1399 		err = -EOPNOTSUPP;
1400 		goto out;
1401 	}
1402 
1403 	if (listener->sk_state != TCP_LISTEN) {
1404 		err = -EINVAL;
1405 		goto out;
1406 	}
1407 
1408 	/* Wait for children sockets to appear; these are the new sockets
1409 	 * created upon connection establishment.
1410 	 */
1411 	timeout = sock_sndtimeo(listener, flags & O_NONBLOCK);
1412 	prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1413 
1414 	while ((connected = vsock_dequeue_accept(listener)) == NULL &&
1415 	       listener->sk_err == 0) {
1416 		release_sock(listener);
1417 		timeout = schedule_timeout(timeout);
1418 		finish_wait(sk_sleep(listener), &wait);
1419 		lock_sock(listener);
1420 
1421 		if (signal_pending(current)) {
1422 			err = sock_intr_errno(timeout);
1423 			goto out;
1424 		} else if (timeout == 0) {
1425 			err = -EAGAIN;
1426 			goto out;
1427 		}
1428 
1429 		prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1430 	}
1431 	finish_wait(sk_sleep(listener), &wait);
1432 
1433 	if (listener->sk_err)
1434 		err = -listener->sk_err;
1435 
1436 	if (connected) {
1437 		sk_acceptq_removed(listener);
1438 
1439 		lock_sock_nested(connected, SINGLE_DEPTH_NESTING);
1440 		vconnected = vsock_sk(connected);
1441 
1442 		/* If the listener socket has received an error, then we should
1443 		 * reject this socket and return.  Note that we simply mark the
1444 		 * socket rejected, drop our reference, and let the cleanup
1445 		 * function handle the cleanup; the fact that we found it in
1446 		 * the listener's accept queue guarantees that the cleanup
1447 		 * function hasn't run yet.
1448 		 */
1449 		if (err) {
1450 			vconnected->rejected = true;
1451 		} else {
1452 			newsock->state = SS_CONNECTED;
1453 			sock_graft(connected, newsock);
1454 		}
1455 
1456 		release_sock(connected);
1457 		sock_put(connected);
1458 	}
1459 
1460 out:
1461 	release_sock(listener);
1462 	return err;
1463 }
1464 
1465 static int vsock_listen(struct socket *sock, int backlog)
1466 {
1467 	int err;
1468 	struct sock *sk;
1469 	struct vsock_sock *vsk;
1470 
1471 	sk = sock->sk;
1472 
1473 	lock_sock(sk);
1474 
1475 	if (sock->type != SOCK_STREAM) {
1476 		err = -EOPNOTSUPP;
1477 		goto out;
1478 	}
1479 
1480 	if (sock->state != SS_UNCONNECTED) {
1481 		err = -EINVAL;
1482 		goto out;
1483 	}
1484 
1485 	vsk = vsock_sk(sk);
1486 
1487 	if (!vsock_addr_bound(&vsk->local_addr)) {
1488 		err = -EINVAL;
1489 		goto out;
1490 	}
1491 
1492 	sk->sk_max_ack_backlog = backlog;
1493 	sk->sk_state = TCP_LISTEN;
1494 
1495 	err = 0;
1496 
1497 out:
1498 	release_sock(sk);
1499 	return err;
1500 }
1501 
1502 static void vsock_update_buffer_size(struct vsock_sock *vsk,
1503 				     const struct vsock_transport *transport,
1504 				     u64 val)
1505 {
1506 	if (val > vsk->buffer_max_size)
1507 		val = vsk->buffer_max_size;
1508 
1509 	if (val < vsk->buffer_min_size)
1510 		val = vsk->buffer_min_size;
1511 
1512 	if (val != vsk->buffer_size &&
1513 	    transport && transport->notify_buffer_size)
1514 		transport->notify_buffer_size(vsk, &val);
1515 
1516 	vsk->buffer_size = val;
1517 }
1518 
1519 static int vsock_stream_setsockopt(struct socket *sock,
1520 				   int level,
1521 				   int optname,
1522 				   char __user *optval,
1523 				   unsigned int optlen)
1524 {
1525 	int err;
1526 	struct sock *sk;
1527 	struct vsock_sock *vsk;
1528 	const struct vsock_transport *transport;
1529 	u64 val;
1530 
1531 	if (level != AF_VSOCK)
1532 		return -ENOPROTOOPT;
1533 
1534 #define COPY_IN(_v)                                       \
1535 	do {						  \
1536 		if (optlen < sizeof(_v)) {		  \
1537 			err = -EINVAL;			  \
1538 			goto exit;			  \
1539 		}					  \
1540 		if (copy_from_user(&_v, optval, sizeof(_v)) != 0) {	\
1541 			err = -EFAULT;					\
1542 			goto exit;					\
1543 		}							\
1544 	} while (0)
1545 
1546 	err = 0;
1547 	sk = sock->sk;
1548 	vsk = vsock_sk(sk);
1549 	transport = vsk->transport;
1550 
1551 	lock_sock(sk);
1552 
1553 	switch (optname) {
1554 	case SO_VM_SOCKETS_BUFFER_SIZE:
1555 		COPY_IN(val);
1556 		vsock_update_buffer_size(vsk, transport, val);
1557 		break;
1558 
1559 	case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1560 		COPY_IN(val);
1561 		vsk->buffer_max_size = val;
1562 		vsock_update_buffer_size(vsk, transport, vsk->buffer_size);
1563 		break;
1564 
1565 	case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1566 		COPY_IN(val);
1567 		vsk->buffer_min_size = val;
1568 		vsock_update_buffer_size(vsk, transport, vsk->buffer_size);
1569 		break;
1570 
1571 	case SO_VM_SOCKETS_CONNECT_TIMEOUT: {
1572 		struct __kernel_old_timeval tv;
1573 		COPY_IN(tv);
1574 		if (tv.tv_sec >= 0 && tv.tv_usec < USEC_PER_SEC &&
1575 		    tv.tv_sec < (MAX_SCHEDULE_TIMEOUT / HZ - 1)) {
1576 			vsk->connect_timeout = tv.tv_sec * HZ +
1577 			    DIV_ROUND_UP(tv.tv_usec, (1000000 / HZ));
1578 			if (vsk->connect_timeout == 0)
1579 				vsk->connect_timeout =
1580 				    VSOCK_DEFAULT_CONNECT_TIMEOUT;
1581 
1582 		} else {
1583 			err = -ERANGE;
1584 		}
1585 		break;
1586 	}
1587 
1588 	default:
1589 		err = -ENOPROTOOPT;
1590 		break;
1591 	}
1592 
1593 #undef COPY_IN
1594 
1595 exit:
1596 	release_sock(sk);
1597 	return err;
1598 }
1599 
1600 static int vsock_stream_getsockopt(struct socket *sock,
1601 				   int level, int optname,
1602 				   char __user *optval,
1603 				   int __user *optlen)
1604 {
1605 	int err;
1606 	int len;
1607 	struct sock *sk;
1608 	struct vsock_sock *vsk;
1609 	u64 val;
1610 
1611 	if (level != AF_VSOCK)
1612 		return -ENOPROTOOPT;
1613 
1614 	err = get_user(len, optlen);
1615 	if (err != 0)
1616 		return err;
1617 
1618 #define COPY_OUT(_v)                            \
1619 	do {					\
1620 		if (len < sizeof(_v))		\
1621 			return -EINVAL;		\
1622 						\
1623 		len = sizeof(_v);		\
1624 		if (copy_to_user(optval, &_v, len) != 0)	\
1625 			return -EFAULT;				\
1626 								\
1627 	} while (0)
1628 
1629 	err = 0;
1630 	sk = sock->sk;
1631 	vsk = vsock_sk(sk);
1632 
1633 	switch (optname) {
1634 	case SO_VM_SOCKETS_BUFFER_SIZE:
1635 		val = vsk->buffer_size;
1636 		COPY_OUT(val);
1637 		break;
1638 
1639 	case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1640 		val = vsk->buffer_max_size;
1641 		COPY_OUT(val);
1642 		break;
1643 
1644 	case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1645 		val = vsk->buffer_min_size;
1646 		COPY_OUT(val);
1647 		break;
1648 
1649 	case SO_VM_SOCKETS_CONNECT_TIMEOUT: {
1650 		struct __kernel_old_timeval tv;
1651 		tv.tv_sec = vsk->connect_timeout / HZ;
1652 		tv.tv_usec =
1653 		    (vsk->connect_timeout -
1654 		     tv.tv_sec * HZ) * (1000000 / HZ);
1655 		COPY_OUT(tv);
1656 		break;
1657 	}
1658 	default:
1659 		return -ENOPROTOOPT;
1660 	}
1661 
1662 	err = put_user(len, optlen);
1663 	if (err != 0)
1664 		return -EFAULT;
1665 
1666 #undef COPY_OUT
1667 
1668 	return 0;
1669 }
1670 
1671 static int vsock_stream_sendmsg(struct socket *sock, struct msghdr *msg,
1672 				size_t len)
1673 {
1674 	struct sock *sk;
1675 	struct vsock_sock *vsk;
1676 	const struct vsock_transport *transport;
1677 	ssize_t total_written;
1678 	long timeout;
1679 	int err;
1680 	struct vsock_transport_send_notify_data send_data;
1681 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
1682 
1683 	sk = sock->sk;
1684 	vsk = vsock_sk(sk);
1685 	transport = vsk->transport;
1686 	total_written = 0;
1687 	err = 0;
1688 
1689 	if (msg->msg_flags & MSG_OOB)
1690 		return -EOPNOTSUPP;
1691 
1692 	lock_sock(sk);
1693 
1694 	/* Callers should not provide a destination with stream sockets. */
1695 	if (msg->msg_namelen) {
1696 		err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP;
1697 		goto out;
1698 	}
1699 
1700 	/* Send data only if both sides are not shutdown in the direction. */
1701 	if (sk->sk_shutdown & SEND_SHUTDOWN ||
1702 	    vsk->peer_shutdown & RCV_SHUTDOWN) {
1703 		err = -EPIPE;
1704 		goto out;
1705 	}
1706 
1707 	if (!transport || sk->sk_state != TCP_ESTABLISHED ||
1708 	    !vsock_addr_bound(&vsk->local_addr)) {
1709 		err = -ENOTCONN;
1710 		goto out;
1711 	}
1712 
1713 	if (!vsock_addr_bound(&vsk->remote_addr)) {
1714 		err = -EDESTADDRREQ;
1715 		goto out;
1716 	}
1717 
1718 	/* Wait for room in the produce queue to enqueue our user's data. */
1719 	timeout = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1720 
1721 	err = transport->notify_send_init(vsk, &send_data);
1722 	if (err < 0)
1723 		goto out;
1724 
1725 	while (total_written < len) {
1726 		ssize_t written;
1727 
1728 		add_wait_queue(sk_sleep(sk), &wait);
1729 		while (vsock_stream_has_space(vsk) == 0 &&
1730 		       sk->sk_err == 0 &&
1731 		       !(sk->sk_shutdown & SEND_SHUTDOWN) &&
1732 		       !(vsk->peer_shutdown & RCV_SHUTDOWN)) {
1733 
1734 			/* Don't wait for non-blocking sockets. */
1735 			if (timeout == 0) {
1736 				err = -EAGAIN;
1737 				remove_wait_queue(sk_sleep(sk), &wait);
1738 				goto out_err;
1739 			}
1740 
1741 			err = transport->notify_send_pre_block(vsk, &send_data);
1742 			if (err < 0) {
1743 				remove_wait_queue(sk_sleep(sk), &wait);
1744 				goto out_err;
1745 			}
1746 
1747 			release_sock(sk);
1748 			timeout = wait_woken(&wait, TASK_INTERRUPTIBLE, timeout);
1749 			lock_sock(sk);
1750 			if (signal_pending(current)) {
1751 				err = sock_intr_errno(timeout);
1752 				remove_wait_queue(sk_sleep(sk), &wait);
1753 				goto out_err;
1754 			} else if (timeout == 0) {
1755 				err = -EAGAIN;
1756 				remove_wait_queue(sk_sleep(sk), &wait);
1757 				goto out_err;
1758 			}
1759 		}
1760 		remove_wait_queue(sk_sleep(sk), &wait);
1761 
1762 		/* These checks occur both as part of and after the loop
1763 		 * conditional since we need to check before and after
1764 		 * sleeping.
1765 		 */
1766 		if (sk->sk_err) {
1767 			err = -sk->sk_err;
1768 			goto out_err;
1769 		} else if ((sk->sk_shutdown & SEND_SHUTDOWN) ||
1770 			   (vsk->peer_shutdown & RCV_SHUTDOWN)) {
1771 			err = -EPIPE;
1772 			goto out_err;
1773 		}
1774 
1775 		err = transport->notify_send_pre_enqueue(vsk, &send_data);
1776 		if (err < 0)
1777 			goto out_err;
1778 
1779 		/* Note that enqueue will only write as many bytes as are free
1780 		 * in the produce queue, so we don't need to ensure len is
1781 		 * smaller than the queue size.  It is the caller's
1782 		 * responsibility to check how many bytes we were able to send.
1783 		 */
1784 
1785 		written = transport->stream_enqueue(
1786 				vsk, msg,
1787 				len - total_written);
1788 		if (written < 0) {
1789 			err = -ENOMEM;
1790 			goto out_err;
1791 		}
1792 
1793 		total_written += written;
1794 
1795 		err = transport->notify_send_post_enqueue(
1796 				vsk, written, &send_data);
1797 		if (err < 0)
1798 			goto out_err;
1799 
1800 	}
1801 
1802 out_err:
1803 	if (total_written > 0)
1804 		err = total_written;
1805 out:
1806 	release_sock(sk);
1807 	return err;
1808 }
1809 
1810 
1811 static int
1812 vsock_stream_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
1813 		     int flags)
1814 {
1815 	struct sock *sk;
1816 	struct vsock_sock *vsk;
1817 	const struct vsock_transport *transport;
1818 	int err;
1819 	size_t target;
1820 	ssize_t copied;
1821 	long timeout;
1822 	struct vsock_transport_recv_notify_data recv_data;
1823 
1824 	DEFINE_WAIT(wait);
1825 
1826 	sk = sock->sk;
1827 	vsk = vsock_sk(sk);
1828 	transport = vsk->transport;
1829 	err = 0;
1830 
1831 	lock_sock(sk);
1832 
1833 	if (!transport || sk->sk_state != TCP_ESTABLISHED) {
1834 		/* Recvmsg is supposed to return 0 if a peer performs an
1835 		 * orderly shutdown. Differentiate between that case and when a
1836 		 * peer has not connected or a local shutdown occured with the
1837 		 * SOCK_DONE flag.
1838 		 */
1839 		if (sock_flag(sk, SOCK_DONE))
1840 			err = 0;
1841 		else
1842 			err = -ENOTCONN;
1843 
1844 		goto out;
1845 	}
1846 
1847 	if (flags & MSG_OOB) {
1848 		err = -EOPNOTSUPP;
1849 		goto out;
1850 	}
1851 
1852 	/* We don't check peer_shutdown flag here since peer may actually shut
1853 	 * down, but there can be data in the queue that a local socket can
1854 	 * receive.
1855 	 */
1856 	if (sk->sk_shutdown & RCV_SHUTDOWN) {
1857 		err = 0;
1858 		goto out;
1859 	}
1860 
1861 	/* It is valid on Linux to pass in a zero-length receive buffer.  This
1862 	 * is not an error.  We may as well bail out now.
1863 	 */
1864 	if (!len) {
1865 		err = 0;
1866 		goto out;
1867 	}
1868 
1869 	/* We must not copy less than target bytes into the user's buffer
1870 	 * before returning successfully, so we wait for the consume queue to
1871 	 * have that much data to consume before dequeueing.  Note that this
1872 	 * makes it impossible to handle cases where target is greater than the
1873 	 * queue size.
1874 	 */
1875 	target = sock_rcvlowat(sk, flags & MSG_WAITALL, len);
1876 	if (target >= transport->stream_rcvhiwat(vsk)) {
1877 		err = -ENOMEM;
1878 		goto out;
1879 	}
1880 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1881 	copied = 0;
1882 
1883 	err = transport->notify_recv_init(vsk, target, &recv_data);
1884 	if (err < 0)
1885 		goto out;
1886 
1887 
1888 	while (1) {
1889 		s64 ready;
1890 
1891 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1892 		ready = vsock_stream_has_data(vsk);
1893 
1894 		if (ready == 0) {
1895 			if (sk->sk_err != 0 ||
1896 			    (sk->sk_shutdown & RCV_SHUTDOWN) ||
1897 			    (vsk->peer_shutdown & SEND_SHUTDOWN)) {
1898 				finish_wait(sk_sleep(sk), &wait);
1899 				break;
1900 			}
1901 			/* Don't wait for non-blocking sockets. */
1902 			if (timeout == 0) {
1903 				err = -EAGAIN;
1904 				finish_wait(sk_sleep(sk), &wait);
1905 				break;
1906 			}
1907 
1908 			err = transport->notify_recv_pre_block(
1909 					vsk, target, &recv_data);
1910 			if (err < 0) {
1911 				finish_wait(sk_sleep(sk), &wait);
1912 				break;
1913 			}
1914 			release_sock(sk);
1915 			timeout = schedule_timeout(timeout);
1916 			lock_sock(sk);
1917 
1918 			if (signal_pending(current)) {
1919 				err = sock_intr_errno(timeout);
1920 				finish_wait(sk_sleep(sk), &wait);
1921 				break;
1922 			} else if (timeout == 0) {
1923 				err = -EAGAIN;
1924 				finish_wait(sk_sleep(sk), &wait);
1925 				break;
1926 			}
1927 		} else {
1928 			ssize_t read;
1929 
1930 			finish_wait(sk_sleep(sk), &wait);
1931 
1932 			if (ready < 0) {
1933 				/* Invalid queue pair content. XXX This should
1934 				* be changed to a connection reset in a later
1935 				* change.
1936 				*/
1937 
1938 				err = -ENOMEM;
1939 				goto out;
1940 			}
1941 
1942 			err = transport->notify_recv_pre_dequeue(
1943 					vsk, target, &recv_data);
1944 			if (err < 0)
1945 				break;
1946 
1947 			read = transport->stream_dequeue(
1948 					vsk, msg,
1949 					len - copied, flags);
1950 			if (read < 0) {
1951 				err = -ENOMEM;
1952 				break;
1953 			}
1954 
1955 			copied += read;
1956 
1957 			err = transport->notify_recv_post_dequeue(
1958 					vsk, target, read,
1959 					!(flags & MSG_PEEK), &recv_data);
1960 			if (err < 0)
1961 				goto out;
1962 
1963 			if (read >= target || flags & MSG_PEEK)
1964 				break;
1965 
1966 			target -= read;
1967 		}
1968 	}
1969 
1970 	if (sk->sk_err)
1971 		err = -sk->sk_err;
1972 	else if (sk->sk_shutdown & RCV_SHUTDOWN)
1973 		err = 0;
1974 
1975 	if (copied > 0)
1976 		err = copied;
1977 
1978 out:
1979 	release_sock(sk);
1980 	return err;
1981 }
1982 
1983 static const struct proto_ops vsock_stream_ops = {
1984 	.family = PF_VSOCK,
1985 	.owner = THIS_MODULE,
1986 	.release = vsock_release,
1987 	.bind = vsock_bind,
1988 	.connect = vsock_stream_connect,
1989 	.socketpair = sock_no_socketpair,
1990 	.accept = vsock_accept,
1991 	.getname = vsock_getname,
1992 	.poll = vsock_poll,
1993 	.ioctl = sock_no_ioctl,
1994 	.listen = vsock_listen,
1995 	.shutdown = vsock_shutdown,
1996 	.setsockopt = vsock_stream_setsockopt,
1997 	.getsockopt = vsock_stream_getsockopt,
1998 	.sendmsg = vsock_stream_sendmsg,
1999 	.recvmsg = vsock_stream_recvmsg,
2000 	.mmap = sock_no_mmap,
2001 	.sendpage = sock_no_sendpage,
2002 };
2003 
2004 static int vsock_create(struct net *net, struct socket *sock,
2005 			int protocol, int kern)
2006 {
2007 	struct vsock_sock *vsk;
2008 	struct sock *sk;
2009 	int ret;
2010 
2011 	if (!sock)
2012 		return -EINVAL;
2013 
2014 	if (protocol && protocol != PF_VSOCK)
2015 		return -EPROTONOSUPPORT;
2016 
2017 	switch (sock->type) {
2018 	case SOCK_DGRAM:
2019 		sock->ops = &vsock_dgram_ops;
2020 		break;
2021 	case SOCK_STREAM:
2022 		sock->ops = &vsock_stream_ops;
2023 		break;
2024 	default:
2025 		return -ESOCKTNOSUPPORT;
2026 	}
2027 
2028 	sock->state = SS_UNCONNECTED;
2029 
2030 	sk = __vsock_create(net, sock, NULL, GFP_KERNEL, 0, kern);
2031 	if (!sk)
2032 		return -ENOMEM;
2033 
2034 	vsk = vsock_sk(sk);
2035 
2036 	if (sock->type == SOCK_DGRAM) {
2037 		ret = vsock_assign_transport(vsk, NULL);
2038 		if (ret < 0) {
2039 			sock_put(sk);
2040 			return ret;
2041 		}
2042 	}
2043 
2044 	vsock_insert_unbound(vsk);
2045 
2046 	return 0;
2047 }
2048 
2049 static const struct net_proto_family vsock_family_ops = {
2050 	.family = AF_VSOCK,
2051 	.create = vsock_create,
2052 	.owner = THIS_MODULE,
2053 };
2054 
2055 static long vsock_dev_do_ioctl(struct file *filp,
2056 			       unsigned int cmd, void __user *ptr)
2057 {
2058 	u32 __user *p = ptr;
2059 	u32 cid = VMADDR_CID_ANY;
2060 	int retval = 0;
2061 
2062 	switch (cmd) {
2063 	case IOCTL_VM_SOCKETS_GET_LOCAL_CID:
2064 		/* To be compatible with the VMCI behavior, we prioritize the
2065 		 * guest CID instead of well-know host CID (VMADDR_CID_HOST).
2066 		 */
2067 		if (transport_g2h)
2068 			cid = transport_g2h->get_local_cid();
2069 		else if (transport_h2g)
2070 			cid = transport_h2g->get_local_cid();
2071 
2072 		if (put_user(cid, p) != 0)
2073 			retval = -EFAULT;
2074 		break;
2075 
2076 	default:
2077 		pr_err("Unknown ioctl %d\n", cmd);
2078 		retval = -EINVAL;
2079 	}
2080 
2081 	return retval;
2082 }
2083 
2084 static long vsock_dev_ioctl(struct file *filp,
2085 			    unsigned int cmd, unsigned long arg)
2086 {
2087 	return vsock_dev_do_ioctl(filp, cmd, (void __user *)arg);
2088 }
2089 
2090 #ifdef CONFIG_COMPAT
2091 static long vsock_dev_compat_ioctl(struct file *filp,
2092 				   unsigned int cmd, unsigned long arg)
2093 {
2094 	return vsock_dev_do_ioctl(filp, cmd, compat_ptr(arg));
2095 }
2096 #endif
2097 
2098 static const struct file_operations vsock_device_ops = {
2099 	.owner		= THIS_MODULE,
2100 	.unlocked_ioctl	= vsock_dev_ioctl,
2101 #ifdef CONFIG_COMPAT
2102 	.compat_ioctl	= vsock_dev_compat_ioctl,
2103 #endif
2104 	.open		= nonseekable_open,
2105 };
2106 
2107 static struct miscdevice vsock_device = {
2108 	.name		= "vsock",
2109 	.fops		= &vsock_device_ops,
2110 };
2111 
2112 static int __init vsock_init(void)
2113 {
2114 	int err = 0;
2115 
2116 	vsock_init_tables();
2117 
2118 	vsock_proto.owner = THIS_MODULE;
2119 	vsock_device.minor = MISC_DYNAMIC_MINOR;
2120 	err = misc_register(&vsock_device);
2121 	if (err) {
2122 		pr_err("Failed to register misc device\n");
2123 		goto err_reset_transport;
2124 	}
2125 
2126 	err = proto_register(&vsock_proto, 1);	/* we want our slab */
2127 	if (err) {
2128 		pr_err("Cannot register vsock protocol\n");
2129 		goto err_deregister_misc;
2130 	}
2131 
2132 	err = sock_register(&vsock_family_ops);
2133 	if (err) {
2134 		pr_err("could not register af_vsock (%d) address family: %d\n",
2135 		       AF_VSOCK, err);
2136 		goto err_unregister_proto;
2137 	}
2138 
2139 	return 0;
2140 
2141 err_unregister_proto:
2142 	proto_unregister(&vsock_proto);
2143 err_deregister_misc:
2144 	misc_deregister(&vsock_device);
2145 err_reset_transport:
2146 	return err;
2147 }
2148 
2149 static void __exit vsock_exit(void)
2150 {
2151 	misc_deregister(&vsock_device);
2152 	sock_unregister(AF_VSOCK);
2153 	proto_unregister(&vsock_proto);
2154 }
2155 
2156 const struct vsock_transport *vsock_core_get_transport(struct vsock_sock *vsk)
2157 {
2158 	return vsk->transport;
2159 }
2160 EXPORT_SYMBOL_GPL(vsock_core_get_transport);
2161 
2162 int vsock_core_register(const struct vsock_transport *t, int features)
2163 {
2164 	const struct vsock_transport *t_h2g, *t_g2h, *t_dgram, *t_local;
2165 	int err = mutex_lock_interruptible(&vsock_register_mutex);
2166 
2167 	if (err)
2168 		return err;
2169 
2170 	t_h2g = transport_h2g;
2171 	t_g2h = transport_g2h;
2172 	t_dgram = transport_dgram;
2173 	t_local = transport_local;
2174 
2175 	if (features & VSOCK_TRANSPORT_F_H2G) {
2176 		if (t_h2g) {
2177 			err = -EBUSY;
2178 			goto err_busy;
2179 		}
2180 		t_h2g = t;
2181 	}
2182 
2183 	if (features & VSOCK_TRANSPORT_F_G2H) {
2184 		if (t_g2h) {
2185 			err = -EBUSY;
2186 			goto err_busy;
2187 		}
2188 		t_g2h = t;
2189 	}
2190 
2191 	if (features & VSOCK_TRANSPORT_F_DGRAM) {
2192 		if (t_dgram) {
2193 			err = -EBUSY;
2194 			goto err_busy;
2195 		}
2196 		t_dgram = t;
2197 	}
2198 
2199 	if (features & VSOCK_TRANSPORT_F_LOCAL) {
2200 		if (t_local) {
2201 			err = -EBUSY;
2202 			goto err_busy;
2203 		}
2204 		t_local = t;
2205 	}
2206 
2207 	transport_h2g = t_h2g;
2208 	transport_g2h = t_g2h;
2209 	transport_dgram = t_dgram;
2210 	transport_local = t_local;
2211 
2212 err_busy:
2213 	mutex_unlock(&vsock_register_mutex);
2214 	return err;
2215 }
2216 EXPORT_SYMBOL_GPL(vsock_core_register);
2217 
2218 void vsock_core_unregister(const struct vsock_transport *t)
2219 {
2220 	mutex_lock(&vsock_register_mutex);
2221 
2222 	if (transport_h2g == t)
2223 		transport_h2g = NULL;
2224 
2225 	if (transport_g2h == t)
2226 		transport_g2h = NULL;
2227 
2228 	if (transport_dgram == t)
2229 		transport_dgram = NULL;
2230 
2231 	if (transport_local == t)
2232 		transport_local = NULL;
2233 
2234 	mutex_unlock(&vsock_register_mutex);
2235 }
2236 EXPORT_SYMBOL_GPL(vsock_core_unregister);
2237 
2238 module_init(vsock_init);
2239 module_exit(vsock_exit);
2240 
2241 MODULE_AUTHOR("VMware, Inc.");
2242 MODULE_DESCRIPTION("VMware Virtual Socket Family");
2243 MODULE_VERSION("1.0.2.0-k");
2244 MODULE_LICENSE("GPL v2");
2245