xref: /linux/net/vmw_vsock/af_vsock.c (revision ea8a163e02d6925773129e2dd86e419e491b791d)
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 connection oriented socket this must be called when vsk->remote_addr
419  * is set (e.g. during the connect() or when a connection request on a listener
420  * socket is received).
421  * The vsk->remote_addr is used to decide which transport to use:
422  *  - remote CID == VMADDR_CID_LOCAL or g2h->local_cid or VMADDR_CID_HOST if
423  *    g2h is not loaded, will use local transport;
424  *  - remote CID <= VMADDR_CID_HOST or h2g is not loaded or remote flags field
425  *    includes VMADDR_FLAG_TO_HOST flag value, will use guest->host transport;
426  *  - remote CID > VMADDR_CID_HOST will use host->guest transport;
427  */
428 int vsock_assign_transport(struct vsock_sock *vsk, struct vsock_sock *psk)
429 {
430 	const struct vsock_transport *new_transport;
431 	struct sock *sk = sk_vsock(vsk);
432 	unsigned int remote_cid = vsk->remote_addr.svm_cid;
433 	__u8 remote_flags;
434 	int ret;
435 
436 	/* If the packet is coming with the source and destination CIDs higher
437 	 * than VMADDR_CID_HOST, then a vsock channel where all the packets are
438 	 * forwarded to the host should be established. Then the host will
439 	 * need to forward the packets to the guest.
440 	 *
441 	 * The flag is set on the (listen) receive path (psk is not NULL). On
442 	 * the connect path the flag can be set by the user space application.
443 	 */
444 	if (psk && vsk->local_addr.svm_cid > VMADDR_CID_HOST &&
445 	    vsk->remote_addr.svm_cid > VMADDR_CID_HOST)
446 		vsk->remote_addr.svm_flags |= VMADDR_FLAG_TO_HOST;
447 
448 	remote_flags = vsk->remote_addr.svm_flags;
449 
450 	switch (sk->sk_type) {
451 	case SOCK_DGRAM:
452 		new_transport = transport_dgram;
453 		break;
454 	case SOCK_STREAM:
455 	case SOCK_SEQPACKET:
456 		if (vsock_use_local_transport(remote_cid))
457 			new_transport = transport_local;
458 		else if (remote_cid <= VMADDR_CID_HOST || !transport_h2g ||
459 			 (remote_flags & VMADDR_FLAG_TO_HOST))
460 			new_transport = transport_g2h;
461 		else
462 			new_transport = transport_h2g;
463 		break;
464 	default:
465 		return -ESOCKTNOSUPPORT;
466 	}
467 
468 	if (vsk->transport) {
469 		if (vsk->transport == new_transport)
470 			return 0;
471 
472 		/* transport->release() must be called with sock lock acquired.
473 		 * This path can only be taken during vsock_connect(), where we
474 		 * have already held the sock lock. In the other cases, this
475 		 * function is called on a new socket which is not assigned to
476 		 * any transport.
477 		 */
478 		vsk->transport->release(vsk);
479 		vsock_deassign_transport(vsk);
480 	}
481 
482 	/* We increase the module refcnt to prevent the transport unloading
483 	 * while there are open sockets assigned to it.
484 	 */
485 	if (!new_transport || !try_module_get(new_transport->module))
486 		return -ENODEV;
487 
488 	if (sk->sk_type == SOCK_SEQPACKET) {
489 		if (!new_transport->seqpacket_allow ||
490 		    !new_transport->seqpacket_allow(remote_cid)) {
491 			module_put(new_transport->module);
492 			return -ESOCKTNOSUPPORT;
493 		}
494 	}
495 
496 	ret = new_transport->init(vsk, psk);
497 	if (ret) {
498 		module_put(new_transport->module);
499 		return ret;
500 	}
501 
502 	vsk->transport = new_transport;
503 
504 	return 0;
505 }
506 EXPORT_SYMBOL_GPL(vsock_assign_transport);
507 
508 bool vsock_find_cid(unsigned int cid)
509 {
510 	if (transport_g2h && cid == transport_g2h->get_local_cid())
511 		return true;
512 
513 	if (transport_h2g && cid == VMADDR_CID_HOST)
514 		return true;
515 
516 	if (transport_local && cid == VMADDR_CID_LOCAL)
517 		return true;
518 
519 	return false;
520 }
521 EXPORT_SYMBOL_GPL(vsock_find_cid);
522 
523 static struct sock *vsock_dequeue_accept(struct sock *listener)
524 {
525 	struct vsock_sock *vlistener;
526 	struct vsock_sock *vconnected;
527 
528 	vlistener = vsock_sk(listener);
529 
530 	if (list_empty(&vlistener->accept_queue))
531 		return NULL;
532 
533 	vconnected = list_entry(vlistener->accept_queue.next,
534 				struct vsock_sock, accept_queue);
535 
536 	list_del_init(&vconnected->accept_queue);
537 	sock_put(listener);
538 	/* The caller will need a reference on the connected socket so we let
539 	 * it call sock_put().
540 	 */
541 
542 	return sk_vsock(vconnected);
543 }
544 
545 static bool vsock_is_accept_queue_empty(struct sock *sk)
546 {
547 	struct vsock_sock *vsk = vsock_sk(sk);
548 	return list_empty(&vsk->accept_queue);
549 }
550 
551 static bool vsock_is_pending(struct sock *sk)
552 {
553 	struct vsock_sock *vsk = vsock_sk(sk);
554 	return !list_empty(&vsk->pending_links);
555 }
556 
557 static int vsock_send_shutdown(struct sock *sk, int mode)
558 {
559 	struct vsock_sock *vsk = vsock_sk(sk);
560 
561 	if (!vsk->transport)
562 		return -ENODEV;
563 
564 	return vsk->transport->shutdown(vsk, mode);
565 }
566 
567 static void vsock_pending_work(struct work_struct *work)
568 {
569 	struct sock *sk;
570 	struct sock *listener;
571 	struct vsock_sock *vsk;
572 	bool cleanup;
573 
574 	vsk = container_of(work, struct vsock_sock, pending_work.work);
575 	sk = sk_vsock(vsk);
576 	listener = vsk->listener;
577 	cleanup = true;
578 
579 	lock_sock(listener);
580 	lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
581 
582 	if (vsock_is_pending(sk)) {
583 		vsock_remove_pending(listener, sk);
584 
585 		sk_acceptq_removed(listener);
586 	} else if (!vsk->rejected) {
587 		/* We are not on the pending list and accept() did not reject
588 		 * us, so we must have been accepted by our user process.  We
589 		 * just need to drop our references to the sockets and be on
590 		 * our way.
591 		 */
592 		cleanup = false;
593 		goto out;
594 	}
595 
596 	/* We need to remove ourself from the global connected sockets list so
597 	 * incoming packets can't find this socket, and to reduce the reference
598 	 * count.
599 	 */
600 	vsock_remove_connected(vsk);
601 
602 	sk->sk_state = TCP_CLOSE;
603 
604 out:
605 	release_sock(sk);
606 	release_sock(listener);
607 	if (cleanup)
608 		sock_put(sk);
609 
610 	sock_put(sk);
611 	sock_put(listener);
612 }
613 
614 /**** SOCKET OPERATIONS ****/
615 
616 static int __vsock_bind_connectible(struct vsock_sock *vsk,
617 				    struct sockaddr_vm *addr)
618 {
619 	static u32 port;
620 	struct sockaddr_vm new_addr;
621 
622 	if (!port)
623 		port = LAST_RESERVED_PORT + 1 +
624 			prandom_u32_max(U32_MAX - LAST_RESERVED_PORT);
625 
626 	vsock_addr_init(&new_addr, addr->svm_cid, addr->svm_port);
627 
628 	if (addr->svm_port == VMADDR_PORT_ANY) {
629 		bool found = false;
630 		unsigned int i;
631 
632 		for (i = 0; i < MAX_PORT_RETRIES; i++) {
633 			if (port <= LAST_RESERVED_PORT)
634 				port = LAST_RESERVED_PORT + 1;
635 
636 			new_addr.svm_port = port++;
637 
638 			if (!__vsock_find_bound_socket(&new_addr)) {
639 				found = true;
640 				break;
641 			}
642 		}
643 
644 		if (!found)
645 			return -EADDRNOTAVAIL;
646 	} else {
647 		/* If port is in reserved range, ensure caller
648 		 * has necessary privileges.
649 		 */
650 		if (addr->svm_port <= LAST_RESERVED_PORT &&
651 		    !capable(CAP_NET_BIND_SERVICE)) {
652 			return -EACCES;
653 		}
654 
655 		if (__vsock_find_bound_socket(&new_addr))
656 			return -EADDRINUSE;
657 	}
658 
659 	vsock_addr_init(&vsk->local_addr, new_addr.svm_cid, new_addr.svm_port);
660 
661 	/* Remove connection oriented sockets from the unbound list and add them
662 	 * to the hash table for easy lookup by its address.  The unbound list
663 	 * is simply an extra entry at the end of the hash table, a trick used
664 	 * by AF_UNIX.
665 	 */
666 	__vsock_remove_bound(vsk);
667 	__vsock_insert_bound(vsock_bound_sockets(&vsk->local_addr), vsk);
668 
669 	return 0;
670 }
671 
672 static int __vsock_bind_dgram(struct vsock_sock *vsk,
673 			      struct sockaddr_vm *addr)
674 {
675 	return vsk->transport->dgram_bind(vsk, addr);
676 }
677 
678 static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr)
679 {
680 	struct vsock_sock *vsk = vsock_sk(sk);
681 	int retval;
682 
683 	/* First ensure this socket isn't already bound. */
684 	if (vsock_addr_bound(&vsk->local_addr))
685 		return -EINVAL;
686 
687 	/* Now bind to the provided address or select appropriate values if
688 	 * none are provided (VMADDR_CID_ANY and VMADDR_PORT_ANY).  Note that
689 	 * like AF_INET prevents binding to a non-local IP address (in most
690 	 * cases), we only allow binding to a local CID.
691 	 */
692 	if (addr->svm_cid != VMADDR_CID_ANY && !vsock_find_cid(addr->svm_cid))
693 		return -EADDRNOTAVAIL;
694 
695 	switch (sk->sk_socket->type) {
696 	case SOCK_STREAM:
697 	case SOCK_SEQPACKET:
698 		spin_lock_bh(&vsock_table_lock);
699 		retval = __vsock_bind_connectible(vsk, addr);
700 		spin_unlock_bh(&vsock_table_lock);
701 		break;
702 
703 	case SOCK_DGRAM:
704 		retval = __vsock_bind_dgram(vsk, addr);
705 		break;
706 
707 	default:
708 		retval = -EINVAL;
709 		break;
710 	}
711 
712 	return retval;
713 }
714 
715 static void vsock_connect_timeout(struct work_struct *work);
716 
717 static struct sock *__vsock_create(struct net *net,
718 				   struct socket *sock,
719 				   struct sock *parent,
720 				   gfp_t priority,
721 				   unsigned short type,
722 				   int kern)
723 {
724 	struct sock *sk;
725 	struct vsock_sock *psk;
726 	struct vsock_sock *vsk;
727 
728 	sk = sk_alloc(net, AF_VSOCK, priority, &vsock_proto, kern);
729 	if (!sk)
730 		return NULL;
731 
732 	sock_init_data(sock, sk);
733 
734 	/* sk->sk_type is normally set in sock_init_data, but only if sock is
735 	 * non-NULL. We make sure that our sockets always have a type by
736 	 * setting it here if needed.
737 	 */
738 	if (!sock)
739 		sk->sk_type = type;
740 
741 	vsk = vsock_sk(sk);
742 	vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
743 	vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
744 
745 	sk->sk_destruct = vsock_sk_destruct;
746 	sk->sk_backlog_rcv = vsock_queue_rcv_skb;
747 	sock_reset_flag(sk, SOCK_DONE);
748 
749 	INIT_LIST_HEAD(&vsk->bound_table);
750 	INIT_LIST_HEAD(&vsk->connected_table);
751 	vsk->listener = NULL;
752 	INIT_LIST_HEAD(&vsk->pending_links);
753 	INIT_LIST_HEAD(&vsk->accept_queue);
754 	vsk->rejected = false;
755 	vsk->sent_request = false;
756 	vsk->ignore_connecting_rst = false;
757 	vsk->peer_shutdown = 0;
758 	INIT_DELAYED_WORK(&vsk->connect_work, vsock_connect_timeout);
759 	INIT_DELAYED_WORK(&vsk->pending_work, vsock_pending_work);
760 
761 	psk = parent ? vsock_sk(parent) : NULL;
762 	if (parent) {
763 		vsk->trusted = psk->trusted;
764 		vsk->owner = get_cred(psk->owner);
765 		vsk->connect_timeout = psk->connect_timeout;
766 		vsk->buffer_size = psk->buffer_size;
767 		vsk->buffer_min_size = psk->buffer_min_size;
768 		vsk->buffer_max_size = psk->buffer_max_size;
769 		security_sk_clone(parent, sk);
770 	} else {
771 		vsk->trusted = ns_capable_noaudit(&init_user_ns, CAP_NET_ADMIN);
772 		vsk->owner = get_current_cred();
773 		vsk->connect_timeout = VSOCK_DEFAULT_CONNECT_TIMEOUT;
774 		vsk->buffer_size = VSOCK_DEFAULT_BUFFER_SIZE;
775 		vsk->buffer_min_size = VSOCK_DEFAULT_BUFFER_MIN_SIZE;
776 		vsk->buffer_max_size = VSOCK_DEFAULT_BUFFER_MAX_SIZE;
777 	}
778 
779 	return sk;
780 }
781 
782 static bool sock_type_connectible(u16 type)
783 {
784 	return (type == SOCK_STREAM) || (type == SOCK_SEQPACKET);
785 }
786 
787 static void __vsock_release(struct sock *sk, int level)
788 {
789 	if (sk) {
790 		struct sock *pending;
791 		struct vsock_sock *vsk;
792 
793 		vsk = vsock_sk(sk);
794 		pending = NULL;	/* Compiler warning. */
795 
796 		/* When "level" is SINGLE_DEPTH_NESTING, use the nested
797 		 * version to avoid the warning "possible recursive locking
798 		 * detected". When "level" is 0, lock_sock_nested(sk, level)
799 		 * is the same as lock_sock(sk).
800 		 */
801 		lock_sock_nested(sk, level);
802 
803 		if (vsk->transport)
804 			vsk->transport->release(vsk);
805 		else if (sock_type_connectible(sk->sk_type))
806 			vsock_remove_sock(vsk);
807 
808 		sock_orphan(sk);
809 		sk->sk_shutdown = SHUTDOWN_MASK;
810 
811 		skb_queue_purge(&sk->sk_receive_queue);
812 
813 		/* Clean up any sockets that never were accepted. */
814 		while ((pending = vsock_dequeue_accept(sk)) != NULL) {
815 			__vsock_release(pending, SINGLE_DEPTH_NESTING);
816 			sock_put(pending);
817 		}
818 
819 		release_sock(sk);
820 		sock_put(sk);
821 	}
822 }
823 
824 static void vsock_sk_destruct(struct sock *sk)
825 {
826 	struct vsock_sock *vsk = vsock_sk(sk);
827 
828 	vsock_deassign_transport(vsk);
829 
830 	/* When clearing these addresses, there's no need to set the family and
831 	 * possibly register the address family with the kernel.
832 	 */
833 	vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
834 	vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
835 
836 	put_cred(vsk->owner);
837 }
838 
839 static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
840 {
841 	int err;
842 
843 	err = sock_queue_rcv_skb(sk, skb);
844 	if (err)
845 		kfree_skb(skb);
846 
847 	return err;
848 }
849 
850 struct sock *vsock_create_connected(struct sock *parent)
851 {
852 	return __vsock_create(sock_net(parent), NULL, parent, GFP_KERNEL,
853 			      parent->sk_type, 0);
854 }
855 EXPORT_SYMBOL_GPL(vsock_create_connected);
856 
857 s64 vsock_stream_has_data(struct vsock_sock *vsk)
858 {
859 	return vsk->transport->stream_has_data(vsk);
860 }
861 EXPORT_SYMBOL_GPL(vsock_stream_has_data);
862 
863 static s64 vsock_connectible_has_data(struct vsock_sock *vsk)
864 {
865 	struct sock *sk = sk_vsock(vsk);
866 
867 	if (sk->sk_type == SOCK_SEQPACKET)
868 		return vsk->transport->seqpacket_has_data(vsk);
869 	else
870 		return vsock_stream_has_data(vsk);
871 }
872 
873 s64 vsock_stream_has_space(struct vsock_sock *vsk)
874 {
875 	return vsk->transport->stream_has_space(vsk);
876 }
877 EXPORT_SYMBOL_GPL(vsock_stream_has_space);
878 
879 static int vsock_release(struct socket *sock)
880 {
881 	__vsock_release(sock->sk, 0);
882 	sock->sk = NULL;
883 	sock->state = SS_FREE;
884 
885 	return 0;
886 }
887 
888 static int
889 vsock_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
890 {
891 	int err;
892 	struct sock *sk;
893 	struct sockaddr_vm *vm_addr;
894 
895 	sk = sock->sk;
896 
897 	if (vsock_addr_cast(addr, addr_len, &vm_addr) != 0)
898 		return -EINVAL;
899 
900 	lock_sock(sk);
901 	err = __vsock_bind(sk, vm_addr);
902 	release_sock(sk);
903 
904 	return err;
905 }
906 
907 static int vsock_getname(struct socket *sock,
908 			 struct sockaddr *addr, int peer)
909 {
910 	int err;
911 	struct sock *sk;
912 	struct vsock_sock *vsk;
913 	struct sockaddr_vm *vm_addr;
914 
915 	sk = sock->sk;
916 	vsk = vsock_sk(sk);
917 	err = 0;
918 
919 	lock_sock(sk);
920 
921 	if (peer) {
922 		if (sock->state != SS_CONNECTED) {
923 			err = -ENOTCONN;
924 			goto out;
925 		}
926 		vm_addr = &vsk->remote_addr;
927 	} else {
928 		vm_addr = &vsk->local_addr;
929 	}
930 
931 	if (!vm_addr) {
932 		err = -EINVAL;
933 		goto out;
934 	}
935 
936 	/* sys_getsockname() and sys_getpeername() pass us a
937 	 * MAX_SOCK_ADDR-sized buffer and don't set addr_len.  Unfortunately
938 	 * that macro is defined in socket.c instead of .h, so we hardcode its
939 	 * value here.
940 	 */
941 	BUILD_BUG_ON(sizeof(*vm_addr) > 128);
942 	memcpy(addr, vm_addr, sizeof(*vm_addr));
943 	err = sizeof(*vm_addr);
944 
945 out:
946 	release_sock(sk);
947 	return err;
948 }
949 
950 static int vsock_shutdown(struct socket *sock, int mode)
951 {
952 	int err;
953 	struct sock *sk;
954 
955 	/* User level uses SHUT_RD (0) and SHUT_WR (1), but the kernel uses
956 	 * RCV_SHUTDOWN (1) and SEND_SHUTDOWN (2), so we must increment mode
957 	 * here like the other address families do.  Note also that the
958 	 * increment makes SHUT_RDWR (2) into RCV_SHUTDOWN | SEND_SHUTDOWN (3),
959 	 * which is what we want.
960 	 */
961 	mode++;
962 
963 	if ((mode & ~SHUTDOWN_MASK) || !mode)
964 		return -EINVAL;
965 
966 	/* If this is a connection oriented socket and it is not connected then
967 	 * bail out immediately.  If it is a DGRAM socket then we must first
968 	 * kick the socket so that it wakes up from any sleeping calls, for
969 	 * example recv(), and then afterwards return the error.
970 	 */
971 
972 	sk = sock->sk;
973 
974 	lock_sock(sk);
975 	if (sock->state == SS_UNCONNECTED) {
976 		err = -ENOTCONN;
977 		if (sock_type_connectible(sk->sk_type))
978 			goto out;
979 	} else {
980 		sock->state = SS_DISCONNECTING;
981 		err = 0;
982 	}
983 
984 	/* Receive and send shutdowns are treated alike. */
985 	mode = mode & (RCV_SHUTDOWN | SEND_SHUTDOWN);
986 	if (mode) {
987 		sk->sk_shutdown |= mode;
988 		sk->sk_state_change(sk);
989 
990 		if (sock_type_connectible(sk->sk_type)) {
991 			sock_reset_flag(sk, SOCK_DONE);
992 			vsock_send_shutdown(sk, mode);
993 		}
994 	}
995 
996 out:
997 	release_sock(sk);
998 	return err;
999 }
1000 
1001 static __poll_t vsock_poll(struct file *file, struct socket *sock,
1002 			       poll_table *wait)
1003 {
1004 	struct sock *sk;
1005 	__poll_t mask;
1006 	struct vsock_sock *vsk;
1007 
1008 	sk = sock->sk;
1009 	vsk = vsock_sk(sk);
1010 
1011 	poll_wait(file, sk_sleep(sk), wait);
1012 	mask = 0;
1013 
1014 	if (sk->sk_err)
1015 		/* Signify that there has been an error on this socket. */
1016 		mask |= EPOLLERR;
1017 
1018 	/* INET sockets treat local write shutdown and peer write shutdown as a
1019 	 * case of EPOLLHUP set.
1020 	 */
1021 	if ((sk->sk_shutdown == SHUTDOWN_MASK) ||
1022 	    ((sk->sk_shutdown & SEND_SHUTDOWN) &&
1023 	     (vsk->peer_shutdown & SEND_SHUTDOWN))) {
1024 		mask |= EPOLLHUP;
1025 	}
1026 
1027 	if (sk->sk_shutdown & RCV_SHUTDOWN ||
1028 	    vsk->peer_shutdown & SEND_SHUTDOWN) {
1029 		mask |= EPOLLRDHUP;
1030 	}
1031 
1032 	if (sock->type == SOCK_DGRAM) {
1033 		/* For datagram sockets we can read if there is something in
1034 		 * the queue and write as long as the socket isn't shutdown for
1035 		 * sending.
1036 		 */
1037 		if (!skb_queue_empty_lockless(&sk->sk_receive_queue) ||
1038 		    (sk->sk_shutdown & RCV_SHUTDOWN)) {
1039 			mask |= EPOLLIN | EPOLLRDNORM;
1040 		}
1041 
1042 		if (!(sk->sk_shutdown & SEND_SHUTDOWN))
1043 			mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND;
1044 
1045 	} else if (sock_type_connectible(sk->sk_type)) {
1046 		const struct vsock_transport *transport;
1047 
1048 		lock_sock(sk);
1049 
1050 		transport = vsk->transport;
1051 
1052 		/* Listening sockets that have connections in their accept
1053 		 * queue can be read.
1054 		 */
1055 		if (sk->sk_state == TCP_LISTEN
1056 		    && !vsock_is_accept_queue_empty(sk))
1057 			mask |= EPOLLIN | EPOLLRDNORM;
1058 
1059 		/* If there is something in the queue then we can read. */
1060 		if (transport && transport->stream_is_active(vsk) &&
1061 		    !(sk->sk_shutdown & RCV_SHUTDOWN)) {
1062 			bool data_ready_now = false;
1063 			int ret = transport->notify_poll_in(
1064 					vsk, 1, &data_ready_now);
1065 			if (ret < 0) {
1066 				mask |= EPOLLERR;
1067 			} else {
1068 				if (data_ready_now)
1069 					mask |= EPOLLIN | EPOLLRDNORM;
1070 
1071 			}
1072 		}
1073 
1074 		/* Sockets whose connections have been closed, reset, or
1075 		 * terminated should also be considered read, and we check the
1076 		 * shutdown flag for that.
1077 		 */
1078 		if (sk->sk_shutdown & RCV_SHUTDOWN ||
1079 		    vsk->peer_shutdown & SEND_SHUTDOWN) {
1080 			mask |= EPOLLIN | EPOLLRDNORM;
1081 		}
1082 
1083 		/* Connected sockets that can produce data can be written. */
1084 		if (transport && sk->sk_state == TCP_ESTABLISHED) {
1085 			if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1086 				bool space_avail_now = false;
1087 				int ret = transport->notify_poll_out(
1088 						vsk, 1, &space_avail_now);
1089 				if (ret < 0) {
1090 					mask |= EPOLLERR;
1091 				} else {
1092 					if (space_avail_now)
1093 						/* Remove EPOLLWRBAND since INET
1094 						 * sockets are not setting it.
1095 						 */
1096 						mask |= EPOLLOUT | EPOLLWRNORM;
1097 
1098 				}
1099 			}
1100 		}
1101 
1102 		/* Simulate INET socket poll behaviors, which sets
1103 		 * EPOLLOUT|EPOLLWRNORM when peer is closed and nothing to read,
1104 		 * but local send is not shutdown.
1105 		 */
1106 		if (sk->sk_state == TCP_CLOSE || sk->sk_state == TCP_CLOSING) {
1107 			if (!(sk->sk_shutdown & SEND_SHUTDOWN))
1108 				mask |= EPOLLOUT | EPOLLWRNORM;
1109 
1110 		}
1111 
1112 		release_sock(sk);
1113 	}
1114 
1115 	return mask;
1116 }
1117 
1118 static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg,
1119 			       size_t len)
1120 {
1121 	int err;
1122 	struct sock *sk;
1123 	struct vsock_sock *vsk;
1124 	struct sockaddr_vm *remote_addr;
1125 	const struct vsock_transport *transport;
1126 
1127 	if (msg->msg_flags & MSG_OOB)
1128 		return -EOPNOTSUPP;
1129 
1130 	/* For now, MSG_DONTWAIT is always assumed... */
1131 	err = 0;
1132 	sk = sock->sk;
1133 	vsk = vsock_sk(sk);
1134 
1135 	lock_sock(sk);
1136 
1137 	transport = vsk->transport;
1138 
1139 	err = vsock_auto_bind(vsk);
1140 	if (err)
1141 		goto out;
1142 
1143 
1144 	/* If the provided message contains an address, use that.  Otherwise
1145 	 * fall back on the socket's remote handle (if it has been connected).
1146 	 */
1147 	if (msg->msg_name &&
1148 	    vsock_addr_cast(msg->msg_name, msg->msg_namelen,
1149 			    &remote_addr) == 0) {
1150 		/* Ensure this address is of the right type and is a valid
1151 		 * destination.
1152 		 */
1153 
1154 		if (remote_addr->svm_cid == VMADDR_CID_ANY)
1155 			remote_addr->svm_cid = transport->get_local_cid();
1156 
1157 		if (!vsock_addr_bound(remote_addr)) {
1158 			err = -EINVAL;
1159 			goto out;
1160 		}
1161 	} else if (sock->state == SS_CONNECTED) {
1162 		remote_addr = &vsk->remote_addr;
1163 
1164 		if (remote_addr->svm_cid == VMADDR_CID_ANY)
1165 			remote_addr->svm_cid = transport->get_local_cid();
1166 
1167 		/* XXX Should connect() or this function ensure remote_addr is
1168 		 * bound?
1169 		 */
1170 		if (!vsock_addr_bound(&vsk->remote_addr)) {
1171 			err = -EINVAL;
1172 			goto out;
1173 		}
1174 	} else {
1175 		err = -EINVAL;
1176 		goto out;
1177 	}
1178 
1179 	if (!transport->dgram_allow(remote_addr->svm_cid,
1180 				    remote_addr->svm_port)) {
1181 		err = -EINVAL;
1182 		goto out;
1183 	}
1184 
1185 	err = transport->dgram_enqueue(vsk, remote_addr, msg, len);
1186 
1187 out:
1188 	release_sock(sk);
1189 	return err;
1190 }
1191 
1192 static int vsock_dgram_connect(struct socket *sock,
1193 			       struct sockaddr *addr, int addr_len, int flags)
1194 {
1195 	int err;
1196 	struct sock *sk;
1197 	struct vsock_sock *vsk;
1198 	struct sockaddr_vm *remote_addr;
1199 
1200 	sk = sock->sk;
1201 	vsk = vsock_sk(sk);
1202 
1203 	err = vsock_addr_cast(addr, addr_len, &remote_addr);
1204 	if (err == -EAFNOSUPPORT && remote_addr->svm_family == AF_UNSPEC) {
1205 		lock_sock(sk);
1206 		vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY,
1207 				VMADDR_PORT_ANY);
1208 		sock->state = SS_UNCONNECTED;
1209 		release_sock(sk);
1210 		return 0;
1211 	} else if (err != 0)
1212 		return -EINVAL;
1213 
1214 	lock_sock(sk);
1215 
1216 	err = vsock_auto_bind(vsk);
1217 	if (err)
1218 		goto out;
1219 
1220 	if (!vsk->transport->dgram_allow(remote_addr->svm_cid,
1221 					 remote_addr->svm_port)) {
1222 		err = -EINVAL;
1223 		goto out;
1224 	}
1225 
1226 	memcpy(&vsk->remote_addr, remote_addr, sizeof(vsk->remote_addr));
1227 	sock->state = SS_CONNECTED;
1228 
1229 out:
1230 	release_sock(sk);
1231 	return err;
1232 }
1233 
1234 static int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg,
1235 			       size_t len, int flags)
1236 {
1237 	struct vsock_sock *vsk = vsock_sk(sock->sk);
1238 
1239 	return vsk->transport->dgram_dequeue(vsk, msg, len, flags);
1240 }
1241 
1242 static const struct proto_ops vsock_dgram_ops = {
1243 	.family = PF_VSOCK,
1244 	.owner = THIS_MODULE,
1245 	.release = vsock_release,
1246 	.bind = vsock_bind,
1247 	.connect = vsock_dgram_connect,
1248 	.socketpair = sock_no_socketpair,
1249 	.accept = sock_no_accept,
1250 	.getname = vsock_getname,
1251 	.poll = vsock_poll,
1252 	.ioctl = sock_no_ioctl,
1253 	.listen = sock_no_listen,
1254 	.shutdown = vsock_shutdown,
1255 	.sendmsg = vsock_dgram_sendmsg,
1256 	.recvmsg = vsock_dgram_recvmsg,
1257 	.mmap = sock_no_mmap,
1258 	.sendpage = sock_no_sendpage,
1259 };
1260 
1261 static int vsock_transport_cancel_pkt(struct vsock_sock *vsk)
1262 {
1263 	const struct vsock_transport *transport = vsk->transport;
1264 
1265 	if (!transport || !transport->cancel_pkt)
1266 		return -EOPNOTSUPP;
1267 
1268 	return transport->cancel_pkt(vsk);
1269 }
1270 
1271 static void vsock_connect_timeout(struct work_struct *work)
1272 {
1273 	struct sock *sk;
1274 	struct vsock_sock *vsk;
1275 
1276 	vsk = container_of(work, struct vsock_sock, connect_work.work);
1277 	sk = sk_vsock(vsk);
1278 
1279 	lock_sock(sk);
1280 	if (sk->sk_state == TCP_SYN_SENT &&
1281 	    (sk->sk_shutdown != SHUTDOWN_MASK)) {
1282 		sk->sk_state = TCP_CLOSE;
1283 		sk->sk_err = ETIMEDOUT;
1284 		sk_error_report(sk);
1285 		vsock_transport_cancel_pkt(vsk);
1286 	}
1287 	release_sock(sk);
1288 
1289 	sock_put(sk);
1290 }
1291 
1292 static int vsock_connect(struct socket *sock, struct sockaddr *addr,
1293 			 int addr_len, int flags)
1294 {
1295 	int err;
1296 	struct sock *sk;
1297 	struct vsock_sock *vsk;
1298 	const struct vsock_transport *transport;
1299 	struct sockaddr_vm *remote_addr;
1300 	long timeout;
1301 	DEFINE_WAIT(wait);
1302 
1303 	err = 0;
1304 	sk = sock->sk;
1305 	vsk = vsock_sk(sk);
1306 
1307 	lock_sock(sk);
1308 
1309 	/* XXX AF_UNSPEC should make us disconnect like AF_INET. */
1310 	switch (sock->state) {
1311 	case SS_CONNECTED:
1312 		err = -EISCONN;
1313 		goto out;
1314 	case SS_DISCONNECTING:
1315 		err = -EINVAL;
1316 		goto out;
1317 	case SS_CONNECTING:
1318 		/* This continues on so we can move sock into the SS_CONNECTED
1319 		 * state once the connection has completed (at which point err
1320 		 * will be set to zero also).  Otherwise, we will either wait
1321 		 * for the connection or return -EALREADY should this be a
1322 		 * non-blocking call.
1323 		 */
1324 		err = -EALREADY;
1325 		if (flags & O_NONBLOCK)
1326 			goto out;
1327 		break;
1328 	default:
1329 		if ((sk->sk_state == TCP_LISTEN) ||
1330 		    vsock_addr_cast(addr, addr_len, &remote_addr) != 0) {
1331 			err = -EINVAL;
1332 			goto out;
1333 		}
1334 
1335 		/* Set the remote address that we are connecting to. */
1336 		memcpy(&vsk->remote_addr, remote_addr,
1337 		       sizeof(vsk->remote_addr));
1338 
1339 		err = vsock_assign_transport(vsk, NULL);
1340 		if (err)
1341 			goto out;
1342 
1343 		transport = vsk->transport;
1344 
1345 		/* The hypervisor and well-known contexts do not have socket
1346 		 * endpoints.
1347 		 */
1348 		if (!transport ||
1349 		    !transport->stream_allow(remote_addr->svm_cid,
1350 					     remote_addr->svm_port)) {
1351 			err = -ENETUNREACH;
1352 			goto out;
1353 		}
1354 
1355 		err = vsock_auto_bind(vsk);
1356 		if (err)
1357 			goto out;
1358 
1359 		sk->sk_state = TCP_SYN_SENT;
1360 
1361 		err = transport->connect(vsk);
1362 		if (err < 0)
1363 			goto out;
1364 
1365 		/* Mark sock as connecting and set the error code to in
1366 		 * progress in case this is a non-blocking connect.
1367 		 */
1368 		sock->state = SS_CONNECTING;
1369 		err = -EINPROGRESS;
1370 	}
1371 
1372 	/* The receive path will handle all communication until we are able to
1373 	 * enter the connected state.  Here we wait for the connection to be
1374 	 * completed or a notification of an error.
1375 	 */
1376 	timeout = vsk->connect_timeout;
1377 	prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1378 
1379 	while (sk->sk_state != TCP_ESTABLISHED && sk->sk_err == 0) {
1380 		if (flags & O_NONBLOCK) {
1381 			/* If we're not going to block, we schedule a timeout
1382 			 * function to generate a timeout on the connection
1383 			 * attempt, in case the peer doesn't respond in a
1384 			 * timely manner. We hold on to the socket until the
1385 			 * timeout fires.
1386 			 */
1387 			sock_hold(sk);
1388 			schedule_delayed_work(&vsk->connect_work, timeout);
1389 
1390 			/* Skip ahead to preserve error code set above. */
1391 			goto out_wait;
1392 		}
1393 
1394 		release_sock(sk);
1395 		timeout = schedule_timeout(timeout);
1396 		lock_sock(sk);
1397 
1398 		if (signal_pending(current)) {
1399 			err = sock_intr_errno(timeout);
1400 			sk->sk_state = sk->sk_state == TCP_ESTABLISHED ? TCP_CLOSING : TCP_CLOSE;
1401 			sock->state = SS_UNCONNECTED;
1402 			vsock_transport_cancel_pkt(vsk);
1403 			goto out_wait;
1404 		} else if (timeout == 0) {
1405 			err = -ETIMEDOUT;
1406 			sk->sk_state = TCP_CLOSE;
1407 			sock->state = SS_UNCONNECTED;
1408 			vsock_transport_cancel_pkt(vsk);
1409 			goto out_wait;
1410 		}
1411 
1412 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1413 	}
1414 
1415 	if (sk->sk_err) {
1416 		err = -sk->sk_err;
1417 		sk->sk_state = TCP_CLOSE;
1418 		sock->state = SS_UNCONNECTED;
1419 	} else {
1420 		err = 0;
1421 	}
1422 
1423 out_wait:
1424 	finish_wait(sk_sleep(sk), &wait);
1425 out:
1426 	release_sock(sk);
1427 	return err;
1428 }
1429 
1430 static int vsock_accept(struct socket *sock, struct socket *newsock, int flags,
1431 			bool kern)
1432 {
1433 	struct sock *listener;
1434 	int err;
1435 	struct sock *connected;
1436 	struct vsock_sock *vconnected;
1437 	long timeout;
1438 	DEFINE_WAIT(wait);
1439 
1440 	err = 0;
1441 	listener = sock->sk;
1442 
1443 	lock_sock(listener);
1444 
1445 	if (!sock_type_connectible(sock->type)) {
1446 		err = -EOPNOTSUPP;
1447 		goto out;
1448 	}
1449 
1450 	if (listener->sk_state != TCP_LISTEN) {
1451 		err = -EINVAL;
1452 		goto out;
1453 	}
1454 
1455 	/* Wait for children sockets to appear; these are the new sockets
1456 	 * created upon connection establishment.
1457 	 */
1458 	timeout = sock_rcvtimeo(listener, flags & O_NONBLOCK);
1459 	prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1460 
1461 	while ((connected = vsock_dequeue_accept(listener)) == NULL &&
1462 	       listener->sk_err == 0) {
1463 		release_sock(listener);
1464 		timeout = schedule_timeout(timeout);
1465 		finish_wait(sk_sleep(listener), &wait);
1466 		lock_sock(listener);
1467 
1468 		if (signal_pending(current)) {
1469 			err = sock_intr_errno(timeout);
1470 			goto out;
1471 		} else if (timeout == 0) {
1472 			err = -EAGAIN;
1473 			goto out;
1474 		}
1475 
1476 		prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1477 	}
1478 	finish_wait(sk_sleep(listener), &wait);
1479 
1480 	if (listener->sk_err)
1481 		err = -listener->sk_err;
1482 
1483 	if (connected) {
1484 		sk_acceptq_removed(listener);
1485 
1486 		lock_sock_nested(connected, SINGLE_DEPTH_NESTING);
1487 		vconnected = vsock_sk(connected);
1488 
1489 		/* If the listener socket has received an error, then we should
1490 		 * reject this socket and return.  Note that we simply mark the
1491 		 * socket rejected, drop our reference, and let the cleanup
1492 		 * function handle the cleanup; the fact that we found it in
1493 		 * the listener's accept queue guarantees that the cleanup
1494 		 * function hasn't run yet.
1495 		 */
1496 		if (err) {
1497 			vconnected->rejected = true;
1498 		} else {
1499 			newsock->state = SS_CONNECTED;
1500 			sock_graft(connected, newsock);
1501 		}
1502 
1503 		release_sock(connected);
1504 		sock_put(connected);
1505 	}
1506 
1507 out:
1508 	release_sock(listener);
1509 	return err;
1510 }
1511 
1512 static int vsock_listen(struct socket *sock, int backlog)
1513 {
1514 	int err;
1515 	struct sock *sk;
1516 	struct vsock_sock *vsk;
1517 
1518 	sk = sock->sk;
1519 
1520 	lock_sock(sk);
1521 
1522 	if (!sock_type_connectible(sk->sk_type)) {
1523 		err = -EOPNOTSUPP;
1524 		goto out;
1525 	}
1526 
1527 	if (sock->state != SS_UNCONNECTED) {
1528 		err = -EINVAL;
1529 		goto out;
1530 	}
1531 
1532 	vsk = vsock_sk(sk);
1533 
1534 	if (!vsock_addr_bound(&vsk->local_addr)) {
1535 		err = -EINVAL;
1536 		goto out;
1537 	}
1538 
1539 	sk->sk_max_ack_backlog = backlog;
1540 	sk->sk_state = TCP_LISTEN;
1541 
1542 	err = 0;
1543 
1544 out:
1545 	release_sock(sk);
1546 	return err;
1547 }
1548 
1549 static void vsock_update_buffer_size(struct vsock_sock *vsk,
1550 				     const struct vsock_transport *transport,
1551 				     u64 val)
1552 {
1553 	if (val > vsk->buffer_max_size)
1554 		val = vsk->buffer_max_size;
1555 
1556 	if (val < vsk->buffer_min_size)
1557 		val = vsk->buffer_min_size;
1558 
1559 	if (val != vsk->buffer_size &&
1560 	    transport && transport->notify_buffer_size)
1561 		transport->notify_buffer_size(vsk, &val);
1562 
1563 	vsk->buffer_size = val;
1564 }
1565 
1566 static int vsock_connectible_setsockopt(struct socket *sock,
1567 					int level,
1568 					int optname,
1569 					sockptr_t optval,
1570 					unsigned int optlen)
1571 {
1572 	int err;
1573 	struct sock *sk;
1574 	struct vsock_sock *vsk;
1575 	const struct vsock_transport *transport;
1576 	u64 val;
1577 
1578 	if (level != AF_VSOCK)
1579 		return -ENOPROTOOPT;
1580 
1581 #define COPY_IN(_v)                                       \
1582 	do {						  \
1583 		if (optlen < sizeof(_v)) {		  \
1584 			err = -EINVAL;			  \
1585 			goto exit;			  \
1586 		}					  \
1587 		if (copy_from_sockptr(&_v, optval, sizeof(_v)) != 0) {	\
1588 			err = -EFAULT;					\
1589 			goto exit;					\
1590 		}							\
1591 	} while (0)
1592 
1593 	err = 0;
1594 	sk = sock->sk;
1595 	vsk = vsock_sk(sk);
1596 
1597 	lock_sock(sk);
1598 
1599 	transport = vsk->transport;
1600 
1601 	switch (optname) {
1602 	case SO_VM_SOCKETS_BUFFER_SIZE:
1603 		COPY_IN(val);
1604 		vsock_update_buffer_size(vsk, transport, val);
1605 		break;
1606 
1607 	case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1608 		COPY_IN(val);
1609 		vsk->buffer_max_size = val;
1610 		vsock_update_buffer_size(vsk, transport, vsk->buffer_size);
1611 		break;
1612 
1613 	case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1614 		COPY_IN(val);
1615 		vsk->buffer_min_size = val;
1616 		vsock_update_buffer_size(vsk, transport, vsk->buffer_size);
1617 		break;
1618 
1619 	case SO_VM_SOCKETS_CONNECT_TIMEOUT_NEW:
1620 	case SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD: {
1621 		struct __kernel_sock_timeval tv;
1622 
1623 		err = sock_copy_user_timeval(&tv, optval, optlen,
1624 					     optname == SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD);
1625 		if (err)
1626 			break;
1627 		if (tv.tv_sec >= 0 && tv.tv_usec < USEC_PER_SEC &&
1628 		    tv.tv_sec < (MAX_SCHEDULE_TIMEOUT / HZ - 1)) {
1629 			vsk->connect_timeout = tv.tv_sec * HZ +
1630 				DIV_ROUND_UP((unsigned long)tv.tv_usec, (USEC_PER_SEC / HZ));
1631 			if (vsk->connect_timeout == 0)
1632 				vsk->connect_timeout =
1633 				    VSOCK_DEFAULT_CONNECT_TIMEOUT;
1634 
1635 		} else {
1636 			err = -ERANGE;
1637 		}
1638 		break;
1639 	}
1640 
1641 	default:
1642 		err = -ENOPROTOOPT;
1643 		break;
1644 	}
1645 
1646 #undef COPY_IN
1647 
1648 exit:
1649 	release_sock(sk);
1650 	return err;
1651 }
1652 
1653 static int vsock_connectible_getsockopt(struct socket *sock,
1654 					int level, int optname,
1655 					char __user *optval,
1656 					int __user *optlen)
1657 {
1658 	struct sock *sk = sock->sk;
1659 	struct vsock_sock *vsk = vsock_sk(sk);
1660 
1661 	union {
1662 		u64 val64;
1663 		struct old_timeval32 tm32;
1664 		struct __kernel_old_timeval tm;
1665 		struct  __kernel_sock_timeval stm;
1666 	} v;
1667 
1668 	int lv = sizeof(v.val64);
1669 	int len;
1670 
1671 	if (level != AF_VSOCK)
1672 		return -ENOPROTOOPT;
1673 
1674 	if (get_user(len, optlen))
1675 		return -EFAULT;
1676 
1677 	memset(&v, 0, sizeof(v));
1678 
1679 	switch (optname) {
1680 	case SO_VM_SOCKETS_BUFFER_SIZE:
1681 		v.val64 = vsk->buffer_size;
1682 		break;
1683 
1684 	case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1685 		v.val64 = vsk->buffer_max_size;
1686 		break;
1687 
1688 	case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1689 		v.val64 = vsk->buffer_min_size;
1690 		break;
1691 
1692 	case SO_VM_SOCKETS_CONNECT_TIMEOUT_NEW:
1693 	case SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD:
1694 		lv = sock_get_timeout(vsk->connect_timeout, &v,
1695 				      optname == SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD);
1696 		break;
1697 
1698 	default:
1699 		return -ENOPROTOOPT;
1700 	}
1701 
1702 	if (len < lv)
1703 		return -EINVAL;
1704 	if (len > lv)
1705 		len = lv;
1706 	if (copy_to_user(optval, &v, len))
1707 		return -EFAULT;
1708 
1709 	if (put_user(len, optlen))
1710 		return -EFAULT;
1711 
1712 	return 0;
1713 }
1714 
1715 static int vsock_connectible_sendmsg(struct socket *sock, struct msghdr *msg,
1716 				     size_t len)
1717 {
1718 	struct sock *sk;
1719 	struct vsock_sock *vsk;
1720 	const struct vsock_transport *transport;
1721 	ssize_t total_written;
1722 	long timeout;
1723 	int err;
1724 	struct vsock_transport_send_notify_data send_data;
1725 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
1726 
1727 	sk = sock->sk;
1728 	vsk = vsock_sk(sk);
1729 	total_written = 0;
1730 	err = 0;
1731 
1732 	if (msg->msg_flags & MSG_OOB)
1733 		return -EOPNOTSUPP;
1734 
1735 	lock_sock(sk);
1736 
1737 	transport = vsk->transport;
1738 
1739 	/* Callers should not provide a destination with connection oriented
1740 	 * sockets.
1741 	 */
1742 	if (msg->msg_namelen) {
1743 		err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP;
1744 		goto out;
1745 	}
1746 
1747 	/* Send data only if both sides are not shutdown in the direction. */
1748 	if (sk->sk_shutdown & SEND_SHUTDOWN ||
1749 	    vsk->peer_shutdown & RCV_SHUTDOWN) {
1750 		err = -EPIPE;
1751 		goto out;
1752 	}
1753 
1754 	if (!transport || sk->sk_state != TCP_ESTABLISHED ||
1755 	    !vsock_addr_bound(&vsk->local_addr)) {
1756 		err = -ENOTCONN;
1757 		goto out;
1758 	}
1759 
1760 	if (!vsock_addr_bound(&vsk->remote_addr)) {
1761 		err = -EDESTADDRREQ;
1762 		goto out;
1763 	}
1764 
1765 	/* Wait for room in the produce queue to enqueue our user's data. */
1766 	timeout = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1767 
1768 	err = transport->notify_send_init(vsk, &send_data);
1769 	if (err < 0)
1770 		goto out;
1771 
1772 	while (total_written < len) {
1773 		ssize_t written;
1774 
1775 		add_wait_queue(sk_sleep(sk), &wait);
1776 		while (vsock_stream_has_space(vsk) == 0 &&
1777 		       sk->sk_err == 0 &&
1778 		       !(sk->sk_shutdown & SEND_SHUTDOWN) &&
1779 		       !(vsk->peer_shutdown & RCV_SHUTDOWN)) {
1780 
1781 			/* Don't wait for non-blocking sockets. */
1782 			if (timeout == 0) {
1783 				err = -EAGAIN;
1784 				remove_wait_queue(sk_sleep(sk), &wait);
1785 				goto out_err;
1786 			}
1787 
1788 			err = transport->notify_send_pre_block(vsk, &send_data);
1789 			if (err < 0) {
1790 				remove_wait_queue(sk_sleep(sk), &wait);
1791 				goto out_err;
1792 			}
1793 
1794 			release_sock(sk);
1795 			timeout = wait_woken(&wait, TASK_INTERRUPTIBLE, timeout);
1796 			lock_sock(sk);
1797 			if (signal_pending(current)) {
1798 				err = sock_intr_errno(timeout);
1799 				remove_wait_queue(sk_sleep(sk), &wait);
1800 				goto out_err;
1801 			} else if (timeout == 0) {
1802 				err = -EAGAIN;
1803 				remove_wait_queue(sk_sleep(sk), &wait);
1804 				goto out_err;
1805 			}
1806 		}
1807 		remove_wait_queue(sk_sleep(sk), &wait);
1808 
1809 		/* These checks occur both as part of and after the loop
1810 		 * conditional since we need to check before and after
1811 		 * sleeping.
1812 		 */
1813 		if (sk->sk_err) {
1814 			err = -sk->sk_err;
1815 			goto out_err;
1816 		} else if ((sk->sk_shutdown & SEND_SHUTDOWN) ||
1817 			   (vsk->peer_shutdown & RCV_SHUTDOWN)) {
1818 			err = -EPIPE;
1819 			goto out_err;
1820 		}
1821 
1822 		err = transport->notify_send_pre_enqueue(vsk, &send_data);
1823 		if (err < 0)
1824 			goto out_err;
1825 
1826 		/* Note that enqueue will only write as many bytes as are free
1827 		 * in the produce queue, so we don't need to ensure len is
1828 		 * smaller than the queue size.  It is the caller's
1829 		 * responsibility to check how many bytes we were able to send.
1830 		 */
1831 
1832 		if (sk->sk_type == SOCK_SEQPACKET) {
1833 			written = transport->seqpacket_enqueue(vsk,
1834 						msg, len - total_written);
1835 		} else {
1836 			written = transport->stream_enqueue(vsk,
1837 					msg, len - total_written);
1838 		}
1839 		if (written < 0) {
1840 			err = -ENOMEM;
1841 			goto out_err;
1842 		}
1843 
1844 		total_written += written;
1845 
1846 		err = transport->notify_send_post_enqueue(
1847 				vsk, written, &send_data);
1848 		if (err < 0)
1849 			goto out_err;
1850 
1851 	}
1852 
1853 out_err:
1854 	if (total_written > 0) {
1855 		/* Return number of written bytes only if:
1856 		 * 1) SOCK_STREAM socket.
1857 		 * 2) SOCK_SEQPACKET socket when whole buffer is sent.
1858 		 */
1859 		if (sk->sk_type == SOCK_STREAM || total_written == len)
1860 			err = total_written;
1861 	}
1862 out:
1863 	release_sock(sk);
1864 	return err;
1865 }
1866 
1867 static int vsock_connectible_wait_data(struct sock *sk,
1868 				       struct wait_queue_entry *wait,
1869 				       long timeout,
1870 				       struct vsock_transport_recv_notify_data *recv_data,
1871 				       size_t target)
1872 {
1873 	const struct vsock_transport *transport;
1874 	struct vsock_sock *vsk;
1875 	s64 data;
1876 	int err;
1877 
1878 	vsk = vsock_sk(sk);
1879 	err = 0;
1880 	transport = vsk->transport;
1881 
1882 	while ((data = vsock_connectible_has_data(vsk)) == 0) {
1883 		prepare_to_wait(sk_sleep(sk), wait, TASK_INTERRUPTIBLE);
1884 
1885 		if (sk->sk_err != 0 ||
1886 		    (sk->sk_shutdown & RCV_SHUTDOWN) ||
1887 		    (vsk->peer_shutdown & SEND_SHUTDOWN)) {
1888 			break;
1889 		}
1890 
1891 		/* Don't wait for non-blocking sockets. */
1892 		if (timeout == 0) {
1893 			err = -EAGAIN;
1894 			break;
1895 		}
1896 
1897 		if (recv_data) {
1898 			err = transport->notify_recv_pre_block(vsk, target, recv_data);
1899 			if (err < 0)
1900 				break;
1901 		}
1902 
1903 		release_sock(sk);
1904 		timeout = schedule_timeout(timeout);
1905 		lock_sock(sk);
1906 
1907 		if (signal_pending(current)) {
1908 			err = sock_intr_errno(timeout);
1909 			break;
1910 		} else if (timeout == 0) {
1911 			err = -EAGAIN;
1912 			break;
1913 		}
1914 	}
1915 
1916 	finish_wait(sk_sleep(sk), wait);
1917 
1918 	if (err)
1919 		return err;
1920 
1921 	/* Internal transport error when checking for available
1922 	 * data. XXX This should be changed to a connection
1923 	 * reset in a later change.
1924 	 */
1925 	if (data < 0)
1926 		return -ENOMEM;
1927 
1928 	return data;
1929 }
1930 
1931 static int __vsock_stream_recvmsg(struct sock *sk, struct msghdr *msg,
1932 				  size_t len, int flags)
1933 {
1934 	struct vsock_transport_recv_notify_data recv_data;
1935 	const struct vsock_transport *transport;
1936 	struct vsock_sock *vsk;
1937 	ssize_t copied;
1938 	size_t target;
1939 	long timeout;
1940 	int err;
1941 
1942 	DEFINE_WAIT(wait);
1943 
1944 	vsk = vsock_sk(sk);
1945 	transport = vsk->transport;
1946 
1947 	/* We must not copy less than target bytes into the user's buffer
1948 	 * before returning successfully, so we wait for the consume queue to
1949 	 * have that much data to consume before dequeueing.  Note that this
1950 	 * makes it impossible to handle cases where target is greater than the
1951 	 * queue size.
1952 	 */
1953 	target = sock_rcvlowat(sk, flags & MSG_WAITALL, len);
1954 	if (target >= transport->stream_rcvhiwat(vsk)) {
1955 		err = -ENOMEM;
1956 		goto out;
1957 	}
1958 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1959 	copied = 0;
1960 
1961 	err = transport->notify_recv_init(vsk, target, &recv_data);
1962 	if (err < 0)
1963 		goto out;
1964 
1965 
1966 	while (1) {
1967 		ssize_t read;
1968 
1969 		err = vsock_connectible_wait_data(sk, &wait, timeout,
1970 						  &recv_data, target);
1971 		if (err <= 0)
1972 			break;
1973 
1974 		err = transport->notify_recv_pre_dequeue(vsk, target,
1975 							 &recv_data);
1976 		if (err < 0)
1977 			break;
1978 
1979 		read = transport->stream_dequeue(vsk, msg, len - copied, flags);
1980 		if (read < 0) {
1981 			err = -ENOMEM;
1982 			break;
1983 		}
1984 
1985 		copied += read;
1986 
1987 		err = transport->notify_recv_post_dequeue(vsk, target, read,
1988 						!(flags & MSG_PEEK), &recv_data);
1989 		if (err < 0)
1990 			goto out;
1991 
1992 		if (read >= target || flags & MSG_PEEK)
1993 			break;
1994 
1995 		target -= read;
1996 	}
1997 
1998 	if (sk->sk_err)
1999 		err = -sk->sk_err;
2000 	else if (sk->sk_shutdown & RCV_SHUTDOWN)
2001 		err = 0;
2002 
2003 	if (copied > 0)
2004 		err = copied;
2005 
2006 out:
2007 	return err;
2008 }
2009 
2010 static int __vsock_seqpacket_recvmsg(struct sock *sk, struct msghdr *msg,
2011 				     size_t len, int flags)
2012 {
2013 	const struct vsock_transport *transport;
2014 	struct vsock_sock *vsk;
2015 	ssize_t msg_len;
2016 	long timeout;
2017 	int err = 0;
2018 	DEFINE_WAIT(wait);
2019 
2020 	vsk = vsock_sk(sk);
2021 	transport = vsk->transport;
2022 
2023 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2024 
2025 	err = vsock_connectible_wait_data(sk, &wait, timeout, NULL, 0);
2026 	if (err <= 0)
2027 		goto out;
2028 
2029 	msg_len = transport->seqpacket_dequeue(vsk, msg, flags);
2030 
2031 	if (msg_len < 0) {
2032 		err = -ENOMEM;
2033 		goto out;
2034 	}
2035 
2036 	if (sk->sk_err) {
2037 		err = -sk->sk_err;
2038 	} else if (sk->sk_shutdown & RCV_SHUTDOWN) {
2039 		err = 0;
2040 	} else {
2041 		/* User sets MSG_TRUNC, so return real length of
2042 		 * packet.
2043 		 */
2044 		if (flags & MSG_TRUNC)
2045 			err = msg_len;
2046 		else
2047 			err = len - msg_data_left(msg);
2048 
2049 		/* Always set MSG_TRUNC if real length of packet is
2050 		 * bigger than user's buffer.
2051 		 */
2052 		if (msg_len > len)
2053 			msg->msg_flags |= MSG_TRUNC;
2054 	}
2055 
2056 out:
2057 	return err;
2058 }
2059 
2060 static int
2061 vsock_connectible_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
2062 			  int flags)
2063 {
2064 	struct sock *sk;
2065 	struct vsock_sock *vsk;
2066 	const struct vsock_transport *transport;
2067 	int err;
2068 
2069 	DEFINE_WAIT(wait);
2070 
2071 	sk = sock->sk;
2072 	vsk = vsock_sk(sk);
2073 	err = 0;
2074 
2075 	lock_sock(sk);
2076 
2077 	transport = vsk->transport;
2078 
2079 	if (!transport || sk->sk_state != TCP_ESTABLISHED) {
2080 		/* Recvmsg is supposed to return 0 if a peer performs an
2081 		 * orderly shutdown. Differentiate between that case and when a
2082 		 * peer has not connected or a local shutdown occurred with the
2083 		 * SOCK_DONE flag.
2084 		 */
2085 		if (sock_flag(sk, SOCK_DONE))
2086 			err = 0;
2087 		else
2088 			err = -ENOTCONN;
2089 
2090 		goto out;
2091 	}
2092 
2093 	if (flags & MSG_OOB) {
2094 		err = -EOPNOTSUPP;
2095 		goto out;
2096 	}
2097 
2098 	/* We don't check peer_shutdown flag here since peer may actually shut
2099 	 * down, but there can be data in the queue that a local socket can
2100 	 * receive.
2101 	 */
2102 	if (sk->sk_shutdown & RCV_SHUTDOWN) {
2103 		err = 0;
2104 		goto out;
2105 	}
2106 
2107 	/* It is valid on Linux to pass in a zero-length receive buffer.  This
2108 	 * is not an error.  We may as well bail out now.
2109 	 */
2110 	if (!len) {
2111 		err = 0;
2112 		goto out;
2113 	}
2114 
2115 	if (sk->sk_type == SOCK_STREAM)
2116 		err = __vsock_stream_recvmsg(sk, msg, len, flags);
2117 	else
2118 		err = __vsock_seqpacket_recvmsg(sk, msg, len, flags);
2119 
2120 out:
2121 	release_sock(sk);
2122 	return err;
2123 }
2124 
2125 static const struct proto_ops vsock_stream_ops = {
2126 	.family = PF_VSOCK,
2127 	.owner = THIS_MODULE,
2128 	.release = vsock_release,
2129 	.bind = vsock_bind,
2130 	.connect = vsock_connect,
2131 	.socketpair = sock_no_socketpair,
2132 	.accept = vsock_accept,
2133 	.getname = vsock_getname,
2134 	.poll = vsock_poll,
2135 	.ioctl = sock_no_ioctl,
2136 	.listen = vsock_listen,
2137 	.shutdown = vsock_shutdown,
2138 	.setsockopt = vsock_connectible_setsockopt,
2139 	.getsockopt = vsock_connectible_getsockopt,
2140 	.sendmsg = vsock_connectible_sendmsg,
2141 	.recvmsg = vsock_connectible_recvmsg,
2142 	.mmap = sock_no_mmap,
2143 	.sendpage = sock_no_sendpage,
2144 };
2145 
2146 static const struct proto_ops vsock_seqpacket_ops = {
2147 	.family = PF_VSOCK,
2148 	.owner = THIS_MODULE,
2149 	.release = vsock_release,
2150 	.bind = vsock_bind,
2151 	.connect = vsock_connect,
2152 	.socketpair = sock_no_socketpair,
2153 	.accept = vsock_accept,
2154 	.getname = vsock_getname,
2155 	.poll = vsock_poll,
2156 	.ioctl = sock_no_ioctl,
2157 	.listen = vsock_listen,
2158 	.shutdown = vsock_shutdown,
2159 	.setsockopt = vsock_connectible_setsockopt,
2160 	.getsockopt = vsock_connectible_getsockopt,
2161 	.sendmsg = vsock_connectible_sendmsg,
2162 	.recvmsg = vsock_connectible_recvmsg,
2163 	.mmap = sock_no_mmap,
2164 	.sendpage = sock_no_sendpage,
2165 };
2166 
2167 static int vsock_create(struct net *net, struct socket *sock,
2168 			int protocol, int kern)
2169 {
2170 	struct vsock_sock *vsk;
2171 	struct sock *sk;
2172 	int ret;
2173 
2174 	if (!sock)
2175 		return -EINVAL;
2176 
2177 	if (protocol && protocol != PF_VSOCK)
2178 		return -EPROTONOSUPPORT;
2179 
2180 	switch (sock->type) {
2181 	case SOCK_DGRAM:
2182 		sock->ops = &vsock_dgram_ops;
2183 		break;
2184 	case SOCK_STREAM:
2185 		sock->ops = &vsock_stream_ops;
2186 		break;
2187 	case SOCK_SEQPACKET:
2188 		sock->ops = &vsock_seqpacket_ops;
2189 		break;
2190 	default:
2191 		return -ESOCKTNOSUPPORT;
2192 	}
2193 
2194 	sock->state = SS_UNCONNECTED;
2195 
2196 	sk = __vsock_create(net, sock, NULL, GFP_KERNEL, 0, kern);
2197 	if (!sk)
2198 		return -ENOMEM;
2199 
2200 	vsk = vsock_sk(sk);
2201 
2202 	if (sock->type == SOCK_DGRAM) {
2203 		ret = vsock_assign_transport(vsk, NULL);
2204 		if (ret < 0) {
2205 			sock_put(sk);
2206 			return ret;
2207 		}
2208 	}
2209 
2210 	vsock_insert_unbound(vsk);
2211 
2212 	return 0;
2213 }
2214 
2215 static const struct net_proto_family vsock_family_ops = {
2216 	.family = AF_VSOCK,
2217 	.create = vsock_create,
2218 	.owner = THIS_MODULE,
2219 };
2220 
2221 static long vsock_dev_do_ioctl(struct file *filp,
2222 			       unsigned int cmd, void __user *ptr)
2223 {
2224 	u32 __user *p = ptr;
2225 	u32 cid = VMADDR_CID_ANY;
2226 	int retval = 0;
2227 
2228 	switch (cmd) {
2229 	case IOCTL_VM_SOCKETS_GET_LOCAL_CID:
2230 		/* To be compatible with the VMCI behavior, we prioritize the
2231 		 * guest CID instead of well-know host CID (VMADDR_CID_HOST).
2232 		 */
2233 		if (transport_g2h)
2234 			cid = transport_g2h->get_local_cid();
2235 		else if (transport_h2g)
2236 			cid = transport_h2g->get_local_cid();
2237 
2238 		if (put_user(cid, p) != 0)
2239 			retval = -EFAULT;
2240 		break;
2241 
2242 	default:
2243 		retval = -ENOIOCTLCMD;
2244 	}
2245 
2246 	return retval;
2247 }
2248 
2249 static long vsock_dev_ioctl(struct file *filp,
2250 			    unsigned int cmd, unsigned long arg)
2251 {
2252 	return vsock_dev_do_ioctl(filp, cmd, (void __user *)arg);
2253 }
2254 
2255 #ifdef CONFIG_COMPAT
2256 static long vsock_dev_compat_ioctl(struct file *filp,
2257 				   unsigned int cmd, unsigned long arg)
2258 {
2259 	return vsock_dev_do_ioctl(filp, cmd, compat_ptr(arg));
2260 }
2261 #endif
2262 
2263 static const struct file_operations vsock_device_ops = {
2264 	.owner		= THIS_MODULE,
2265 	.unlocked_ioctl	= vsock_dev_ioctl,
2266 #ifdef CONFIG_COMPAT
2267 	.compat_ioctl	= vsock_dev_compat_ioctl,
2268 #endif
2269 	.open		= nonseekable_open,
2270 };
2271 
2272 static struct miscdevice vsock_device = {
2273 	.name		= "vsock",
2274 	.fops		= &vsock_device_ops,
2275 };
2276 
2277 static int __init vsock_init(void)
2278 {
2279 	int err = 0;
2280 
2281 	vsock_init_tables();
2282 
2283 	vsock_proto.owner = THIS_MODULE;
2284 	vsock_device.minor = MISC_DYNAMIC_MINOR;
2285 	err = misc_register(&vsock_device);
2286 	if (err) {
2287 		pr_err("Failed to register misc device\n");
2288 		goto err_reset_transport;
2289 	}
2290 
2291 	err = proto_register(&vsock_proto, 1);	/* we want our slab */
2292 	if (err) {
2293 		pr_err("Cannot register vsock protocol\n");
2294 		goto err_deregister_misc;
2295 	}
2296 
2297 	err = sock_register(&vsock_family_ops);
2298 	if (err) {
2299 		pr_err("could not register af_vsock (%d) address family: %d\n",
2300 		       AF_VSOCK, err);
2301 		goto err_unregister_proto;
2302 	}
2303 
2304 	return 0;
2305 
2306 err_unregister_proto:
2307 	proto_unregister(&vsock_proto);
2308 err_deregister_misc:
2309 	misc_deregister(&vsock_device);
2310 err_reset_transport:
2311 	return err;
2312 }
2313 
2314 static void __exit vsock_exit(void)
2315 {
2316 	misc_deregister(&vsock_device);
2317 	sock_unregister(AF_VSOCK);
2318 	proto_unregister(&vsock_proto);
2319 }
2320 
2321 const struct vsock_transport *vsock_core_get_transport(struct vsock_sock *vsk)
2322 {
2323 	return vsk->transport;
2324 }
2325 EXPORT_SYMBOL_GPL(vsock_core_get_transport);
2326 
2327 int vsock_core_register(const struct vsock_transport *t, int features)
2328 {
2329 	const struct vsock_transport *t_h2g, *t_g2h, *t_dgram, *t_local;
2330 	int err = mutex_lock_interruptible(&vsock_register_mutex);
2331 
2332 	if (err)
2333 		return err;
2334 
2335 	t_h2g = transport_h2g;
2336 	t_g2h = transport_g2h;
2337 	t_dgram = transport_dgram;
2338 	t_local = transport_local;
2339 
2340 	if (features & VSOCK_TRANSPORT_F_H2G) {
2341 		if (t_h2g) {
2342 			err = -EBUSY;
2343 			goto err_busy;
2344 		}
2345 		t_h2g = t;
2346 	}
2347 
2348 	if (features & VSOCK_TRANSPORT_F_G2H) {
2349 		if (t_g2h) {
2350 			err = -EBUSY;
2351 			goto err_busy;
2352 		}
2353 		t_g2h = t;
2354 	}
2355 
2356 	if (features & VSOCK_TRANSPORT_F_DGRAM) {
2357 		if (t_dgram) {
2358 			err = -EBUSY;
2359 			goto err_busy;
2360 		}
2361 		t_dgram = t;
2362 	}
2363 
2364 	if (features & VSOCK_TRANSPORT_F_LOCAL) {
2365 		if (t_local) {
2366 			err = -EBUSY;
2367 			goto err_busy;
2368 		}
2369 		t_local = t;
2370 	}
2371 
2372 	transport_h2g = t_h2g;
2373 	transport_g2h = t_g2h;
2374 	transport_dgram = t_dgram;
2375 	transport_local = t_local;
2376 
2377 err_busy:
2378 	mutex_unlock(&vsock_register_mutex);
2379 	return err;
2380 }
2381 EXPORT_SYMBOL_GPL(vsock_core_register);
2382 
2383 void vsock_core_unregister(const struct vsock_transport *t)
2384 {
2385 	mutex_lock(&vsock_register_mutex);
2386 
2387 	if (transport_h2g == t)
2388 		transport_h2g = NULL;
2389 
2390 	if (transport_g2h == t)
2391 		transport_g2h = NULL;
2392 
2393 	if (transport_dgram == t)
2394 		transport_dgram = NULL;
2395 
2396 	if (transport_local == t)
2397 		transport_local = NULL;
2398 
2399 	mutex_unlock(&vsock_register_mutex);
2400 }
2401 EXPORT_SYMBOL_GPL(vsock_core_unregister);
2402 
2403 module_init(vsock_init);
2404 module_exit(vsock_exit);
2405 
2406 MODULE_AUTHOR("VMware, Inc.");
2407 MODULE_DESCRIPTION("VMware Virtual Socket Family");
2408 MODULE_VERSION("1.0.2.0-k");
2409 MODULE_LICENSE("GPL v2");
2410