xref: /openbmc/linux/net/vmw_vsock/af_vsock.c (revision bbb774d9)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * VMware vSockets Driver
4  *
5  * Copyright (C) 2007-2013 VMware, Inc. All rights reserved.
6  */
7 
8 /* Implementation notes:
9  *
10  * - There are two kinds of sockets: those created by user action (such as
11  * calling socket(2)) and those created by incoming connection request packets.
12  *
13  * - There are two "global" tables, one for bound sockets (sockets that have
14  * specified an address that they are responsible for) and one for connected
15  * sockets (sockets that have established a connection with another socket).
16  * These tables are "global" in that all sockets on the system are placed
17  * within them. - Note, though, that the bound table contains an extra entry
18  * for a list of unbound sockets and SOCK_DGRAM sockets will always remain in
19  * that list. The bound table is used solely for lookup of sockets when packets
20  * are received and that's not necessary for SOCK_DGRAM sockets since we create
21  * a datagram handle for each and need not perform a lookup.  Keeping SOCK_DGRAM
22  * sockets out of the bound hash buckets will reduce the chance of collisions
23  * when looking for SOCK_STREAM sockets and prevents us from having to check the
24  * socket type in the hash table lookups.
25  *
26  * - Sockets created by user action will either be "client" sockets that
27  * initiate a connection or "server" sockets that listen for connections; we do
28  * not support simultaneous connects (two "client" sockets connecting).
29  *
30  * - "Server" sockets are referred to as listener sockets throughout this
31  * implementation because they are in the TCP_LISTEN state.  When a
32  * connection request is received (the second kind of socket mentioned above),
33  * we create a new socket and refer to it as a pending socket.  These pending
34  * sockets are placed on the pending connection list of the listener socket.
35  * When future packets are received for the address the listener socket is
36  * bound to, we check if the source of the packet is from one that has an
37  * existing pending connection.  If it does, we process the packet for the
38  * pending socket.  When that socket reaches the connected state, it is removed
39  * from the listener socket's pending list and enqueued in the listener
40  * socket's accept queue.  Callers of accept(2) will accept connected sockets
41  * from the listener socket's accept queue.  If the socket cannot be accepted
42  * for some reason then it is marked rejected.  Once the connection is
43  * accepted, it is owned by the user process and the responsibility for cleanup
44  * falls with that user process.
45  *
46  * - It is possible that these pending sockets will never reach the connected
47  * state; in fact, we may never receive another packet after the connection
48  * request.  Because of this, we must schedule a cleanup function to run in the
49  * future, after some amount of time passes where a connection should have been
50  * established.  This function ensures that the socket is off all lists so it
51  * cannot be retrieved, then drops all references to the socket so it is cleaned
52  * up (sock_put() -> sk_free() -> our sk_destruct implementation).  Note this
53  * function will also cleanup rejected sockets, those that reach the connected
54  * state but leave it before they have been accepted.
55  *
56  * - Lock ordering for pending or accept queue sockets is:
57  *
58  *     lock_sock(listener);
59  *     lock_sock_nested(pending, SINGLE_DEPTH_NESTING);
60  *
61  * Using explicit nested locking keeps lockdep happy since normally only one
62  * lock of a given class may be taken at a time.
63  *
64  * - Sockets created by user action will be cleaned up when the user process
65  * calls close(2), causing our release implementation to be called. Our release
66  * implementation will perform some cleanup then drop the last reference so our
67  * sk_destruct implementation is invoked.  Our sk_destruct implementation will
68  * perform additional cleanup that's common for both types of sockets.
69  *
70  * - A socket's reference count is what ensures that the structure won't be
71  * freed.  Each entry in a list (such as the "global" bound and connected tables
72  * and the listener socket's pending list and connected queue) ensures a
73  * reference.  When we defer work until process context and pass a socket as our
74  * argument, we must ensure the reference count is increased to ensure the
75  * socket isn't freed before the function is run; the deferred function will
76  * then drop the reference.
77  *
78  * - sk->sk_state uses the TCP state constants because they are widely used by
79  * other address families and exposed to userspace tools like ss(8):
80  *
81  *   TCP_CLOSE - unconnected
82  *   TCP_SYN_SENT - connecting
83  *   TCP_ESTABLISHED - connected
84  *   TCP_CLOSING - disconnecting
85  *   TCP_LISTEN - listening
86  */
87 
88 #include <linux/compat.h>
89 #include <linux/types.h>
90 #include <linux/bitops.h>
91 #include <linux/cred.h>
92 #include <linux/init.h>
93 #include <linux/io.h>
94 #include <linux/kernel.h>
95 #include <linux/sched/signal.h>
96 #include <linux/kmod.h>
97 #include <linux/list.h>
98 #include <linux/miscdevice.h>
99 #include <linux/module.h>
100 #include <linux/mutex.h>
101 #include <linux/net.h>
102 #include <linux/poll.h>
103 #include <linux/random.h>
104 #include <linux/skbuff.h>
105 #include <linux/smp.h>
106 #include <linux/socket.h>
107 #include <linux/stddef.h>
108 #include <linux/unistd.h>
109 #include <linux/wait.h>
110 #include <linux/workqueue.h>
111 #include <net/sock.h>
112 #include <net/af_vsock.h>
113 
114 static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr);
115 static void vsock_sk_destruct(struct sock *sk);
116 static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
117 
118 /* Protocol family. */
119 static struct proto vsock_proto = {
120 	.name = "AF_VSOCK",
121 	.owner = THIS_MODULE,
122 	.obj_size = sizeof(struct vsock_sock),
123 };
124 
125 /* The default peer timeout indicates how long we will wait for a peer response
126  * to a control message.
127  */
128 #define VSOCK_DEFAULT_CONNECT_TIMEOUT (2 * HZ)
129 
130 #define VSOCK_DEFAULT_BUFFER_SIZE     (1024 * 256)
131 #define VSOCK_DEFAULT_BUFFER_MAX_SIZE (1024 * 256)
132 #define VSOCK_DEFAULT_BUFFER_MIN_SIZE 128
133 
134 /* Transport used for host->guest communication */
135 static const struct vsock_transport *transport_h2g;
136 /* Transport used for guest->host communication */
137 static const struct vsock_transport *transport_g2h;
138 /* Transport used for DGRAM communication */
139 static const struct vsock_transport *transport_dgram;
140 /* Transport used for local communication */
141 static const struct vsock_transport *transport_local;
142 static DEFINE_MUTEX(vsock_register_mutex);
143 
144 /**** UTILS ****/
145 
146 /* Each bound VSocket is stored in the bind hash table and each connected
147  * VSocket is stored in the connected hash table.
148  *
149  * Unbound sockets are all put on the same list attached to the end of the hash
150  * table (vsock_unbound_sockets).  Bound sockets are added to the hash table in
151  * the bucket that their local address hashes to (vsock_bound_sockets(addr)
152  * represents the list that addr hashes to).
153  *
154  * Specifically, we initialize the vsock_bind_table array to a size of
155  * VSOCK_HASH_SIZE + 1 so that vsock_bind_table[0] through
156  * vsock_bind_table[VSOCK_HASH_SIZE - 1] are for bound sockets and
157  * vsock_bind_table[VSOCK_HASH_SIZE] is for unbound sockets.  The hash function
158  * mods with VSOCK_HASH_SIZE to ensure this.
159  */
160 #define MAX_PORT_RETRIES        24
161 
162 #define VSOCK_HASH(addr)        ((addr)->svm_port % VSOCK_HASH_SIZE)
163 #define vsock_bound_sockets(addr) (&vsock_bind_table[VSOCK_HASH(addr)])
164 #define vsock_unbound_sockets     (&vsock_bind_table[VSOCK_HASH_SIZE])
165 
166 /* XXX This can probably be implemented in a better way. */
167 #define VSOCK_CONN_HASH(src, dst)				\
168 	(((src)->svm_cid ^ (dst)->svm_port) % VSOCK_HASH_SIZE)
169 #define vsock_connected_sockets(src, dst)		\
170 	(&vsock_connected_table[VSOCK_CONN_HASH(src, dst)])
171 #define vsock_connected_sockets_vsk(vsk)				\
172 	vsock_connected_sockets(&(vsk)->remote_addr, &(vsk)->local_addr)
173 
174 struct list_head vsock_bind_table[VSOCK_HASH_SIZE + 1];
175 EXPORT_SYMBOL_GPL(vsock_bind_table);
176 struct list_head vsock_connected_table[VSOCK_HASH_SIZE];
177 EXPORT_SYMBOL_GPL(vsock_connected_table);
178 DEFINE_SPINLOCK(vsock_table_lock);
179 EXPORT_SYMBOL_GPL(vsock_table_lock);
180 
181 /* Autobind this socket to the local address if necessary. */
182 static int vsock_auto_bind(struct vsock_sock *vsk)
183 {
184 	struct sock *sk = sk_vsock(vsk);
185 	struct sockaddr_vm local_addr;
186 
187 	if (vsock_addr_bound(&vsk->local_addr))
188 		return 0;
189 	vsock_addr_init(&local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
190 	return __vsock_bind(sk, &local_addr);
191 }
192 
193 static void vsock_init_tables(void)
194 {
195 	int i;
196 
197 	for (i = 0; i < ARRAY_SIZE(vsock_bind_table); i++)
198 		INIT_LIST_HEAD(&vsock_bind_table[i]);
199 
200 	for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++)
201 		INIT_LIST_HEAD(&vsock_connected_table[i]);
202 }
203 
204 static void __vsock_insert_bound(struct list_head *list,
205 				 struct vsock_sock *vsk)
206 {
207 	sock_hold(&vsk->sk);
208 	list_add(&vsk->bound_table, list);
209 }
210 
211 static void __vsock_insert_connected(struct list_head *list,
212 				     struct vsock_sock *vsk)
213 {
214 	sock_hold(&vsk->sk);
215 	list_add(&vsk->connected_table, list);
216 }
217 
218 static void __vsock_remove_bound(struct vsock_sock *vsk)
219 {
220 	list_del_init(&vsk->bound_table);
221 	sock_put(&vsk->sk);
222 }
223 
224 static void __vsock_remove_connected(struct vsock_sock *vsk)
225 {
226 	list_del_init(&vsk->connected_table);
227 	sock_put(&vsk->sk);
228 }
229 
230 static struct sock *__vsock_find_bound_socket(struct sockaddr_vm *addr)
231 {
232 	struct vsock_sock *vsk;
233 
234 	list_for_each_entry(vsk, vsock_bound_sockets(addr), bound_table) {
235 		if (vsock_addr_equals_addr(addr, &vsk->local_addr))
236 			return sk_vsock(vsk);
237 
238 		if (addr->svm_port == vsk->local_addr.svm_port &&
239 		    (vsk->local_addr.svm_cid == VMADDR_CID_ANY ||
240 		     addr->svm_cid == VMADDR_CID_ANY))
241 			return sk_vsock(vsk);
242 	}
243 
244 	return NULL;
245 }
246 
247 static struct sock *__vsock_find_connected_socket(struct sockaddr_vm *src,
248 						  struct sockaddr_vm *dst)
249 {
250 	struct vsock_sock *vsk;
251 
252 	list_for_each_entry(vsk, vsock_connected_sockets(src, dst),
253 			    connected_table) {
254 		if (vsock_addr_equals_addr(src, &vsk->remote_addr) &&
255 		    dst->svm_port == vsk->local_addr.svm_port) {
256 			return sk_vsock(vsk);
257 		}
258 	}
259 
260 	return NULL;
261 }
262 
263 static void vsock_insert_unbound(struct vsock_sock *vsk)
264 {
265 	spin_lock_bh(&vsock_table_lock);
266 	__vsock_insert_bound(vsock_unbound_sockets, vsk);
267 	spin_unlock_bh(&vsock_table_lock);
268 }
269 
270 void vsock_insert_connected(struct vsock_sock *vsk)
271 {
272 	struct list_head *list = vsock_connected_sockets(
273 		&vsk->remote_addr, &vsk->local_addr);
274 
275 	spin_lock_bh(&vsock_table_lock);
276 	__vsock_insert_connected(list, vsk);
277 	spin_unlock_bh(&vsock_table_lock);
278 }
279 EXPORT_SYMBOL_GPL(vsock_insert_connected);
280 
281 void vsock_remove_bound(struct vsock_sock *vsk)
282 {
283 	spin_lock_bh(&vsock_table_lock);
284 	if (__vsock_in_bound_table(vsk))
285 		__vsock_remove_bound(vsk);
286 	spin_unlock_bh(&vsock_table_lock);
287 }
288 EXPORT_SYMBOL_GPL(vsock_remove_bound);
289 
290 void vsock_remove_connected(struct vsock_sock *vsk)
291 {
292 	spin_lock_bh(&vsock_table_lock);
293 	if (__vsock_in_connected_table(vsk))
294 		__vsock_remove_connected(vsk);
295 	spin_unlock_bh(&vsock_table_lock);
296 }
297 EXPORT_SYMBOL_GPL(vsock_remove_connected);
298 
299 struct sock *vsock_find_bound_socket(struct sockaddr_vm *addr)
300 {
301 	struct sock *sk;
302 
303 	spin_lock_bh(&vsock_table_lock);
304 	sk = __vsock_find_bound_socket(addr);
305 	if (sk)
306 		sock_hold(sk);
307 
308 	spin_unlock_bh(&vsock_table_lock);
309 
310 	return sk;
311 }
312 EXPORT_SYMBOL_GPL(vsock_find_bound_socket);
313 
314 struct sock *vsock_find_connected_socket(struct sockaddr_vm *src,
315 					 struct sockaddr_vm *dst)
316 {
317 	struct sock *sk;
318 
319 	spin_lock_bh(&vsock_table_lock);
320 	sk = __vsock_find_connected_socket(src, dst);
321 	if (sk)
322 		sock_hold(sk);
323 
324 	spin_unlock_bh(&vsock_table_lock);
325 
326 	return sk;
327 }
328 EXPORT_SYMBOL_GPL(vsock_find_connected_socket);
329 
330 void vsock_remove_sock(struct vsock_sock *vsk)
331 {
332 	vsock_remove_bound(vsk);
333 	vsock_remove_connected(vsk);
334 }
335 EXPORT_SYMBOL_GPL(vsock_remove_sock);
336 
337 void vsock_for_each_connected_socket(struct vsock_transport *transport,
338 				     void (*fn)(struct sock *sk))
339 {
340 	int i;
341 
342 	spin_lock_bh(&vsock_table_lock);
343 
344 	for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) {
345 		struct vsock_sock *vsk;
346 		list_for_each_entry(vsk, &vsock_connected_table[i],
347 				    connected_table) {
348 			if (vsk->transport != transport)
349 				continue;
350 
351 			fn(sk_vsock(vsk));
352 		}
353 	}
354 
355 	spin_unlock_bh(&vsock_table_lock);
356 }
357 EXPORT_SYMBOL_GPL(vsock_for_each_connected_socket);
358 
359 void vsock_add_pending(struct sock *listener, struct sock *pending)
360 {
361 	struct vsock_sock *vlistener;
362 	struct vsock_sock *vpending;
363 
364 	vlistener = vsock_sk(listener);
365 	vpending = vsock_sk(pending);
366 
367 	sock_hold(pending);
368 	sock_hold(listener);
369 	list_add_tail(&vpending->pending_links, &vlistener->pending_links);
370 }
371 EXPORT_SYMBOL_GPL(vsock_add_pending);
372 
373 void vsock_remove_pending(struct sock *listener, struct sock *pending)
374 {
375 	struct vsock_sock *vpending = vsock_sk(pending);
376 
377 	list_del_init(&vpending->pending_links);
378 	sock_put(listener);
379 	sock_put(pending);
380 }
381 EXPORT_SYMBOL_GPL(vsock_remove_pending);
382 
383 void vsock_enqueue_accept(struct sock *listener, struct sock *connected)
384 {
385 	struct vsock_sock *vlistener;
386 	struct vsock_sock *vconnected;
387 
388 	vlistener = vsock_sk(listener);
389 	vconnected = vsock_sk(connected);
390 
391 	sock_hold(connected);
392 	sock_hold(listener);
393 	list_add_tail(&vconnected->accept_queue, &vlistener->accept_queue);
394 }
395 EXPORT_SYMBOL_GPL(vsock_enqueue_accept);
396 
397 static bool vsock_use_local_transport(unsigned int remote_cid)
398 {
399 	if (!transport_local)
400 		return false;
401 
402 	if (remote_cid == VMADDR_CID_LOCAL)
403 		return true;
404 
405 	if (transport_g2h) {
406 		return remote_cid == transport_g2h->get_local_cid();
407 	} else {
408 		return remote_cid == VMADDR_CID_HOST;
409 	}
410 }
411 
412 static void vsock_deassign_transport(struct vsock_sock *vsk)
413 {
414 	if (!vsk->transport)
415 		return;
416 
417 	vsk->transport->destruct(vsk);
418 	module_put(vsk->transport->module);
419 	vsk->transport = NULL;
420 }
421 
422 /* Assign a transport to a socket and call the .init transport callback.
423  *
424  * Note: for connection oriented socket this must be called when vsk->remote_addr
425  * is set (e.g. during the connect() or when a connection request on a listener
426  * socket is received).
427  * The vsk->remote_addr is used to decide which transport to use:
428  *  - remote CID == VMADDR_CID_LOCAL or g2h->local_cid or VMADDR_CID_HOST if
429  *    g2h is not loaded, will use local transport;
430  *  - remote CID <= VMADDR_CID_HOST or h2g is not loaded or remote flags field
431  *    includes VMADDR_FLAG_TO_HOST flag value, will use guest->host transport;
432  *  - remote CID > VMADDR_CID_HOST will use host->guest transport;
433  */
434 int vsock_assign_transport(struct vsock_sock *vsk, struct vsock_sock *psk)
435 {
436 	const struct vsock_transport *new_transport;
437 	struct sock *sk = sk_vsock(vsk);
438 	unsigned int remote_cid = vsk->remote_addr.svm_cid;
439 	__u8 remote_flags;
440 	int ret;
441 
442 	/* If the packet is coming with the source and destination CIDs higher
443 	 * than VMADDR_CID_HOST, then a vsock channel where all the packets are
444 	 * forwarded to the host should be established. Then the host will
445 	 * need to forward the packets to the guest.
446 	 *
447 	 * The flag is set on the (listen) receive path (psk is not NULL). On
448 	 * the connect path the flag can be set by the user space application.
449 	 */
450 	if (psk && vsk->local_addr.svm_cid > VMADDR_CID_HOST &&
451 	    vsk->remote_addr.svm_cid > VMADDR_CID_HOST)
452 		vsk->remote_addr.svm_flags |= VMADDR_FLAG_TO_HOST;
453 
454 	remote_flags = vsk->remote_addr.svm_flags;
455 
456 	switch (sk->sk_type) {
457 	case SOCK_DGRAM:
458 		new_transport = transport_dgram;
459 		break;
460 	case SOCK_STREAM:
461 	case SOCK_SEQPACKET:
462 		if (vsock_use_local_transport(remote_cid))
463 			new_transport = transport_local;
464 		else if (remote_cid <= VMADDR_CID_HOST || !transport_h2g ||
465 			 (remote_flags & VMADDR_FLAG_TO_HOST))
466 			new_transport = transport_g2h;
467 		else
468 			new_transport = transport_h2g;
469 		break;
470 	default:
471 		return -ESOCKTNOSUPPORT;
472 	}
473 
474 	if (vsk->transport) {
475 		if (vsk->transport == new_transport)
476 			return 0;
477 
478 		/* transport->release() must be called with sock lock acquired.
479 		 * This path can only be taken during vsock_connect(), where we
480 		 * have already held the sock lock. In the other cases, this
481 		 * function is called on a new socket which is not assigned to
482 		 * any transport.
483 		 */
484 		vsk->transport->release(vsk);
485 		vsock_deassign_transport(vsk);
486 	}
487 
488 	/* We increase the module refcnt to prevent the transport unloading
489 	 * while there are open sockets assigned to it.
490 	 */
491 	if (!new_transport || !try_module_get(new_transport->module))
492 		return -ENODEV;
493 
494 	if (sk->sk_type == SOCK_SEQPACKET) {
495 		if (!new_transport->seqpacket_allow ||
496 		    !new_transport->seqpacket_allow(remote_cid)) {
497 			module_put(new_transport->module);
498 			return -ESOCKTNOSUPPORT;
499 		}
500 	}
501 
502 	ret = new_transport->init(vsk, psk);
503 	if (ret) {
504 		module_put(new_transport->module);
505 		return ret;
506 	}
507 
508 	vsk->transport = new_transport;
509 
510 	return 0;
511 }
512 EXPORT_SYMBOL_GPL(vsock_assign_transport);
513 
514 bool vsock_find_cid(unsigned int cid)
515 {
516 	if (transport_g2h && cid == transport_g2h->get_local_cid())
517 		return true;
518 
519 	if (transport_h2g && cid == VMADDR_CID_HOST)
520 		return true;
521 
522 	if (transport_local && cid == VMADDR_CID_LOCAL)
523 		return true;
524 
525 	return false;
526 }
527 EXPORT_SYMBOL_GPL(vsock_find_cid);
528 
529 static struct sock *vsock_dequeue_accept(struct sock *listener)
530 {
531 	struct vsock_sock *vlistener;
532 	struct vsock_sock *vconnected;
533 
534 	vlistener = vsock_sk(listener);
535 
536 	if (list_empty(&vlistener->accept_queue))
537 		return NULL;
538 
539 	vconnected = list_entry(vlistener->accept_queue.next,
540 				struct vsock_sock, accept_queue);
541 
542 	list_del_init(&vconnected->accept_queue);
543 	sock_put(listener);
544 	/* The caller will need a reference on the connected socket so we let
545 	 * it call sock_put().
546 	 */
547 
548 	return sk_vsock(vconnected);
549 }
550 
551 static bool vsock_is_accept_queue_empty(struct sock *sk)
552 {
553 	struct vsock_sock *vsk = vsock_sk(sk);
554 	return list_empty(&vsk->accept_queue);
555 }
556 
557 static bool vsock_is_pending(struct sock *sk)
558 {
559 	struct vsock_sock *vsk = vsock_sk(sk);
560 	return !list_empty(&vsk->pending_links);
561 }
562 
563 static int vsock_send_shutdown(struct sock *sk, int mode)
564 {
565 	struct vsock_sock *vsk = vsock_sk(sk);
566 
567 	if (!vsk->transport)
568 		return -ENODEV;
569 
570 	return vsk->transport->shutdown(vsk, mode);
571 }
572 
573 static void vsock_pending_work(struct work_struct *work)
574 {
575 	struct sock *sk;
576 	struct sock *listener;
577 	struct vsock_sock *vsk;
578 	bool cleanup;
579 
580 	vsk = container_of(work, struct vsock_sock, pending_work.work);
581 	sk = sk_vsock(vsk);
582 	listener = vsk->listener;
583 	cleanup = true;
584 
585 	lock_sock(listener);
586 	lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
587 
588 	if (vsock_is_pending(sk)) {
589 		vsock_remove_pending(listener, sk);
590 
591 		sk_acceptq_removed(listener);
592 	} else if (!vsk->rejected) {
593 		/* We are not on the pending list and accept() did not reject
594 		 * us, so we must have been accepted by our user process.  We
595 		 * just need to drop our references to the sockets and be on
596 		 * our way.
597 		 */
598 		cleanup = false;
599 		goto out;
600 	}
601 
602 	/* We need to remove ourself from the global connected sockets list so
603 	 * incoming packets can't find this socket, and to reduce the reference
604 	 * count.
605 	 */
606 	vsock_remove_connected(vsk);
607 
608 	sk->sk_state = TCP_CLOSE;
609 
610 out:
611 	release_sock(sk);
612 	release_sock(listener);
613 	if (cleanup)
614 		sock_put(sk);
615 
616 	sock_put(sk);
617 	sock_put(listener);
618 }
619 
620 /**** SOCKET OPERATIONS ****/
621 
622 static int __vsock_bind_connectible(struct vsock_sock *vsk,
623 				    struct sockaddr_vm *addr)
624 {
625 	static u32 port;
626 	struct sockaddr_vm new_addr;
627 
628 	if (!port)
629 		port = LAST_RESERVED_PORT + 1 +
630 			prandom_u32_max(U32_MAX - LAST_RESERVED_PORT);
631 
632 	vsock_addr_init(&new_addr, addr->svm_cid, addr->svm_port);
633 
634 	if (addr->svm_port == VMADDR_PORT_ANY) {
635 		bool found = false;
636 		unsigned int i;
637 
638 		for (i = 0; i < MAX_PORT_RETRIES; i++) {
639 			if (port <= LAST_RESERVED_PORT)
640 				port = LAST_RESERVED_PORT + 1;
641 
642 			new_addr.svm_port = port++;
643 
644 			if (!__vsock_find_bound_socket(&new_addr)) {
645 				found = true;
646 				break;
647 			}
648 		}
649 
650 		if (!found)
651 			return -EADDRNOTAVAIL;
652 	} else {
653 		/* If port is in reserved range, ensure caller
654 		 * has necessary privileges.
655 		 */
656 		if (addr->svm_port <= LAST_RESERVED_PORT &&
657 		    !capable(CAP_NET_BIND_SERVICE)) {
658 			return -EACCES;
659 		}
660 
661 		if (__vsock_find_bound_socket(&new_addr))
662 			return -EADDRINUSE;
663 	}
664 
665 	vsock_addr_init(&vsk->local_addr, new_addr.svm_cid, new_addr.svm_port);
666 
667 	/* Remove connection oriented sockets from the unbound list and add them
668 	 * to the hash table for easy lookup by its address.  The unbound list
669 	 * is simply an extra entry at the end of the hash table, a trick used
670 	 * by AF_UNIX.
671 	 */
672 	__vsock_remove_bound(vsk);
673 	__vsock_insert_bound(vsock_bound_sockets(&vsk->local_addr), vsk);
674 
675 	return 0;
676 }
677 
678 static int __vsock_bind_dgram(struct vsock_sock *vsk,
679 			      struct sockaddr_vm *addr)
680 {
681 	return vsk->transport->dgram_bind(vsk, addr);
682 }
683 
684 static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr)
685 {
686 	struct vsock_sock *vsk = vsock_sk(sk);
687 	int retval;
688 
689 	/* First ensure this socket isn't already bound. */
690 	if (vsock_addr_bound(&vsk->local_addr))
691 		return -EINVAL;
692 
693 	/* Now bind to the provided address or select appropriate values if
694 	 * none are provided (VMADDR_CID_ANY and VMADDR_PORT_ANY).  Note that
695 	 * like AF_INET prevents binding to a non-local IP address (in most
696 	 * cases), we only allow binding to a local CID.
697 	 */
698 	if (addr->svm_cid != VMADDR_CID_ANY && !vsock_find_cid(addr->svm_cid))
699 		return -EADDRNOTAVAIL;
700 
701 	switch (sk->sk_socket->type) {
702 	case SOCK_STREAM:
703 	case SOCK_SEQPACKET:
704 		spin_lock_bh(&vsock_table_lock);
705 		retval = __vsock_bind_connectible(vsk, addr);
706 		spin_unlock_bh(&vsock_table_lock);
707 		break;
708 
709 	case SOCK_DGRAM:
710 		retval = __vsock_bind_dgram(vsk, addr);
711 		break;
712 
713 	default:
714 		retval = -EINVAL;
715 		break;
716 	}
717 
718 	return retval;
719 }
720 
721 static void vsock_connect_timeout(struct work_struct *work);
722 
723 static struct sock *__vsock_create(struct net *net,
724 				   struct socket *sock,
725 				   struct sock *parent,
726 				   gfp_t priority,
727 				   unsigned short type,
728 				   int kern)
729 {
730 	struct sock *sk;
731 	struct vsock_sock *psk;
732 	struct vsock_sock *vsk;
733 
734 	sk = sk_alloc(net, AF_VSOCK, priority, &vsock_proto, kern);
735 	if (!sk)
736 		return NULL;
737 
738 	sock_init_data(sock, sk);
739 
740 	/* sk->sk_type is normally set in sock_init_data, but only if sock is
741 	 * non-NULL. We make sure that our sockets always have a type by
742 	 * setting it here if needed.
743 	 */
744 	if (!sock)
745 		sk->sk_type = type;
746 
747 	vsk = vsock_sk(sk);
748 	vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
749 	vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
750 
751 	sk->sk_destruct = vsock_sk_destruct;
752 	sk->sk_backlog_rcv = vsock_queue_rcv_skb;
753 	sock_reset_flag(sk, SOCK_DONE);
754 
755 	INIT_LIST_HEAD(&vsk->bound_table);
756 	INIT_LIST_HEAD(&vsk->connected_table);
757 	vsk->listener = NULL;
758 	INIT_LIST_HEAD(&vsk->pending_links);
759 	INIT_LIST_HEAD(&vsk->accept_queue);
760 	vsk->rejected = false;
761 	vsk->sent_request = false;
762 	vsk->ignore_connecting_rst = false;
763 	vsk->peer_shutdown = 0;
764 	INIT_DELAYED_WORK(&vsk->connect_work, vsock_connect_timeout);
765 	INIT_DELAYED_WORK(&vsk->pending_work, vsock_pending_work);
766 
767 	psk = parent ? vsock_sk(parent) : NULL;
768 	if (parent) {
769 		vsk->trusted = psk->trusted;
770 		vsk->owner = get_cred(psk->owner);
771 		vsk->connect_timeout = psk->connect_timeout;
772 		vsk->buffer_size = psk->buffer_size;
773 		vsk->buffer_min_size = psk->buffer_min_size;
774 		vsk->buffer_max_size = psk->buffer_max_size;
775 		security_sk_clone(parent, sk);
776 	} else {
777 		vsk->trusted = ns_capable_noaudit(&init_user_ns, CAP_NET_ADMIN);
778 		vsk->owner = get_current_cred();
779 		vsk->connect_timeout = VSOCK_DEFAULT_CONNECT_TIMEOUT;
780 		vsk->buffer_size = VSOCK_DEFAULT_BUFFER_SIZE;
781 		vsk->buffer_min_size = VSOCK_DEFAULT_BUFFER_MIN_SIZE;
782 		vsk->buffer_max_size = VSOCK_DEFAULT_BUFFER_MAX_SIZE;
783 	}
784 
785 	return sk;
786 }
787 
788 static bool sock_type_connectible(u16 type)
789 {
790 	return (type == SOCK_STREAM) || (type == SOCK_SEQPACKET);
791 }
792 
793 static void __vsock_release(struct sock *sk, int level)
794 {
795 	if (sk) {
796 		struct sock *pending;
797 		struct vsock_sock *vsk;
798 
799 		vsk = vsock_sk(sk);
800 		pending = NULL;	/* Compiler warning. */
801 
802 		/* When "level" is SINGLE_DEPTH_NESTING, use the nested
803 		 * version to avoid the warning "possible recursive locking
804 		 * detected". When "level" is 0, lock_sock_nested(sk, level)
805 		 * is the same as lock_sock(sk).
806 		 */
807 		lock_sock_nested(sk, level);
808 
809 		if (vsk->transport)
810 			vsk->transport->release(vsk);
811 		else if (sock_type_connectible(sk->sk_type))
812 			vsock_remove_sock(vsk);
813 
814 		sock_orphan(sk);
815 		sk->sk_shutdown = SHUTDOWN_MASK;
816 
817 		skb_queue_purge(&sk->sk_receive_queue);
818 
819 		/* Clean up any sockets that never were accepted. */
820 		while ((pending = vsock_dequeue_accept(sk)) != NULL) {
821 			__vsock_release(pending, SINGLE_DEPTH_NESTING);
822 			sock_put(pending);
823 		}
824 
825 		release_sock(sk);
826 		sock_put(sk);
827 	}
828 }
829 
830 static void vsock_sk_destruct(struct sock *sk)
831 {
832 	struct vsock_sock *vsk = vsock_sk(sk);
833 
834 	vsock_deassign_transport(vsk);
835 
836 	/* When clearing these addresses, there's no need to set the family and
837 	 * possibly register the address family with the kernel.
838 	 */
839 	vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
840 	vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
841 
842 	put_cred(vsk->owner);
843 }
844 
845 static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
846 {
847 	int err;
848 
849 	err = sock_queue_rcv_skb(sk, skb);
850 	if (err)
851 		kfree_skb(skb);
852 
853 	return err;
854 }
855 
856 struct sock *vsock_create_connected(struct sock *parent)
857 {
858 	return __vsock_create(sock_net(parent), NULL, parent, GFP_KERNEL,
859 			      parent->sk_type, 0);
860 }
861 EXPORT_SYMBOL_GPL(vsock_create_connected);
862 
863 s64 vsock_stream_has_data(struct vsock_sock *vsk)
864 {
865 	return vsk->transport->stream_has_data(vsk);
866 }
867 EXPORT_SYMBOL_GPL(vsock_stream_has_data);
868 
869 static s64 vsock_connectible_has_data(struct vsock_sock *vsk)
870 {
871 	struct sock *sk = sk_vsock(vsk);
872 
873 	if (sk->sk_type == SOCK_SEQPACKET)
874 		return vsk->transport->seqpacket_has_data(vsk);
875 	else
876 		return vsock_stream_has_data(vsk);
877 }
878 
879 s64 vsock_stream_has_space(struct vsock_sock *vsk)
880 {
881 	return vsk->transport->stream_has_space(vsk);
882 }
883 EXPORT_SYMBOL_GPL(vsock_stream_has_space);
884 
885 static int vsock_release(struct socket *sock)
886 {
887 	__vsock_release(sock->sk, 0);
888 	sock->sk = NULL;
889 	sock->state = SS_FREE;
890 
891 	return 0;
892 }
893 
894 static int
895 vsock_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
896 {
897 	int err;
898 	struct sock *sk;
899 	struct sockaddr_vm *vm_addr;
900 
901 	sk = sock->sk;
902 
903 	if (vsock_addr_cast(addr, addr_len, &vm_addr) != 0)
904 		return -EINVAL;
905 
906 	lock_sock(sk);
907 	err = __vsock_bind(sk, vm_addr);
908 	release_sock(sk);
909 
910 	return err;
911 }
912 
913 static int vsock_getname(struct socket *sock,
914 			 struct sockaddr *addr, int peer)
915 {
916 	int err;
917 	struct sock *sk;
918 	struct vsock_sock *vsk;
919 	struct sockaddr_vm *vm_addr;
920 
921 	sk = sock->sk;
922 	vsk = vsock_sk(sk);
923 	err = 0;
924 
925 	lock_sock(sk);
926 
927 	if (peer) {
928 		if (sock->state != SS_CONNECTED) {
929 			err = -ENOTCONN;
930 			goto out;
931 		}
932 		vm_addr = &vsk->remote_addr;
933 	} else {
934 		vm_addr = &vsk->local_addr;
935 	}
936 
937 	if (!vm_addr) {
938 		err = -EINVAL;
939 		goto out;
940 	}
941 
942 	/* sys_getsockname() and sys_getpeername() pass us a
943 	 * MAX_SOCK_ADDR-sized buffer and don't set addr_len.  Unfortunately
944 	 * that macro is defined in socket.c instead of .h, so we hardcode its
945 	 * value here.
946 	 */
947 	BUILD_BUG_ON(sizeof(*vm_addr) > 128);
948 	memcpy(addr, vm_addr, sizeof(*vm_addr));
949 	err = sizeof(*vm_addr);
950 
951 out:
952 	release_sock(sk);
953 	return err;
954 }
955 
956 static int vsock_shutdown(struct socket *sock, int mode)
957 {
958 	int err;
959 	struct sock *sk;
960 
961 	/* User level uses SHUT_RD (0) and SHUT_WR (1), but the kernel uses
962 	 * RCV_SHUTDOWN (1) and SEND_SHUTDOWN (2), so we must increment mode
963 	 * here like the other address families do.  Note also that the
964 	 * increment makes SHUT_RDWR (2) into RCV_SHUTDOWN | SEND_SHUTDOWN (3),
965 	 * which is what we want.
966 	 */
967 	mode++;
968 
969 	if ((mode & ~SHUTDOWN_MASK) || !mode)
970 		return -EINVAL;
971 
972 	/* If this is a connection oriented socket and it is not connected then
973 	 * bail out immediately.  If it is a DGRAM socket then we must first
974 	 * kick the socket so that it wakes up from any sleeping calls, for
975 	 * example recv(), and then afterwards return the error.
976 	 */
977 
978 	sk = sock->sk;
979 
980 	lock_sock(sk);
981 	if (sock->state == SS_UNCONNECTED) {
982 		err = -ENOTCONN;
983 		if (sock_type_connectible(sk->sk_type))
984 			goto out;
985 	} else {
986 		sock->state = SS_DISCONNECTING;
987 		err = 0;
988 	}
989 
990 	/* Receive and send shutdowns are treated alike. */
991 	mode = mode & (RCV_SHUTDOWN | SEND_SHUTDOWN);
992 	if (mode) {
993 		sk->sk_shutdown |= mode;
994 		sk->sk_state_change(sk);
995 
996 		if (sock_type_connectible(sk->sk_type)) {
997 			sock_reset_flag(sk, SOCK_DONE);
998 			vsock_send_shutdown(sk, mode);
999 		}
1000 	}
1001 
1002 out:
1003 	release_sock(sk);
1004 	return err;
1005 }
1006 
1007 static __poll_t vsock_poll(struct file *file, struct socket *sock,
1008 			       poll_table *wait)
1009 {
1010 	struct sock *sk;
1011 	__poll_t mask;
1012 	struct vsock_sock *vsk;
1013 
1014 	sk = sock->sk;
1015 	vsk = vsock_sk(sk);
1016 
1017 	poll_wait(file, sk_sleep(sk), wait);
1018 	mask = 0;
1019 
1020 	if (sk->sk_err)
1021 		/* Signify that there has been an error on this socket. */
1022 		mask |= EPOLLERR;
1023 
1024 	/* INET sockets treat local write shutdown and peer write shutdown as a
1025 	 * case of EPOLLHUP set.
1026 	 */
1027 	if ((sk->sk_shutdown == SHUTDOWN_MASK) ||
1028 	    ((sk->sk_shutdown & SEND_SHUTDOWN) &&
1029 	     (vsk->peer_shutdown & SEND_SHUTDOWN))) {
1030 		mask |= EPOLLHUP;
1031 	}
1032 
1033 	if (sk->sk_shutdown & RCV_SHUTDOWN ||
1034 	    vsk->peer_shutdown & SEND_SHUTDOWN) {
1035 		mask |= EPOLLRDHUP;
1036 	}
1037 
1038 	if (sock->type == SOCK_DGRAM) {
1039 		/* For datagram sockets we can read if there is something in
1040 		 * the queue and write as long as the socket isn't shutdown for
1041 		 * sending.
1042 		 */
1043 		if (!skb_queue_empty_lockless(&sk->sk_receive_queue) ||
1044 		    (sk->sk_shutdown & RCV_SHUTDOWN)) {
1045 			mask |= EPOLLIN | EPOLLRDNORM;
1046 		}
1047 
1048 		if (!(sk->sk_shutdown & SEND_SHUTDOWN))
1049 			mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND;
1050 
1051 	} else if (sock_type_connectible(sk->sk_type)) {
1052 		const struct vsock_transport *transport;
1053 
1054 		lock_sock(sk);
1055 
1056 		transport = vsk->transport;
1057 
1058 		/* Listening sockets that have connections in their accept
1059 		 * queue can be read.
1060 		 */
1061 		if (sk->sk_state == TCP_LISTEN
1062 		    && !vsock_is_accept_queue_empty(sk))
1063 			mask |= EPOLLIN | EPOLLRDNORM;
1064 
1065 		/* If there is something in the queue then we can read. */
1066 		if (transport && transport->stream_is_active(vsk) &&
1067 		    !(sk->sk_shutdown & RCV_SHUTDOWN)) {
1068 			bool data_ready_now = false;
1069 			int ret = transport->notify_poll_in(
1070 					vsk, 1, &data_ready_now);
1071 			if (ret < 0) {
1072 				mask |= EPOLLERR;
1073 			} else {
1074 				if (data_ready_now)
1075 					mask |= EPOLLIN | EPOLLRDNORM;
1076 
1077 			}
1078 		}
1079 
1080 		/* Sockets whose connections have been closed, reset, or
1081 		 * terminated should also be considered read, and we check the
1082 		 * shutdown flag for that.
1083 		 */
1084 		if (sk->sk_shutdown & RCV_SHUTDOWN ||
1085 		    vsk->peer_shutdown & SEND_SHUTDOWN) {
1086 			mask |= EPOLLIN | EPOLLRDNORM;
1087 		}
1088 
1089 		/* Connected sockets that can produce data can be written. */
1090 		if (transport && sk->sk_state == TCP_ESTABLISHED) {
1091 			if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1092 				bool space_avail_now = false;
1093 				int ret = transport->notify_poll_out(
1094 						vsk, 1, &space_avail_now);
1095 				if (ret < 0) {
1096 					mask |= EPOLLERR;
1097 				} else {
1098 					if (space_avail_now)
1099 						/* Remove EPOLLWRBAND since INET
1100 						 * sockets are not setting it.
1101 						 */
1102 						mask |= EPOLLOUT | EPOLLWRNORM;
1103 
1104 				}
1105 			}
1106 		}
1107 
1108 		/* Simulate INET socket poll behaviors, which sets
1109 		 * EPOLLOUT|EPOLLWRNORM when peer is closed and nothing to read,
1110 		 * but local send is not shutdown.
1111 		 */
1112 		if (sk->sk_state == TCP_CLOSE || sk->sk_state == TCP_CLOSING) {
1113 			if (!(sk->sk_shutdown & SEND_SHUTDOWN))
1114 				mask |= EPOLLOUT | EPOLLWRNORM;
1115 
1116 		}
1117 
1118 		release_sock(sk);
1119 	}
1120 
1121 	return mask;
1122 }
1123 
1124 static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg,
1125 			       size_t len)
1126 {
1127 	int err;
1128 	struct sock *sk;
1129 	struct vsock_sock *vsk;
1130 	struct sockaddr_vm *remote_addr;
1131 	const struct vsock_transport *transport;
1132 
1133 	if (msg->msg_flags & MSG_OOB)
1134 		return -EOPNOTSUPP;
1135 
1136 	/* For now, MSG_DONTWAIT is always assumed... */
1137 	err = 0;
1138 	sk = sock->sk;
1139 	vsk = vsock_sk(sk);
1140 
1141 	lock_sock(sk);
1142 
1143 	transport = vsk->transport;
1144 
1145 	err = vsock_auto_bind(vsk);
1146 	if (err)
1147 		goto out;
1148 
1149 
1150 	/* If the provided message contains an address, use that.  Otherwise
1151 	 * fall back on the socket's remote handle (if it has been connected).
1152 	 */
1153 	if (msg->msg_name &&
1154 	    vsock_addr_cast(msg->msg_name, msg->msg_namelen,
1155 			    &remote_addr) == 0) {
1156 		/* Ensure this address is of the right type and is a valid
1157 		 * destination.
1158 		 */
1159 
1160 		if (remote_addr->svm_cid == VMADDR_CID_ANY)
1161 			remote_addr->svm_cid = transport->get_local_cid();
1162 
1163 		if (!vsock_addr_bound(remote_addr)) {
1164 			err = -EINVAL;
1165 			goto out;
1166 		}
1167 	} else if (sock->state == SS_CONNECTED) {
1168 		remote_addr = &vsk->remote_addr;
1169 
1170 		if (remote_addr->svm_cid == VMADDR_CID_ANY)
1171 			remote_addr->svm_cid = transport->get_local_cid();
1172 
1173 		/* XXX Should connect() or this function ensure remote_addr is
1174 		 * bound?
1175 		 */
1176 		if (!vsock_addr_bound(&vsk->remote_addr)) {
1177 			err = -EINVAL;
1178 			goto out;
1179 		}
1180 	} else {
1181 		err = -EINVAL;
1182 		goto out;
1183 	}
1184 
1185 	if (!transport->dgram_allow(remote_addr->svm_cid,
1186 				    remote_addr->svm_port)) {
1187 		err = -EINVAL;
1188 		goto out;
1189 	}
1190 
1191 	err = transport->dgram_enqueue(vsk, remote_addr, msg, len);
1192 
1193 out:
1194 	release_sock(sk);
1195 	return err;
1196 }
1197 
1198 static int vsock_dgram_connect(struct socket *sock,
1199 			       struct sockaddr *addr, int addr_len, int flags)
1200 {
1201 	int err;
1202 	struct sock *sk;
1203 	struct vsock_sock *vsk;
1204 	struct sockaddr_vm *remote_addr;
1205 
1206 	sk = sock->sk;
1207 	vsk = vsock_sk(sk);
1208 
1209 	err = vsock_addr_cast(addr, addr_len, &remote_addr);
1210 	if (err == -EAFNOSUPPORT && remote_addr->svm_family == AF_UNSPEC) {
1211 		lock_sock(sk);
1212 		vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY,
1213 				VMADDR_PORT_ANY);
1214 		sock->state = SS_UNCONNECTED;
1215 		release_sock(sk);
1216 		return 0;
1217 	} else if (err != 0)
1218 		return -EINVAL;
1219 
1220 	lock_sock(sk);
1221 
1222 	err = vsock_auto_bind(vsk);
1223 	if (err)
1224 		goto out;
1225 
1226 	if (!vsk->transport->dgram_allow(remote_addr->svm_cid,
1227 					 remote_addr->svm_port)) {
1228 		err = -EINVAL;
1229 		goto out;
1230 	}
1231 
1232 	memcpy(&vsk->remote_addr, remote_addr, sizeof(vsk->remote_addr));
1233 	sock->state = SS_CONNECTED;
1234 
1235 out:
1236 	release_sock(sk);
1237 	return err;
1238 }
1239 
1240 static int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg,
1241 			       size_t len, int flags)
1242 {
1243 	struct vsock_sock *vsk = vsock_sk(sock->sk);
1244 
1245 	return vsk->transport->dgram_dequeue(vsk, msg, len, flags);
1246 }
1247 
1248 static const struct proto_ops vsock_dgram_ops = {
1249 	.family = PF_VSOCK,
1250 	.owner = THIS_MODULE,
1251 	.release = vsock_release,
1252 	.bind = vsock_bind,
1253 	.connect = vsock_dgram_connect,
1254 	.socketpair = sock_no_socketpair,
1255 	.accept = sock_no_accept,
1256 	.getname = vsock_getname,
1257 	.poll = vsock_poll,
1258 	.ioctl = sock_no_ioctl,
1259 	.listen = sock_no_listen,
1260 	.shutdown = vsock_shutdown,
1261 	.sendmsg = vsock_dgram_sendmsg,
1262 	.recvmsg = vsock_dgram_recvmsg,
1263 	.mmap = sock_no_mmap,
1264 	.sendpage = sock_no_sendpage,
1265 };
1266 
1267 static int vsock_transport_cancel_pkt(struct vsock_sock *vsk)
1268 {
1269 	const struct vsock_transport *transport = vsk->transport;
1270 
1271 	if (!transport || !transport->cancel_pkt)
1272 		return -EOPNOTSUPP;
1273 
1274 	return transport->cancel_pkt(vsk);
1275 }
1276 
1277 static void vsock_connect_timeout(struct work_struct *work)
1278 {
1279 	struct sock *sk;
1280 	struct vsock_sock *vsk;
1281 
1282 	vsk = container_of(work, struct vsock_sock, connect_work.work);
1283 	sk = sk_vsock(vsk);
1284 
1285 	lock_sock(sk);
1286 	if (sk->sk_state == TCP_SYN_SENT &&
1287 	    (sk->sk_shutdown != SHUTDOWN_MASK)) {
1288 		sk->sk_state = TCP_CLOSE;
1289 		sk->sk_socket->state = SS_UNCONNECTED;
1290 		sk->sk_err = ETIMEDOUT;
1291 		sk_error_report(sk);
1292 		vsock_transport_cancel_pkt(vsk);
1293 	}
1294 	release_sock(sk);
1295 
1296 	sock_put(sk);
1297 }
1298 
1299 static int vsock_connect(struct socket *sock, struct sockaddr *addr,
1300 			 int addr_len, int flags)
1301 {
1302 	int err;
1303 	struct sock *sk;
1304 	struct vsock_sock *vsk;
1305 	const struct vsock_transport *transport;
1306 	struct sockaddr_vm *remote_addr;
1307 	long timeout;
1308 	DEFINE_WAIT(wait);
1309 
1310 	err = 0;
1311 	sk = sock->sk;
1312 	vsk = vsock_sk(sk);
1313 
1314 	lock_sock(sk);
1315 
1316 	/* XXX AF_UNSPEC should make us disconnect like AF_INET. */
1317 	switch (sock->state) {
1318 	case SS_CONNECTED:
1319 		err = -EISCONN;
1320 		goto out;
1321 	case SS_DISCONNECTING:
1322 		err = -EINVAL;
1323 		goto out;
1324 	case SS_CONNECTING:
1325 		/* This continues on so we can move sock into the SS_CONNECTED
1326 		 * state once the connection has completed (at which point err
1327 		 * will be set to zero also).  Otherwise, we will either wait
1328 		 * for the connection or return -EALREADY should this be a
1329 		 * non-blocking call.
1330 		 */
1331 		err = -EALREADY;
1332 		if (flags & O_NONBLOCK)
1333 			goto out;
1334 		break;
1335 	default:
1336 		if ((sk->sk_state == TCP_LISTEN) ||
1337 		    vsock_addr_cast(addr, addr_len, &remote_addr) != 0) {
1338 			err = -EINVAL;
1339 			goto out;
1340 		}
1341 
1342 		/* Set the remote address that we are connecting to. */
1343 		memcpy(&vsk->remote_addr, remote_addr,
1344 		       sizeof(vsk->remote_addr));
1345 
1346 		err = vsock_assign_transport(vsk, NULL);
1347 		if (err)
1348 			goto out;
1349 
1350 		transport = vsk->transport;
1351 
1352 		/* The hypervisor and well-known contexts do not have socket
1353 		 * endpoints.
1354 		 */
1355 		if (!transport ||
1356 		    !transport->stream_allow(remote_addr->svm_cid,
1357 					     remote_addr->svm_port)) {
1358 			err = -ENETUNREACH;
1359 			goto out;
1360 		}
1361 
1362 		err = vsock_auto_bind(vsk);
1363 		if (err)
1364 			goto out;
1365 
1366 		sk->sk_state = TCP_SYN_SENT;
1367 
1368 		err = transport->connect(vsk);
1369 		if (err < 0)
1370 			goto out;
1371 
1372 		/* Mark sock as connecting and set the error code to in
1373 		 * progress in case this is a non-blocking connect.
1374 		 */
1375 		sock->state = SS_CONNECTING;
1376 		err = -EINPROGRESS;
1377 	}
1378 
1379 	/* The receive path will handle all communication until we are able to
1380 	 * enter the connected state.  Here we wait for the connection to be
1381 	 * completed or a notification of an error.
1382 	 */
1383 	timeout = vsk->connect_timeout;
1384 	prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1385 
1386 	while (sk->sk_state != TCP_ESTABLISHED && sk->sk_err == 0) {
1387 		if (flags & O_NONBLOCK) {
1388 			/* If we're not going to block, we schedule a timeout
1389 			 * function to generate a timeout on the connection
1390 			 * attempt, in case the peer doesn't respond in a
1391 			 * timely manner. We hold on to the socket until the
1392 			 * timeout fires.
1393 			 */
1394 			sock_hold(sk);
1395 
1396 			/* If the timeout function is already scheduled,
1397 			 * reschedule it, then ungrab the socket refcount to
1398 			 * keep it balanced.
1399 			 */
1400 			if (mod_delayed_work(system_wq, &vsk->connect_work,
1401 					     timeout))
1402 				sock_put(sk);
1403 
1404 			/* Skip ahead to preserve error code set above. */
1405 			goto out_wait;
1406 		}
1407 
1408 		release_sock(sk);
1409 		timeout = schedule_timeout(timeout);
1410 		lock_sock(sk);
1411 
1412 		if (signal_pending(current)) {
1413 			err = sock_intr_errno(timeout);
1414 			sk->sk_state = sk->sk_state == TCP_ESTABLISHED ? TCP_CLOSING : TCP_CLOSE;
1415 			sock->state = SS_UNCONNECTED;
1416 			vsock_transport_cancel_pkt(vsk);
1417 			vsock_remove_connected(vsk);
1418 			goto out_wait;
1419 		} else if (timeout == 0) {
1420 			err = -ETIMEDOUT;
1421 			sk->sk_state = TCP_CLOSE;
1422 			sock->state = SS_UNCONNECTED;
1423 			vsock_transport_cancel_pkt(vsk);
1424 			goto out_wait;
1425 		}
1426 
1427 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1428 	}
1429 
1430 	if (sk->sk_err) {
1431 		err = -sk->sk_err;
1432 		sk->sk_state = TCP_CLOSE;
1433 		sock->state = SS_UNCONNECTED;
1434 	} else {
1435 		err = 0;
1436 	}
1437 
1438 out_wait:
1439 	finish_wait(sk_sleep(sk), &wait);
1440 out:
1441 	release_sock(sk);
1442 	return err;
1443 }
1444 
1445 static int vsock_accept(struct socket *sock, struct socket *newsock, int flags,
1446 			bool kern)
1447 {
1448 	struct sock *listener;
1449 	int err;
1450 	struct sock *connected;
1451 	struct vsock_sock *vconnected;
1452 	long timeout;
1453 	DEFINE_WAIT(wait);
1454 
1455 	err = 0;
1456 	listener = sock->sk;
1457 
1458 	lock_sock(listener);
1459 
1460 	if (!sock_type_connectible(sock->type)) {
1461 		err = -EOPNOTSUPP;
1462 		goto out;
1463 	}
1464 
1465 	if (listener->sk_state != TCP_LISTEN) {
1466 		err = -EINVAL;
1467 		goto out;
1468 	}
1469 
1470 	/* Wait for children sockets to appear; these are the new sockets
1471 	 * created upon connection establishment.
1472 	 */
1473 	timeout = sock_rcvtimeo(listener, flags & O_NONBLOCK);
1474 	prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1475 
1476 	while ((connected = vsock_dequeue_accept(listener)) == NULL &&
1477 	       listener->sk_err == 0) {
1478 		release_sock(listener);
1479 		timeout = schedule_timeout(timeout);
1480 		finish_wait(sk_sleep(listener), &wait);
1481 		lock_sock(listener);
1482 
1483 		if (signal_pending(current)) {
1484 			err = sock_intr_errno(timeout);
1485 			goto out;
1486 		} else if (timeout == 0) {
1487 			err = -EAGAIN;
1488 			goto out;
1489 		}
1490 
1491 		prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1492 	}
1493 	finish_wait(sk_sleep(listener), &wait);
1494 
1495 	if (listener->sk_err)
1496 		err = -listener->sk_err;
1497 
1498 	if (connected) {
1499 		sk_acceptq_removed(listener);
1500 
1501 		lock_sock_nested(connected, SINGLE_DEPTH_NESTING);
1502 		vconnected = vsock_sk(connected);
1503 
1504 		/* If the listener socket has received an error, then we should
1505 		 * reject this socket and return.  Note that we simply mark the
1506 		 * socket rejected, drop our reference, and let the cleanup
1507 		 * function handle the cleanup; the fact that we found it in
1508 		 * the listener's accept queue guarantees that the cleanup
1509 		 * function hasn't run yet.
1510 		 */
1511 		if (err) {
1512 			vconnected->rejected = true;
1513 		} else {
1514 			newsock->state = SS_CONNECTED;
1515 			sock_graft(connected, newsock);
1516 		}
1517 
1518 		release_sock(connected);
1519 		sock_put(connected);
1520 	}
1521 
1522 out:
1523 	release_sock(listener);
1524 	return err;
1525 }
1526 
1527 static int vsock_listen(struct socket *sock, int backlog)
1528 {
1529 	int err;
1530 	struct sock *sk;
1531 	struct vsock_sock *vsk;
1532 
1533 	sk = sock->sk;
1534 
1535 	lock_sock(sk);
1536 
1537 	if (!sock_type_connectible(sk->sk_type)) {
1538 		err = -EOPNOTSUPP;
1539 		goto out;
1540 	}
1541 
1542 	if (sock->state != SS_UNCONNECTED) {
1543 		err = -EINVAL;
1544 		goto out;
1545 	}
1546 
1547 	vsk = vsock_sk(sk);
1548 
1549 	if (!vsock_addr_bound(&vsk->local_addr)) {
1550 		err = -EINVAL;
1551 		goto out;
1552 	}
1553 
1554 	sk->sk_max_ack_backlog = backlog;
1555 	sk->sk_state = TCP_LISTEN;
1556 
1557 	err = 0;
1558 
1559 out:
1560 	release_sock(sk);
1561 	return err;
1562 }
1563 
1564 static void vsock_update_buffer_size(struct vsock_sock *vsk,
1565 				     const struct vsock_transport *transport,
1566 				     u64 val)
1567 {
1568 	if (val > vsk->buffer_max_size)
1569 		val = vsk->buffer_max_size;
1570 
1571 	if (val < vsk->buffer_min_size)
1572 		val = vsk->buffer_min_size;
1573 
1574 	if (val != vsk->buffer_size &&
1575 	    transport && transport->notify_buffer_size)
1576 		transport->notify_buffer_size(vsk, &val);
1577 
1578 	vsk->buffer_size = val;
1579 }
1580 
1581 static int vsock_connectible_setsockopt(struct socket *sock,
1582 					int level,
1583 					int optname,
1584 					sockptr_t optval,
1585 					unsigned int optlen)
1586 {
1587 	int err;
1588 	struct sock *sk;
1589 	struct vsock_sock *vsk;
1590 	const struct vsock_transport *transport;
1591 	u64 val;
1592 
1593 	if (level != AF_VSOCK)
1594 		return -ENOPROTOOPT;
1595 
1596 #define COPY_IN(_v)                                       \
1597 	do {						  \
1598 		if (optlen < sizeof(_v)) {		  \
1599 			err = -EINVAL;			  \
1600 			goto exit;			  \
1601 		}					  \
1602 		if (copy_from_sockptr(&_v, optval, sizeof(_v)) != 0) {	\
1603 			err = -EFAULT;					\
1604 			goto exit;					\
1605 		}							\
1606 	} while (0)
1607 
1608 	err = 0;
1609 	sk = sock->sk;
1610 	vsk = vsock_sk(sk);
1611 
1612 	lock_sock(sk);
1613 
1614 	transport = vsk->transport;
1615 
1616 	switch (optname) {
1617 	case SO_VM_SOCKETS_BUFFER_SIZE:
1618 		COPY_IN(val);
1619 		vsock_update_buffer_size(vsk, transport, val);
1620 		break;
1621 
1622 	case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1623 		COPY_IN(val);
1624 		vsk->buffer_max_size = val;
1625 		vsock_update_buffer_size(vsk, transport, vsk->buffer_size);
1626 		break;
1627 
1628 	case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1629 		COPY_IN(val);
1630 		vsk->buffer_min_size = val;
1631 		vsock_update_buffer_size(vsk, transport, vsk->buffer_size);
1632 		break;
1633 
1634 	case SO_VM_SOCKETS_CONNECT_TIMEOUT_NEW:
1635 	case SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD: {
1636 		struct __kernel_sock_timeval tv;
1637 
1638 		err = sock_copy_user_timeval(&tv, optval, optlen,
1639 					     optname == SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD);
1640 		if (err)
1641 			break;
1642 		if (tv.tv_sec >= 0 && tv.tv_usec < USEC_PER_SEC &&
1643 		    tv.tv_sec < (MAX_SCHEDULE_TIMEOUT / HZ - 1)) {
1644 			vsk->connect_timeout = tv.tv_sec * HZ +
1645 				DIV_ROUND_UP((unsigned long)tv.tv_usec, (USEC_PER_SEC / HZ));
1646 			if (vsk->connect_timeout == 0)
1647 				vsk->connect_timeout =
1648 				    VSOCK_DEFAULT_CONNECT_TIMEOUT;
1649 
1650 		} else {
1651 			err = -ERANGE;
1652 		}
1653 		break;
1654 	}
1655 
1656 	default:
1657 		err = -ENOPROTOOPT;
1658 		break;
1659 	}
1660 
1661 #undef COPY_IN
1662 
1663 exit:
1664 	release_sock(sk);
1665 	return err;
1666 }
1667 
1668 static int vsock_connectible_getsockopt(struct socket *sock,
1669 					int level, int optname,
1670 					char __user *optval,
1671 					int __user *optlen)
1672 {
1673 	struct sock *sk = sock->sk;
1674 	struct vsock_sock *vsk = vsock_sk(sk);
1675 
1676 	union {
1677 		u64 val64;
1678 		struct old_timeval32 tm32;
1679 		struct __kernel_old_timeval tm;
1680 		struct  __kernel_sock_timeval stm;
1681 	} v;
1682 
1683 	int lv = sizeof(v.val64);
1684 	int len;
1685 
1686 	if (level != AF_VSOCK)
1687 		return -ENOPROTOOPT;
1688 
1689 	if (get_user(len, optlen))
1690 		return -EFAULT;
1691 
1692 	memset(&v, 0, sizeof(v));
1693 
1694 	switch (optname) {
1695 	case SO_VM_SOCKETS_BUFFER_SIZE:
1696 		v.val64 = vsk->buffer_size;
1697 		break;
1698 
1699 	case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1700 		v.val64 = vsk->buffer_max_size;
1701 		break;
1702 
1703 	case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1704 		v.val64 = vsk->buffer_min_size;
1705 		break;
1706 
1707 	case SO_VM_SOCKETS_CONNECT_TIMEOUT_NEW:
1708 	case SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD:
1709 		lv = sock_get_timeout(vsk->connect_timeout, &v,
1710 				      optname == SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD);
1711 		break;
1712 
1713 	default:
1714 		return -ENOPROTOOPT;
1715 	}
1716 
1717 	if (len < lv)
1718 		return -EINVAL;
1719 	if (len > lv)
1720 		len = lv;
1721 	if (copy_to_user(optval, &v, len))
1722 		return -EFAULT;
1723 
1724 	if (put_user(len, optlen))
1725 		return -EFAULT;
1726 
1727 	return 0;
1728 }
1729 
1730 static int vsock_connectible_sendmsg(struct socket *sock, struct msghdr *msg,
1731 				     size_t len)
1732 {
1733 	struct sock *sk;
1734 	struct vsock_sock *vsk;
1735 	const struct vsock_transport *transport;
1736 	ssize_t total_written;
1737 	long timeout;
1738 	int err;
1739 	struct vsock_transport_send_notify_data send_data;
1740 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
1741 
1742 	sk = sock->sk;
1743 	vsk = vsock_sk(sk);
1744 	total_written = 0;
1745 	err = 0;
1746 
1747 	if (msg->msg_flags & MSG_OOB)
1748 		return -EOPNOTSUPP;
1749 
1750 	lock_sock(sk);
1751 
1752 	transport = vsk->transport;
1753 
1754 	/* Callers should not provide a destination with connection oriented
1755 	 * sockets.
1756 	 */
1757 	if (msg->msg_namelen) {
1758 		err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP;
1759 		goto out;
1760 	}
1761 
1762 	/* Send data only if both sides are not shutdown in the direction. */
1763 	if (sk->sk_shutdown & SEND_SHUTDOWN ||
1764 	    vsk->peer_shutdown & RCV_SHUTDOWN) {
1765 		err = -EPIPE;
1766 		goto out;
1767 	}
1768 
1769 	if (!transport || sk->sk_state != TCP_ESTABLISHED ||
1770 	    !vsock_addr_bound(&vsk->local_addr)) {
1771 		err = -ENOTCONN;
1772 		goto out;
1773 	}
1774 
1775 	if (!vsock_addr_bound(&vsk->remote_addr)) {
1776 		err = -EDESTADDRREQ;
1777 		goto out;
1778 	}
1779 
1780 	/* Wait for room in the produce queue to enqueue our user's data. */
1781 	timeout = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1782 
1783 	err = transport->notify_send_init(vsk, &send_data);
1784 	if (err < 0)
1785 		goto out;
1786 
1787 	while (total_written < len) {
1788 		ssize_t written;
1789 
1790 		add_wait_queue(sk_sleep(sk), &wait);
1791 		while (vsock_stream_has_space(vsk) == 0 &&
1792 		       sk->sk_err == 0 &&
1793 		       !(sk->sk_shutdown & SEND_SHUTDOWN) &&
1794 		       !(vsk->peer_shutdown & RCV_SHUTDOWN)) {
1795 
1796 			/* Don't wait for non-blocking sockets. */
1797 			if (timeout == 0) {
1798 				err = -EAGAIN;
1799 				remove_wait_queue(sk_sleep(sk), &wait);
1800 				goto out_err;
1801 			}
1802 
1803 			err = transport->notify_send_pre_block(vsk, &send_data);
1804 			if (err < 0) {
1805 				remove_wait_queue(sk_sleep(sk), &wait);
1806 				goto out_err;
1807 			}
1808 
1809 			release_sock(sk);
1810 			timeout = wait_woken(&wait, TASK_INTERRUPTIBLE, timeout);
1811 			lock_sock(sk);
1812 			if (signal_pending(current)) {
1813 				err = sock_intr_errno(timeout);
1814 				remove_wait_queue(sk_sleep(sk), &wait);
1815 				goto out_err;
1816 			} else if (timeout == 0) {
1817 				err = -EAGAIN;
1818 				remove_wait_queue(sk_sleep(sk), &wait);
1819 				goto out_err;
1820 			}
1821 		}
1822 		remove_wait_queue(sk_sleep(sk), &wait);
1823 
1824 		/* These checks occur both as part of and after the loop
1825 		 * conditional since we need to check before and after
1826 		 * sleeping.
1827 		 */
1828 		if (sk->sk_err) {
1829 			err = -sk->sk_err;
1830 			goto out_err;
1831 		} else if ((sk->sk_shutdown & SEND_SHUTDOWN) ||
1832 			   (vsk->peer_shutdown & RCV_SHUTDOWN)) {
1833 			err = -EPIPE;
1834 			goto out_err;
1835 		}
1836 
1837 		err = transport->notify_send_pre_enqueue(vsk, &send_data);
1838 		if (err < 0)
1839 			goto out_err;
1840 
1841 		/* Note that enqueue will only write as many bytes as are free
1842 		 * in the produce queue, so we don't need to ensure len is
1843 		 * smaller than the queue size.  It is the caller's
1844 		 * responsibility to check how many bytes we were able to send.
1845 		 */
1846 
1847 		if (sk->sk_type == SOCK_SEQPACKET) {
1848 			written = transport->seqpacket_enqueue(vsk,
1849 						msg, len - total_written);
1850 		} else {
1851 			written = transport->stream_enqueue(vsk,
1852 					msg, len - total_written);
1853 		}
1854 		if (written < 0) {
1855 			err = -ENOMEM;
1856 			goto out_err;
1857 		}
1858 
1859 		total_written += written;
1860 
1861 		err = transport->notify_send_post_enqueue(
1862 				vsk, written, &send_data);
1863 		if (err < 0)
1864 			goto out_err;
1865 
1866 	}
1867 
1868 out_err:
1869 	if (total_written > 0) {
1870 		/* Return number of written bytes only if:
1871 		 * 1) SOCK_STREAM socket.
1872 		 * 2) SOCK_SEQPACKET socket when whole buffer is sent.
1873 		 */
1874 		if (sk->sk_type == SOCK_STREAM || total_written == len)
1875 			err = total_written;
1876 	}
1877 out:
1878 	release_sock(sk);
1879 	return err;
1880 }
1881 
1882 static int vsock_connectible_wait_data(struct sock *sk,
1883 				       struct wait_queue_entry *wait,
1884 				       long timeout,
1885 				       struct vsock_transport_recv_notify_data *recv_data,
1886 				       size_t target)
1887 {
1888 	const struct vsock_transport *transport;
1889 	struct vsock_sock *vsk;
1890 	s64 data;
1891 	int err;
1892 
1893 	vsk = vsock_sk(sk);
1894 	err = 0;
1895 	transport = vsk->transport;
1896 
1897 	while ((data = vsock_connectible_has_data(vsk)) == 0) {
1898 		prepare_to_wait(sk_sleep(sk), wait, TASK_INTERRUPTIBLE);
1899 
1900 		if (sk->sk_err != 0 ||
1901 		    (sk->sk_shutdown & RCV_SHUTDOWN) ||
1902 		    (vsk->peer_shutdown & SEND_SHUTDOWN)) {
1903 			break;
1904 		}
1905 
1906 		/* Don't wait for non-blocking sockets. */
1907 		if (timeout == 0) {
1908 			err = -EAGAIN;
1909 			break;
1910 		}
1911 
1912 		if (recv_data) {
1913 			err = transport->notify_recv_pre_block(vsk, target, recv_data);
1914 			if (err < 0)
1915 				break;
1916 		}
1917 
1918 		release_sock(sk);
1919 		timeout = schedule_timeout(timeout);
1920 		lock_sock(sk);
1921 
1922 		if (signal_pending(current)) {
1923 			err = sock_intr_errno(timeout);
1924 			break;
1925 		} else if (timeout == 0) {
1926 			err = -EAGAIN;
1927 			break;
1928 		}
1929 	}
1930 
1931 	finish_wait(sk_sleep(sk), wait);
1932 
1933 	if (err)
1934 		return err;
1935 
1936 	/* Internal transport error when checking for available
1937 	 * data. XXX This should be changed to a connection
1938 	 * reset in a later change.
1939 	 */
1940 	if (data < 0)
1941 		return -ENOMEM;
1942 
1943 	return data;
1944 }
1945 
1946 static int __vsock_stream_recvmsg(struct sock *sk, struct msghdr *msg,
1947 				  size_t len, int flags)
1948 {
1949 	struct vsock_transport_recv_notify_data recv_data;
1950 	const struct vsock_transport *transport;
1951 	struct vsock_sock *vsk;
1952 	ssize_t copied;
1953 	size_t target;
1954 	long timeout;
1955 	int err;
1956 
1957 	DEFINE_WAIT(wait);
1958 
1959 	vsk = vsock_sk(sk);
1960 	transport = vsk->transport;
1961 
1962 	/* We must not copy less than target bytes into the user's buffer
1963 	 * before returning successfully, so we wait for the consume queue to
1964 	 * have that much data to consume before dequeueing.  Note that this
1965 	 * makes it impossible to handle cases where target is greater than the
1966 	 * queue size.
1967 	 */
1968 	target = sock_rcvlowat(sk, flags & MSG_WAITALL, len);
1969 	if (target >= transport->stream_rcvhiwat(vsk)) {
1970 		err = -ENOMEM;
1971 		goto out;
1972 	}
1973 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1974 	copied = 0;
1975 
1976 	err = transport->notify_recv_init(vsk, target, &recv_data);
1977 	if (err < 0)
1978 		goto out;
1979 
1980 
1981 	while (1) {
1982 		ssize_t read;
1983 
1984 		err = vsock_connectible_wait_data(sk, &wait, timeout,
1985 						  &recv_data, target);
1986 		if (err <= 0)
1987 			break;
1988 
1989 		err = transport->notify_recv_pre_dequeue(vsk, target,
1990 							 &recv_data);
1991 		if (err < 0)
1992 			break;
1993 
1994 		read = transport->stream_dequeue(vsk, msg, len - copied, flags);
1995 		if (read < 0) {
1996 			err = -ENOMEM;
1997 			break;
1998 		}
1999 
2000 		copied += read;
2001 
2002 		err = transport->notify_recv_post_dequeue(vsk, target, read,
2003 						!(flags & MSG_PEEK), &recv_data);
2004 		if (err < 0)
2005 			goto out;
2006 
2007 		if (read >= target || flags & MSG_PEEK)
2008 			break;
2009 
2010 		target -= read;
2011 	}
2012 
2013 	if (sk->sk_err)
2014 		err = -sk->sk_err;
2015 	else if (sk->sk_shutdown & RCV_SHUTDOWN)
2016 		err = 0;
2017 
2018 	if (copied > 0)
2019 		err = copied;
2020 
2021 out:
2022 	return err;
2023 }
2024 
2025 static int __vsock_seqpacket_recvmsg(struct sock *sk, struct msghdr *msg,
2026 				     size_t len, int flags)
2027 {
2028 	const struct vsock_transport *transport;
2029 	struct vsock_sock *vsk;
2030 	ssize_t msg_len;
2031 	long timeout;
2032 	int err = 0;
2033 	DEFINE_WAIT(wait);
2034 
2035 	vsk = vsock_sk(sk);
2036 	transport = vsk->transport;
2037 
2038 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2039 
2040 	err = vsock_connectible_wait_data(sk, &wait, timeout, NULL, 0);
2041 	if (err <= 0)
2042 		goto out;
2043 
2044 	msg_len = transport->seqpacket_dequeue(vsk, msg, flags);
2045 
2046 	if (msg_len < 0) {
2047 		err = -ENOMEM;
2048 		goto out;
2049 	}
2050 
2051 	if (sk->sk_err) {
2052 		err = -sk->sk_err;
2053 	} else if (sk->sk_shutdown & RCV_SHUTDOWN) {
2054 		err = 0;
2055 	} else {
2056 		/* User sets MSG_TRUNC, so return real length of
2057 		 * packet.
2058 		 */
2059 		if (flags & MSG_TRUNC)
2060 			err = msg_len;
2061 		else
2062 			err = len - msg_data_left(msg);
2063 
2064 		/* Always set MSG_TRUNC if real length of packet is
2065 		 * bigger than user's buffer.
2066 		 */
2067 		if (msg_len > len)
2068 			msg->msg_flags |= MSG_TRUNC;
2069 	}
2070 
2071 out:
2072 	return err;
2073 }
2074 
2075 static int
2076 vsock_connectible_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
2077 			  int flags)
2078 {
2079 	struct sock *sk;
2080 	struct vsock_sock *vsk;
2081 	const struct vsock_transport *transport;
2082 	int err;
2083 
2084 	DEFINE_WAIT(wait);
2085 
2086 	sk = sock->sk;
2087 	vsk = vsock_sk(sk);
2088 	err = 0;
2089 
2090 	lock_sock(sk);
2091 
2092 	transport = vsk->transport;
2093 
2094 	if (!transport || sk->sk_state != TCP_ESTABLISHED) {
2095 		/* Recvmsg is supposed to return 0 if a peer performs an
2096 		 * orderly shutdown. Differentiate between that case and when a
2097 		 * peer has not connected or a local shutdown occurred with the
2098 		 * SOCK_DONE flag.
2099 		 */
2100 		if (sock_flag(sk, SOCK_DONE))
2101 			err = 0;
2102 		else
2103 			err = -ENOTCONN;
2104 
2105 		goto out;
2106 	}
2107 
2108 	if (flags & MSG_OOB) {
2109 		err = -EOPNOTSUPP;
2110 		goto out;
2111 	}
2112 
2113 	/* We don't check peer_shutdown flag here since peer may actually shut
2114 	 * down, but there can be data in the queue that a local socket can
2115 	 * receive.
2116 	 */
2117 	if (sk->sk_shutdown & RCV_SHUTDOWN) {
2118 		err = 0;
2119 		goto out;
2120 	}
2121 
2122 	/* It is valid on Linux to pass in a zero-length receive buffer.  This
2123 	 * is not an error.  We may as well bail out now.
2124 	 */
2125 	if (!len) {
2126 		err = 0;
2127 		goto out;
2128 	}
2129 
2130 	if (sk->sk_type == SOCK_STREAM)
2131 		err = __vsock_stream_recvmsg(sk, msg, len, flags);
2132 	else
2133 		err = __vsock_seqpacket_recvmsg(sk, msg, len, flags);
2134 
2135 out:
2136 	release_sock(sk);
2137 	return err;
2138 }
2139 
2140 static const struct proto_ops vsock_stream_ops = {
2141 	.family = PF_VSOCK,
2142 	.owner = THIS_MODULE,
2143 	.release = vsock_release,
2144 	.bind = vsock_bind,
2145 	.connect = vsock_connect,
2146 	.socketpair = sock_no_socketpair,
2147 	.accept = vsock_accept,
2148 	.getname = vsock_getname,
2149 	.poll = vsock_poll,
2150 	.ioctl = sock_no_ioctl,
2151 	.listen = vsock_listen,
2152 	.shutdown = vsock_shutdown,
2153 	.setsockopt = vsock_connectible_setsockopt,
2154 	.getsockopt = vsock_connectible_getsockopt,
2155 	.sendmsg = vsock_connectible_sendmsg,
2156 	.recvmsg = vsock_connectible_recvmsg,
2157 	.mmap = sock_no_mmap,
2158 	.sendpage = sock_no_sendpage,
2159 };
2160 
2161 static const struct proto_ops vsock_seqpacket_ops = {
2162 	.family = PF_VSOCK,
2163 	.owner = THIS_MODULE,
2164 	.release = vsock_release,
2165 	.bind = vsock_bind,
2166 	.connect = vsock_connect,
2167 	.socketpair = sock_no_socketpair,
2168 	.accept = vsock_accept,
2169 	.getname = vsock_getname,
2170 	.poll = vsock_poll,
2171 	.ioctl = sock_no_ioctl,
2172 	.listen = vsock_listen,
2173 	.shutdown = vsock_shutdown,
2174 	.setsockopt = vsock_connectible_setsockopt,
2175 	.getsockopt = vsock_connectible_getsockopt,
2176 	.sendmsg = vsock_connectible_sendmsg,
2177 	.recvmsg = vsock_connectible_recvmsg,
2178 	.mmap = sock_no_mmap,
2179 	.sendpage = sock_no_sendpage,
2180 };
2181 
2182 static int vsock_create(struct net *net, struct socket *sock,
2183 			int protocol, int kern)
2184 {
2185 	struct vsock_sock *vsk;
2186 	struct sock *sk;
2187 	int ret;
2188 
2189 	if (!sock)
2190 		return -EINVAL;
2191 
2192 	if (protocol && protocol != PF_VSOCK)
2193 		return -EPROTONOSUPPORT;
2194 
2195 	switch (sock->type) {
2196 	case SOCK_DGRAM:
2197 		sock->ops = &vsock_dgram_ops;
2198 		break;
2199 	case SOCK_STREAM:
2200 		sock->ops = &vsock_stream_ops;
2201 		break;
2202 	case SOCK_SEQPACKET:
2203 		sock->ops = &vsock_seqpacket_ops;
2204 		break;
2205 	default:
2206 		return -ESOCKTNOSUPPORT;
2207 	}
2208 
2209 	sock->state = SS_UNCONNECTED;
2210 
2211 	sk = __vsock_create(net, sock, NULL, GFP_KERNEL, 0, kern);
2212 	if (!sk)
2213 		return -ENOMEM;
2214 
2215 	vsk = vsock_sk(sk);
2216 
2217 	if (sock->type == SOCK_DGRAM) {
2218 		ret = vsock_assign_transport(vsk, NULL);
2219 		if (ret < 0) {
2220 			sock_put(sk);
2221 			return ret;
2222 		}
2223 	}
2224 
2225 	vsock_insert_unbound(vsk);
2226 
2227 	return 0;
2228 }
2229 
2230 static const struct net_proto_family vsock_family_ops = {
2231 	.family = AF_VSOCK,
2232 	.create = vsock_create,
2233 	.owner = THIS_MODULE,
2234 };
2235 
2236 static long vsock_dev_do_ioctl(struct file *filp,
2237 			       unsigned int cmd, void __user *ptr)
2238 {
2239 	u32 __user *p = ptr;
2240 	u32 cid = VMADDR_CID_ANY;
2241 	int retval = 0;
2242 
2243 	switch (cmd) {
2244 	case IOCTL_VM_SOCKETS_GET_LOCAL_CID:
2245 		/* To be compatible with the VMCI behavior, we prioritize the
2246 		 * guest CID instead of well-know host CID (VMADDR_CID_HOST).
2247 		 */
2248 		if (transport_g2h)
2249 			cid = transport_g2h->get_local_cid();
2250 		else if (transport_h2g)
2251 			cid = transport_h2g->get_local_cid();
2252 
2253 		if (put_user(cid, p) != 0)
2254 			retval = -EFAULT;
2255 		break;
2256 
2257 	default:
2258 		retval = -ENOIOCTLCMD;
2259 	}
2260 
2261 	return retval;
2262 }
2263 
2264 static long vsock_dev_ioctl(struct file *filp,
2265 			    unsigned int cmd, unsigned long arg)
2266 {
2267 	return vsock_dev_do_ioctl(filp, cmd, (void __user *)arg);
2268 }
2269 
2270 #ifdef CONFIG_COMPAT
2271 static long vsock_dev_compat_ioctl(struct file *filp,
2272 				   unsigned int cmd, unsigned long arg)
2273 {
2274 	return vsock_dev_do_ioctl(filp, cmd, compat_ptr(arg));
2275 }
2276 #endif
2277 
2278 static const struct file_operations vsock_device_ops = {
2279 	.owner		= THIS_MODULE,
2280 	.unlocked_ioctl	= vsock_dev_ioctl,
2281 #ifdef CONFIG_COMPAT
2282 	.compat_ioctl	= vsock_dev_compat_ioctl,
2283 #endif
2284 	.open		= nonseekable_open,
2285 };
2286 
2287 static struct miscdevice vsock_device = {
2288 	.name		= "vsock",
2289 	.fops		= &vsock_device_ops,
2290 };
2291 
2292 static int __init vsock_init(void)
2293 {
2294 	int err = 0;
2295 
2296 	vsock_init_tables();
2297 
2298 	vsock_proto.owner = THIS_MODULE;
2299 	vsock_device.minor = MISC_DYNAMIC_MINOR;
2300 	err = misc_register(&vsock_device);
2301 	if (err) {
2302 		pr_err("Failed to register misc device\n");
2303 		goto err_reset_transport;
2304 	}
2305 
2306 	err = proto_register(&vsock_proto, 1);	/* we want our slab */
2307 	if (err) {
2308 		pr_err("Cannot register vsock protocol\n");
2309 		goto err_deregister_misc;
2310 	}
2311 
2312 	err = sock_register(&vsock_family_ops);
2313 	if (err) {
2314 		pr_err("could not register af_vsock (%d) address family: %d\n",
2315 		       AF_VSOCK, err);
2316 		goto err_unregister_proto;
2317 	}
2318 
2319 	return 0;
2320 
2321 err_unregister_proto:
2322 	proto_unregister(&vsock_proto);
2323 err_deregister_misc:
2324 	misc_deregister(&vsock_device);
2325 err_reset_transport:
2326 	return err;
2327 }
2328 
2329 static void __exit vsock_exit(void)
2330 {
2331 	misc_deregister(&vsock_device);
2332 	sock_unregister(AF_VSOCK);
2333 	proto_unregister(&vsock_proto);
2334 }
2335 
2336 const struct vsock_transport *vsock_core_get_transport(struct vsock_sock *vsk)
2337 {
2338 	return vsk->transport;
2339 }
2340 EXPORT_SYMBOL_GPL(vsock_core_get_transport);
2341 
2342 int vsock_core_register(const struct vsock_transport *t, int features)
2343 {
2344 	const struct vsock_transport *t_h2g, *t_g2h, *t_dgram, *t_local;
2345 	int err = mutex_lock_interruptible(&vsock_register_mutex);
2346 
2347 	if (err)
2348 		return err;
2349 
2350 	t_h2g = transport_h2g;
2351 	t_g2h = transport_g2h;
2352 	t_dgram = transport_dgram;
2353 	t_local = transport_local;
2354 
2355 	if (features & VSOCK_TRANSPORT_F_H2G) {
2356 		if (t_h2g) {
2357 			err = -EBUSY;
2358 			goto err_busy;
2359 		}
2360 		t_h2g = t;
2361 	}
2362 
2363 	if (features & VSOCK_TRANSPORT_F_G2H) {
2364 		if (t_g2h) {
2365 			err = -EBUSY;
2366 			goto err_busy;
2367 		}
2368 		t_g2h = t;
2369 	}
2370 
2371 	if (features & VSOCK_TRANSPORT_F_DGRAM) {
2372 		if (t_dgram) {
2373 			err = -EBUSY;
2374 			goto err_busy;
2375 		}
2376 		t_dgram = t;
2377 	}
2378 
2379 	if (features & VSOCK_TRANSPORT_F_LOCAL) {
2380 		if (t_local) {
2381 			err = -EBUSY;
2382 			goto err_busy;
2383 		}
2384 		t_local = t;
2385 	}
2386 
2387 	transport_h2g = t_h2g;
2388 	transport_g2h = t_g2h;
2389 	transport_dgram = t_dgram;
2390 	transport_local = t_local;
2391 
2392 err_busy:
2393 	mutex_unlock(&vsock_register_mutex);
2394 	return err;
2395 }
2396 EXPORT_SYMBOL_GPL(vsock_core_register);
2397 
2398 void vsock_core_unregister(const struct vsock_transport *t)
2399 {
2400 	mutex_lock(&vsock_register_mutex);
2401 
2402 	if (transport_h2g == t)
2403 		transport_h2g = NULL;
2404 
2405 	if (transport_g2h == t)
2406 		transport_g2h = NULL;
2407 
2408 	if (transport_dgram == t)
2409 		transport_dgram = NULL;
2410 
2411 	if (transport_local == t)
2412 		transport_local = NULL;
2413 
2414 	mutex_unlock(&vsock_register_mutex);
2415 }
2416 EXPORT_SYMBOL_GPL(vsock_core_unregister);
2417 
2418 module_init(vsock_init);
2419 module_exit(vsock_exit);
2420 
2421 MODULE_AUTHOR("VMware, Inc.");
2422 MODULE_DESCRIPTION("VMware Virtual Socket Family");
2423 MODULE_VERSION("1.0.2.0-k");
2424 MODULE_LICENSE("GPL v2");
2425