xref: /openbmc/linux/net/vmw_vsock/af_vsock.c (revision 519b58bb)
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 void vsock_data_ready(struct sock *sk)
886 {
887 	struct vsock_sock *vsk = vsock_sk(sk);
888 
889 	if (vsock_stream_has_data(vsk) >= sk->sk_rcvlowat ||
890 	    sock_flag(sk, SOCK_DONE))
891 		sk->sk_data_ready(sk);
892 }
893 EXPORT_SYMBOL_GPL(vsock_data_ready);
894 
895 static int vsock_release(struct socket *sock)
896 {
897 	__vsock_release(sock->sk, 0);
898 	sock->sk = NULL;
899 	sock->state = SS_FREE;
900 
901 	return 0;
902 }
903 
904 static int
905 vsock_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
906 {
907 	int err;
908 	struct sock *sk;
909 	struct sockaddr_vm *vm_addr;
910 
911 	sk = sock->sk;
912 
913 	if (vsock_addr_cast(addr, addr_len, &vm_addr) != 0)
914 		return -EINVAL;
915 
916 	lock_sock(sk);
917 	err = __vsock_bind(sk, vm_addr);
918 	release_sock(sk);
919 
920 	return err;
921 }
922 
923 static int vsock_getname(struct socket *sock,
924 			 struct sockaddr *addr, int peer)
925 {
926 	int err;
927 	struct sock *sk;
928 	struct vsock_sock *vsk;
929 	struct sockaddr_vm *vm_addr;
930 
931 	sk = sock->sk;
932 	vsk = vsock_sk(sk);
933 	err = 0;
934 
935 	lock_sock(sk);
936 
937 	if (peer) {
938 		if (sock->state != SS_CONNECTED) {
939 			err = -ENOTCONN;
940 			goto out;
941 		}
942 		vm_addr = &vsk->remote_addr;
943 	} else {
944 		vm_addr = &vsk->local_addr;
945 	}
946 
947 	if (!vm_addr) {
948 		err = -EINVAL;
949 		goto out;
950 	}
951 
952 	/* sys_getsockname() and sys_getpeername() pass us a
953 	 * MAX_SOCK_ADDR-sized buffer and don't set addr_len.  Unfortunately
954 	 * that macro is defined in socket.c instead of .h, so we hardcode its
955 	 * value here.
956 	 */
957 	BUILD_BUG_ON(sizeof(*vm_addr) > 128);
958 	memcpy(addr, vm_addr, sizeof(*vm_addr));
959 	err = sizeof(*vm_addr);
960 
961 out:
962 	release_sock(sk);
963 	return err;
964 }
965 
966 static int vsock_shutdown(struct socket *sock, int mode)
967 {
968 	int err;
969 	struct sock *sk;
970 
971 	/* User level uses SHUT_RD (0) and SHUT_WR (1), but the kernel uses
972 	 * RCV_SHUTDOWN (1) and SEND_SHUTDOWN (2), so we must increment mode
973 	 * here like the other address families do.  Note also that the
974 	 * increment makes SHUT_RDWR (2) into RCV_SHUTDOWN | SEND_SHUTDOWN (3),
975 	 * which is what we want.
976 	 */
977 	mode++;
978 
979 	if ((mode & ~SHUTDOWN_MASK) || !mode)
980 		return -EINVAL;
981 
982 	/* If this is a connection oriented socket and it is not connected then
983 	 * bail out immediately.  If it is a DGRAM socket then we must first
984 	 * kick the socket so that it wakes up from any sleeping calls, for
985 	 * example recv(), and then afterwards return the error.
986 	 */
987 
988 	sk = sock->sk;
989 
990 	lock_sock(sk);
991 	if (sock->state == SS_UNCONNECTED) {
992 		err = -ENOTCONN;
993 		if (sock_type_connectible(sk->sk_type))
994 			goto out;
995 	} else {
996 		sock->state = SS_DISCONNECTING;
997 		err = 0;
998 	}
999 
1000 	/* Receive and send shutdowns are treated alike. */
1001 	mode = mode & (RCV_SHUTDOWN | SEND_SHUTDOWN);
1002 	if (mode) {
1003 		sk->sk_shutdown |= mode;
1004 		sk->sk_state_change(sk);
1005 
1006 		if (sock_type_connectible(sk->sk_type)) {
1007 			sock_reset_flag(sk, SOCK_DONE);
1008 			vsock_send_shutdown(sk, mode);
1009 		}
1010 	}
1011 
1012 out:
1013 	release_sock(sk);
1014 	return err;
1015 }
1016 
1017 static __poll_t vsock_poll(struct file *file, struct socket *sock,
1018 			       poll_table *wait)
1019 {
1020 	struct sock *sk;
1021 	__poll_t mask;
1022 	struct vsock_sock *vsk;
1023 
1024 	sk = sock->sk;
1025 	vsk = vsock_sk(sk);
1026 
1027 	poll_wait(file, sk_sleep(sk), wait);
1028 	mask = 0;
1029 
1030 	if (sk->sk_err)
1031 		/* Signify that there has been an error on this socket. */
1032 		mask |= EPOLLERR;
1033 
1034 	/* INET sockets treat local write shutdown and peer write shutdown as a
1035 	 * case of EPOLLHUP set.
1036 	 */
1037 	if ((sk->sk_shutdown == SHUTDOWN_MASK) ||
1038 	    ((sk->sk_shutdown & SEND_SHUTDOWN) &&
1039 	     (vsk->peer_shutdown & SEND_SHUTDOWN))) {
1040 		mask |= EPOLLHUP;
1041 	}
1042 
1043 	if (sk->sk_shutdown & RCV_SHUTDOWN ||
1044 	    vsk->peer_shutdown & SEND_SHUTDOWN) {
1045 		mask |= EPOLLRDHUP;
1046 	}
1047 
1048 	if (sock->type == SOCK_DGRAM) {
1049 		/* For datagram sockets we can read if there is something in
1050 		 * the queue and write as long as the socket isn't shutdown for
1051 		 * sending.
1052 		 */
1053 		if (!skb_queue_empty_lockless(&sk->sk_receive_queue) ||
1054 		    (sk->sk_shutdown & RCV_SHUTDOWN)) {
1055 			mask |= EPOLLIN | EPOLLRDNORM;
1056 		}
1057 
1058 		if (!(sk->sk_shutdown & SEND_SHUTDOWN))
1059 			mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND;
1060 
1061 	} else if (sock_type_connectible(sk->sk_type)) {
1062 		const struct vsock_transport *transport;
1063 
1064 		lock_sock(sk);
1065 
1066 		transport = vsk->transport;
1067 
1068 		/* Listening sockets that have connections in their accept
1069 		 * queue can be read.
1070 		 */
1071 		if (sk->sk_state == TCP_LISTEN
1072 		    && !vsock_is_accept_queue_empty(sk))
1073 			mask |= EPOLLIN | EPOLLRDNORM;
1074 
1075 		/* If there is something in the queue then we can read. */
1076 		if (transport && transport->stream_is_active(vsk) &&
1077 		    !(sk->sk_shutdown & RCV_SHUTDOWN)) {
1078 			bool data_ready_now = false;
1079 			int target = sock_rcvlowat(sk, 0, INT_MAX);
1080 			int ret = transport->notify_poll_in(
1081 					vsk, target, &data_ready_now);
1082 			if (ret < 0) {
1083 				mask |= EPOLLERR;
1084 			} else {
1085 				if (data_ready_now)
1086 					mask |= EPOLLIN | EPOLLRDNORM;
1087 
1088 			}
1089 		}
1090 
1091 		/* Sockets whose connections have been closed, reset, or
1092 		 * terminated should also be considered read, and we check the
1093 		 * shutdown flag for that.
1094 		 */
1095 		if (sk->sk_shutdown & RCV_SHUTDOWN ||
1096 		    vsk->peer_shutdown & SEND_SHUTDOWN) {
1097 			mask |= EPOLLIN | EPOLLRDNORM;
1098 		}
1099 
1100 		/* Connected sockets that can produce data can be written. */
1101 		if (transport && sk->sk_state == TCP_ESTABLISHED) {
1102 			if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1103 				bool space_avail_now = false;
1104 				int ret = transport->notify_poll_out(
1105 						vsk, 1, &space_avail_now);
1106 				if (ret < 0) {
1107 					mask |= EPOLLERR;
1108 				} else {
1109 					if (space_avail_now)
1110 						/* Remove EPOLLWRBAND since INET
1111 						 * sockets are not setting it.
1112 						 */
1113 						mask |= EPOLLOUT | EPOLLWRNORM;
1114 
1115 				}
1116 			}
1117 		}
1118 
1119 		/* Simulate INET socket poll behaviors, which sets
1120 		 * EPOLLOUT|EPOLLWRNORM when peer is closed and nothing to read,
1121 		 * but local send is not shutdown.
1122 		 */
1123 		if (sk->sk_state == TCP_CLOSE || sk->sk_state == TCP_CLOSING) {
1124 			if (!(sk->sk_shutdown & SEND_SHUTDOWN))
1125 				mask |= EPOLLOUT | EPOLLWRNORM;
1126 
1127 		}
1128 
1129 		release_sock(sk);
1130 	}
1131 
1132 	return mask;
1133 }
1134 
1135 static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg,
1136 			       size_t len)
1137 {
1138 	int err;
1139 	struct sock *sk;
1140 	struct vsock_sock *vsk;
1141 	struct sockaddr_vm *remote_addr;
1142 	const struct vsock_transport *transport;
1143 
1144 	if (msg->msg_flags & MSG_OOB)
1145 		return -EOPNOTSUPP;
1146 
1147 	/* For now, MSG_DONTWAIT is always assumed... */
1148 	err = 0;
1149 	sk = sock->sk;
1150 	vsk = vsock_sk(sk);
1151 
1152 	lock_sock(sk);
1153 
1154 	transport = vsk->transport;
1155 
1156 	err = vsock_auto_bind(vsk);
1157 	if (err)
1158 		goto out;
1159 
1160 
1161 	/* If the provided message contains an address, use that.  Otherwise
1162 	 * fall back on the socket's remote handle (if it has been connected).
1163 	 */
1164 	if (msg->msg_name &&
1165 	    vsock_addr_cast(msg->msg_name, msg->msg_namelen,
1166 			    &remote_addr) == 0) {
1167 		/* Ensure this address is of the right type and is a valid
1168 		 * destination.
1169 		 */
1170 
1171 		if (remote_addr->svm_cid == VMADDR_CID_ANY)
1172 			remote_addr->svm_cid = transport->get_local_cid();
1173 
1174 		if (!vsock_addr_bound(remote_addr)) {
1175 			err = -EINVAL;
1176 			goto out;
1177 		}
1178 	} else if (sock->state == SS_CONNECTED) {
1179 		remote_addr = &vsk->remote_addr;
1180 
1181 		if (remote_addr->svm_cid == VMADDR_CID_ANY)
1182 			remote_addr->svm_cid = transport->get_local_cid();
1183 
1184 		/* XXX Should connect() or this function ensure remote_addr is
1185 		 * bound?
1186 		 */
1187 		if (!vsock_addr_bound(&vsk->remote_addr)) {
1188 			err = -EINVAL;
1189 			goto out;
1190 		}
1191 	} else {
1192 		err = -EINVAL;
1193 		goto out;
1194 	}
1195 
1196 	if (!transport->dgram_allow(remote_addr->svm_cid,
1197 				    remote_addr->svm_port)) {
1198 		err = -EINVAL;
1199 		goto out;
1200 	}
1201 
1202 	err = transport->dgram_enqueue(vsk, remote_addr, msg, len);
1203 
1204 out:
1205 	release_sock(sk);
1206 	return err;
1207 }
1208 
1209 static int vsock_dgram_connect(struct socket *sock,
1210 			       struct sockaddr *addr, int addr_len, int flags)
1211 {
1212 	int err;
1213 	struct sock *sk;
1214 	struct vsock_sock *vsk;
1215 	struct sockaddr_vm *remote_addr;
1216 
1217 	sk = sock->sk;
1218 	vsk = vsock_sk(sk);
1219 
1220 	err = vsock_addr_cast(addr, addr_len, &remote_addr);
1221 	if (err == -EAFNOSUPPORT && remote_addr->svm_family == AF_UNSPEC) {
1222 		lock_sock(sk);
1223 		vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY,
1224 				VMADDR_PORT_ANY);
1225 		sock->state = SS_UNCONNECTED;
1226 		release_sock(sk);
1227 		return 0;
1228 	} else if (err != 0)
1229 		return -EINVAL;
1230 
1231 	lock_sock(sk);
1232 
1233 	err = vsock_auto_bind(vsk);
1234 	if (err)
1235 		goto out;
1236 
1237 	if (!vsk->transport->dgram_allow(remote_addr->svm_cid,
1238 					 remote_addr->svm_port)) {
1239 		err = -EINVAL;
1240 		goto out;
1241 	}
1242 
1243 	memcpy(&vsk->remote_addr, remote_addr, sizeof(vsk->remote_addr));
1244 	sock->state = SS_CONNECTED;
1245 
1246 out:
1247 	release_sock(sk);
1248 	return err;
1249 }
1250 
1251 static int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg,
1252 			       size_t len, int flags)
1253 {
1254 	struct vsock_sock *vsk = vsock_sk(sock->sk);
1255 
1256 	return vsk->transport->dgram_dequeue(vsk, msg, len, flags);
1257 }
1258 
1259 static const struct proto_ops vsock_dgram_ops = {
1260 	.family = PF_VSOCK,
1261 	.owner = THIS_MODULE,
1262 	.release = vsock_release,
1263 	.bind = vsock_bind,
1264 	.connect = vsock_dgram_connect,
1265 	.socketpair = sock_no_socketpair,
1266 	.accept = sock_no_accept,
1267 	.getname = vsock_getname,
1268 	.poll = vsock_poll,
1269 	.ioctl = sock_no_ioctl,
1270 	.listen = sock_no_listen,
1271 	.shutdown = vsock_shutdown,
1272 	.sendmsg = vsock_dgram_sendmsg,
1273 	.recvmsg = vsock_dgram_recvmsg,
1274 	.mmap = sock_no_mmap,
1275 	.sendpage = sock_no_sendpage,
1276 };
1277 
1278 static int vsock_transport_cancel_pkt(struct vsock_sock *vsk)
1279 {
1280 	const struct vsock_transport *transport = vsk->transport;
1281 
1282 	if (!transport || !transport->cancel_pkt)
1283 		return -EOPNOTSUPP;
1284 
1285 	return transport->cancel_pkt(vsk);
1286 }
1287 
1288 static void vsock_connect_timeout(struct work_struct *work)
1289 {
1290 	struct sock *sk;
1291 	struct vsock_sock *vsk;
1292 
1293 	vsk = container_of(work, struct vsock_sock, connect_work.work);
1294 	sk = sk_vsock(vsk);
1295 
1296 	lock_sock(sk);
1297 	if (sk->sk_state == TCP_SYN_SENT &&
1298 	    (sk->sk_shutdown != SHUTDOWN_MASK)) {
1299 		sk->sk_state = TCP_CLOSE;
1300 		sk->sk_socket->state = SS_UNCONNECTED;
1301 		sk->sk_err = ETIMEDOUT;
1302 		sk_error_report(sk);
1303 		vsock_transport_cancel_pkt(vsk);
1304 	}
1305 	release_sock(sk);
1306 
1307 	sock_put(sk);
1308 }
1309 
1310 static int vsock_connect(struct socket *sock, struct sockaddr *addr,
1311 			 int addr_len, int flags)
1312 {
1313 	int err;
1314 	struct sock *sk;
1315 	struct vsock_sock *vsk;
1316 	const struct vsock_transport *transport;
1317 	struct sockaddr_vm *remote_addr;
1318 	long timeout;
1319 	DEFINE_WAIT(wait);
1320 
1321 	err = 0;
1322 	sk = sock->sk;
1323 	vsk = vsock_sk(sk);
1324 
1325 	lock_sock(sk);
1326 
1327 	/* XXX AF_UNSPEC should make us disconnect like AF_INET. */
1328 	switch (sock->state) {
1329 	case SS_CONNECTED:
1330 		err = -EISCONN;
1331 		goto out;
1332 	case SS_DISCONNECTING:
1333 		err = -EINVAL;
1334 		goto out;
1335 	case SS_CONNECTING:
1336 		/* This continues on so we can move sock into the SS_CONNECTED
1337 		 * state once the connection has completed (at which point err
1338 		 * will be set to zero also).  Otherwise, we will either wait
1339 		 * for the connection or return -EALREADY should this be a
1340 		 * non-blocking call.
1341 		 */
1342 		err = -EALREADY;
1343 		if (flags & O_NONBLOCK)
1344 			goto out;
1345 		break;
1346 	default:
1347 		if ((sk->sk_state == TCP_LISTEN) ||
1348 		    vsock_addr_cast(addr, addr_len, &remote_addr) != 0) {
1349 			err = -EINVAL;
1350 			goto out;
1351 		}
1352 
1353 		/* Set the remote address that we are connecting to. */
1354 		memcpy(&vsk->remote_addr, remote_addr,
1355 		       sizeof(vsk->remote_addr));
1356 
1357 		err = vsock_assign_transport(vsk, NULL);
1358 		if (err)
1359 			goto out;
1360 
1361 		transport = vsk->transport;
1362 
1363 		/* The hypervisor and well-known contexts do not have socket
1364 		 * endpoints.
1365 		 */
1366 		if (!transport ||
1367 		    !transport->stream_allow(remote_addr->svm_cid,
1368 					     remote_addr->svm_port)) {
1369 			err = -ENETUNREACH;
1370 			goto out;
1371 		}
1372 
1373 		err = vsock_auto_bind(vsk);
1374 		if (err)
1375 			goto out;
1376 
1377 		sk->sk_state = TCP_SYN_SENT;
1378 
1379 		err = transport->connect(vsk);
1380 		if (err < 0)
1381 			goto out;
1382 
1383 		/* Mark sock as connecting and set the error code to in
1384 		 * progress in case this is a non-blocking connect.
1385 		 */
1386 		sock->state = SS_CONNECTING;
1387 		err = -EINPROGRESS;
1388 	}
1389 
1390 	/* The receive path will handle all communication until we are able to
1391 	 * enter the connected state.  Here we wait for the connection to be
1392 	 * completed or a notification of an error.
1393 	 */
1394 	timeout = vsk->connect_timeout;
1395 	prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1396 
1397 	while (sk->sk_state != TCP_ESTABLISHED && sk->sk_err == 0) {
1398 		if (flags & O_NONBLOCK) {
1399 			/* If we're not going to block, we schedule a timeout
1400 			 * function to generate a timeout on the connection
1401 			 * attempt, in case the peer doesn't respond in a
1402 			 * timely manner. We hold on to the socket until the
1403 			 * timeout fires.
1404 			 */
1405 			sock_hold(sk);
1406 
1407 			/* If the timeout function is already scheduled,
1408 			 * reschedule it, then ungrab the socket refcount to
1409 			 * keep it balanced.
1410 			 */
1411 			if (mod_delayed_work(system_wq, &vsk->connect_work,
1412 					     timeout))
1413 				sock_put(sk);
1414 
1415 			/* Skip ahead to preserve error code set above. */
1416 			goto out_wait;
1417 		}
1418 
1419 		release_sock(sk);
1420 		timeout = schedule_timeout(timeout);
1421 		lock_sock(sk);
1422 
1423 		if (signal_pending(current)) {
1424 			err = sock_intr_errno(timeout);
1425 			sk->sk_state = sk->sk_state == TCP_ESTABLISHED ? TCP_CLOSING : TCP_CLOSE;
1426 			sock->state = SS_UNCONNECTED;
1427 			vsock_transport_cancel_pkt(vsk);
1428 			vsock_remove_connected(vsk);
1429 			goto out_wait;
1430 		} else if (timeout == 0) {
1431 			err = -ETIMEDOUT;
1432 			sk->sk_state = TCP_CLOSE;
1433 			sock->state = SS_UNCONNECTED;
1434 			vsock_transport_cancel_pkt(vsk);
1435 			goto out_wait;
1436 		}
1437 
1438 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1439 	}
1440 
1441 	if (sk->sk_err) {
1442 		err = -sk->sk_err;
1443 		sk->sk_state = TCP_CLOSE;
1444 		sock->state = SS_UNCONNECTED;
1445 	} else {
1446 		err = 0;
1447 	}
1448 
1449 out_wait:
1450 	finish_wait(sk_sleep(sk), &wait);
1451 out:
1452 	release_sock(sk);
1453 	return err;
1454 }
1455 
1456 static int vsock_accept(struct socket *sock, struct socket *newsock, int flags,
1457 			bool kern)
1458 {
1459 	struct sock *listener;
1460 	int err;
1461 	struct sock *connected;
1462 	struct vsock_sock *vconnected;
1463 	long timeout;
1464 	DEFINE_WAIT(wait);
1465 
1466 	err = 0;
1467 	listener = sock->sk;
1468 
1469 	lock_sock(listener);
1470 
1471 	if (!sock_type_connectible(sock->type)) {
1472 		err = -EOPNOTSUPP;
1473 		goto out;
1474 	}
1475 
1476 	if (listener->sk_state != TCP_LISTEN) {
1477 		err = -EINVAL;
1478 		goto out;
1479 	}
1480 
1481 	/* Wait for children sockets to appear; these are the new sockets
1482 	 * created upon connection establishment.
1483 	 */
1484 	timeout = sock_rcvtimeo(listener, flags & O_NONBLOCK);
1485 	prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1486 
1487 	while ((connected = vsock_dequeue_accept(listener)) == NULL &&
1488 	       listener->sk_err == 0) {
1489 		release_sock(listener);
1490 		timeout = schedule_timeout(timeout);
1491 		finish_wait(sk_sleep(listener), &wait);
1492 		lock_sock(listener);
1493 
1494 		if (signal_pending(current)) {
1495 			err = sock_intr_errno(timeout);
1496 			goto out;
1497 		} else if (timeout == 0) {
1498 			err = -EAGAIN;
1499 			goto out;
1500 		}
1501 
1502 		prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1503 	}
1504 	finish_wait(sk_sleep(listener), &wait);
1505 
1506 	if (listener->sk_err)
1507 		err = -listener->sk_err;
1508 
1509 	if (connected) {
1510 		sk_acceptq_removed(listener);
1511 
1512 		lock_sock_nested(connected, SINGLE_DEPTH_NESTING);
1513 		vconnected = vsock_sk(connected);
1514 
1515 		/* If the listener socket has received an error, then we should
1516 		 * reject this socket and return.  Note that we simply mark the
1517 		 * socket rejected, drop our reference, and let the cleanup
1518 		 * function handle the cleanup; the fact that we found it in
1519 		 * the listener's accept queue guarantees that the cleanup
1520 		 * function hasn't run yet.
1521 		 */
1522 		if (err) {
1523 			vconnected->rejected = true;
1524 		} else {
1525 			newsock->state = SS_CONNECTED;
1526 			sock_graft(connected, newsock);
1527 		}
1528 
1529 		release_sock(connected);
1530 		sock_put(connected);
1531 	}
1532 
1533 out:
1534 	release_sock(listener);
1535 	return err;
1536 }
1537 
1538 static int vsock_listen(struct socket *sock, int backlog)
1539 {
1540 	int err;
1541 	struct sock *sk;
1542 	struct vsock_sock *vsk;
1543 
1544 	sk = sock->sk;
1545 
1546 	lock_sock(sk);
1547 
1548 	if (!sock_type_connectible(sk->sk_type)) {
1549 		err = -EOPNOTSUPP;
1550 		goto out;
1551 	}
1552 
1553 	if (sock->state != SS_UNCONNECTED) {
1554 		err = -EINVAL;
1555 		goto out;
1556 	}
1557 
1558 	vsk = vsock_sk(sk);
1559 
1560 	if (!vsock_addr_bound(&vsk->local_addr)) {
1561 		err = -EINVAL;
1562 		goto out;
1563 	}
1564 
1565 	sk->sk_max_ack_backlog = backlog;
1566 	sk->sk_state = TCP_LISTEN;
1567 
1568 	err = 0;
1569 
1570 out:
1571 	release_sock(sk);
1572 	return err;
1573 }
1574 
1575 static void vsock_update_buffer_size(struct vsock_sock *vsk,
1576 				     const struct vsock_transport *transport,
1577 				     u64 val)
1578 {
1579 	if (val > vsk->buffer_max_size)
1580 		val = vsk->buffer_max_size;
1581 
1582 	if (val < vsk->buffer_min_size)
1583 		val = vsk->buffer_min_size;
1584 
1585 	if (val != vsk->buffer_size &&
1586 	    transport && transport->notify_buffer_size)
1587 		transport->notify_buffer_size(vsk, &val);
1588 
1589 	vsk->buffer_size = val;
1590 }
1591 
1592 static int vsock_connectible_setsockopt(struct socket *sock,
1593 					int level,
1594 					int optname,
1595 					sockptr_t optval,
1596 					unsigned int optlen)
1597 {
1598 	int err;
1599 	struct sock *sk;
1600 	struct vsock_sock *vsk;
1601 	const struct vsock_transport *transport;
1602 	u64 val;
1603 
1604 	if (level != AF_VSOCK)
1605 		return -ENOPROTOOPT;
1606 
1607 #define COPY_IN(_v)                                       \
1608 	do {						  \
1609 		if (optlen < sizeof(_v)) {		  \
1610 			err = -EINVAL;			  \
1611 			goto exit;			  \
1612 		}					  \
1613 		if (copy_from_sockptr(&_v, optval, sizeof(_v)) != 0) {	\
1614 			err = -EFAULT;					\
1615 			goto exit;					\
1616 		}							\
1617 	} while (0)
1618 
1619 	err = 0;
1620 	sk = sock->sk;
1621 	vsk = vsock_sk(sk);
1622 
1623 	lock_sock(sk);
1624 
1625 	transport = vsk->transport;
1626 
1627 	switch (optname) {
1628 	case SO_VM_SOCKETS_BUFFER_SIZE:
1629 		COPY_IN(val);
1630 		vsock_update_buffer_size(vsk, transport, val);
1631 		break;
1632 
1633 	case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1634 		COPY_IN(val);
1635 		vsk->buffer_max_size = val;
1636 		vsock_update_buffer_size(vsk, transport, vsk->buffer_size);
1637 		break;
1638 
1639 	case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1640 		COPY_IN(val);
1641 		vsk->buffer_min_size = val;
1642 		vsock_update_buffer_size(vsk, transport, vsk->buffer_size);
1643 		break;
1644 
1645 	case SO_VM_SOCKETS_CONNECT_TIMEOUT_NEW:
1646 	case SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD: {
1647 		struct __kernel_sock_timeval tv;
1648 
1649 		err = sock_copy_user_timeval(&tv, optval, optlen,
1650 					     optname == SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD);
1651 		if (err)
1652 			break;
1653 		if (tv.tv_sec >= 0 && tv.tv_usec < USEC_PER_SEC &&
1654 		    tv.tv_sec < (MAX_SCHEDULE_TIMEOUT / HZ - 1)) {
1655 			vsk->connect_timeout = tv.tv_sec * HZ +
1656 				DIV_ROUND_UP((unsigned long)tv.tv_usec, (USEC_PER_SEC / HZ));
1657 			if (vsk->connect_timeout == 0)
1658 				vsk->connect_timeout =
1659 				    VSOCK_DEFAULT_CONNECT_TIMEOUT;
1660 
1661 		} else {
1662 			err = -ERANGE;
1663 		}
1664 		break;
1665 	}
1666 
1667 	default:
1668 		err = -ENOPROTOOPT;
1669 		break;
1670 	}
1671 
1672 #undef COPY_IN
1673 
1674 exit:
1675 	release_sock(sk);
1676 	return err;
1677 }
1678 
1679 static int vsock_connectible_getsockopt(struct socket *sock,
1680 					int level, int optname,
1681 					char __user *optval,
1682 					int __user *optlen)
1683 {
1684 	struct sock *sk = sock->sk;
1685 	struct vsock_sock *vsk = vsock_sk(sk);
1686 
1687 	union {
1688 		u64 val64;
1689 		struct old_timeval32 tm32;
1690 		struct __kernel_old_timeval tm;
1691 		struct  __kernel_sock_timeval stm;
1692 	} v;
1693 
1694 	int lv = sizeof(v.val64);
1695 	int len;
1696 
1697 	if (level != AF_VSOCK)
1698 		return -ENOPROTOOPT;
1699 
1700 	if (get_user(len, optlen))
1701 		return -EFAULT;
1702 
1703 	memset(&v, 0, sizeof(v));
1704 
1705 	switch (optname) {
1706 	case SO_VM_SOCKETS_BUFFER_SIZE:
1707 		v.val64 = vsk->buffer_size;
1708 		break;
1709 
1710 	case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1711 		v.val64 = vsk->buffer_max_size;
1712 		break;
1713 
1714 	case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1715 		v.val64 = vsk->buffer_min_size;
1716 		break;
1717 
1718 	case SO_VM_SOCKETS_CONNECT_TIMEOUT_NEW:
1719 	case SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD:
1720 		lv = sock_get_timeout(vsk->connect_timeout, &v,
1721 				      optname == SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD);
1722 		break;
1723 
1724 	default:
1725 		return -ENOPROTOOPT;
1726 	}
1727 
1728 	if (len < lv)
1729 		return -EINVAL;
1730 	if (len > lv)
1731 		len = lv;
1732 	if (copy_to_user(optval, &v, len))
1733 		return -EFAULT;
1734 
1735 	if (put_user(len, optlen))
1736 		return -EFAULT;
1737 
1738 	return 0;
1739 }
1740 
1741 static int vsock_connectible_sendmsg(struct socket *sock, struct msghdr *msg,
1742 				     size_t len)
1743 {
1744 	struct sock *sk;
1745 	struct vsock_sock *vsk;
1746 	const struct vsock_transport *transport;
1747 	ssize_t total_written;
1748 	long timeout;
1749 	int err;
1750 	struct vsock_transport_send_notify_data send_data;
1751 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
1752 
1753 	sk = sock->sk;
1754 	vsk = vsock_sk(sk);
1755 	total_written = 0;
1756 	err = 0;
1757 
1758 	if (msg->msg_flags & MSG_OOB)
1759 		return -EOPNOTSUPP;
1760 
1761 	lock_sock(sk);
1762 
1763 	transport = vsk->transport;
1764 
1765 	/* Callers should not provide a destination with connection oriented
1766 	 * sockets.
1767 	 */
1768 	if (msg->msg_namelen) {
1769 		err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP;
1770 		goto out;
1771 	}
1772 
1773 	/* Send data only if both sides are not shutdown in the direction. */
1774 	if (sk->sk_shutdown & SEND_SHUTDOWN ||
1775 	    vsk->peer_shutdown & RCV_SHUTDOWN) {
1776 		err = -EPIPE;
1777 		goto out;
1778 	}
1779 
1780 	if (!transport || sk->sk_state != TCP_ESTABLISHED ||
1781 	    !vsock_addr_bound(&vsk->local_addr)) {
1782 		err = -ENOTCONN;
1783 		goto out;
1784 	}
1785 
1786 	if (!vsock_addr_bound(&vsk->remote_addr)) {
1787 		err = -EDESTADDRREQ;
1788 		goto out;
1789 	}
1790 
1791 	/* Wait for room in the produce queue to enqueue our user's data. */
1792 	timeout = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1793 
1794 	err = transport->notify_send_init(vsk, &send_data);
1795 	if (err < 0)
1796 		goto out;
1797 
1798 	while (total_written < len) {
1799 		ssize_t written;
1800 
1801 		add_wait_queue(sk_sleep(sk), &wait);
1802 		while (vsock_stream_has_space(vsk) == 0 &&
1803 		       sk->sk_err == 0 &&
1804 		       !(sk->sk_shutdown & SEND_SHUTDOWN) &&
1805 		       !(vsk->peer_shutdown & RCV_SHUTDOWN)) {
1806 
1807 			/* Don't wait for non-blocking sockets. */
1808 			if (timeout == 0) {
1809 				err = -EAGAIN;
1810 				remove_wait_queue(sk_sleep(sk), &wait);
1811 				goto out_err;
1812 			}
1813 
1814 			err = transport->notify_send_pre_block(vsk, &send_data);
1815 			if (err < 0) {
1816 				remove_wait_queue(sk_sleep(sk), &wait);
1817 				goto out_err;
1818 			}
1819 
1820 			release_sock(sk);
1821 			timeout = wait_woken(&wait, TASK_INTERRUPTIBLE, timeout);
1822 			lock_sock(sk);
1823 			if (signal_pending(current)) {
1824 				err = sock_intr_errno(timeout);
1825 				remove_wait_queue(sk_sleep(sk), &wait);
1826 				goto out_err;
1827 			} else if (timeout == 0) {
1828 				err = -EAGAIN;
1829 				remove_wait_queue(sk_sleep(sk), &wait);
1830 				goto out_err;
1831 			}
1832 		}
1833 		remove_wait_queue(sk_sleep(sk), &wait);
1834 
1835 		/* These checks occur both as part of and after the loop
1836 		 * conditional since we need to check before and after
1837 		 * sleeping.
1838 		 */
1839 		if (sk->sk_err) {
1840 			err = -sk->sk_err;
1841 			goto out_err;
1842 		} else if ((sk->sk_shutdown & SEND_SHUTDOWN) ||
1843 			   (vsk->peer_shutdown & RCV_SHUTDOWN)) {
1844 			err = -EPIPE;
1845 			goto out_err;
1846 		}
1847 
1848 		err = transport->notify_send_pre_enqueue(vsk, &send_data);
1849 		if (err < 0)
1850 			goto out_err;
1851 
1852 		/* Note that enqueue will only write as many bytes as are free
1853 		 * in the produce queue, so we don't need to ensure len is
1854 		 * smaller than the queue size.  It is the caller's
1855 		 * responsibility to check how many bytes we were able to send.
1856 		 */
1857 
1858 		if (sk->sk_type == SOCK_SEQPACKET) {
1859 			written = transport->seqpacket_enqueue(vsk,
1860 						msg, len - total_written);
1861 		} else {
1862 			written = transport->stream_enqueue(vsk,
1863 					msg, len - total_written);
1864 		}
1865 		if (written < 0) {
1866 			err = -ENOMEM;
1867 			goto out_err;
1868 		}
1869 
1870 		total_written += written;
1871 
1872 		err = transport->notify_send_post_enqueue(
1873 				vsk, written, &send_data);
1874 		if (err < 0)
1875 			goto out_err;
1876 
1877 	}
1878 
1879 out_err:
1880 	if (total_written > 0) {
1881 		/* Return number of written bytes only if:
1882 		 * 1) SOCK_STREAM socket.
1883 		 * 2) SOCK_SEQPACKET socket when whole buffer is sent.
1884 		 */
1885 		if (sk->sk_type == SOCK_STREAM || total_written == len)
1886 			err = total_written;
1887 	}
1888 out:
1889 	release_sock(sk);
1890 	return err;
1891 }
1892 
1893 static int vsock_connectible_wait_data(struct sock *sk,
1894 				       struct wait_queue_entry *wait,
1895 				       long timeout,
1896 				       struct vsock_transport_recv_notify_data *recv_data,
1897 				       size_t target)
1898 {
1899 	const struct vsock_transport *transport;
1900 	struct vsock_sock *vsk;
1901 	s64 data;
1902 	int err;
1903 
1904 	vsk = vsock_sk(sk);
1905 	err = 0;
1906 	transport = vsk->transport;
1907 
1908 	while (1) {
1909 		prepare_to_wait(sk_sleep(sk), wait, TASK_INTERRUPTIBLE);
1910 		data = vsock_connectible_has_data(vsk);
1911 		if (data != 0)
1912 			break;
1913 
1914 		if (sk->sk_err != 0 ||
1915 		    (sk->sk_shutdown & RCV_SHUTDOWN) ||
1916 		    (vsk->peer_shutdown & SEND_SHUTDOWN)) {
1917 			break;
1918 		}
1919 
1920 		/* Don't wait for non-blocking sockets. */
1921 		if (timeout == 0) {
1922 			err = -EAGAIN;
1923 			break;
1924 		}
1925 
1926 		if (recv_data) {
1927 			err = transport->notify_recv_pre_block(vsk, target, recv_data);
1928 			if (err < 0)
1929 				break;
1930 		}
1931 
1932 		release_sock(sk);
1933 		timeout = schedule_timeout(timeout);
1934 		lock_sock(sk);
1935 
1936 		if (signal_pending(current)) {
1937 			err = sock_intr_errno(timeout);
1938 			break;
1939 		} else if (timeout == 0) {
1940 			err = -EAGAIN;
1941 			break;
1942 		}
1943 	}
1944 
1945 	finish_wait(sk_sleep(sk), wait);
1946 
1947 	if (err)
1948 		return err;
1949 
1950 	/* Internal transport error when checking for available
1951 	 * data. XXX This should be changed to a connection
1952 	 * reset in a later change.
1953 	 */
1954 	if (data < 0)
1955 		return -ENOMEM;
1956 
1957 	return data;
1958 }
1959 
1960 static int __vsock_stream_recvmsg(struct sock *sk, struct msghdr *msg,
1961 				  size_t len, int flags)
1962 {
1963 	struct vsock_transport_recv_notify_data recv_data;
1964 	const struct vsock_transport *transport;
1965 	struct vsock_sock *vsk;
1966 	ssize_t copied;
1967 	size_t target;
1968 	long timeout;
1969 	int err;
1970 
1971 	DEFINE_WAIT(wait);
1972 
1973 	vsk = vsock_sk(sk);
1974 	transport = vsk->transport;
1975 
1976 	/* We must not copy less than target bytes into the user's buffer
1977 	 * before returning successfully, so we wait for the consume queue to
1978 	 * have that much data to consume before dequeueing.  Note that this
1979 	 * makes it impossible to handle cases where target is greater than the
1980 	 * queue size.
1981 	 */
1982 	target = sock_rcvlowat(sk, flags & MSG_WAITALL, len);
1983 	if (target >= transport->stream_rcvhiwat(vsk)) {
1984 		err = -ENOMEM;
1985 		goto out;
1986 	}
1987 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1988 	copied = 0;
1989 
1990 	err = transport->notify_recv_init(vsk, target, &recv_data);
1991 	if (err < 0)
1992 		goto out;
1993 
1994 
1995 	while (1) {
1996 		ssize_t read;
1997 
1998 		err = vsock_connectible_wait_data(sk, &wait, timeout,
1999 						  &recv_data, target);
2000 		if (err <= 0)
2001 			break;
2002 
2003 		err = transport->notify_recv_pre_dequeue(vsk, target,
2004 							 &recv_data);
2005 		if (err < 0)
2006 			break;
2007 
2008 		read = transport->stream_dequeue(vsk, msg, len - copied, flags);
2009 		if (read < 0) {
2010 			err = -ENOMEM;
2011 			break;
2012 		}
2013 
2014 		copied += read;
2015 
2016 		err = transport->notify_recv_post_dequeue(vsk, target, read,
2017 						!(flags & MSG_PEEK), &recv_data);
2018 		if (err < 0)
2019 			goto out;
2020 
2021 		if (read >= target || flags & MSG_PEEK)
2022 			break;
2023 
2024 		target -= read;
2025 	}
2026 
2027 	if (sk->sk_err)
2028 		err = -sk->sk_err;
2029 	else if (sk->sk_shutdown & RCV_SHUTDOWN)
2030 		err = 0;
2031 
2032 	if (copied > 0)
2033 		err = copied;
2034 
2035 out:
2036 	return err;
2037 }
2038 
2039 static int __vsock_seqpacket_recvmsg(struct sock *sk, struct msghdr *msg,
2040 				     size_t len, int flags)
2041 {
2042 	const struct vsock_transport *transport;
2043 	struct vsock_sock *vsk;
2044 	ssize_t msg_len;
2045 	long timeout;
2046 	int err = 0;
2047 	DEFINE_WAIT(wait);
2048 
2049 	vsk = vsock_sk(sk);
2050 	transport = vsk->transport;
2051 
2052 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2053 
2054 	err = vsock_connectible_wait_data(sk, &wait, timeout, NULL, 0);
2055 	if (err <= 0)
2056 		goto out;
2057 
2058 	msg_len = transport->seqpacket_dequeue(vsk, msg, flags);
2059 
2060 	if (msg_len < 0) {
2061 		err = -ENOMEM;
2062 		goto out;
2063 	}
2064 
2065 	if (sk->sk_err) {
2066 		err = -sk->sk_err;
2067 	} else if (sk->sk_shutdown & RCV_SHUTDOWN) {
2068 		err = 0;
2069 	} else {
2070 		/* User sets MSG_TRUNC, so return real length of
2071 		 * packet.
2072 		 */
2073 		if (flags & MSG_TRUNC)
2074 			err = msg_len;
2075 		else
2076 			err = len - msg_data_left(msg);
2077 
2078 		/* Always set MSG_TRUNC if real length of packet is
2079 		 * bigger than user's buffer.
2080 		 */
2081 		if (msg_len > len)
2082 			msg->msg_flags |= MSG_TRUNC;
2083 	}
2084 
2085 out:
2086 	return err;
2087 }
2088 
2089 static int
2090 vsock_connectible_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
2091 			  int flags)
2092 {
2093 	struct sock *sk;
2094 	struct vsock_sock *vsk;
2095 	const struct vsock_transport *transport;
2096 	int err;
2097 
2098 	sk = sock->sk;
2099 	vsk = vsock_sk(sk);
2100 	err = 0;
2101 
2102 	lock_sock(sk);
2103 
2104 	transport = vsk->transport;
2105 
2106 	if (!transport || sk->sk_state != TCP_ESTABLISHED) {
2107 		/* Recvmsg is supposed to return 0 if a peer performs an
2108 		 * orderly shutdown. Differentiate between that case and when a
2109 		 * peer has not connected or a local shutdown occurred with the
2110 		 * SOCK_DONE flag.
2111 		 */
2112 		if (sock_flag(sk, SOCK_DONE))
2113 			err = 0;
2114 		else
2115 			err = -ENOTCONN;
2116 
2117 		goto out;
2118 	}
2119 
2120 	if (flags & MSG_OOB) {
2121 		err = -EOPNOTSUPP;
2122 		goto out;
2123 	}
2124 
2125 	/* We don't check peer_shutdown flag here since peer may actually shut
2126 	 * down, but there can be data in the queue that a local socket can
2127 	 * receive.
2128 	 */
2129 	if (sk->sk_shutdown & RCV_SHUTDOWN) {
2130 		err = 0;
2131 		goto out;
2132 	}
2133 
2134 	/* It is valid on Linux to pass in a zero-length receive buffer.  This
2135 	 * is not an error.  We may as well bail out now.
2136 	 */
2137 	if (!len) {
2138 		err = 0;
2139 		goto out;
2140 	}
2141 
2142 	if (sk->sk_type == SOCK_STREAM)
2143 		err = __vsock_stream_recvmsg(sk, msg, len, flags);
2144 	else
2145 		err = __vsock_seqpacket_recvmsg(sk, msg, len, flags);
2146 
2147 out:
2148 	release_sock(sk);
2149 	return err;
2150 }
2151 
2152 static int vsock_set_rcvlowat(struct sock *sk, int val)
2153 {
2154 	const struct vsock_transport *transport;
2155 	struct vsock_sock *vsk;
2156 
2157 	vsk = vsock_sk(sk);
2158 
2159 	if (val > vsk->buffer_size)
2160 		return -EINVAL;
2161 
2162 	transport = vsk->transport;
2163 
2164 	if (transport && transport->set_rcvlowat)
2165 		return transport->set_rcvlowat(vsk, val);
2166 
2167 	WRITE_ONCE(sk->sk_rcvlowat, val ? : 1);
2168 	return 0;
2169 }
2170 
2171 static const struct proto_ops vsock_stream_ops = {
2172 	.family = PF_VSOCK,
2173 	.owner = THIS_MODULE,
2174 	.release = vsock_release,
2175 	.bind = vsock_bind,
2176 	.connect = vsock_connect,
2177 	.socketpair = sock_no_socketpair,
2178 	.accept = vsock_accept,
2179 	.getname = vsock_getname,
2180 	.poll = vsock_poll,
2181 	.ioctl = sock_no_ioctl,
2182 	.listen = vsock_listen,
2183 	.shutdown = vsock_shutdown,
2184 	.setsockopt = vsock_connectible_setsockopt,
2185 	.getsockopt = vsock_connectible_getsockopt,
2186 	.sendmsg = vsock_connectible_sendmsg,
2187 	.recvmsg = vsock_connectible_recvmsg,
2188 	.mmap = sock_no_mmap,
2189 	.sendpage = sock_no_sendpage,
2190 	.set_rcvlowat = vsock_set_rcvlowat,
2191 };
2192 
2193 static const struct proto_ops vsock_seqpacket_ops = {
2194 	.family = PF_VSOCK,
2195 	.owner = THIS_MODULE,
2196 	.release = vsock_release,
2197 	.bind = vsock_bind,
2198 	.connect = vsock_connect,
2199 	.socketpair = sock_no_socketpair,
2200 	.accept = vsock_accept,
2201 	.getname = vsock_getname,
2202 	.poll = vsock_poll,
2203 	.ioctl = sock_no_ioctl,
2204 	.listen = vsock_listen,
2205 	.shutdown = vsock_shutdown,
2206 	.setsockopt = vsock_connectible_setsockopt,
2207 	.getsockopt = vsock_connectible_getsockopt,
2208 	.sendmsg = vsock_connectible_sendmsg,
2209 	.recvmsg = vsock_connectible_recvmsg,
2210 	.mmap = sock_no_mmap,
2211 	.sendpage = sock_no_sendpage,
2212 };
2213 
2214 static int vsock_create(struct net *net, struct socket *sock,
2215 			int protocol, int kern)
2216 {
2217 	struct vsock_sock *vsk;
2218 	struct sock *sk;
2219 	int ret;
2220 
2221 	if (!sock)
2222 		return -EINVAL;
2223 
2224 	if (protocol && protocol != PF_VSOCK)
2225 		return -EPROTONOSUPPORT;
2226 
2227 	switch (sock->type) {
2228 	case SOCK_DGRAM:
2229 		sock->ops = &vsock_dgram_ops;
2230 		break;
2231 	case SOCK_STREAM:
2232 		sock->ops = &vsock_stream_ops;
2233 		break;
2234 	case SOCK_SEQPACKET:
2235 		sock->ops = &vsock_seqpacket_ops;
2236 		break;
2237 	default:
2238 		return -ESOCKTNOSUPPORT;
2239 	}
2240 
2241 	sock->state = SS_UNCONNECTED;
2242 
2243 	sk = __vsock_create(net, sock, NULL, GFP_KERNEL, 0, kern);
2244 	if (!sk)
2245 		return -ENOMEM;
2246 
2247 	vsk = vsock_sk(sk);
2248 
2249 	if (sock->type == SOCK_DGRAM) {
2250 		ret = vsock_assign_transport(vsk, NULL);
2251 		if (ret < 0) {
2252 			sock_put(sk);
2253 			return ret;
2254 		}
2255 	}
2256 
2257 	vsock_insert_unbound(vsk);
2258 
2259 	return 0;
2260 }
2261 
2262 static const struct net_proto_family vsock_family_ops = {
2263 	.family = AF_VSOCK,
2264 	.create = vsock_create,
2265 	.owner = THIS_MODULE,
2266 };
2267 
2268 static long vsock_dev_do_ioctl(struct file *filp,
2269 			       unsigned int cmd, void __user *ptr)
2270 {
2271 	u32 __user *p = ptr;
2272 	u32 cid = VMADDR_CID_ANY;
2273 	int retval = 0;
2274 
2275 	switch (cmd) {
2276 	case IOCTL_VM_SOCKETS_GET_LOCAL_CID:
2277 		/* To be compatible with the VMCI behavior, we prioritize the
2278 		 * guest CID instead of well-know host CID (VMADDR_CID_HOST).
2279 		 */
2280 		if (transport_g2h)
2281 			cid = transport_g2h->get_local_cid();
2282 		else if (transport_h2g)
2283 			cid = transport_h2g->get_local_cid();
2284 
2285 		if (put_user(cid, p) != 0)
2286 			retval = -EFAULT;
2287 		break;
2288 
2289 	default:
2290 		retval = -ENOIOCTLCMD;
2291 	}
2292 
2293 	return retval;
2294 }
2295 
2296 static long vsock_dev_ioctl(struct file *filp,
2297 			    unsigned int cmd, unsigned long arg)
2298 {
2299 	return vsock_dev_do_ioctl(filp, cmd, (void __user *)arg);
2300 }
2301 
2302 #ifdef CONFIG_COMPAT
2303 static long vsock_dev_compat_ioctl(struct file *filp,
2304 				   unsigned int cmd, unsigned long arg)
2305 {
2306 	return vsock_dev_do_ioctl(filp, cmd, compat_ptr(arg));
2307 }
2308 #endif
2309 
2310 static const struct file_operations vsock_device_ops = {
2311 	.owner		= THIS_MODULE,
2312 	.unlocked_ioctl	= vsock_dev_ioctl,
2313 #ifdef CONFIG_COMPAT
2314 	.compat_ioctl	= vsock_dev_compat_ioctl,
2315 #endif
2316 	.open		= nonseekable_open,
2317 };
2318 
2319 static struct miscdevice vsock_device = {
2320 	.name		= "vsock",
2321 	.fops		= &vsock_device_ops,
2322 };
2323 
2324 static int __init vsock_init(void)
2325 {
2326 	int err = 0;
2327 
2328 	vsock_init_tables();
2329 
2330 	vsock_proto.owner = THIS_MODULE;
2331 	vsock_device.minor = MISC_DYNAMIC_MINOR;
2332 	err = misc_register(&vsock_device);
2333 	if (err) {
2334 		pr_err("Failed to register misc device\n");
2335 		goto err_reset_transport;
2336 	}
2337 
2338 	err = proto_register(&vsock_proto, 1);	/* we want our slab */
2339 	if (err) {
2340 		pr_err("Cannot register vsock protocol\n");
2341 		goto err_deregister_misc;
2342 	}
2343 
2344 	err = sock_register(&vsock_family_ops);
2345 	if (err) {
2346 		pr_err("could not register af_vsock (%d) address family: %d\n",
2347 		       AF_VSOCK, err);
2348 		goto err_unregister_proto;
2349 	}
2350 
2351 	return 0;
2352 
2353 err_unregister_proto:
2354 	proto_unregister(&vsock_proto);
2355 err_deregister_misc:
2356 	misc_deregister(&vsock_device);
2357 err_reset_transport:
2358 	return err;
2359 }
2360 
2361 static void __exit vsock_exit(void)
2362 {
2363 	misc_deregister(&vsock_device);
2364 	sock_unregister(AF_VSOCK);
2365 	proto_unregister(&vsock_proto);
2366 }
2367 
2368 const struct vsock_transport *vsock_core_get_transport(struct vsock_sock *vsk)
2369 {
2370 	return vsk->transport;
2371 }
2372 EXPORT_SYMBOL_GPL(vsock_core_get_transport);
2373 
2374 int vsock_core_register(const struct vsock_transport *t, int features)
2375 {
2376 	const struct vsock_transport *t_h2g, *t_g2h, *t_dgram, *t_local;
2377 	int err = mutex_lock_interruptible(&vsock_register_mutex);
2378 
2379 	if (err)
2380 		return err;
2381 
2382 	t_h2g = transport_h2g;
2383 	t_g2h = transport_g2h;
2384 	t_dgram = transport_dgram;
2385 	t_local = transport_local;
2386 
2387 	if (features & VSOCK_TRANSPORT_F_H2G) {
2388 		if (t_h2g) {
2389 			err = -EBUSY;
2390 			goto err_busy;
2391 		}
2392 		t_h2g = t;
2393 	}
2394 
2395 	if (features & VSOCK_TRANSPORT_F_G2H) {
2396 		if (t_g2h) {
2397 			err = -EBUSY;
2398 			goto err_busy;
2399 		}
2400 		t_g2h = t;
2401 	}
2402 
2403 	if (features & VSOCK_TRANSPORT_F_DGRAM) {
2404 		if (t_dgram) {
2405 			err = -EBUSY;
2406 			goto err_busy;
2407 		}
2408 		t_dgram = t;
2409 	}
2410 
2411 	if (features & VSOCK_TRANSPORT_F_LOCAL) {
2412 		if (t_local) {
2413 			err = -EBUSY;
2414 			goto err_busy;
2415 		}
2416 		t_local = t;
2417 	}
2418 
2419 	transport_h2g = t_h2g;
2420 	transport_g2h = t_g2h;
2421 	transport_dgram = t_dgram;
2422 	transport_local = t_local;
2423 
2424 err_busy:
2425 	mutex_unlock(&vsock_register_mutex);
2426 	return err;
2427 }
2428 EXPORT_SYMBOL_GPL(vsock_core_register);
2429 
2430 void vsock_core_unregister(const struct vsock_transport *t)
2431 {
2432 	mutex_lock(&vsock_register_mutex);
2433 
2434 	if (transport_h2g == t)
2435 		transport_h2g = NULL;
2436 
2437 	if (transport_g2h == t)
2438 		transport_g2h = NULL;
2439 
2440 	if (transport_dgram == t)
2441 		transport_dgram = NULL;
2442 
2443 	if (transport_local == t)
2444 		transport_local = NULL;
2445 
2446 	mutex_unlock(&vsock_register_mutex);
2447 }
2448 EXPORT_SYMBOL_GPL(vsock_core_unregister);
2449 
2450 module_init(vsock_init);
2451 module_exit(vsock_exit);
2452 
2453 MODULE_AUTHOR("VMware, Inc.");
2454 MODULE_DESCRIPTION("VMware Virtual Socket Family");
2455 MODULE_VERSION("1.0.2.0-k");
2456 MODULE_LICENSE("GPL v2");
2457