xref: /openbmc/linux/net/vmw_vsock/af_vsock.c (revision e65e175b)
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 = get_random_u32_above(LAST_RESERVED_PORT);
630 
631 	vsock_addr_init(&new_addr, addr->svm_cid, addr->svm_port);
632 
633 	if (addr->svm_port == VMADDR_PORT_ANY) {
634 		bool found = false;
635 		unsigned int i;
636 
637 		for (i = 0; i < MAX_PORT_RETRIES; i++) {
638 			if (port <= LAST_RESERVED_PORT)
639 				port = LAST_RESERVED_PORT + 1;
640 
641 			new_addr.svm_port = port++;
642 
643 			if (!__vsock_find_bound_socket(&new_addr)) {
644 				found = true;
645 				break;
646 			}
647 		}
648 
649 		if (!found)
650 			return -EADDRNOTAVAIL;
651 	} else {
652 		/* If port is in reserved range, ensure caller
653 		 * has necessary privileges.
654 		 */
655 		if (addr->svm_port <= LAST_RESERVED_PORT &&
656 		    !capable(CAP_NET_BIND_SERVICE)) {
657 			return -EACCES;
658 		}
659 
660 		if (__vsock_find_bound_socket(&new_addr))
661 			return -EADDRINUSE;
662 	}
663 
664 	vsock_addr_init(&vsk->local_addr, new_addr.svm_cid, new_addr.svm_port);
665 
666 	/* Remove connection oriented sockets from the unbound list and add them
667 	 * to the hash table for easy lookup by its address.  The unbound list
668 	 * is simply an extra entry at the end of the hash table, a trick used
669 	 * by AF_UNIX.
670 	 */
671 	__vsock_remove_bound(vsk);
672 	__vsock_insert_bound(vsock_bound_sockets(&vsk->local_addr), vsk);
673 
674 	return 0;
675 }
676 
677 static int __vsock_bind_dgram(struct vsock_sock *vsk,
678 			      struct sockaddr_vm *addr)
679 {
680 	return vsk->transport->dgram_bind(vsk, addr);
681 }
682 
683 static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr)
684 {
685 	struct vsock_sock *vsk = vsock_sk(sk);
686 	int retval;
687 
688 	/* First ensure this socket isn't already bound. */
689 	if (vsock_addr_bound(&vsk->local_addr))
690 		return -EINVAL;
691 
692 	/* Now bind to the provided address or select appropriate values if
693 	 * none are provided (VMADDR_CID_ANY and VMADDR_PORT_ANY).  Note that
694 	 * like AF_INET prevents binding to a non-local IP address (in most
695 	 * cases), we only allow binding to a local CID.
696 	 */
697 	if (addr->svm_cid != VMADDR_CID_ANY && !vsock_find_cid(addr->svm_cid))
698 		return -EADDRNOTAVAIL;
699 
700 	switch (sk->sk_socket->type) {
701 	case SOCK_STREAM:
702 	case SOCK_SEQPACKET:
703 		spin_lock_bh(&vsock_table_lock);
704 		retval = __vsock_bind_connectible(vsk, addr);
705 		spin_unlock_bh(&vsock_table_lock);
706 		break;
707 
708 	case SOCK_DGRAM:
709 		retval = __vsock_bind_dgram(vsk, addr);
710 		break;
711 
712 	default:
713 		retval = -EINVAL;
714 		break;
715 	}
716 
717 	return retval;
718 }
719 
720 static void vsock_connect_timeout(struct work_struct *work);
721 
722 static struct sock *__vsock_create(struct net *net,
723 				   struct socket *sock,
724 				   struct sock *parent,
725 				   gfp_t priority,
726 				   unsigned short type,
727 				   int kern)
728 {
729 	struct sock *sk;
730 	struct vsock_sock *psk;
731 	struct vsock_sock *vsk;
732 
733 	sk = sk_alloc(net, AF_VSOCK, priority, &vsock_proto, kern);
734 	if (!sk)
735 		return NULL;
736 
737 	sock_init_data(sock, sk);
738 
739 	/* sk->sk_type is normally set in sock_init_data, but only if sock is
740 	 * non-NULL. We make sure that our sockets always have a type by
741 	 * setting it here if needed.
742 	 */
743 	if (!sock)
744 		sk->sk_type = type;
745 
746 	vsk = vsock_sk(sk);
747 	vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
748 	vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
749 
750 	sk->sk_destruct = vsock_sk_destruct;
751 	sk->sk_backlog_rcv = vsock_queue_rcv_skb;
752 	sock_reset_flag(sk, SOCK_DONE);
753 
754 	INIT_LIST_HEAD(&vsk->bound_table);
755 	INIT_LIST_HEAD(&vsk->connected_table);
756 	vsk->listener = NULL;
757 	INIT_LIST_HEAD(&vsk->pending_links);
758 	INIT_LIST_HEAD(&vsk->accept_queue);
759 	vsk->rejected = false;
760 	vsk->sent_request = false;
761 	vsk->ignore_connecting_rst = false;
762 	vsk->peer_shutdown = 0;
763 	INIT_DELAYED_WORK(&vsk->connect_work, vsock_connect_timeout);
764 	INIT_DELAYED_WORK(&vsk->pending_work, vsock_pending_work);
765 
766 	psk = parent ? vsock_sk(parent) : NULL;
767 	if (parent) {
768 		vsk->trusted = psk->trusted;
769 		vsk->owner = get_cred(psk->owner);
770 		vsk->connect_timeout = psk->connect_timeout;
771 		vsk->buffer_size = psk->buffer_size;
772 		vsk->buffer_min_size = psk->buffer_min_size;
773 		vsk->buffer_max_size = psk->buffer_max_size;
774 		security_sk_clone(parent, sk);
775 	} else {
776 		vsk->trusted = ns_capable_noaudit(&init_user_ns, CAP_NET_ADMIN);
777 		vsk->owner = get_current_cred();
778 		vsk->connect_timeout = VSOCK_DEFAULT_CONNECT_TIMEOUT;
779 		vsk->buffer_size = VSOCK_DEFAULT_BUFFER_SIZE;
780 		vsk->buffer_min_size = VSOCK_DEFAULT_BUFFER_MIN_SIZE;
781 		vsk->buffer_max_size = VSOCK_DEFAULT_BUFFER_MAX_SIZE;
782 	}
783 
784 	return sk;
785 }
786 
787 static bool sock_type_connectible(u16 type)
788 {
789 	return (type == SOCK_STREAM) || (type == SOCK_SEQPACKET);
790 }
791 
792 static void __vsock_release(struct sock *sk, int level)
793 {
794 	if (sk) {
795 		struct sock *pending;
796 		struct vsock_sock *vsk;
797 
798 		vsk = vsock_sk(sk);
799 		pending = NULL;	/* Compiler warning. */
800 
801 		/* When "level" is SINGLE_DEPTH_NESTING, use the nested
802 		 * version to avoid the warning "possible recursive locking
803 		 * detected". When "level" is 0, lock_sock_nested(sk, level)
804 		 * is the same as lock_sock(sk).
805 		 */
806 		lock_sock_nested(sk, level);
807 
808 		if (vsk->transport)
809 			vsk->transport->release(vsk);
810 		else if (sock_type_connectible(sk->sk_type))
811 			vsock_remove_sock(vsk);
812 
813 		sock_orphan(sk);
814 		sk->sk_shutdown = SHUTDOWN_MASK;
815 
816 		skb_queue_purge(&sk->sk_receive_queue);
817 
818 		/* Clean up any sockets that never were accepted. */
819 		while ((pending = vsock_dequeue_accept(sk)) != NULL) {
820 			__vsock_release(pending, SINGLE_DEPTH_NESTING);
821 			sock_put(pending);
822 		}
823 
824 		release_sock(sk);
825 		sock_put(sk);
826 	}
827 }
828 
829 static void vsock_sk_destruct(struct sock *sk)
830 {
831 	struct vsock_sock *vsk = vsock_sk(sk);
832 
833 	vsock_deassign_transport(vsk);
834 
835 	/* When clearing these addresses, there's no need to set the family and
836 	 * possibly register the address family with the kernel.
837 	 */
838 	vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
839 	vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
840 
841 	put_cred(vsk->owner);
842 }
843 
844 static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
845 {
846 	int err;
847 
848 	err = sock_queue_rcv_skb(sk, skb);
849 	if (err)
850 		kfree_skb(skb);
851 
852 	return err;
853 }
854 
855 struct sock *vsock_create_connected(struct sock *parent)
856 {
857 	return __vsock_create(sock_net(parent), NULL, parent, GFP_KERNEL,
858 			      parent->sk_type, 0);
859 }
860 EXPORT_SYMBOL_GPL(vsock_create_connected);
861 
862 s64 vsock_stream_has_data(struct vsock_sock *vsk)
863 {
864 	return vsk->transport->stream_has_data(vsk);
865 }
866 EXPORT_SYMBOL_GPL(vsock_stream_has_data);
867 
868 static s64 vsock_connectible_has_data(struct vsock_sock *vsk)
869 {
870 	struct sock *sk = sk_vsock(vsk);
871 
872 	if (sk->sk_type == SOCK_SEQPACKET)
873 		return vsk->transport->seqpacket_has_data(vsk);
874 	else
875 		return vsock_stream_has_data(vsk);
876 }
877 
878 s64 vsock_stream_has_space(struct vsock_sock *vsk)
879 {
880 	return vsk->transport->stream_has_space(vsk);
881 }
882 EXPORT_SYMBOL_GPL(vsock_stream_has_space);
883 
884 void vsock_data_ready(struct sock *sk)
885 {
886 	struct vsock_sock *vsk = vsock_sk(sk);
887 
888 	if (vsock_stream_has_data(vsk) >= sk->sk_rcvlowat ||
889 	    sock_flag(sk, SOCK_DONE))
890 		sk->sk_data_ready(sk);
891 }
892 EXPORT_SYMBOL_GPL(vsock_data_ready);
893 
894 static int vsock_release(struct socket *sock)
895 {
896 	__vsock_release(sock->sk, 0);
897 	sock->sk = NULL;
898 	sock->state = SS_FREE;
899 
900 	return 0;
901 }
902 
903 static int
904 vsock_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
905 {
906 	int err;
907 	struct sock *sk;
908 	struct sockaddr_vm *vm_addr;
909 
910 	sk = sock->sk;
911 
912 	if (vsock_addr_cast(addr, addr_len, &vm_addr) != 0)
913 		return -EINVAL;
914 
915 	lock_sock(sk);
916 	err = __vsock_bind(sk, vm_addr);
917 	release_sock(sk);
918 
919 	return err;
920 }
921 
922 static int vsock_getname(struct socket *sock,
923 			 struct sockaddr *addr, int peer)
924 {
925 	int err;
926 	struct sock *sk;
927 	struct vsock_sock *vsk;
928 	struct sockaddr_vm *vm_addr;
929 
930 	sk = sock->sk;
931 	vsk = vsock_sk(sk);
932 	err = 0;
933 
934 	lock_sock(sk);
935 
936 	if (peer) {
937 		if (sock->state != SS_CONNECTED) {
938 			err = -ENOTCONN;
939 			goto out;
940 		}
941 		vm_addr = &vsk->remote_addr;
942 	} else {
943 		vm_addr = &vsk->local_addr;
944 	}
945 
946 	if (!vm_addr) {
947 		err = -EINVAL;
948 		goto out;
949 	}
950 
951 	/* sys_getsockname() and sys_getpeername() pass us a
952 	 * MAX_SOCK_ADDR-sized buffer and don't set addr_len.  Unfortunately
953 	 * that macro is defined in socket.c instead of .h, so we hardcode its
954 	 * value here.
955 	 */
956 	BUILD_BUG_ON(sizeof(*vm_addr) > 128);
957 	memcpy(addr, vm_addr, sizeof(*vm_addr));
958 	err = sizeof(*vm_addr);
959 
960 out:
961 	release_sock(sk);
962 	return err;
963 }
964 
965 static int vsock_shutdown(struct socket *sock, int mode)
966 {
967 	int err;
968 	struct sock *sk;
969 
970 	/* User level uses SHUT_RD (0) and SHUT_WR (1), but the kernel uses
971 	 * RCV_SHUTDOWN (1) and SEND_SHUTDOWN (2), so we must increment mode
972 	 * here like the other address families do.  Note also that the
973 	 * increment makes SHUT_RDWR (2) into RCV_SHUTDOWN | SEND_SHUTDOWN (3),
974 	 * which is what we want.
975 	 */
976 	mode++;
977 
978 	if ((mode & ~SHUTDOWN_MASK) || !mode)
979 		return -EINVAL;
980 
981 	/* If this is a connection oriented socket and it is not connected then
982 	 * bail out immediately.  If it is a DGRAM socket then we must first
983 	 * kick the socket so that it wakes up from any sleeping calls, for
984 	 * example recv(), and then afterwards return the error.
985 	 */
986 
987 	sk = sock->sk;
988 
989 	lock_sock(sk);
990 	if (sock->state == SS_UNCONNECTED) {
991 		err = -ENOTCONN;
992 		if (sock_type_connectible(sk->sk_type))
993 			goto out;
994 	} else {
995 		sock->state = SS_DISCONNECTING;
996 		err = 0;
997 	}
998 
999 	/* Receive and send shutdowns are treated alike. */
1000 	mode = mode & (RCV_SHUTDOWN | SEND_SHUTDOWN);
1001 	if (mode) {
1002 		sk->sk_shutdown |= mode;
1003 		sk->sk_state_change(sk);
1004 
1005 		if (sock_type_connectible(sk->sk_type)) {
1006 			sock_reset_flag(sk, SOCK_DONE);
1007 			vsock_send_shutdown(sk, mode);
1008 		}
1009 	}
1010 
1011 out:
1012 	release_sock(sk);
1013 	return err;
1014 }
1015 
1016 static __poll_t vsock_poll(struct file *file, struct socket *sock,
1017 			       poll_table *wait)
1018 {
1019 	struct sock *sk;
1020 	__poll_t mask;
1021 	struct vsock_sock *vsk;
1022 
1023 	sk = sock->sk;
1024 	vsk = vsock_sk(sk);
1025 
1026 	poll_wait(file, sk_sleep(sk), wait);
1027 	mask = 0;
1028 
1029 	if (sk->sk_err)
1030 		/* Signify that there has been an error on this socket. */
1031 		mask |= EPOLLERR;
1032 
1033 	/* INET sockets treat local write shutdown and peer write shutdown as a
1034 	 * case of EPOLLHUP set.
1035 	 */
1036 	if ((sk->sk_shutdown == SHUTDOWN_MASK) ||
1037 	    ((sk->sk_shutdown & SEND_SHUTDOWN) &&
1038 	     (vsk->peer_shutdown & SEND_SHUTDOWN))) {
1039 		mask |= EPOLLHUP;
1040 	}
1041 
1042 	if (sk->sk_shutdown & RCV_SHUTDOWN ||
1043 	    vsk->peer_shutdown & SEND_SHUTDOWN) {
1044 		mask |= EPOLLRDHUP;
1045 	}
1046 
1047 	if (sock->type == SOCK_DGRAM) {
1048 		/* For datagram sockets we can read if there is something in
1049 		 * the queue and write as long as the socket isn't shutdown for
1050 		 * sending.
1051 		 */
1052 		if (!skb_queue_empty_lockless(&sk->sk_receive_queue) ||
1053 		    (sk->sk_shutdown & RCV_SHUTDOWN)) {
1054 			mask |= EPOLLIN | EPOLLRDNORM;
1055 		}
1056 
1057 		if (!(sk->sk_shutdown & SEND_SHUTDOWN))
1058 			mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND;
1059 
1060 	} else if (sock_type_connectible(sk->sk_type)) {
1061 		const struct vsock_transport *transport;
1062 
1063 		lock_sock(sk);
1064 
1065 		transport = vsk->transport;
1066 
1067 		/* Listening sockets that have connections in their accept
1068 		 * queue can be read.
1069 		 */
1070 		if (sk->sk_state == TCP_LISTEN
1071 		    && !vsock_is_accept_queue_empty(sk))
1072 			mask |= EPOLLIN | EPOLLRDNORM;
1073 
1074 		/* If there is something in the queue then we can read. */
1075 		if (transport && transport->stream_is_active(vsk) &&
1076 		    !(sk->sk_shutdown & RCV_SHUTDOWN)) {
1077 			bool data_ready_now = false;
1078 			int target = sock_rcvlowat(sk, 0, INT_MAX);
1079 			int ret = transport->notify_poll_in(
1080 					vsk, target, &data_ready_now);
1081 			if (ret < 0) {
1082 				mask |= EPOLLERR;
1083 			} else {
1084 				if (data_ready_now)
1085 					mask |= EPOLLIN | EPOLLRDNORM;
1086 
1087 			}
1088 		}
1089 
1090 		/* Sockets whose connections have been closed, reset, or
1091 		 * terminated should also be considered read, and we check the
1092 		 * shutdown flag for that.
1093 		 */
1094 		if (sk->sk_shutdown & RCV_SHUTDOWN ||
1095 		    vsk->peer_shutdown & SEND_SHUTDOWN) {
1096 			mask |= EPOLLIN | EPOLLRDNORM;
1097 		}
1098 
1099 		/* Connected sockets that can produce data can be written. */
1100 		if (transport && sk->sk_state == TCP_ESTABLISHED) {
1101 			if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1102 				bool space_avail_now = false;
1103 				int ret = transport->notify_poll_out(
1104 						vsk, 1, &space_avail_now);
1105 				if (ret < 0) {
1106 					mask |= EPOLLERR;
1107 				} else {
1108 					if (space_avail_now)
1109 						/* Remove EPOLLWRBAND since INET
1110 						 * sockets are not setting it.
1111 						 */
1112 						mask |= EPOLLOUT | EPOLLWRNORM;
1113 
1114 				}
1115 			}
1116 		}
1117 
1118 		/* Simulate INET socket poll behaviors, which sets
1119 		 * EPOLLOUT|EPOLLWRNORM when peer is closed and nothing to read,
1120 		 * but local send is not shutdown.
1121 		 */
1122 		if (sk->sk_state == TCP_CLOSE || sk->sk_state == TCP_CLOSING) {
1123 			if (!(sk->sk_shutdown & SEND_SHUTDOWN))
1124 				mask |= EPOLLOUT | EPOLLWRNORM;
1125 
1126 		}
1127 
1128 		release_sock(sk);
1129 	}
1130 
1131 	return mask;
1132 }
1133 
1134 static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg,
1135 			       size_t len)
1136 {
1137 	int err;
1138 	struct sock *sk;
1139 	struct vsock_sock *vsk;
1140 	struct sockaddr_vm *remote_addr;
1141 	const struct vsock_transport *transport;
1142 
1143 	if (msg->msg_flags & MSG_OOB)
1144 		return -EOPNOTSUPP;
1145 
1146 	/* For now, MSG_DONTWAIT is always assumed... */
1147 	err = 0;
1148 	sk = sock->sk;
1149 	vsk = vsock_sk(sk);
1150 
1151 	lock_sock(sk);
1152 
1153 	transport = vsk->transport;
1154 
1155 	err = vsock_auto_bind(vsk);
1156 	if (err)
1157 		goto out;
1158 
1159 
1160 	/* If the provided message contains an address, use that.  Otherwise
1161 	 * fall back on the socket's remote handle (if it has been connected).
1162 	 */
1163 	if (msg->msg_name &&
1164 	    vsock_addr_cast(msg->msg_name, msg->msg_namelen,
1165 			    &remote_addr) == 0) {
1166 		/* Ensure this address is of the right type and is a valid
1167 		 * destination.
1168 		 */
1169 
1170 		if (remote_addr->svm_cid == VMADDR_CID_ANY)
1171 			remote_addr->svm_cid = transport->get_local_cid();
1172 
1173 		if (!vsock_addr_bound(remote_addr)) {
1174 			err = -EINVAL;
1175 			goto out;
1176 		}
1177 	} else if (sock->state == SS_CONNECTED) {
1178 		remote_addr = &vsk->remote_addr;
1179 
1180 		if (remote_addr->svm_cid == VMADDR_CID_ANY)
1181 			remote_addr->svm_cid = transport->get_local_cid();
1182 
1183 		/* XXX Should connect() or this function ensure remote_addr is
1184 		 * bound?
1185 		 */
1186 		if (!vsock_addr_bound(&vsk->remote_addr)) {
1187 			err = -EINVAL;
1188 			goto out;
1189 		}
1190 	} else {
1191 		err = -EINVAL;
1192 		goto out;
1193 	}
1194 
1195 	if (!transport->dgram_allow(remote_addr->svm_cid,
1196 				    remote_addr->svm_port)) {
1197 		err = -EINVAL;
1198 		goto out;
1199 	}
1200 
1201 	err = transport->dgram_enqueue(vsk, remote_addr, msg, len);
1202 
1203 out:
1204 	release_sock(sk);
1205 	return err;
1206 }
1207 
1208 static int vsock_dgram_connect(struct socket *sock,
1209 			       struct sockaddr *addr, int addr_len, int flags)
1210 {
1211 	int err;
1212 	struct sock *sk;
1213 	struct vsock_sock *vsk;
1214 	struct sockaddr_vm *remote_addr;
1215 
1216 	sk = sock->sk;
1217 	vsk = vsock_sk(sk);
1218 
1219 	err = vsock_addr_cast(addr, addr_len, &remote_addr);
1220 	if (err == -EAFNOSUPPORT && remote_addr->svm_family == AF_UNSPEC) {
1221 		lock_sock(sk);
1222 		vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY,
1223 				VMADDR_PORT_ANY);
1224 		sock->state = SS_UNCONNECTED;
1225 		release_sock(sk);
1226 		return 0;
1227 	} else if (err != 0)
1228 		return -EINVAL;
1229 
1230 	lock_sock(sk);
1231 
1232 	err = vsock_auto_bind(vsk);
1233 	if (err)
1234 		goto out;
1235 
1236 	if (!vsk->transport->dgram_allow(remote_addr->svm_cid,
1237 					 remote_addr->svm_port)) {
1238 		err = -EINVAL;
1239 		goto out;
1240 	}
1241 
1242 	memcpy(&vsk->remote_addr, remote_addr, sizeof(vsk->remote_addr));
1243 	sock->state = SS_CONNECTED;
1244 
1245 out:
1246 	release_sock(sk);
1247 	return err;
1248 }
1249 
1250 static int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg,
1251 			       size_t len, int flags)
1252 {
1253 	struct vsock_sock *vsk = vsock_sk(sock->sk);
1254 
1255 	return vsk->transport->dgram_dequeue(vsk, msg, len, flags);
1256 }
1257 
1258 static const struct proto_ops vsock_dgram_ops = {
1259 	.family = PF_VSOCK,
1260 	.owner = THIS_MODULE,
1261 	.release = vsock_release,
1262 	.bind = vsock_bind,
1263 	.connect = vsock_dgram_connect,
1264 	.socketpair = sock_no_socketpair,
1265 	.accept = sock_no_accept,
1266 	.getname = vsock_getname,
1267 	.poll = vsock_poll,
1268 	.ioctl = sock_no_ioctl,
1269 	.listen = sock_no_listen,
1270 	.shutdown = vsock_shutdown,
1271 	.sendmsg = vsock_dgram_sendmsg,
1272 	.recvmsg = vsock_dgram_recvmsg,
1273 	.mmap = sock_no_mmap,
1274 	.sendpage = sock_no_sendpage,
1275 };
1276 
1277 static int vsock_transport_cancel_pkt(struct vsock_sock *vsk)
1278 {
1279 	const struct vsock_transport *transport = vsk->transport;
1280 
1281 	if (!transport || !transport->cancel_pkt)
1282 		return -EOPNOTSUPP;
1283 
1284 	return transport->cancel_pkt(vsk);
1285 }
1286 
1287 static void vsock_connect_timeout(struct work_struct *work)
1288 {
1289 	struct sock *sk;
1290 	struct vsock_sock *vsk;
1291 
1292 	vsk = container_of(work, struct vsock_sock, connect_work.work);
1293 	sk = sk_vsock(vsk);
1294 
1295 	lock_sock(sk);
1296 	if (sk->sk_state == TCP_SYN_SENT &&
1297 	    (sk->sk_shutdown != SHUTDOWN_MASK)) {
1298 		sk->sk_state = TCP_CLOSE;
1299 		sk->sk_socket->state = SS_UNCONNECTED;
1300 		sk->sk_err = ETIMEDOUT;
1301 		sk_error_report(sk);
1302 		vsock_transport_cancel_pkt(vsk);
1303 	}
1304 	release_sock(sk);
1305 
1306 	sock_put(sk);
1307 }
1308 
1309 static int vsock_connect(struct socket *sock, struct sockaddr *addr,
1310 			 int addr_len, int flags)
1311 {
1312 	int err;
1313 	struct sock *sk;
1314 	struct vsock_sock *vsk;
1315 	const struct vsock_transport *transport;
1316 	struct sockaddr_vm *remote_addr;
1317 	long timeout;
1318 	DEFINE_WAIT(wait);
1319 
1320 	err = 0;
1321 	sk = sock->sk;
1322 	vsk = vsock_sk(sk);
1323 
1324 	lock_sock(sk);
1325 
1326 	/* XXX AF_UNSPEC should make us disconnect like AF_INET. */
1327 	switch (sock->state) {
1328 	case SS_CONNECTED:
1329 		err = -EISCONN;
1330 		goto out;
1331 	case SS_DISCONNECTING:
1332 		err = -EINVAL;
1333 		goto out;
1334 	case SS_CONNECTING:
1335 		/* This continues on so we can move sock into the SS_CONNECTED
1336 		 * state once the connection has completed (at which point err
1337 		 * will be set to zero also).  Otherwise, we will either wait
1338 		 * for the connection or return -EALREADY should this be a
1339 		 * non-blocking call.
1340 		 */
1341 		err = -EALREADY;
1342 		if (flags & O_NONBLOCK)
1343 			goto out;
1344 		break;
1345 	default:
1346 		if ((sk->sk_state == TCP_LISTEN) ||
1347 		    vsock_addr_cast(addr, addr_len, &remote_addr) != 0) {
1348 			err = -EINVAL;
1349 			goto out;
1350 		}
1351 
1352 		/* Set the remote address that we are connecting to. */
1353 		memcpy(&vsk->remote_addr, remote_addr,
1354 		       sizeof(vsk->remote_addr));
1355 
1356 		err = vsock_assign_transport(vsk, NULL);
1357 		if (err)
1358 			goto out;
1359 
1360 		transport = vsk->transport;
1361 
1362 		/* The hypervisor and well-known contexts do not have socket
1363 		 * endpoints.
1364 		 */
1365 		if (!transport ||
1366 		    !transport->stream_allow(remote_addr->svm_cid,
1367 					     remote_addr->svm_port)) {
1368 			err = -ENETUNREACH;
1369 			goto out;
1370 		}
1371 
1372 		err = vsock_auto_bind(vsk);
1373 		if (err)
1374 			goto out;
1375 
1376 		sk->sk_state = TCP_SYN_SENT;
1377 
1378 		err = transport->connect(vsk);
1379 		if (err < 0)
1380 			goto out;
1381 
1382 		/* Mark sock as connecting and set the error code to in
1383 		 * progress in case this is a non-blocking connect.
1384 		 */
1385 		sock->state = SS_CONNECTING;
1386 		err = -EINPROGRESS;
1387 	}
1388 
1389 	/* The receive path will handle all communication until we are able to
1390 	 * enter the connected state.  Here we wait for the connection to be
1391 	 * completed or a notification of an error.
1392 	 */
1393 	timeout = vsk->connect_timeout;
1394 	prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1395 
1396 	while (sk->sk_state != TCP_ESTABLISHED && sk->sk_err == 0) {
1397 		if (flags & O_NONBLOCK) {
1398 			/* If we're not going to block, we schedule a timeout
1399 			 * function to generate a timeout on the connection
1400 			 * attempt, in case the peer doesn't respond in a
1401 			 * timely manner. We hold on to the socket until the
1402 			 * timeout fires.
1403 			 */
1404 			sock_hold(sk);
1405 
1406 			/* If the timeout function is already scheduled,
1407 			 * reschedule it, then ungrab the socket refcount to
1408 			 * keep it balanced.
1409 			 */
1410 			if (mod_delayed_work(system_wq, &vsk->connect_work,
1411 					     timeout))
1412 				sock_put(sk);
1413 
1414 			/* Skip ahead to preserve error code set above. */
1415 			goto out_wait;
1416 		}
1417 
1418 		release_sock(sk);
1419 		timeout = schedule_timeout(timeout);
1420 		lock_sock(sk);
1421 
1422 		if (signal_pending(current)) {
1423 			err = sock_intr_errno(timeout);
1424 			sk->sk_state = sk->sk_state == TCP_ESTABLISHED ? TCP_CLOSING : TCP_CLOSE;
1425 			sock->state = SS_UNCONNECTED;
1426 			vsock_transport_cancel_pkt(vsk);
1427 			vsock_remove_connected(vsk);
1428 			goto out_wait;
1429 		} else if (timeout == 0) {
1430 			err = -ETIMEDOUT;
1431 			sk->sk_state = TCP_CLOSE;
1432 			sock->state = SS_UNCONNECTED;
1433 			vsock_transport_cancel_pkt(vsk);
1434 			goto out_wait;
1435 		}
1436 
1437 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1438 	}
1439 
1440 	if (sk->sk_err) {
1441 		err = -sk->sk_err;
1442 		sk->sk_state = TCP_CLOSE;
1443 		sock->state = SS_UNCONNECTED;
1444 	} else {
1445 		err = 0;
1446 	}
1447 
1448 out_wait:
1449 	finish_wait(sk_sleep(sk), &wait);
1450 out:
1451 	release_sock(sk);
1452 	return err;
1453 }
1454 
1455 static int vsock_accept(struct socket *sock, struct socket *newsock, int flags,
1456 			bool kern)
1457 {
1458 	struct sock *listener;
1459 	int err;
1460 	struct sock *connected;
1461 	struct vsock_sock *vconnected;
1462 	long timeout;
1463 	DEFINE_WAIT(wait);
1464 
1465 	err = 0;
1466 	listener = sock->sk;
1467 
1468 	lock_sock(listener);
1469 
1470 	if (!sock_type_connectible(sock->type)) {
1471 		err = -EOPNOTSUPP;
1472 		goto out;
1473 	}
1474 
1475 	if (listener->sk_state != TCP_LISTEN) {
1476 		err = -EINVAL;
1477 		goto out;
1478 	}
1479 
1480 	/* Wait for children sockets to appear; these are the new sockets
1481 	 * created upon connection establishment.
1482 	 */
1483 	timeout = sock_rcvtimeo(listener, flags & O_NONBLOCK);
1484 	prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1485 
1486 	while ((connected = vsock_dequeue_accept(listener)) == NULL &&
1487 	       listener->sk_err == 0) {
1488 		release_sock(listener);
1489 		timeout = schedule_timeout(timeout);
1490 		finish_wait(sk_sleep(listener), &wait);
1491 		lock_sock(listener);
1492 
1493 		if (signal_pending(current)) {
1494 			err = sock_intr_errno(timeout);
1495 			goto out;
1496 		} else if (timeout == 0) {
1497 			err = -EAGAIN;
1498 			goto out;
1499 		}
1500 
1501 		prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1502 	}
1503 	finish_wait(sk_sleep(listener), &wait);
1504 
1505 	if (listener->sk_err)
1506 		err = -listener->sk_err;
1507 
1508 	if (connected) {
1509 		sk_acceptq_removed(listener);
1510 
1511 		lock_sock_nested(connected, SINGLE_DEPTH_NESTING);
1512 		vconnected = vsock_sk(connected);
1513 
1514 		/* If the listener socket has received an error, then we should
1515 		 * reject this socket and return.  Note that we simply mark the
1516 		 * socket rejected, drop our reference, and let the cleanup
1517 		 * function handle the cleanup; the fact that we found it in
1518 		 * the listener's accept queue guarantees that the cleanup
1519 		 * function hasn't run yet.
1520 		 */
1521 		if (err) {
1522 			vconnected->rejected = true;
1523 		} else {
1524 			newsock->state = SS_CONNECTED;
1525 			sock_graft(connected, newsock);
1526 		}
1527 
1528 		release_sock(connected);
1529 		sock_put(connected);
1530 	}
1531 
1532 out:
1533 	release_sock(listener);
1534 	return err;
1535 }
1536 
1537 static int vsock_listen(struct socket *sock, int backlog)
1538 {
1539 	int err;
1540 	struct sock *sk;
1541 	struct vsock_sock *vsk;
1542 
1543 	sk = sock->sk;
1544 
1545 	lock_sock(sk);
1546 
1547 	if (!sock_type_connectible(sk->sk_type)) {
1548 		err = -EOPNOTSUPP;
1549 		goto out;
1550 	}
1551 
1552 	if (sock->state != SS_UNCONNECTED) {
1553 		err = -EINVAL;
1554 		goto out;
1555 	}
1556 
1557 	vsk = vsock_sk(sk);
1558 
1559 	if (!vsock_addr_bound(&vsk->local_addr)) {
1560 		err = -EINVAL;
1561 		goto out;
1562 	}
1563 
1564 	sk->sk_max_ack_backlog = backlog;
1565 	sk->sk_state = TCP_LISTEN;
1566 
1567 	err = 0;
1568 
1569 out:
1570 	release_sock(sk);
1571 	return err;
1572 }
1573 
1574 static void vsock_update_buffer_size(struct vsock_sock *vsk,
1575 				     const struct vsock_transport *transport,
1576 				     u64 val)
1577 {
1578 	if (val > vsk->buffer_max_size)
1579 		val = vsk->buffer_max_size;
1580 
1581 	if (val < vsk->buffer_min_size)
1582 		val = vsk->buffer_min_size;
1583 
1584 	if (val != vsk->buffer_size &&
1585 	    transport && transport->notify_buffer_size)
1586 		transport->notify_buffer_size(vsk, &val);
1587 
1588 	vsk->buffer_size = val;
1589 }
1590 
1591 static int vsock_connectible_setsockopt(struct socket *sock,
1592 					int level,
1593 					int optname,
1594 					sockptr_t optval,
1595 					unsigned int optlen)
1596 {
1597 	int err;
1598 	struct sock *sk;
1599 	struct vsock_sock *vsk;
1600 	const struct vsock_transport *transport;
1601 	u64 val;
1602 
1603 	if (level != AF_VSOCK)
1604 		return -ENOPROTOOPT;
1605 
1606 #define COPY_IN(_v)                                       \
1607 	do {						  \
1608 		if (optlen < sizeof(_v)) {		  \
1609 			err = -EINVAL;			  \
1610 			goto exit;			  \
1611 		}					  \
1612 		if (copy_from_sockptr(&_v, optval, sizeof(_v)) != 0) {	\
1613 			err = -EFAULT;					\
1614 			goto exit;					\
1615 		}							\
1616 	} while (0)
1617 
1618 	err = 0;
1619 	sk = sock->sk;
1620 	vsk = vsock_sk(sk);
1621 
1622 	lock_sock(sk);
1623 
1624 	transport = vsk->transport;
1625 
1626 	switch (optname) {
1627 	case SO_VM_SOCKETS_BUFFER_SIZE:
1628 		COPY_IN(val);
1629 		vsock_update_buffer_size(vsk, transport, val);
1630 		break;
1631 
1632 	case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1633 		COPY_IN(val);
1634 		vsk->buffer_max_size = val;
1635 		vsock_update_buffer_size(vsk, transport, vsk->buffer_size);
1636 		break;
1637 
1638 	case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1639 		COPY_IN(val);
1640 		vsk->buffer_min_size = val;
1641 		vsock_update_buffer_size(vsk, transport, vsk->buffer_size);
1642 		break;
1643 
1644 	case SO_VM_SOCKETS_CONNECT_TIMEOUT_NEW:
1645 	case SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD: {
1646 		struct __kernel_sock_timeval tv;
1647 
1648 		err = sock_copy_user_timeval(&tv, optval, optlen,
1649 					     optname == SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD);
1650 		if (err)
1651 			break;
1652 		if (tv.tv_sec >= 0 && tv.tv_usec < USEC_PER_SEC &&
1653 		    tv.tv_sec < (MAX_SCHEDULE_TIMEOUT / HZ - 1)) {
1654 			vsk->connect_timeout = tv.tv_sec * HZ +
1655 				DIV_ROUND_UP((unsigned long)tv.tv_usec, (USEC_PER_SEC / HZ));
1656 			if (vsk->connect_timeout == 0)
1657 				vsk->connect_timeout =
1658 				    VSOCK_DEFAULT_CONNECT_TIMEOUT;
1659 
1660 		} else {
1661 			err = -ERANGE;
1662 		}
1663 		break;
1664 	}
1665 
1666 	default:
1667 		err = -ENOPROTOOPT;
1668 		break;
1669 	}
1670 
1671 #undef COPY_IN
1672 
1673 exit:
1674 	release_sock(sk);
1675 	return err;
1676 }
1677 
1678 static int vsock_connectible_getsockopt(struct socket *sock,
1679 					int level, int optname,
1680 					char __user *optval,
1681 					int __user *optlen)
1682 {
1683 	struct sock *sk = sock->sk;
1684 	struct vsock_sock *vsk = vsock_sk(sk);
1685 
1686 	union {
1687 		u64 val64;
1688 		struct old_timeval32 tm32;
1689 		struct __kernel_old_timeval tm;
1690 		struct  __kernel_sock_timeval stm;
1691 	} v;
1692 
1693 	int lv = sizeof(v.val64);
1694 	int len;
1695 
1696 	if (level != AF_VSOCK)
1697 		return -ENOPROTOOPT;
1698 
1699 	if (get_user(len, optlen))
1700 		return -EFAULT;
1701 
1702 	memset(&v, 0, sizeof(v));
1703 
1704 	switch (optname) {
1705 	case SO_VM_SOCKETS_BUFFER_SIZE:
1706 		v.val64 = vsk->buffer_size;
1707 		break;
1708 
1709 	case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1710 		v.val64 = vsk->buffer_max_size;
1711 		break;
1712 
1713 	case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1714 		v.val64 = vsk->buffer_min_size;
1715 		break;
1716 
1717 	case SO_VM_SOCKETS_CONNECT_TIMEOUT_NEW:
1718 	case SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD:
1719 		lv = sock_get_timeout(vsk->connect_timeout, &v,
1720 				      optname == SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD);
1721 		break;
1722 
1723 	default:
1724 		return -ENOPROTOOPT;
1725 	}
1726 
1727 	if (len < lv)
1728 		return -EINVAL;
1729 	if (len > lv)
1730 		len = lv;
1731 	if (copy_to_user(optval, &v, len))
1732 		return -EFAULT;
1733 
1734 	if (put_user(len, optlen))
1735 		return -EFAULT;
1736 
1737 	return 0;
1738 }
1739 
1740 static int vsock_connectible_sendmsg(struct socket *sock, struct msghdr *msg,
1741 				     size_t len)
1742 {
1743 	struct sock *sk;
1744 	struct vsock_sock *vsk;
1745 	const struct vsock_transport *transport;
1746 	ssize_t total_written;
1747 	long timeout;
1748 	int err;
1749 	struct vsock_transport_send_notify_data send_data;
1750 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
1751 
1752 	sk = sock->sk;
1753 	vsk = vsock_sk(sk);
1754 	total_written = 0;
1755 	err = 0;
1756 
1757 	if (msg->msg_flags & MSG_OOB)
1758 		return -EOPNOTSUPP;
1759 
1760 	lock_sock(sk);
1761 
1762 	transport = vsk->transport;
1763 
1764 	/* Callers should not provide a destination with connection oriented
1765 	 * sockets.
1766 	 */
1767 	if (msg->msg_namelen) {
1768 		err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP;
1769 		goto out;
1770 	}
1771 
1772 	/* Send data only if both sides are not shutdown in the direction. */
1773 	if (sk->sk_shutdown & SEND_SHUTDOWN ||
1774 	    vsk->peer_shutdown & RCV_SHUTDOWN) {
1775 		err = -EPIPE;
1776 		goto out;
1777 	}
1778 
1779 	if (!transport || sk->sk_state != TCP_ESTABLISHED ||
1780 	    !vsock_addr_bound(&vsk->local_addr)) {
1781 		err = -ENOTCONN;
1782 		goto out;
1783 	}
1784 
1785 	if (!vsock_addr_bound(&vsk->remote_addr)) {
1786 		err = -EDESTADDRREQ;
1787 		goto out;
1788 	}
1789 
1790 	/* Wait for room in the produce queue to enqueue our user's data. */
1791 	timeout = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1792 
1793 	err = transport->notify_send_init(vsk, &send_data);
1794 	if (err < 0)
1795 		goto out;
1796 
1797 	while (total_written < len) {
1798 		ssize_t written;
1799 
1800 		add_wait_queue(sk_sleep(sk), &wait);
1801 		while (vsock_stream_has_space(vsk) == 0 &&
1802 		       sk->sk_err == 0 &&
1803 		       !(sk->sk_shutdown & SEND_SHUTDOWN) &&
1804 		       !(vsk->peer_shutdown & RCV_SHUTDOWN)) {
1805 
1806 			/* Don't wait for non-blocking sockets. */
1807 			if (timeout == 0) {
1808 				err = -EAGAIN;
1809 				remove_wait_queue(sk_sleep(sk), &wait);
1810 				goto out_err;
1811 			}
1812 
1813 			err = transport->notify_send_pre_block(vsk, &send_data);
1814 			if (err < 0) {
1815 				remove_wait_queue(sk_sleep(sk), &wait);
1816 				goto out_err;
1817 			}
1818 
1819 			release_sock(sk);
1820 			timeout = wait_woken(&wait, TASK_INTERRUPTIBLE, timeout);
1821 			lock_sock(sk);
1822 			if (signal_pending(current)) {
1823 				err = sock_intr_errno(timeout);
1824 				remove_wait_queue(sk_sleep(sk), &wait);
1825 				goto out_err;
1826 			} else if (timeout == 0) {
1827 				err = -EAGAIN;
1828 				remove_wait_queue(sk_sleep(sk), &wait);
1829 				goto out_err;
1830 			}
1831 		}
1832 		remove_wait_queue(sk_sleep(sk), &wait);
1833 
1834 		/* These checks occur both as part of and after the loop
1835 		 * conditional since we need to check before and after
1836 		 * sleeping.
1837 		 */
1838 		if (sk->sk_err) {
1839 			err = -sk->sk_err;
1840 			goto out_err;
1841 		} else if ((sk->sk_shutdown & SEND_SHUTDOWN) ||
1842 			   (vsk->peer_shutdown & RCV_SHUTDOWN)) {
1843 			err = -EPIPE;
1844 			goto out_err;
1845 		}
1846 
1847 		err = transport->notify_send_pre_enqueue(vsk, &send_data);
1848 		if (err < 0)
1849 			goto out_err;
1850 
1851 		/* Note that enqueue will only write as many bytes as are free
1852 		 * in the produce queue, so we don't need to ensure len is
1853 		 * smaller than the queue size.  It is the caller's
1854 		 * responsibility to check how many bytes we were able to send.
1855 		 */
1856 
1857 		if (sk->sk_type == SOCK_SEQPACKET) {
1858 			written = transport->seqpacket_enqueue(vsk,
1859 						msg, len - total_written);
1860 		} else {
1861 			written = transport->stream_enqueue(vsk,
1862 					msg, len - total_written);
1863 		}
1864 		if (written < 0) {
1865 			err = -ENOMEM;
1866 			goto out_err;
1867 		}
1868 
1869 		total_written += written;
1870 
1871 		err = transport->notify_send_post_enqueue(
1872 				vsk, written, &send_data);
1873 		if (err < 0)
1874 			goto out_err;
1875 
1876 	}
1877 
1878 out_err:
1879 	if (total_written > 0) {
1880 		/* Return number of written bytes only if:
1881 		 * 1) SOCK_STREAM socket.
1882 		 * 2) SOCK_SEQPACKET socket when whole buffer is sent.
1883 		 */
1884 		if (sk->sk_type == SOCK_STREAM || total_written == len)
1885 			err = total_written;
1886 	}
1887 out:
1888 	release_sock(sk);
1889 	return err;
1890 }
1891 
1892 static int vsock_connectible_wait_data(struct sock *sk,
1893 				       struct wait_queue_entry *wait,
1894 				       long timeout,
1895 				       struct vsock_transport_recv_notify_data *recv_data,
1896 				       size_t target)
1897 {
1898 	const struct vsock_transport *transport;
1899 	struct vsock_sock *vsk;
1900 	s64 data;
1901 	int err;
1902 
1903 	vsk = vsock_sk(sk);
1904 	err = 0;
1905 	transport = vsk->transport;
1906 
1907 	while (1) {
1908 		prepare_to_wait(sk_sleep(sk), wait, TASK_INTERRUPTIBLE);
1909 		data = vsock_connectible_has_data(vsk);
1910 		if (data != 0)
1911 			break;
1912 
1913 		if (sk->sk_err != 0 ||
1914 		    (sk->sk_shutdown & RCV_SHUTDOWN) ||
1915 		    (vsk->peer_shutdown & SEND_SHUTDOWN)) {
1916 			break;
1917 		}
1918 
1919 		/* Don't wait for non-blocking sockets. */
1920 		if (timeout == 0) {
1921 			err = -EAGAIN;
1922 			break;
1923 		}
1924 
1925 		if (recv_data) {
1926 			err = transport->notify_recv_pre_block(vsk, target, recv_data);
1927 			if (err < 0)
1928 				break;
1929 		}
1930 
1931 		release_sock(sk);
1932 		timeout = schedule_timeout(timeout);
1933 		lock_sock(sk);
1934 
1935 		if (signal_pending(current)) {
1936 			err = sock_intr_errno(timeout);
1937 			break;
1938 		} else if (timeout == 0) {
1939 			err = -EAGAIN;
1940 			break;
1941 		}
1942 	}
1943 
1944 	finish_wait(sk_sleep(sk), wait);
1945 
1946 	if (err)
1947 		return err;
1948 
1949 	/* Internal transport error when checking for available
1950 	 * data. XXX This should be changed to a connection
1951 	 * reset in a later change.
1952 	 */
1953 	if (data < 0)
1954 		return -ENOMEM;
1955 
1956 	return data;
1957 }
1958 
1959 static int __vsock_stream_recvmsg(struct sock *sk, struct msghdr *msg,
1960 				  size_t len, int flags)
1961 {
1962 	struct vsock_transport_recv_notify_data recv_data;
1963 	const struct vsock_transport *transport;
1964 	struct vsock_sock *vsk;
1965 	ssize_t copied;
1966 	size_t target;
1967 	long timeout;
1968 	int err;
1969 
1970 	DEFINE_WAIT(wait);
1971 
1972 	vsk = vsock_sk(sk);
1973 	transport = vsk->transport;
1974 
1975 	/* We must not copy less than target bytes into the user's buffer
1976 	 * before returning successfully, so we wait for the consume queue to
1977 	 * have that much data to consume before dequeueing.  Note that this
1978 	 * makes it impossible to handle cases where target is greater than the
1979 	 * queue size.
1980 	 */
1981 	target = sock_rcvlowat(sk, flags & MSG_WAITALL, len);
1982 	if (target >= transport->stream_rcvhiwat(vsk)) {
1983 		err = -ENOMEM;
1984 		goto out;
1985 	}
1986 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1987 	copied = 0;
1988 
1989 	err = transport->notify_recv_init(vsk, target, &recv_data);
1990 	if (err < 0)
1991 		goto out;
1992 
1993 
1994 	while (1) {
1995 		ssize_t read;
1996 
1997 		err = vsock_connectible_wait_data(sk, &wait, timeout,
1998 						  &recv_data, target);
1999 		if (err <= 0)
2000 			break;
2001 
2002 		err = transport->notify_recv_pre_dequeue(vsk, target,
2003 							 &recv_data);
2004 		if (err < 0)
2005 			break;
2006 
2007 		read = transport->stream_dequeue(vsk, msg, len - copied, flags);
2008 		if (read < 0) {
2009 			err = -ENOMEM;
2010 			break;
2011 		}
2012 
2013 		copied += read;
2014 
2015 		err = transport->notify_recv_post_dequeue(vsk, target, read,
2016 						!(flags & MSG_PEEK), &recv_data);
2017 		if (err < 0)
2018 			goto out;
2019 
2020 		if (read >= target || flags & MSG_PEEK)
2021 			break;
2022 
2023 		target -= read;
2024 	}
2025 
2026 	if (sk->sk_err)
2027 		err = -sk->sk_err;
2028 	else if (sk->sk_shutdown & RCV_SHUTDOWN)
2029 		err = 0;
2030 
2031 	if (copied > 0)
2032 		err = copied;
2033 
2034 out:
2035 	return err;
2036 }
2037 
2038 static int __vsock_seqpacket_recvmsg(struct sock *sk, struct msghdr *msg,
2039 				     size_t len, int flags)
2040 {
2041 	const struct vsock_transport *transport;
2042 	struct vsock_sock *vsk;
2043 	ssize_t msg_len;
2044 	long timeout;
2045 	int err = 0;
2046 	DEFINE_WAIT(wait);
2047 
2048 	vsk = vsock_sk(sk);
2049 	transport = vsk->transport;
2050 
2051 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2052 
2053 	err = vsock_connectible_wait_data(sk, &wait, timeout, NULL, 0);
2054 	if (err <= 0)
2055 		goto out;
2056 
2057 	msg_len = transport->seqpacket_dequeue(vsk, msg, flags);
2058 
2059 	if (msg_len < 0) {
2060 		err = -ENOMEM;
2061 		goto out;
2062 	}
2063 
2064 	if (sk->sk_err) {
2065 		err = -sk->sk_err;
2066 	} else if (sk->sk_shutdown & RCV_SHUTDOWN) {
2067 		err = 0;
2068 	} else {
2069 		/* User sets MSG_TRUNC, so return real length of
2070 		 * packet.
2071 		 */
2072 		if (flags & MSG_TRUNC)
2073 			err = msg_len;
2074 		else
2075 			err = len - msg_data_left(msg);
2076 
2077 		/* Always set MSG_TRUNC if real length of packet is
2078 		 * bigger than user's buffer.
2079 		 */
2080 		if (msg_len > len)
2081 			msg->msg_flags |= MSG_TRUNC;
2082 	}
2083 
2084 out:
2085 	return err;
2086 }
2087 
2088 static int
2089 vsock_connectible_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
2090 			  int flags)
2091 {
2092 	struct sock *sk;
2093 	struct vsock_sock *vsk;
2094 	const struct vsock_transport *transport;
2095 	int err;
2096 
2097 	sk = sock->sk;
2098 	vsk = vsock_sk(sk);
2099 	err = 0;
2100 
2101 	lock_sock(sk);
2102 
2103 	transport = vsk->transport;
2104 
2105 	if (!transport || sk->sk_state != TCP_ESTABLISHED) {
2106 		/* Recvmsg is supposed to return 0 if a peer performs an
2107 		 * orderly shutdown. Differentiate between that case and when a
2108 		 * peer has not connected or a local shutdown occurred with the
2109 		 * SOCK_DONE flag.
2110 		 */
2111 		if (sock_flag(sk, SOCK_DONE))
2112 			err = 0;
2113 		else
2114 			err = -ENOTCONN;
2115 
2116 		goto out;
2117 	}
2118 
2119 	if (flags & MSG_OOB) {
2120 		err = -EOPNOTSUPP;
2121 		goto out;
2122 	}
2123 
2124 	/* We don't check peer_shutdown flag here since peer may actually shut
2125 	 * down, but there can be data in the queue that a local socket can
2126 	 * receive.
2127 	 */
2128 	if (sk->sk_shutdown & RCV_SHUTDOWN) {
2129 		err = 0;
2130 		goto out;
2131 	}
2132 
2133 	/* It is valid on Linux to pass in a zero-length receive buffer.  This
2134 	 * is not an error.  We may as well bail out now.
2135 	 */
2136 	if (!len) {
2137 		err = 0;
2138 		goto out;
2139 	}
2140 
2141 	if (sk->sk_type == SOCK_STREAM)
2142 		err = __vsock_stream_recvmsg(sk, msg, len, flags);
2143 	else
2144 		err = __vsock_seqpacket_recvmsg(sk, msg, len, flags);
2145 
2146 out:
2147 	release_sock(sk);
2148 	return err;
2149 }
2150 
2151 static int vsock_set_rcvlowat(struct sock *sk, int val)
2152 {
2153 	const struct vsock_transport *transport;
2154 	struct vsock_sock *vsk;
2155 
2156 	vsk = vsock_sk(sk);
2157 
2158 	if (val > vsk->buffer_size)
2159 		return -EINVAL;
2160 
2161 	transport = vsk->transport;
2162 
2163 	if (transport && transport->set_rcvlowat)
2164 		return transport->set_rcvlowat(vsk, val);
2165 
2166 	WRITE_ONCE(sk->sk_rcvlowat, val ? : 1);
2167 	return 0;
2168 }
2169 
2170 static const struct proto_ops vsock_stream_ops = {
2171 	.family = PF_VSOCK,
2172 	.owner = THIS_MODULE,
2173 	.release = vsock_release,
2174 	.bind = vsock_bind,
2175 	.connect = vsock_connect,
2176 	.socketpair = sock_no_socketpair,
2177 	.accept = vsock_accept,
2178 	.getname = vsock_getname,
2179 	.poll = vsock_poll,
2180 	.ioctl = sock_no_ioctl,
2181 	.listen = vsock_listen,
2182 	.shutdown = vsock_shutdown,
2183 	.setsockopt = vsock_connectible_setsockopt,
2184 	.getsockopt = vsock_connectible_getsockopt,
2185 	.sendmsg = vsock_connectible_sendmsg,
2186 	.recvmsg = vsock_connectible_recvmsg,
2187 	.mmap = sock_no_mmap,
2188 	.sendpage = sock_no_sendpage,
2189 	.set_rcvlowat = vsock_set_rcvlowat,
2190 };
2191 
2192 static const struct proto_ops vsock_seqpacket_ops = {
2193 	.family = PF_VSOCK,
2194 	.owner = THIS_MODULE,
2195 	.release = vsock_release,
2196 	.bind = vsock_bind,
2197 	.connect = vsock_connect,
2198 	.socketpair = sock_no_socketpair,
2199 	.accept = vsock_accept,
2200 	.getname = vsock_getname,
2201 	.poll = vsock_poll,
2202 	.ioctl = sock_no_ioctl,
2203 	.listen = vsock_listen,
2204 	.shutdown = vsock_shutdown,
2205 	.setsockopt = vsock_connectible_setsockopt,
2206 	.getsockopt = vsock_connectible_getsockopt,
2207 	.sendmsg = vsock_connectible_sendmsg,
2208 	.recvmsg = vsock_connectible_recvmsg,
2209 	.mmap = sock_no_mmap,
2210 	.sendpage = sock_no_sendpage,
2211 };
2212 
2213 static int vsock_create(struct net *net, struct socket *sock,
2214 			int protocol, int kern)
2215 {
2216 	struct vsock_sock *vsk;
2217 	struct sock *sk;
2218 	int ret;
2219 
2220 	if (!sock)
2221 		return -EINVAL;
2222 
2223 	if (protocol && protocol != PF_VSOCK)
2224 		return -EPROTONOSUPPORT;
2225 
2226 	switch (sock->type) {
2227 	case SOCK_DGRAM:
2228 		sock->ops = &vsock_dgram_ops;
2229 		break;
2230 	case SOCK_STREAM:
2231 		sock->ops = &vsock_stream_ops;
2232 		break;
2233 	case SOCK_SEQPACKET:
2234 		sock->ops = &vsock_seqpacket_ops;
2235 		break;
2236 	default:
2237 		return -ESOCKTNOSUPPORT;
2238 	}
2239 
2240 	sock->state = SS_UNCONNECTED;
2241 
2242 	sk = __vsock_create(net, sock, NULL, GFP_KERNEL, 0, kern);
2243 	if (!sk)
2244 		return -ENOMEM;
2245 
2246 	vsk = vsock_sk(sk);
2247 
2248 	if (sock->type == SOCK_DGRAM) {
2249 		ret = vsock_assign_transport(vsk, NULL);
2250 		if (ret < 0) {
2251 			sock_put(sk);
2252 			return ret;
2253 		}
2254 	}
2255 
2256 	vsock_insert_unbound(vsk);
2257 
2258 	return 0;
2259 }
2260 
2261 static const struct net_proto_family vsock_family_ops = {
2262 	.family = AF_VSOCK,
2263 	.create = vsock_create,
2264 	.owner = THIS_MODULE,
2265 };
2266 
2267 static long vsock_dev_do_ioctl(struct file *filp,
2268 			       unsigned int cmd, void __user *ptr)
2269 {
2270 	u32 __user *p = ptr;
2271 	u32 cid = VMADDR_CID_ANY;
2272 	int retval = 0;
2273 
2274 	switch (cmd) {
2275 	case IOCTL_VM_SOCKETS_GET_LOCAL_CID:
2276 		/* To be compatible with the VMCI behavior, we prioritize the
2277 		 * guest CID instead of well-know host CID (VMADDR_CID_HOST).
2278 		 */
2279 		if (transport_g2h)
2280 			cid = transport_g2h->get_local_cid();
2281 		else if (transport_h2g)
2282 			cid = transport_h2g->get_local_cid();
2283 
2284 		if (put_user(cid, p) != 0)
2285 			retval = -EFAULT;
2286 		break;
2287 
2288 	default:
2289 		retval = -ENOIOCTLCMD;
2290 	}
2291 
2292 	return retval;
2293 }
2294 
2295 static long vsock_dev_ioctl(struct file *filp,
2296 			    unsigned int cmd, unsigned long arg)
2297 {
2298 	return vsock_dev_do_ioctl(filp, cmd, (void __user *)arg);
2299 }
2300 
2301 #ifdef CONFIG_COMPAT
2302 static long vsock_dev_compat_ioctl(struct file *filp,
2303 				   unsigned int cmd, unsigned long arg)
2304 {
2305 	return vsock_dev_do_ioctl(filp, cmd, compat_ptr(arg));
2306 }
2307 #endif
2308 
2309 static const struct file_operations vsock_device_ops = {
2310 	.owner		= THIS_MODULE,
2311 	.unlocked_ioctl	= vsock_dev_ioctl,
2312 #ifdef CONFIG_COMPAT
2313 	.compat_ioctl	= vsock_dev_compat_ioctl,
2314 #endif
2315 	.open		= nonseekable_open,
2316 };
2317 
2318 static struct miscdevice vsock_device = {
2319 	.name		= "vsock",
2320 	.fops		= &vsock_device_ops,
2321 };
2322 
2323 static int __init vsock_init(void)
2324 {
2325 	int err = 0;
2326 
2327 	vsock_init_tables();
2328 
2329 	vsock_proto.owner = THIS_MODULE;
2330 	vsock_device.minor = MISC_DYNAMIC_MINOR;
2331 	err = misc_register(&vsock_device);
2332 	if (err) {
2333 		pr_err("Failed to register misc device\n");
2334 		goto err_reset_transport;
2335 	}
2336 
2337 	err = proto_register(&vsock_proto, 1);	/* we want our slab */
2338 	if (err) {
2339 		pr_err("Cannot register vsock protocol\n");
2340 		goto err_deregister_misc;
2341 	}
2342 
2343 	err = sock_register(&vsock_family_ops);
2344 	if (err) {
2345 		pr_err("could not register af_vsock (%d) address family: %d\n",
2346 		       AF_VSOCK, err);
2347 		goto err_unregister_proto;
2348 	}
2349 
2350 	return 0;
2351 
2352 err_unregister_proto:
2353 	proto_unregister(&vsock_proto);
2354 err_deregister_misc:
2355 	misc_deregister(&vsock_device);
2356 err_reset_transport:
2357 	return err;
2358 }
2359 
2360 static void __exit vsock_exit(void)
2361 {
2362 	misc_deregister(&vsock_device);
2363 	sock_unregister(AF_VSOCK);
2364 	proto_unregister(&vsock_proto);
2365 }
2366 
2367 const struct vsock_transport *vsock_core_get_transport(struct vsock_sock *vsk)
2368 {
2369 	return vsk->transport;
2370 }
2371 EXPORT_SYMBOL_GPL(vsock_core_get_transport);
2372 
2373 int vsock_core_register(const struct vsock_transport *t, int features)
2374 {
2375 	const struct vsock_transport *t_h2g, *t_g2h, *t_dgram, *t_local;
2376 	int err = mutex_lock_interruptible(&vsock_register_mutex);
2377 
2378 	if (err)
2379 		return err;
2380 
2381 	t_h2g = transport_h2g;
2382 	t_g2h = transport_g2h;
2383 	t_dgram = transport_dgram;
2384 	t_local = transport_local;
2385 
2386 	if (features & VSOCK_TRANSPORT_F_H2G) {
2387 		if (t_h2g) {
2388 			err = -EBUSY;
2389 			goto err_busy;
2390 		}
2391 		t_h2g = t;
2392 	}
2393 
2394 	if (features & VSOCK_TRANSPORT_F_G2H) {
2395 		if (t_g2h) {
2396 			err = -EBUSY;
2397 			goto err_busy;
2398 		}
2399 		t_g2h = t;
2400 	}
2401 
2402 	if (features & VSOCK_TRANSPORT_F_DGRAM) {
2403 		if (t_dgram) {
2404 			err = -EBUSY;
2405 			goto err_busy;
2406 		}
2407 		t_dgram = t;
2408 	}
2409 
2410 	if (features & VSOCK_TRANSPORT_F_LOCAL) {
2411 		if (t_local) {
2412 			err = -EBUSY;
2413 			goto err_busy;
2414 		}
2415 		t_local = t;
2416 	}
2417 
2418 	transport_h2g = t_h2g;
2419 	transport_g2h = t_g2h;
2420 	transport_dgram = t_dgram;
2421 	transport_local = t_local;
2422 
2423 err_busy:
2424 	mutex_unlock(&vsock_register_mutex);
2425 	return err;
2426 }
2427 EXPORT_SYMBOL_GPL(vsock_core_register);
2428 
2429 void vsock_core_unregister(const struct vsock_transport *t)
2430 {
2431 	mutex_lock(&vsock_register_mutex);
2432 
2433 	if (transport_h2g == t)
2434 		transport_h2g = NULL;
2435 
2436 	if (transport_g2h == t)
2437 		transport_g2h = NULL;
2438 
2439 	if (transport_dgram == t)
2440 		transport_dgram = NULL;
2441 
2442 	if (transport_local == t)
2443 		transport_local = NULL;
2444 
2445 	mutex_unlock(&vsock_register_mutex);
2446 }
2447 EXPORT_SYMBOL_GPL(vsock_core_unregister);
2448 
2449 module_init(vsock_init);
2450 module_exit(vsock_exit);
2451 
2452 MODULE_AUTHOR("VMware, Inc.");
2453 MODULE_DESCRIPTION("VMware Virtual Socket Family");
2454 MODULE_VERSION("1.0.2.0-k");
2455 MODULE_LICENSE("GPL v2");
2456