xref: /openbmc/linux/net/vmw_vsock/af_vsock.c (revision d4295e12)
1 /*
2  * VMware vSockets Driver
3  *
4  * Copyright (C) 2007-2013 VMware, Inc. All rights reserved.
5  *
6  * This program is free software; you can redistribute it and/or modify it
7  * under the terms of the GNU General Public License as published by the Free
8  * Software Foundation version 2 and no later version.
9  *
10  * This program is distributed in the hope that it will be useful, but WITHOUT
11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13  * more details.
14  */
15 
16 /* Implementation notes:
17  *
18  * - There are two kinds of sockets: those created by user action (such as
19  * calling socket(2)) and those created by incoming connection request packets.
20  *
21  * - There are two "global" tables, one for bound sockets (sockets that have
22  * specified an address that they are responsible for) and one for connected
23  * sockets (sockets that have established a connection with another socket).
24  * These tables are "global" in that all sockets on the system are placed
25  * within them. - Note, though, that the bound table contains an extra entry
26  * for a list of unbound sockets and SOCK_DGRAM sockets will always remain in
27  * that list. The bound table is used solely for lookup of sockets when packets
28  * are received and that's not necessary for SOCK_DGRAM sockets since we create
29  * a datagram handle for each and need not perform a lookup.  Keeping SOCK_DGRAM
30  * sockets out of the bound hash buckets will reduce the chance of collisions
31  * when looking for SOCK_STREAM sockets and prevents us from having to check the
32  * socket type in the hash table lookups.
33  *
34  * - Sockets created by user action will either be "client" sockets that
35  * initiate a connection or "server" sockets that listen for connections; we do
36  * not support simultaneous connects (two "client" sockets connecting).
37  *
38  * - "Server" sockets are referred to as listener sockets throughout this
39  * implementation because they are in the TCP_LISTEN state.  When a
40  * connection request is received (the second kind of socket mentioned above),
41  * we create a new socket and refer to it as a pending socket.  These pending
42  * sockets are placed on the pending connection list of the listener socket.
43  * When future packets are received for the address the listener socket is
44  * bound to, we check if the source of the packet is from one that has an
45  * existing pending connection.  If it does, we process the packet for the
46  * pending socket.  When that socket reaches the connected state, it is removed
47  * from the listener socket's pending list and enqueued in the listener
48  * socket's accept queue.  Callers of accept(2) will accept connected sockets
49  * from the listener socket's accept queue.  If the socket cannot be accepted
50  * for some reason then it is marked rejected.  Once the connection is
51  * accepted, it is owned by the user process and the responsibility for cleanup
52  * falls with that user process.
53  *
54  * - It is possible that these pending sockets will never reach the connected
55  * state; in fact, we may never receive another packet after the connection
56  * request.  Because of this, we must schedule a cleanup function to run in the
57  * future, after some amount of time passes where a connection should have been
58  * established.  This function ensures that the socket is off all lists so it
59  * cannot be retrieved, then drops all references to the socket so it is cleaned
60  * up (sock_put() -> sk_free() -> our sk_destruct implementation).  Note this
61  * function will also cleanup rejected sockets, those that reach the connected
62  * state but leave it before they have been accepted.
63  *
64  * - Lock ordering for pending or accept queue sockets is:
65  *
66  *     lock_sock(listener);
67  *     lock_sock_nested(pending, SINGLE_DEPTH_NESTING);
68  *
69  * Using explicit nested locking keeps lockdep happy since normally only one
70  * lock of a given class may be taken at a time.
71  *
72  * - Sockets created by user action will be cleaned up when the user process
73  * calls close(2), causing our release implementation to be called. Our release
74  * implementation will perform some cleanup then drop the last reference so our
75  * sk_destruct implementation is invoked.  Our sk_destruct implementation will
76  * perform additional cleanup that's common for both types of sockets.
77  *
78  * - A socket's reference count is what ensures that the structure won't be
79  * freed.  Each entry in a list (such as the "global" bound and connected tables
80  * and the listener socket's pending list and connected queue) ensures a
81  * reference.  When we defer work until process context and pass a socket as our
82  * argument, we must ensure the reference count is increased to ensure the
83  * socket isn't freed before the function is run; the deferred function will
84  * then drop the reference.
85  *
86  * - sk->sk_state uses the TCP state constants because they are widely used by
87  * other address families and exposed to userspace tools like ss(8):
88  *
89  *   TCP_CLOSE - unconnected
90  *   TCP_SYN_SENT - connecting
91  *   TCP_ESTABLISHED - connected
92  *   TCP_CLOSING - disconnecting
93  *   TCP_LISTEN - listening
94  */
95 
96 #include <linux/types.h>
97 #include <linux/bitops.h>
98 #include <linux/cred.h>
99 #include <linux/init.h>
100 #include <linux/io.h>
101 #include <linux/kernel.h>
102 #include <linux/sched/signal.h>
103 #include <linux/kmod.h>
104 #include <linux/list.h>
105 #include <linux/miscdevice.h>
106 #include <linux/module.h>
107 #include <linux/mutex.h>
108 #include <linux/net.h>
109 #include <linux/poll.h>
110 #include <linux/skbuff.h>
111 #include <linux/smp.h>
112 #include <linux/socket.h>
113 #include <linux/stddef.h>
114 #include <linux/unistd.h>
115 #include <linux/wait.h>
116 #include <linux/workqueue.h>
117 #include <net/sock.h>
118 #include <net/af_vsock.h>
119 
120 static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr);
121 static void vsock_sk_destruct(struct sock *sk);
122 static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
123 
124 /* Protocol family. */
125 static struct proto vsock_proto = {
126 	.name = "AF_VSOCK",
127 	.owner = THIS_MODULE,
128 	.obj_size = sizeof(struct vsock_sock),
129 };
130 
131 /* The default peer timeout indicates how long we will wait for a peer response
132  * to a control message.
133  */
134 #define VSOCK_DEFAULT_CONNECT_TIMEOUT (2 * HZ)
135 
136 static const struct vsock_transport *transport;
137 static DEFINE_MUTEX(vsock_register_mutex);
138 
139 /**** EXPORTS ****/
140 
141 /* Get the ID of the local context.  This is transport dependent. */
142 
143 int vm_sockets_get_local_cid(void)
144 {
145 	return transport->get_local_cid();
146 }
147 EXPORT_SYMBOL_GPL(vm_sockets_get_local_cid);
148 
149 /**** UTILS ****/
150 
151 /* Each bound VSocket is stored in the bind hash table and each connected
152  * VSocket is stored in the connected hash table.
153  *
154  * Unbound sockets are all put on the same list attached to the end of the hash
155  * table (vsock_unbound_sockets).  Bound sockets are added to the hash table in
156  * the bucket that their local address hashes to (vsock_bound_sockets(addr)
157  * represents the list that addr hashes to).
158  *
159  * Specifically, we initialize the vsock_bind_table array to a size of
160  * VSOCK_HASH_SIZE + 1 so that vsock_bind_table[0] through
161  * vsock_bind_table[VSOCK_HASH_SIZE - 1] are for bound sockets and
162  * vsock_bind_table[VSOCK_HASH_SIZE] is for unbound sockets.  The hash function
163  * mods with VSOCK_HASH_SIZE to ensure this.
164  */
165 #define MAX_PORT_RETRIES        24
166 
167 #define VSOCK_HASH(addr)        ((addr)->svm_port % VSOCK_HASH_SIZE)
168 #define vsock_bound_sockets(addr) (&vsock_bind_table[VSOCK_HASH(addr)])
169 #define vsock_unbound_sockets     (&vsock_bind_table[VSOCK_HASH_SIZE])
170 
171 /* XXX This can probably be implemented in a better way. */
172 #define VSOCK_CONN_HASH(src, dst)				\
173 	(((src)->svm_cid ^ (dst)->svm_port) % VSOCK_HASH_SIZE)
174 #define vsock_connected_sockets(src, dst)		\
175 	(&vsock_connected_table[VSOCK_CONN_HASH(src, dst)])
176 #define vsock_connected_sockets_vsk(vsk)				\
177 	vsock_connected_sockets(&(vsk)->remote_addr, &(vsk)->local_addr)
178 
179 struct list_head vsock_bind_table[VSOCK_HASH_SIZE + 1];
180 EXPORT_SYMBOL_GPL(vsock_bind_table);
181 struct list_head vsock_connected_table[VSOCK_HASH_SIZE];
182 EXPORT_SYMBOL_GPL(vsock_connected_table);
183 DEFINE_SPINLOCK(vsock_table_lock);
184 EXPORT_SYMBOL_GPL(vsock_table_lock);
185 
186 /* Autobind this socket to the local address if necessary. */
187 static int vsock_auto_bind(struct vsock_sock *vsk)
188 {
189 	struct sock *sk = sk_vsock(vsk);
190 	struct sockaddr_vm local_addr;
191 
192 	if (vsock_addr_bound(&vsk->local_addr))
193 		return 0;
194 	vsock_addr_init(&local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
195 	return __vsock_bind(sk, &local_addr);
196 }
197 
198 static int __init vsock_init_tables(void)
199 {
200 	int i;
201 
202 	for (i = 0; i < ARRAY_SIZE(vsock_bind_table); i++)
203 		INIT_LIST_HEAD(&vsock_bind_table[i]);
204 
205 	for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++)
206 		INIT_LIST_HEAD(&vsock_connected_table[i]);
207 	return 0;
208 }
209 
210 static void __vsock_insert_bound(struct list_head *list,
211 				 struct vsock_sock *vsk)
212 {
213 	sock_hold(&vsk->sk);
214 	list_add(&vsk->bound_table, list);
215 }
216 
217 static void __vsock_insert_connected(struct list_head *list,
218 				     struct vsock_sock *vsk)
219 {
220 	sock_hold(&vsk->sk);
221 	list_add(&vsk->connected_table, list);
222 }
223 
224 static void __vsock_remove_bound(struct vsock_sock *vsk)
225 {
226 	list_del_init(&vsk->bound_table);
227 	sock_put(&vsk->sk);
228 }
229 
230 static void __vsock_remove_connected(struct vsock_sock *vsk)
231 {
232 	list_del_init(&vsk->connected_table);
233 	sock_put(&vsk->sk);
234 }
235 
236 static struct sock *__vsock_find_bound_socket(struct sockaddr_vm *addr)
237 {
238 	struct vsock_sock *vsk;
239 
240 	list_for_each_entry(vsk, vsock_bound_sockets(addr), bound_table)
241 		if (addr->svm_port == vsk->local_addr.svm_port)
242 			return sk_vsock(vsk);
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 	__vsock_remove_bound(vsk);
285 	spin_unlock_bh(&vsock_table_lock);
286 }
287 EXPORT_SYMBOL_GPL(vsock_remove_bound);
288 
289 void vsock_remove_connected(struct vsock_sock *vsk)
290 {
291 	spin_lock_bh(&vsock_table_lock);
292 	__vsock_remove_connected(vsk);
293 	spin_unlock_bh(&vsock_table_lock);
294 }
295 EXPORT_SYMBOL_GPL(vsock_remove_connected);
296 
297 struct sock *vsock_find_bound_socket(struct sockaddr_vm *addr)
298 {
299 	struct sock *sk;
300 
301 	spin_lock_bh(&vsock_table_lock);
302 	sk = __vsock_find_bound_socket(addr);
303 	if (sk)
304 		sock_hold(sk);
305 
306 	spin_unlock_bh(&vsock_table_lock);
307 
308 	return sk;
309 }
310 EXPORT_SYMBOL_GPL(vsock_find_bound_socket);
311 
312 struct sock *vsock_find_connected_socket(struct sockaddr_vm *src,
313 					 struct sockaddr_vm *dst)
314 {
315 	struct sock *sk;
316 
317 	spin_lock_bh(&vsock_table_lock);
318 	sk = __vsock_find_connected_socket(src, dst);
319 	if (sk)
320 		sock_hold(sk);
321 
322 	spin_unlock_bh(&vsock_table_lock);
323 
324 	return sk;
325 }
326 EXPORT_SYMBOL_GPL(vsock_find_connected_socket);
327 
328 static bool vsock_in_bound_table(struct vsock_sock *vsk)
329 {
330 	bool ret;
331 
332 	spin_lock_bh(&vsock_table_lock);
333 	ret = __vsock_in_bound_table(vsk);
334 	spin_unlock_bh(&vsock_table_lock);
335 
336 	return ret;
337 }
338 
339 static bool vsock_in_connected_table(struct vsock_sock *vsk)
340 {
341 	bool ret;
342 
343 	spin_lock_bh(&vsock_table_lock);
344 	ret = __vsock_in_connected_table(vsk);
345 	spin_unlock_bh(&vsock_table_lock);
346 
347 	return ret;
348 }
349 
350 void vsock_remove_sock(struct vsock_sock *vsk)
351 {
352 	if (vsock_in_bound_table(vsk))
353 		vsock_remove_bound(vsk);
354 
355 	if (vsock_in_connected_table(vsk))
356 		vsock_remove_connected(vsk);
357 }
358 EXPORT_SYMBOL_GPL(vsock_remove_sock);
359 
360 void vsock_for_each_connected_socket(void (*fn)(struct sock *sk))
361 {
362 	int i;
363 
364 	spin_lock_bh(&vsock_table_lock);
365 
366 	for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) {
367 		struct vsock_sock *vsk;
368 		list_for_each_entry(vsk, &vsock_connected_table[i],
369 				    connected_table)
370 			fn(sk_vsock(vsk));
371 	}
372 
373 	spin_unlock_bh(&vsock_table_lock);
374 }
375 EXPORT_SYMBOL_GPL(vsock_for_each_connected_socket);
376 
377 void vsock_add_pending(struct sock *listener, struct sock *pending)
378 {
379 	struct vsock_sock *vlistener;
380 	struct vsock_sock *vpending;
381 
382 	vlistener = vsock_sk(listener);
383 	vpending = vsock_sk(pending);
384 
385 	sock_hold(pending);
386 	sock_hold(listener);
387 	list_add_tail(&vpending->pending_links, &vlistener->pending_links);
388 }
389 EXPORT_SYMBOL_GPL(vsock_add_pending);
390 
391 void vsock_remove_pending(struct sock *listener, struct sock *pending)
392 {
393 	struct vsock_sock *vpending = vsock_sk(pending);
394 
395 	list_del_init(&vpending->pending_links);
396 	sock_put(listener);
397 	sock_put(pending);
398 }
399 EXPORT_SYMBOL_GPL(vsock_remove_pending);
400 
401 void vsock_enqueue_accept(struct sock *listener, struct sock *connected)
402 {
403 	struct vsock_sock *vlistener;
404 	struct vsock_sock *vconnected;
405 
406 	vlistener = vsock_sk(listener);
407 	vconnected = vsock_sk(connected);
408 
409 	sock_hold(connected);
410 	sock_hold(listener);
411 	list_add_tail(&vconnected->accept_queue, &vlistener->accept_queue);
412 }
413 EXPORT_SYMBOL_GPL(vsock_enqueue_accept);
414 
415 static struct sock *vsock_dequeue_accept(struct sock *listener)
416 {
417 	struct vsock_sock *vlistener;
418 	struct vsock_sock *vconnected;
419 
420 	vlistener = vsock_sk(listener);
421 
422 	if (list_empty(&vlistener->accept_queue))
423 		return NULL;
424 
425 	vconnected = list_entry(vlistener->accept_queue.next,
426 				struct vsock_sock, accept_queue);
427 
428 	list_del_init(&vconnected->accept_queue);
429 	sock_put(listener);
430 	/* The caller will need a reference on the connected socket so we let
431 	 * it call sock_put().
432 	 */
433 
434 	return sk_vsock(vconnected);
435 }
436 
437 static bool vsock_is_accept_queue_empty(struct sock *sk)
438 {
439 	struct vsock_sock *vsk = vsock_sk(sk);
440 	return list_empty(&vsk->accept_queue);
441 }
442 
443 static bool vsock_is_pending(struct sock *sk)
444 {
445 	struct vsock_sock *vsk = vsock_sk(sk);
446 	return !list_empty(&vsk->pending_links);
447 }
448 
449 static int vsock_send_shutdown(struct sock *sk, int mode)
450 {
451 	return transport->shutdown(vsock_sk(sk), mode);
452 }
453 
454 static void vsock_pending_work(struct work_struct *work)
455 {
456 	struct sock *sk;
457 	struct sock *listener;
458 	struct vsock_sock *vsk;
459 	bool cleanup;
460 
461 	vsk = container_of(work, struct vsock_sock, pending_work.work);
462 	sk = sk_vsock(vsk);
463 	listener = vsk->listener;
464 	cleanup = true;
465 
466 	lock_sock(listener);
467 	lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
468 
469 	if (vsock_is_pending(sk)) {
470 		vsock_remove_pending(listener, sk);
471 
472 		listener->sk_ack_backlog--;
473 	} else if (!vsk->rejected) {
474 		/* We are not on the pending list and accept() did not reject
475 		 * us, so we must have been accepted by our user process.  We
476 		 * just need to drop our references to the sockets and be on
477 		 * our way.
478 		 */
479 		cleanup = false;
480 		goto out;
481 	}
482 
483 	/* We need to remove ourself from the global connected sockets list so
484 	 * incoming packets can't find this socket, and to reduce the reference
485 	 * count.
486 	 */
487 	if (vsock_in_connected_table(vsk))
488 		vsock_remove_connected(vsk);
489 
490 	sk->sk_state = TCP_CLOSE;
491 
492 out:
493 	release_sock(sk);
494 	release_sock(listener);
495 	if (cleanup)
496 		sock_put(sk);
497 
498 	sock_put(sk);
499 	sock_put(listener);
500 }
501 
502 /**** SOCKET OPERATIONS ****/
503 
504 static int __vsock_bind_stream(struct vsock_sock *vsk,
505 			       struct sockaddr_vm *addr)
506 {
507 	static u32 port = LAST_RESERVED_PORT + 1;
508 	struct sockaddr_vm new_addr;
509 
510 	vsock_addr_init(&new_addr, addr->svm_cid, addr->svm_port);
511 
512 	if (addr->svm_port == VMADDR_PORT_ANY) {
513 		bool found = false;
514 		unsigned int i;
515 
516 		for (i = 0; i < MAX_PORT_RETRIES; i++) {
517 			if (port <= LAST_RESERVED_PORT)
518 				port = LAST_RESERVED_PORT + 1;
519 
520 			new_addr.svm_port = port++;
521 
522 			if (!__vsock_find_bound_socket(&new_addr)) {
523 				found = true;
524 				break;
525 			}
526 		}
527 
528 		if (!found)
529 			return -EADDRNOTAVAIL;
530 	} else {
531 		/* If port is in reserved range, ensure caller
532 		 * has necessary privileges.
533 		 */
534 		if (addr->svm_port <= LAST_RESERVED_PORT &&
535 		    !capable(CAP_NET_BIND_SERVICE)) {
536 			return -EACCES;
537 		}
538 
539 		if (__vsock_find_bound_socket(&new_addr))
540 			return -EADDRINUSE;
541 	}
542 
543 	vsock_addr_init(&vsk->local_addr, new_addr.svm_cid, new_addr.svm_port);
544 
545 	/* Remove stream sockets from the unbound list and add them to the hash
546 	 * table for easy lookup by its address.  The unbound list is simply an
547 	 * extra entry at the end of the hash table, a trick used by AF_UNIX.
548 	 */
549 	__vsock_remove_bound(vsk);
550 	__vsock_insert_bound(vsock_bound_sockets(&vsk->local_addr), vsk);
551 
552 	return 0;
553 }
554 
555 static int __vsock_bind_dgram(struct vsock_sock *vsk,
556 			      struct sockaddr_vm *addr)
557 {
558 	return transport->dgram_bind(vsk, addr);
559 }
560 
561 static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr)
562 {
563 	struct vsock_sock *vsk = vsock_sk(sk);
564 	u32 cid;
565 	int retval;
566 
567 	/* First ensure this socket isn't already bound. */
568 	if (vsock_addr_bound(&vsk->local_addr))
569 		return -EINVAL;
570 
571 	/* Now bind to the provided address or select appropriate values if
572 	 * none are provided (VMADDR_CID_ANY and VMADDR_PORT_ANY).  Note that
573 	 * like AF_INET prevents binding to a non-local IP address (in most
574 	 * cases), we only allow binding to the local CID.
575 	 */
576 	cid = transport->get_local_cid();
577 	if (addr->svm_cid != cid && addr->svm_cid != VMADDR_CID_ANY)
578 		return -EADDRNOTAVAIL;
579 
580 	switch (sk->sk_socket->type) {
581 	case SOCK_STREAM:
582 		spin_lock_bh(&vsock_table_lock);
583 		retval = __vsock_bind_stream(vsk, addr);
584 		spin_unlock_bh(&vsock_table_lock);
585 		break;
586 
587 	case SOCK_DGRAM:
588 		retval = __vsock_bind_dgram(vsk, addr);
589 		break;
590 
591 	default:
592 		retval = -EINVAL;
593 		break;
594 	}
595 
596 	return retval;
597 }
598 
599 static void vsock_connect_timeout(struct work_struct *work);
600 
601 struct sock *__vsock_create(struct net *net,
602 			    struct socket *sock,
603 			    struct sock *parent,
604 			    gfp_t priority,
605 			    unsigned short type,
606 			    int kern)
607 {
608 	struct sock *sk;
609 	struct vsock_sock *psk;
610 	struct vsock_sock *vsk;
611 
612 	sk = sk_alloc(net, AF_VSOCK, priority, &vsock_proto, kern);
613 	if (!sk)
614 		return NULL;
615 
616 	sock_init_data(sock, sk);
617 
618 	/* sk->sk_type is normally set in sock_init_data, but only if sock is
619 	 * non-NULL. We make sure that our sockets always have a type by
620 	 * setting it here if needed.
621 	 */
622 	if (!sock)
623 		sk->sk_type = type;
624 
625 	vsk = vsock_sk(sk);
626 	vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
627 	vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
628 
629 	sk->sk_destruct = vsock_sk_destruct;
630 	sk->sk_backlog_rcv = vsock_queue_rcv_skb;
631 	sock_reset_flag(sk, SOCK_DONE);
632 
633 	INIT_LIST_HEAD(&vsk->bound_table);
634 	INIT_LIST_HEAD(&vsk->connected_table);
635 	vsk->listener = NULL;
636 	INIT_LIST_HEAD(&vsk->pending_links);
637 	INIT_LIST_HEAD(&vsk->accept_queue);
638 	vsk->rejected = false;
639 	vsk->sent_request = false;
640 	vsk->ignore_connecting_rst = false;
641 	vsk->peer_shutdown = 0;
642 	INIT_DELAYED_WORK(&vsk->connect_work, vsock_connect_timeout);
643 	INIT_DELAYED_WORK(&vsk->pending_work, vsock_pending_work);
644 
645 	psk = parent ? vsock_sk(parent) : NULL;
646 	if (parent) {
647 		vsk->trusted = psk->trusted;
648 		vsk->owner = get_cred(psk->owner);
649 		vsk->connect_timeout = psk->connect_timeout;
650 	} else {
651 		vsk->trusted = capable(CAP_NET_ADMIN);
652 		vsk->owner = get_current_cred();
653 		vsk->connect_timeout = VSOCK_DEFAULT_CONNECT_TIMEOUT;
654 	}
655 
656 	if (transport->init(vsk, psk) < 0) {
657 		sk_free(sk);
658 		return NULL;
659 	}
660 
661 	if (sock)
662 		vsock_insert_unbound(vsk);
663 
664 	return sk;
665 }
666 EXPORT_SYMBOL_GPL(__vsock_create);
667 
668 static void __vsock_release(struct sock *sk)
669 {
670 	if (sk) {
671 		struct sk_buff *skb;
672 		struct sock *pending;
673 		struct vsock_sock *vsk;
674 
675 		vsk = vsock_sk(sk);
676 		pending = NULL;	/* Compiler warning. */
677 
678 		transport->release(vsk);
679 
680 		lock_sock(sk);
681 		sock_orphan(sk);
682 		sk->sk_shutdown = SHUTDOWN_MASK;
683 
684 		while ((skb = skb_dequeue(&sk->sk_receive_queue)))
685 			kfree_skb(skb);
686 
687 		/* Clean up any sockets that never were accepted. */
688 		while ((pending = vsock_dequeue_accept(sk)) != NULL) {
689 			__vsock_release(pending);
690 			sock_put(pending);
691 		}
692 
693 		release_sock(sk);
694 		sock_put(sk);
695 	}
696 }
697 
698 static void vsock_sk_destruct(struct sock *sk)
699 {
700 	struct vsock_sock *vsk = vsock_sk(sk);
701 
702 	transport->destruct(vsk);
703 
704 	/* When clearing these addresses, there's no need to set the family and
705 	 * possibly register the address family with the kernel.
706 	 */
707 	vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
708 	vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
709 
710 	put_cred(vsk->owner);
711 }
712 
713 static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
714 {
715 	int err;
716 
717 	err = sock_queue_rcv_skb(sk, skb);
718 	if (err)
719 		kfree_skb(skb);
720 
721 	return err;
722 }
723 
724 s64 vsock_stream_has_data(struct vsock_sock *vsk)
725 {
726 	return transport->stream_has_data(vsk);
727 }
728 EXPORT_SYMBOL_GPL(vsock_stream_has_data);
729 
730 s64 vsock_stream_has_space(struct vsock_sock *vsk)
731 {
732 	return transport->stream_has_space(vsk);
733 }
734 EXPORT_SYMBOL_GPL(vsock_stream_has_space);
735 
736 static int vsock_release(struct socket *sock)
737 {
738 	__vsock_release(sock->sk);
739 	sock->sk = NULL;
740 	sock->state = SS_FREE;
741 
742 	return 0;
743 }
744 
745 static int
746 vsock_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
747 {
748 	int err;
749 	struct sock *sk;
750 	struct sockaddr_vm *vm_addr;
751 
752 	sk = sock->sk;
753 
754 	if (vsock_addr_cast(addr, addr_len, &vm_addr) != 0)
755 		return -EINVAL;
756 
757 	lock_sock(sk);
758 	err = __vsock_bind(sk, vm_addr);
759 	release_sock(sk);
760 
761 	return err;
762 }
763 
764 static int vsock_getname(struct socket *sock,
765 			 struct sockaddr *addr, int peer)
766 {
767 	int err;
768 	struct sock *sk;
769 	struct vsock_sock *vsk;
770 	struct sockaddr_vm *vm_addr;
771 
772 	sk = sock->sk;
773 	vsk = vsock_sk(sk);
774 	err = 0;
775 
776 	lock_sock(sk);
777 
778 	if (peer) {
779 		if (sock->state != SS_CONNECTED) {
780 			err = -ENOTCONN;
781 			goto out;
782 		}
783 		vm_addr = &vsk->remote_addr;
784 	} else {
785 		vm_addr = &vsk->local_addr;
786 	}
787 
788 	if (!vm_addr) {
789 		err = -EINVAL;
790 		goto out;
791 	}
792 
793 	/* sys_getsockname() and sys_getpeername() pass us a
794 	 * MAX_SOCK_ADDR-sized buffer and don't set addr_len.  Unfortunately
795 	 * that macro is defined in socket.c instead of .h, so we hardcode its
796 	 * value here.
797 	 */
798 	BUILD_BUG_ON(sizeof(*vm_addr) > 128);
799 	memcpy(addr, vm_addr, sizeof(*vm_addr));
800 	err = sizeof(*vm_addr);
801 
802 out:
803 	release_sock(sk);
804 	return err;
805 }
806 
807 static int vsock_shutdown(struct socket *sock, int mode)
808 {
809 	int err;
810 	struct sock *sk;
811 
812 	/* User level uses SHUT_RD (0) and SHUT_WR (1), but the kernel uses
813 	 * RCV_SHUTDOWN (1) and SEND_SHUTDOWN (2), so we must increment mode
814 	 * here like the other address families do.  Note also that the
815 	 * increment makes SHUT_RDWR (2) into RCV_SHUTDOWN | SEND_SHUTDOWN (3),
816 	 * which is what we want.
817 	 */
818 	mode++;
819 
820 	if ((mode & ~SHUTDOWN_MASK) || !mode)
821 		return -EINVAL;
822 
823 	/* If this is a STREAM socket and it is not connected then bail out
824 	 * immediately.  If it is a DGRAM socket then we must first kick the
825 	 * socket so that it wakes up from any sleeping calls, for example
826 	 * recv(), and then afterwards return the error.
827 	 */
828 
829 	sk = sock->sk;
830 	if (sock->state == SS_UNCONNECTED) {
831 		err = -ENOTCONN;
832 		if (sk->sk_type == SOCK_STREAM)
833 			return err;
834 	} else {
835 		sock->state = SS_DISCONNECTING;
836 		err = 0;
837 	}
838 
839 	/* Receive and send shutdowns are treated alike. */
840 	mode = mode & (RCV_SHUTDOWN | SEND_SHUTDOWN);
841 	if (mode) {
842 		lock_sock(sk);
843 		sk->sk_shutdown |= mode;
844 		sk->sk_state_change(sk);
845 		release_sock(sk);
846 
847 		if (sk->sk_type == SOCK_STREAM) {
848 			sock_reset_flag(sk, SOCK_DONE);
849 			vsock_send_shutdown(sk, mode);
850 		}
851 	}
852 
853 	return err;
854 }
855 
856 static __poll_t vsock_poll(struct file *file, struct socket *sock,
857 			       poll_table *wait)
858 {
859 	struct sock *sk;
860 	__poll_t mask;
861 	struct vsock_sock *vsk;
862 
863 	sk = sock->sk;
864 	vsk = vsock_sk(sk);
865 
866 	poll_wait(file, sk_sleep(sk), wait);
867 	mask = 0;
868 
869 	if (sk->sk_err)
870 		/* Signify that there has been an error on this socket. */
871 		mask |= EPOLLERR;
872 
873 	/* INET sockets treat local write shutdown and peer write shutdown as a
874 	 * case of EPOLLHUP set.
875 	 */
876 	if ((sk->sk_shutdown == SHUTDOWN_MASK) ||
877 	    ((sk->sk_shutdown & SEND_SHUTDOWN) &&
878 	     (vsk->peer_shutdown & SEND_SHUTDOWN))) {
879 		mask |= EPOLLHUP;
880 	}
881 
882 	if (sk->sk_shutdown & RCV_SHUTDOWN ||
883 	    vsk->peer_shutdown & SEND_SHUTDOWN) {
884 		mask |= EPOLLRDHUP;
885 	}
886 
887 	if (sock->type == SOCK_DGRAM) {
888 		/* For datagram sockets we can read if there is something in
889 		 * the queue and write as long as the socket isn't shutdown for
890 		 * sending.
891 		 */
892 		if (!skb_queue_empty(&sk->sk_receive_queue) ||
893 		    (sk->sk_shutdown & RCV_SHUTDOWN)) {
894 			mask |= EPOLLIN | EPOLLRDNORM;
895 		}
896 
897 		if (!(sk->sk_shutdown & SEND_SHUTDOWN))
898 			mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND;
899 
900 	} else if (sock->type == SOCK_STREAM) {
901 		lock_sock(sk);
902 
903 		/* Listening sockets that have connections in their accept
904 		 * queue can be read.
905 		 */
906 		if (sk->sk_state == TCP_LISTEN
907 		    && !vsock_is_accept_queue_empty(sk))
908 			mask |= EPOLLIN | EPOLLRDNORM;
909 
910 		/* If there is something in the queue then we can read. */
911 		if (transport->stream_is_active(vsk) &&
912 		    !(sk->sk_shutdown & RCV_SHUTDOWN)) {
913 			bool data_ready_now = false;
914 			int ret = transport->notify_poll_in(
915 					vsk, 1, &data_ready_now);
916 			if (ret < 0) {
917 				mask |= EPOLLERR;
918 			} else {
919 				if (data_ready_now)
920 					mask |= EPOLLIN | EPOLLRDNORM;
921 
922 			}
923 		}
924 
925 		/* Sockets whose connections have been closed, reset, or
926 		 * terminated should also be considered read, and we check the
927 		 * shutdown flag for that.
928 		 */
929 		if (sk->sk_shutdown & RCV_SHUTDOWN ||
930 		    vsk->peer_shutdown & SEND_SHUTDOWN) {
931 			mask |= EPOLLIN | EPOLLRDNORM;
932 		}
933 
934 		/* Connected sockets that can produce data can be written. */
935 		if (sk->sk_state == TCP_ESTABLISHED) {
936 			if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
937 				bool space_avail_now = false;
938 				int ret = transport->notify_poll_out(
939 						vsk, 1, &space_avail_now);
940 				if (ret < 0) {
941 					mask |= EPOLLERR;
942 				} else {
943 					if (space_avail_now)
944 						/* Remove EPOLLWRBAND since INET
945 						 * sockets are not setting it.
946 						 */
947 						mask |= EPOLLOUT | EPOLLWRNORM;
948 
949 				}
950 			}
951 		}
952 
953 		/* Simulate INET socket poll behaviors, which sets
954 		 * EPOLLOUT|EPOLLWRNORM when peer is closed and nothing to read,
955 		 * but local send is not shutdown.
956 		 */
957 		if (sk->sk_state == TCP_CLOSE || sk->sk_state == TCP_CLOSING) {
958 			if (!(sk->sk_shutdown & SEND_SHUTDOWN))
959 				mask |= EPOLLOUT | EPOLLWRNORM;
960 
961 		}
962 
963 		release_sock(sk);
964 	}
965 
966 	return mask;
967 }
968 
969 static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg,
970 			       size_t len)
971 {
972 	int err;
973 	struct sock *sk;
974 	struct vsock_sock *vsk;
975 	struct sockaddr_vm *remote_addr;
976 
977 	if (msg->msg_flags & MSG_OOB)
978 		return -EOPNOTSUPP;
979 
980 	/* For now, MSG_DONTWAIT is always assumed... */
981 	err = 0;
982 	sk = sock->sk;
983 	vsk = vsock_sk(sk);
984 
985 	lock_sock(sk);
986 
987 	err = vsock_auto_bind(vsk);
988 	if (err)
989 		goto out;
990 
991 
992 	/* If the provided message contains an address, use that.  Otherwise
993 	 * fall back on the socket's remote handle (if it has been connected).
994 	 */
995 	if (msg->msg_name &&
996 	    vsock_addr_cast(msg->msg_name, msg->msg_namelen,
997 			    &remote_addr) == 0) {
998 		/* Ensure this address is of the right type and is a valid
999 		 * destination.
1000 		 */
1001 
1002 		if (remote_addr->svm_cid == VMADDR_CID_ANY)
1003 			remote_addr->svm_cid = transport->get_local_cid();
1004 
1005 		if (!vsock_addr_bound(remote_addr)) {
1006 			err = -EINVAL;
1007 			goto out;
1008 		}
1009 	} else if (sock->state == SS_CONNECTED) {
1010 		remote_addr = &vsk->remote_addr;
1011 
1012 		if (remote_addr->svm_cid == VMADDR_CID_ANY)
1013 			remote_addr->svm_cid = transport->get_local_cid();
1014 
1015 		/* XXX Should connect() or this function ensure remote_addr is
1016 		 * bound?
1017 		 */
1018 		if (!vsock_addr_bound(&vsk->remote_addr)) {
1019 			err = -EINVAL;
1020 			goto out;
1021 		}
1022 	} else {
1023 		err = -EINVAL;
1024 		goto out;
1025 	}
1026 
1027 	if (!transport->dgram_allow(remote_addr->svm_cid,
1028 				    remote_addr->svm_port)) {
1029 		err = -EINVAL;
1030 		goto out;
1031 	}
1032 
1033 	err = transport->dgram_enqueue(vsk, remote_addr, msg, len);
1034 
1035 out:
1036 	release_sock(sk);
1037 	return err;
1038 }
1039 
1040 static int vsock_dgram_connect(struct socket *sock,
1041 			       struct sockaddr *addr, int addr_len, int flags)
1042 {
1043 	int err;
1044 	struct sock *sk;
1045 	struct vsock_sock *vsk;
1046 	struct sockaddr_vm *remote_addr;
1047 
1048 	sk = sock->sk;
1049 	vsk = vsock_sk(sk);
1050 
1051 	err = vsock_addr_cast(addr, addr_len, &remote_addr);
1052 	if (err == -EAFNOSUPPORT && remote_addr->svm_family == AF_UNSPEC) {
1053 		lock_sock(sk);
1054 		vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY,
1055 				VMADDR_PORT_ANY);
1056 		sock->state = SS_UNCONNECTED;
1057 		release_sock(sk);
1058 		return 0;
1059 	} else if (err != 0)
1060 		return -EINVAL;
1061 
1062 	lock_sock(sk);
1063 
1064 	err = vsock_auto_bind(vsk);
1065 	if (err)
1066 		goto out;
1067 
1068 	if (!transport->dgram_allow(remote_addr->svm_cid,
1069 				    remote_addr->svm_port)) {
1070 		err = -EINVAL;
1071 		goto out;
1072 	}
1073 
1074 	memcpy(&vsk->remote_addr, remote_addr, sizeof(vsk->remote_addr));
1075 	sock->state = SS_CONNECTED;
1076 
1077 out:
1078 	release_sock(sk);
1079 	return err;
1080 }
1081 
1082 static int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg,
1083 			       size_t len, int flags)
1084 {
1085 	return transport->dgram_dequeue(vsock_sk(sock->sk), msg, len, flags);
1086 }
1087 
1088 static const struct proto_ops vsock_dgram_ops = {
1089 	.family = PF_VSOCK,
1090 	.owner = THIS_MODULE,
1091 	.release = vsock_release,
1092 	.bind = vsock_bind,
1093 	.connect = vsock_dgram_connect,
1094 	.socketpair = sock_no_socketpair,
1095 	.accept = sock_no_accept,
1096 	.getname = vsock_getname,
1097 	.poll = vsock_poll,
1098 	.ioctl = sock_no_ioctl,
1099 	.listen = sock_no_listen,
1100 	.shutdown = vsock_shutdown,
1101 	.setsockopt = sock_no_setsockopt,
1102 	.getsockopt = sock_no_getsockopt,
1103 	.sendmsg = vsock_dgram_sendmsg,
1104 	.recvmsg = vsock_dgram_recvmsg,
1105 	.mmap = sock_no_mmap,
1106 	.sendpage = sock_no_sendpage,
1107 };
1108 
1109 static int vsock_transport_cancel_pkt(struct vsock_sock *vsk)
1110 {
1111 	if (!transport->cancel_pkt)
1112 		return -EOPNOTSUPP;
1113 
1114 	return transport->cancel_pkt(vsk);
1115 }
1116 
1117 static void vsock_connect_timeout(struct work_struct *work)
1118 {
1119 	struct sock *sk;
1120 	struct vsock_sock *vsk;
1121 	int cancel = 0;
1122 
1123 	vsk = container_of(work, struct vsock_sock, connect_work.work);
1124 	sk = sk_vsock(vsk);
1125 
1126 	lock_sock(sk);
1127 	if (sk->sk_state == TCP_SYN_SENT &&
1128 	    (sk->sk_shutdown != SHUTDOWN_MASK)) {
1129 		sk->sk_state = TCP_CLOSE;
1130 		sk->sk_err = ETIMEDOUT;
1131 		sk->sk_error_report(sk);
1132 		cancel = 1;
1133 	}
1134 	release_sock(sk);
1135 	if (cancel)
1136 		vsock_transport_cancel_pkt(vsk);
1137 
1138 	sock_put(sk);
1139 }
1140 
1141 static int vsock_stream_connect(struct socket *sock, struct sockaddr *addr,
1142 				int addr_len, int flags)
1143 {
1144 	int err;
1145 	struct sock *sk;
1146 	struct vsock_sock *vsk;
1147 	struct sockaddr_vm *remote_addr;
1148 	long timeout;
1149 	DEFINE_WAIT(wait);
1150 
1151 	err = 0;
1152 	sk = sock->sk;
1153 	vsk = vsock_sk(sk);
1154 
1155 	lock_sock(sk);
1156 
1157 	/* XXX AF_UNSPEC should make us disconnect like AF_INET. */
1158 	switch (sock->state) {
1159 	case SS_CONNECTED:
1160 		err = -EISCONN;
1161 		goto out;
1162 	case SS_DISCONNECTING:
1163 		err = -EINVAL;
1164 		goto out;
1165 	case SS_CONNECTING:
1166 		/* This continues on so we can move sock into the SS_CONNECTED
1167 		 * state once the connection has completed (at which point err
1168 		 * will be set to zero also).  Otherwise, we will either wait
1169 		 * for the connection or return -EALREADY should this be a
1170 		 * non-blocking call.
1171 		 */
1172 		err = -EALREADY;
1173 		break;
1174 	default:
1175 		if ((sk->sk_state == TCP_LISTEN) ||
1176 		    vsock_addr_cast(addr, addr_len, &remote_addr) != 0) {
1177 			err = -EINVAL;
1178 			goto out;
1179 		}
1180 
1181 		/* The hypervisor and well-known contexts do not have socket
1182 		 * endpoints.
1183 		 */
1184 		if (!transport->stream_allow(remote_addr->svm_cid,
1185 					     remote_addr->svm_port)) {
1186 			err = -ENETUNREACH;
1187 			goto out;
1188 		}
1189 
1190 		/* Set the remote address that we are connecting to. */
1191 		memcpy(&vsk->remote_addr, remote_addr,
1192 		       sizeof(vsk->remote_addr));
1193 
1194 		err = vsock_auto_bind(vsk);
1195 		if (err)
1196 			goto out;
1197 
1198 		sk->sk_state = TCP_SYN_SENT;
1199 
1200 		err = transport->connect(vsk);
1201 		if (err < 0)
1202 			goto out;
1203 
1204 		/* Mark sock as connecting and set the error code to in
1205 		 * progress in case this is a non-blocking connect.
1206 		 */
1207 		sock->state = SS_CONNECTING;
1208 		err = -EINPROGRESS;
1209 	}
1210 
1211 	/* The receive path will handle all communication until we are able to
1212 	 * enter the connected state.  Here we wait for the connection to be
1213 	 * completed or a notification of an error.
1214 	 */
1215 	timeout = vsk->connect_timeout;
1216 	prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1217 
1218 	while (sk->sk_state != TCP_ESTABLISHED && sk->sk_err == 0) {
1219 		if (flags & O_NONBLOCK) {
1220 			/* If we're not going to block, we schedule a timeout
1221 			 * function to generate a timeout on the connection
1222 			 * attempt, in case the peer doesn't respond in a
1223 			 * timely manner. We hold on to the socket until the
1224 			 * timeout fires.
1225 			 */
1226 			sock_hold(sk);
1227 			schedule_delayed_work(&vsk->connect_work, timeout);
1228 
1229 			/* Skip ahead to preserve error code set above. */
1230 			goto out_wait;
1231 		}
1232 
1233 		release_sock(sk);
1234 		timeout = schedule_timeout(timeout);
1235 		lock_sock(sk);
1236 
1237 		if (signal_pending(current)) {
1238 			err = sock_intr_errno(timeout);
1239 			sk->sk_state = TCP_CLOSE;
1240 			sock->state = SS_UNCONNECTED;
1241 			vsock_transport_cancel_pkt(vsk);
1242 			goto out_wait;
1243 		} else if (timeout == 0) {
1244 			err = -ETIMEDOUT;
1245 			sk->sk_state = TCP_CLOSE;
1246 			sock->state = SS_UNCONNECTED;
1247 			vsock_transport_cancel_pkt(vsk);
1248 			goto out_wait;
1249 		}
1250 
1251 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1252 	}
1253 
1254 	if (sk->sk_err) {
1255 		err = -sk->sk_err;
1256 		sk->sk_state = TCP_CLOSE;
1257 		sock->state = SS_UNCONNECTED;
1258 	} else {
1259 		err = 0;
1260 	}
1261 
1262 out_wait:
1263 	finish_wait(sk_sleep(sk), &wait);
1264 out:
1265 	release_sock(sk);
1266 	return err;
1267 }
1268 
1269 static int vsock_accept(struct socket *sock, struct socket *newsock, int flags,
1270 			bool kern)
1271 {
1272 	struct sock *listener;
1273 	int err;
1274 	struct sock *connected;
1275 	struct vsock_sock *vconnected;
1276 	long timeout;
1277 	DEFINE_WAIT(wait);
1278 
1279 	err = 0;
1280 	listener = sock->sk;
1281 
1282 	lock_sock(listener);
1283 
1284 	if (sock->type != SOCK_STREAM) {
1285 		err = -EOPNOTSUPP;
1286 		goto out;
1287 	}
1288 
1289 	if (listener->sk_state != TCP_LISTEN) {
1290 		err = -EINVAL;
1291 		goto out;
1292 	}
1293 
1294 	/* Wait for children sockets to appear; these are the new sockets
1295 	 * created upon connection establishment.
1296 	 */
1297 	timeout = sock_sndtimeo(listener, flags & O_NONBLOCK);
1298 	prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1299 
1300 	while ((connected = vsock_dequeue_accept(listener)) == NULL &&
1301 	       listener->sk_err == 0) {
1302 		release_sock(listener);
1303 		timeout = schedule_timeout(timeout);
1304 		finish_wait(sk_sleep(listener), &wait);
1305 		lock_sock(listener);
1306 
1307 		if (signal_pending(current)) {
1308 			err = sock_intr_errno(timeout);
1309 			goto out;
1310 		} else if (timeout == 0) {
1311 			err = -EAGAIN;
1312 			goto out;
1313 		}
1314 
1315 		prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1316 	}
1317 	finish_wait(sk_sleep(listener), &wait);
1318 
1319 	if (listener->sk_err)
1320 		err = -listener->sk_err;
1321 
1322 	if (connected) {
1323 		listener->sk_ack_backlog--;
1324 
1325 		lock_sock_nested(connected, SINGLE_DEPTH_NESTING);
1326 		vconnected = vsock_sk(connected);
1327 
1328 		/* If the listener socket has received an error, then we should
1329 		 * reject this socket and return.  Note that we simply mark the
1330 		 * socket rejected, drop our reference, and let the cleanup
1331 		 * function handle the cleanup; the fact that we found it in
1332 		 * the listener's accept queue guarantees that the cleanup
1333 		 * function hasn't run yet.
1334 		 */
1335 		if (err) {
1336 			vconnected->rejected = true;
1337 		} else {
1338 			newsock->state = SS_CONNECTED;
1339 			sock_graft(connected, newsock);
1340 		}
1341 
1342 		release_sock(connected);
1343 		sock_put(connected);
1344 	}
1345 
1346 out:
1347 	release_sock(listener);
1348 	return err;
1349 }
1350 
1351 static int vsock_listen(struct socket *sock, int backlog)
1352 {
1353 	int err;
1354 	struct sock *sk;
1355 	struct vsock_sock *vsk;
1356 
1357 	sk = sock->sk;
1358 
1359 	lock_sock(sk);
1360 
1361 	if (sock->type != SOCK_STREAM) {
1362 		err = -EOPNOTSUPP;
1363 		goto out;
1364 	}
1365 
1366 	if (sock->state != SS_UNCONNECTED) {
1367 		err = -EINVAL;
1368 		goto out;
1369 	}
1370 
1371 	vsk = vsock_sk(sk);
1372 
1373 	if (!vsock_addr_bound(&vsk->local_addr)) {
1374 		err = -EINVAL;
1375 		goto out;
1376 	}
1377 
1378 	sk->sk_max_ack_backlog = backlog;
1379 	sk->sk_state = TCP_LISTEN;
1380 
1381 	err = 0;
1382 
1383 out:
1384 	release_sock(sk);
1385 	return err;
1386 }
1387 
1388 static int vsock_stream_setsockopt(struct socket *sock,
1389 				   int level,
1390 				   int optname,
1391 				   char __user *optval,
1392 				   unsigned int optlen)
1393 {
1394 	int err;
1395 	struct sock *sk;
1396 	struct vsock_sock *vsk;
1397 	u64 val;
1398 
1399 	if (level != AF_VSOCK)
1400 		return -ENOPROTOOPT;
1401 
1402 #define COPY_IN(_v)                                       \
1403 	do {						  \
1404 		if (optlen < sizeof(_v)) {		  \
1405 			err = -EINVAL;			  \
1406 			goto exit;			  \
1407 		}					  \
1408 		if (copy_from_user(&_v, optval, sizeof(_v)) != 0) {	\
1409 			err = -EFAULT;					\
1410 			goto exit;					\
1411 		}							\
1412 	} while (0)
1413 
1414 	err = 0;
1415 	sk = sock->sk;
1416 	vsk = vsock_sk(sk);
1417 
1418 	lock_sock(sk);
1419 
1420 	switch (optname) {
1421 	case SO_VM_SOCKETS_BUFFER_SIZE:
1422 		COPY_IN(val);
1423 		transport->set_buffer_size(vsk, val);
1424 		break;
1425 
1426 	case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1427 		COPY_IN(val);
1428 		transport->set_max_buffer_size(vsk, val);
1429 		break;
1430 
1431 	case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1432 		COPY_IN(val);
1433 		transport->set_min_buffer_size(vsk, val);
1434 		break;
1435 
1436 	case SO_VM_SOCKETS_CONNECT_TIMEOUT: {
1437 		struct timeval tv;
1438 		COPY_IN(tv);
1439 		if (tv.tv_sec >= 0 && tv.tv_usec < USEC_PER_SEC &&
1440 		    tv.tv_sec < (MAX_SCHEDULE_TIMEOUT / HZ - 1)) {
1441 			vsk->connect_timeout = tv.tv_sec * HZ +
1442 			    DIV_ROUND_UP(tv.tv_usec, (1000000 / HZ));
1443 			if (vsk->connect_timeout == 0)
1444 				vsk->connect_timeout =
1445 				    VSOCK_DEFAULT_CONNECT_TIMEOUT;
1446 
1447 		} else {
1448 			err = -ERANGE;
1449 		}
1450 		break;
1451 	}
1452 
1453 	default:
1454 		err = -ENOPROTOOPT;
1455 		break;
1456 	}
1457 
1458 #undef COPY_IN
1459 
1460 exit:
1461 	release_sock(sk);
1462 	return err;
1463 }
1464 
1465 static int vsock_stream_getsockopt(struct socket *sock,
1466 				   int level, int optname,
1467 				   char __user *optval,
1468 				   int __user *optlen)
1469 {
1470 	int err;
1471 	int len;
1472 	struct sock *sk;
1473 	struct vsock_sock *vsk;
1474 	u64 val;
1475 
1476 	if (level != AF_VSOCK)
1477 		return -ENOPROTOOPT;
1478 
1479 	err = get_user(len, optlen);
1480 	if (err != 0)
1481 		return err;
1482 
1483 #define COPY_OUT(_v)                            \
1484 	do {					\
1485 		if (len < sizeof(_v))		\
1486 			return -EINVAL;		\
1487 						\
1488 		len = sizeof(_v);		\
1489 		if (copy_to_user(optval, &_v, len) != 0)	\
1490 			return -EFAULT;				\
1491 								\
1492 	} while (0)
1493 
1494 	err = 0;
1495 	sk = sock->sk;
1496 	vsk = vsock_sk(sk);
1497 
1498 	switch (optname) {
1499 	case SO_VM_SOCKETS_BUFFER_SIZE:
1500 		val = transport->get_buffer_size(vsk);
1501 		COPY_OUT(val);
1502 		break;
1503 
1504 	case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1505 		val = transport->get_max_buffer_size(vsk);
1506 		COPY_OUT(val);
1507 		break;
1508 
1509 	case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1510 		val = transport->get_min_buffer_size(vsk);
1511 		COPY_OUT(val);
1512 		break;
1513 
1514 	case SO_VM_SOCKETS_CONNECT_TIMEOUT: {
1515 		struct timeval tv;
1516 		tv.tv_sec = vsk->connect_timeout / HZ;
1517 		tv.tv_usec =
1518 		    (vsk->connect_timeout -
1519 		     tv.tv_sec * HZ) * (1000000 / HZ);
1520 		COPY_OUT(tv);
1521 		break;
1522 	}
1523 	default:
1524 		return -ENOPROTOOPT;
1525 	}
1526 
1527 	err = put_user(len, optlen);
1528 	if (err != 0)
1529 		return -EFAULT;
1530 
1531 #undef COPY_OUT
1532 
1533 	return 0;
1534 }
1535 
1536 static int vsock_stream_sendmsg(struct socket *sock, struct msghdr *msg,
1537 				size_t len)
1538 {
1539 	struct sock *sk;
1540 	struct vsock_sock *vsk;
1541 	ssize_t total_written;
1542 	long timeout;
1543 	int err;
1544 	struct vsock_transport_send_notify_data send_data;
1545 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
1546 
1547 	sk = sock->sk;
1548 	vsk = vsock_sk(sk);
1549 	total_written = 0;
1550 	err = 0;
1551 
1552 	if (msg->msg_flags & MSG_OOB)
1553 		return -EOPNOTSUPP;
1554 
1555 	lock_sock(sk);
1556 
1557 	/* Callers should not provide a destination with stream sockets. */
1558 	if (msg->msg_namelen) {
1559 		err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP;
1560 		goto out;
1561 	}
1562 
1563 	/* Send data only if both sides are not shutdown in the direction. */
1564 	if (sk->sk_shutdown & SEND_SHUTDOWN ||
1565 	    vsk->peer_shutdown & RCV_SHUTDOWN) {
1566 		err = -EPIPE;
1567 		goto out;
1568 	}
1569 
1570 	if (sk->sk_state != TCP_ESTABLISHED ||
1571 	    !vsock_addr_bound(&vsk->local_addr)) {
1572 		err = -ENOTCONN;
1573 		goto out;
1574 	}
1575 
1576 	if (!vsock_addr_bound(&vsk->remote_addr)) {
1577 		err = -EDESTADDRREQ;
1578 		goto out;
1579 	}
1580 
1581 	/* Wait for room in the produce queue to enqueue our user's data. */
1582 	timeout = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1583 
1584 	err = transport->notify_send_init(vsk, &send_data);
1585 	if (err < 0)
1586 		goto out;
1587 
1588 	while (total_written < len) {
1589 		ssize_t written;
1590 
1591 		add_wait_queue(sk_sleep(sk), &wait);
1592 		while (vsock_stream_has_space(vsk) == 0 &&
1593 		       sk->sk_err == 0 &&
1594 		       !(sk->sk_shutdown & SEND_SHUTDOWN) &&
1595 		       !(vsk->peer_shutdown & RCV_SHUTDOWN)) {
1596 
1597 			/* Don't wait for non-blocking sockets. */
1598 			if (timeout == 0) {
1599 				err = -EAGAIN;
1600 				remove_wait_queue(sk_sleep(sk), &wait);
1601 				goto out_err;
1602 			}
1603 
1604 			err = transport->notify_send_pre_block(vsk, &send_data);
1605 			if (err < 0) {
1606 				remove_wait_queue(sk_sleep(sk), &wait);
1607 				goto out_err;
1608 			}
1609 
1610 			release_sock(sk);
1611 			timeout = wait_woken(&wait, TASK_INTERRUPTIBLE, timeout);
1612 			lock_sock(sk);
1613 			if (signal_pending(current)) {
1614 				err = sock_intr_errno(timeout);
1615 				remove_wait_queue(sk_sleep(sk), &wait);
1616 				goto out_err;
1617 			} else if (timeout == 0) {
1618 				err = -EAGAIN;
1619 				remove_wait_queue(sk_sleep(sk), &wait);
1620 				goto out_err;
1621 			}
1622 		}
1623 		remove_wait_queue(sk_sleep(sk), &wait);
1624 
1625 		/* These checks occur both as part of and after the loop
1626 		 * conditional since we need to check before and after
1627 		 * sleeping.
1628 		 */
1629 		if (sk->sk_err) {
1630 			err = -sk->sk_err;
1631 			goto out_err;
1632 		} else if ((sk->sk_shutdown & SEND_SHUTDOWN) ||
1633 			   (vsk->peer_shutdown & RCV_SHUTDOWN)) {
1634 			err = -EPIPE;
1635 			goto out_err;
1636 		}
1637 
1638 		err = transport->notify_send_pre_enqueue(vsk, &send_data);
1639 		if (err < 0)
1640 			goto out_err;
1641 
1642 		/* Note that enqueue will only write as many bytes as are free
1643 		 * in the produce queue, so we don't need to ensure len is
1644 		 * smaller than the queue size.  It is the caller's
1645 		 * responsibility to check how many bytes we were able to send.
1646 		 */
1647 
1648 		written = transport->stream_enqueue(
1649 				vsk, msg,
1650 				len - total_written);
1651 		if (written < 0) {
1652 			err = -ENOMEM;
1653 			goto out_err;
1654 		}
1655 
1656 		total_written += written;
1657 
1658 		err = transport->notify_send_post_enqueue(
1659 				vsk, written, &send_data);
1660 		if (err < 0)
1661 			goto out_err;
1662 
1663 	}
1664 
1665 out_err:
1666 	if (total_written > 0)
1667 		err = total_written;
1668 out:
1669 	release_sock(sk);
1670 	return err;
1671 }
1672 
1673 
1674 static int
1675 vsock_stream_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
1676 		     int flags)
1677 {
1678 	struct sock *sk;
1679 	struct vsock_sock *vsk;
1680 	int err;
1681 	size_t target;
1682 	ssize_t copied;
1683 	long timeout;
1684 	struct vsock_transport_recv_notify_data recv_data;
1685 
1686 	DEFINE_WAIT(wait);
1687 
1688 	sk = sock->sk;
1689 	vsk = vsock_sk(sk);
1690 	err = 0;
1691 
1692 	lock_sock(sk);
1693 
1694 	if (sk->sk_state != TCP_ESTABLISHED) {
1695 		/* Recvmsg is supposed to return 0 if a peer performs an
1696 		 * orderly shutdown. Differentiate between that case and when a
1697 		 * peer has not connected or a local shutdown occured with the
1698 		 * SOCK_DONE flag.
1699 		 */
1700 		if (sock_flag(sk, SOCK_DONE))
1701 			err = 0;
1702 		else
1703 			err = -ENOTCONN;
1704 
1705 		goto out;
1706 	}
1707 
1708 	if (flags & MSG_OOB) {
1709 		err = -EOPNOTSUPP;
1710 		goto out;
1711 	}
1712 
1713 	/* We don't check peer_shutdown flag here since peer may actually shut
1714 	 * down, but there can be data in the queue that a local socket can
1715 	 * receive.
1716 	 */
1717 	if (sk->sk_shutdown & RCV_SHUTDOWN) {
1718 		err = 0;
1719 		goto out;
1720 	}
1721 
1722 	/* It is valid on Linux to pass in a zero-length receive buffer.  This
1723 	 * is not an error.  We may as well bail out now.
1724 	 */
1725 	if (!len) {
1726 		err = 0;
1727 		goto out;
1728 	}
1729 
1730 	/* We must not copy less than target bytes into the user's buffer
1731 	 * before returning successfully, so we wait for the consume queue to
1732 	 * have that much data to consume before dequeueing.  Note that this
1733 	 * makes it impossible to handle cases where target is greater than the
1734 	 * queue size.
1735 	 */
1736 	target = sock_rcvlowat(sk, flags & MSG_WAITALL, len);
1737 	if (target >= transport->stream_rcvhiwat(vsk)) {
1738 		err = -ENOMEM;
1739 		goto out;
1740 	}
1741 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1742 	copied = 0;
1743 
1744 	err = transport->notify_recv_init(vsk, target, &recv_data);
1745 	if (err < 0)
1746 		goto out;
1747 
1748 
1749 	while (1) {
1750 		s64 ready;
1751 
1752 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1753 		ready = vsock_stream_has_data(vsk);
1754 
1755 		if (ready == 0) {
1756 			if (sk->sk_err != 0 ||
1757 			    (sk->sk_shutdown & RCV_SHUTDOWN) ||
1758 			    (vsk->peer_shutdown & SEND_SHUTDOWN)) {
1759 				finish_wait(sk_sleep(sk), &wait);
1760 				break;
1761 			}
1762 			/* Don't wait for non-blocking sockets. */
1763 			if (timeout == 0) {
1764 				err = -EAGAIN;
1765 				finish_wait(sk_sleep(sk), &wait);
1766 				break;
1767 			}
1768 
1769 			err = transport->notify_recv_pre_block(
1770 					vsk, target, &recv_data);
1771 			if (err < 0) {
1772 				finish_wait(sk_sleep(sk), &wait);
1773 				break;
1774 			}
1775 			release_sock(sk);
1776 			timeout = schedule_timeout(timeout);
1777 			lock_sock(sk);
1778 
1779 			if (signal_pending(current)) {
1780 				err = sock_intr_errno(timeout);
1781 				finish_wait(sk_sleep(sk), &wait);
1782 				break;
1783 			} else if (timeout == 0) {
1784 				err = -EAGAIN;
1785 				finish_wait(sk_sleep(sk), &wait);
1786 				break;
1787 			}
1788 		} else {
1789 			ssize_t read;
1790 
1791 			finish_wait(sk_sleep(sk), &wait);
1792 
1793 			if (ready < 0) {
1794 				/* Invalid queue pair content. XXX This should
1795 				* be changed to a connection reset in a later
1796 				* change.
1797 				*/
1798 
1799 				err = -ENOMEM;
1800 				goto out;
1801 			}
1802 
1803 			err = transport->notify_recv_pre_dequeue(
1804 					vsk, target, &recv_data);
1805 			if (err < 0)
1806 				break;
1807 
1808 			read = transport->stream_dequeue(
1809 					vsk, msg,
1810 					len - copied, flags);
1811 			if (read < 0) {
1812 				err = -ENOMEM;
1813 				break;
1814 			}
1815 
1816 			copied += read;
1817 
1818 			err = transport->notify_recv_post_dequeue(
1819 					vsk, target, read,
1820 					!(flags & MSG_PEEK), &recv_data);
1821 			if (err < 0)
1822 				goto out;
1823 
1824 			if (read >= target || flags & MSG_PEEK)
1825 				break;
1826 
1827 			target -= read;
1828 		}
1829 	}
1830 
1831 	if (sk->sk_err)
1832 		err = -sk->sk_err;
1833 	else if (sk->sk_shutdown & RCV_SHUTDOWN)
1834 		err = 0;
1835 
1836 	if (copied > 0)
1837 		err = copied;
1838 
1839 out:
1840 	release_sock(sk);
1841 	return err;
1842 }
1843 
1844 static const struct proto_ops vsock_stream_ops = {
1845 	.family = PF_VSOCK,
1846 	.owner = THIS_MODULE,
1847 	.release = vsock_release,
1848 	.bind = vsock_bind,
1849 	.connect = vsock_stream_connect,
1850 	.socketpair = sock_no_socketpair,
1851 	.accept = vsock_accept,
1852 	.getname = vsock_getname,
1853 	.poll = vsock_poll,
1854 	.ioctl = sock_no_ioctl,
1855 	.listen = vsock_listen,
1856 	.shutdown = vsock_shutdown,
1857 	.setsockopt = vsock_stream_setsockopt,
1858 	.getsockopt = vsock_stream_getsockopt,
1859 	.sendmsg = vsock_stream_sendmsg,
1860 	.recvmsg = vsock_stream_recvmsg,
1861 	.mmap = sock_no_mmap,
1862 	.sendpage = sock_no_sendpage,
1863 };
1864 
1865 static int vsock_create(struct net *net, struct socket *sock,
1866 			int protocol, int kern)
1867 {
1868 	if (!sock)
1869 		return -EINVAL;
1870 
1871 	if (protocol && protocol != PF_VSOCK)
1872 		return -EPROTONOSUPPORT;
1873 
1874 	switch (sock->type) {
1875 	case SOCK_DGRAM:
1876 		sock->ops = &vsock_dgram_ops;
1877 		break;
1878 	case SOCK_STREAM:
1879 		sock->ops = &vsock_stream_ops;
1880 		break;
1881 	default:
1882 		return -ESOCKTNOSUPPORT;
1883 	}
1884 
1885 	sock->state = SS_UNCONNECTED;
1886 
1887 	return __vsock_create(net, sock, NULL, GFP_KERNEL, 0, kern) ? 0 : -ENOMEM;
1888 }
1889 
1890 static const struct net_proto_family vsock_family_ops = {
1891 	.family = AF_VSOCK,
1892 	.create = vsock_create,
1893 	.owner = THIS_MODULE,
1894 };
1895 
1896 static long vsock_dev_do_ioctl(struct file *filp,
1897 			       unsigned int cmd, void __user *ptr)
1898 {
1899 	u32 __user *p = ptr;
1900 	int retval = 0;
1901 
1902 	switch (cmd) {
1903 	case IOCTL_VM_SOCKETS_GET_LOCAL_CID:
1904 		if (put_user(transport->get_local_cid(), p) != 0)
1905 			retval = -EFAULT;
1906 		break;
1907 
1908 	default:
1909 		pr_err("Unknown ioctl %d\n", cmd);
1910 		retval = -EINVAL;
1911 	}
1912 
1913 	return retval;
1914 }
1915 
1916 static long vsock_dev_ioctl(struct file *filp,
1917 			    unsigned int cmd, unsigned long arg)
1918 {
1919 	return vsock_dev_do_ioctl(filp, cmd, (void __user *)arg);
1920 }
1921 
1922 #ifdef CONFIG_COMPAT
1923 static long vsock_dev_compat_ioctl(struct file *filp,
1924 				   unsigned int cmd, unsigned long arg)
1925 {
1926 	return vsock_dev_do_ioctl(filp, cmd, compat_ptr(arg));
1927 }
1928 #endif
1929 
1930 static const struct file_operations vsock_device_ops = {
1931 	.owner		= THIS_MODULE,
1932 	.unlocked_ioctl	= vsock_dev_ioctl,
1933 #ifdef CONFIG_COMPAT
1934 	.compat_ioctl	= vsock_dev_compat_ioctl,
1935 #endif
1936 	.open		= nonseekable_open,
1937 };
1938 
1939 static struct miscdevice vsock_device = {
1940 	.name		= "vsock",
1941 	.fops		= &vsock_device_ops,
1942 };
1943 
1944 int __vsock_core_init(const struct vsock_transport *t, struct module *owner)
1945 {
1946 	int err = mutex_lock_interruptible(&vsock_register_mutex);
1947 
1948 	if (err)
1949 		return err;
1950 
1951 	if (transport) {
1952 		err = -EBUSY;
1953 		goto err_busy;
1954 	}
1955 
1956 	/* Transport must be the owner of the protocol so that it can't
1957 	 * unload while there are open sockets.
1958 	 */
1959 	vsock_proto.owner = owner;
1960 	transport = t;
1961 
1962 	vsock_device.minor = MISC_DYNAMIC_MINOR;
1963 	err = misc_register(&vsock_device);
1964 	if (err) {
1965 		pr_err("Failed to register misc device\n");
1966 		goto err_reset_transport;
1967 	}
1968 
1969 	err = proto_register(&vsock_proto, 1);	/* we want our slab */
1970 	if (err) {
1971 		pr_err("Cannot register vsock protocol\n");
1972 		goto err_deregister_misc;
1973 	}
1974 
1975 	err = sock_register(&vsock_family_ops);
1976 	if (err) {
1977 		pr_err("could not register af_vsock (%d) address family: %d\n",
1978 		       AF_VSOCK, err);
1979 		goto err_unregister_proto;
1980 	}
1981 
1982 	mutex_unlock(&vsock_register_mutex);
1983 	return 0;
1984 
1985 err_unregister_proto:
1986 	proto_unregister(&vsock_proto);
1987 err_deregister_misc:
1988 	misc_deregister(&vsock_device);
1989 err_reset_transport:
1990 	transport = NULL;
1991 err_busy:
1992 	mutex_unlock(&vsock_register_mutex);
1993 	return err;
1994 }
1995 EXPORT_SYMBOL_GPL(__vsock_core_init);
1996 
1997 void vsock_core_exit(void)
1998 {
1999 	mutex_lock(&vsock_register_mutex);
2000 
2001 	misc_deregister(&vsock_device);
2002 	sock_unregister(AF_VSOCK);
2003 	proto_unregister(&vsock_proto);
2004 
2005 	/* We do not want the assignment below re-ordered. */
2006 	mb();
2007 	transport = NULL;
2008 
2009 	mutex_unlock(&vsock_register_mutex);
2010 }
2011 EXPORT_SYMBOL_GPL(vsock_core_exit);
2012 
2013 const struct vsock_transport *vsock_core_get_transport(void)
2014 {
2015 	/* vsock_register_mutex not taken since only the transport uses this
2016 	 * function and only while registered.
2017 	 */
2018 	return transport;
2019 }
2020 EXPORT_SYMBOL_GPL(vsock_core_get_transport);
2021 
2022 static void __exit vsock_exit(void)
2023 {
2024 	/* Do nothing.  This function makes this module removable. */
2025 }
2026 
2027 module_init(vsock_init_tables);
2028 module_exit(vsock_exit);
2029 
2030 MODULE_AUTHOR("VMware, Inc.");
2031 MODULE_DESCRIPTION("VMware Virtual Socket Family");
2032 MODULE_VERSION("1.0.2.0-k");
2033 MODULE_LICENSE("GPL v2");
2034