xref: /linux/net/vmw_vsock/af_vsock.c (revision 4b132aacb0768ac1e652cf517097ea6f237214b9)
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/errqueue.h>
93 #include <linux/init.h>
94 #include <linux/io.h>
95 #include <linux/kernel.h>
96 #include <linux/sched/signal.h>
97 #include <linux/kmod.h>
98 #include <linux/list.h>
99 #include <linux/miscdevice.h>
100 #include <linux/module.h>
101 #include <linux/mutex.h>
102 #include <linux/net.h>
103 #include <linux/poll.h>
104 #include <linux/random.h>
105 #include <linux/skbuff.h>
106 #include <linux/smp.h>
107 #include <linux/socket.h>
108 #include <linux/stddef.h>
109 #include <linux/unistd.h>
110 #include <linux/wait.h>
111 #include <linux/workqueue.h>
112 #include <net/sock.h>
113 #include <net/af_vsock.h>
114 #include <uapi/linux/vm_sockets.h>
115 
116 static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr);
117 static void vsock_sk_destruct(struct sock *sk);
118 static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
119 
120 /* Protocol family. */
121 struct proto vsock_proto = {
122 	.name = "AF_VSOCK",
123 	.owner = THIS_MODULE,
124 	.obj_size = sizeof(struct vsock_sock),
125 #ifdef CONFIG_BPF_SYSCALL
126 	.psock_update_sk_prot = vsock_bpf_update_proto,
127 #endif
128 };
129 
130 /* The default peer timeout indicates how long we will wait for a peer response
131  * to a control message.
132  */
133 #define VSOCK_DEFAULT_CONNECT_TIMEOUT (2 * HZ)
134 
135 #define VSOCK_DEFAULT_BUFFER_SIZE     (1024 * 256)
136 #define VSOCK_DEFAULT_BUFFER_MAX_SIZE (1024 * 256)
137 #define VSOCK_DEFAULT_BUFFER_MIN_SIZE 128
138 
139 /* Transport used for host->guest communication */
140 static const struct vsock_transport *transport_h2g;
141 /* Transport used for guest->host communication */
142 static const struct vsock_transport *transport_g2h;
143 /* Transport used for DGRAM communication */
144 static const struct vsock_transport *transport_dgram;
145 /* Transport used for local communication */
146 static const struct vsock_transport *transport_local;
147 static DEFINE_MUTEX(vsock_register_mutex);
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 void 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 }
208 
209 static void __vsock_insert_bound(struct list_head *list,
210 				 struct vsock_sock *vsk)
211 {
212 	sock_hold(&vsk->sk);
213 	list_add(&vsk->bound_table, list);
214 }
215 
216 static void __vsock_insert_connected(struct list_head *list,
217 				     struct vsock_sock *vsk)
218 {
219 	sock_hold(&vsk->sk);
220 	list_add(&vsk->connected_table, list);
221 }
222 
223 static void __vsock_remove_bound(struct vsock_sock *vsk)
224 {
225 	list_del_init(&vsk->bound_table);
226 	sock_put(&vsk->sk);
227 }
228 
229 static void __vsock_remove_connected(struct vsock_sock *vsk)
230 {
231 	list_del_init(&vsk->connected_table);
232 	sock_put(&vsk->sk);
233 }
234 
235 static struct sock *__vsock_find_bound_socket(struct sockaddr_vm *addr)
236 {
237 	struct vsock_sock *vsk;
238 
239 	list_for_each_entry(vsk, vsock_bound_sockets(addr), bound_table) {
240 		if (vsock_addr_equals_addr(addr, &vsk->local_addr))
241 			return sk_vsock(vsk);
242 
243 		if (addr->svm_port == vsk->local_addr.svm_port &&
244 		    (vsk->local_addr.svm_cid == VMADDR_CID_ANY ||
245 		     addr->svm_cid == VMADDR_CID_ANY))
246 			return sk_vsock(vsk);
247 	}
248 
249 	return NULL;
250 }
251 
252 static struct sock *__vsock_find_connected_socket(struct sockaddr_vm *src,
253 						  struct sockaddr_vm *dst)
254 {
255 	struct vsock_sock *vsk;
256 
257 	list_for_each_entry(vsk, vsock_connected_sockets(src, dst),
258 			    connected_table) {
259 		if (vsock_addr_equals_addr(src, &vsk->remote_addr) &&
260 		    dst->svm_port == vsk->local_addr.svm_port) {
261 			return sk_vsock(vsk);
262 		}
263 	}
264 
265 	return NULL;
266 }
267 
268 static void vsock_insert_unbound(struct vsock_sock *vsk)
269 {
270 	spin_lock_bh(&vsock_table_lock);
271 	__vsock_insert_bound(vsock_unbound_sockets, vsk);
272 	spin_unlock_bh(&vsock_table_lock);
273 }
274 
275 void vsock_insert_connected(struct vsock_sock *vsk)
276 {
277 	struct list_head *list = vsock_connected_sockets(
278 		&vsk->remote_addr, &vsk->local_addr);
279 
280 	spin_lock_bh(&vsock_table_lock);
281 	__vsock_insert_connected(list, vsk);
282 	spin_unlock_bh(&vsock_table_lock);
283 }
284 EXPORT_SYMBOL_GPL(vsock_insert_connected);
285 
286 void vsock_remove_bound(struct vsock_sock *vsk)
287 {
288 	spin_lock_bh(&vsock_table_lock);
289 	if (__vsock_in_bound_table(vsk))
290 		__vsock_remove_bound(vsk);
291 	spin_unlock_bh(&vsock_table_lock);
292 }
293 EXPORT_SYMBOL_GPL(vsock_remove_bound);
294 
295 void vsock_remove_connected(struct vsock_sock *vsk)
296 {
297 	spin_lock_bh(&vsock_table_lock);
298 	if (__vsock_in_connected_table(vsk))
299 		__vsock_remove_connected(vsk);
300 	spin_unlock_bh(&vsock_table_lock);
301 }
302 EXPORT_SYMBOL_GPL(vsock_remove_connected);
303 
304 struct sock *vsock_find_bound_socket(struct sockaddr_vm *addr)
305 {
306 	struct sock *sk;
307 
308 	spin_lock_bh(&vsock_table_lock);
309 	sk = __vsock_find_bound_socket(addr);
310 	if (sk)
311 		sock_hold(sk);
312 
313 	spin_unlock_bh(&vsock_table_lock);
314 
315 	return sk;
316 }
317 EXPORT_SYMBOL_GPL(vsock_find_bound_socket);
318 
319 struct sock *vsock_find_connected_socket(struct sockaddr_vm *src,
320 					 struct sockaddr_vm *dst)
321 {
322 	struct sock *sk;
323 
324 	spin_lock_bh(&vsock_table_lock);
325 	sk = __vsock_find_connected_socket(src, dst);
326 	if (sk)
327 		sock_hold(sk);
328 
329 	spin_unlock_bh(&vsock_table_lock);
330 
331 	return sk;
332 }
333 EXPORT_SYMBOL_GPL(vsock_find_connected_socket);
334 
335 void vsock_remove_sock(struct vsock_sock *vsk)
336 {
337 	vsock_remove_bound(vsk);
338 	vsock_remove_connected(vsk);
339 }
340 EXPORT_SYMBOL_GPL(vsock_remove_sock);
341 
342 void vsock_for_each_connected_socket(struct vsock_transport *transport,
343 				     void (*fn)(struct sock *sk))
344 {
345 	int i;
346 
347 	spin_lock_bh(&vsock_table_lock);
348 
349 	for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) {
350 		struct vsock_sock *vsk;
351 		list_for_each_entry(vsk, &vsock_connected_table[i],
352 				    connected_table) {
353 			if (vsk->transport != transport)
354 				continue;
355 
356 			fn(sk_vsock(vsk));
357 		}
358 	}
359 
360 	spin_unlock_bh(&vsock_table_lock);
361 }
362 EXPORT_SYMBOL_GPL(vsock_for_each_connected_socket);
363 
364 void vsock_add_pending(struct sock *listener, struct sock *pending)
365 {
366 	struct vsock_sock *vlistener;
367 	struct vsock_sock *vpending;
368 
369 	vlistener = vsock_sk(listener);
370 	vpending = vsock_sk(pending);
371 
372 	sock_hold(pending);
373 	sock_hold(listener);
374 	list_add_tail(&vpending->pending_links, &vlistener->pending_links);
375 }
376 EXPORT_SYMBOL_GPL(vsock_add_pending);
377 
378 void vsock_remove_pending(struct sock *listener, struct sock *pending)
379 {
380 	struct vsock_sock *vpending = vsock_sk(pending);
381 
382 	list_del_init(&vpending->pending_links);
383 	sock_put(listener);
384 	sock_put(pending);
385 }
386 EXPORT_SYMBOL_GPL(vsock_remove_pending);
387 
388 void vsock_enqueue_accept(struct sock *listener, struct sock *connected)
389 {
390 	struct vsock_sock *vlistener;
391 	struct vsock_sock *vconnected;
392 
393 	vlistener = vsock_sk(listener);
394 	vconnected = vsock_sk(connected);
395 
396 	sock_hold(connected);
397 	sock_hold(listener);
398 	list_add_tail(&vconnected->accept_queue, &vlistener->accept_queue);
399 }
400 EXPORT_SYMBOL_GPL(vsock_enqueue_accept);
401 
402 static bool vsock_use_local_transport(unsigned int remote_cid)
403 {
404 	if (!transport_local)
405 		return false;
406 
407 	if (remote_cid == VMADDR_CID_LOCAL)
408 		return true;
409 
410 	if (transport_g2h) {
411 		return remote_cid == transport_g2h->get_local_cid();
412 	} else {
413 		return remote_cid == VMADDR_CID_HOST;
414 	}
415 }
416 
417 static void vsock_deassign_transport(struct vsock_sock *vsk)
418 {
419 	if (!vsk->transport)
420 		return;
421 
422 	vsk->transport->destruct(vsk);
423 	module_put(vsk->transport->module);
424 	vsk->transport = NULL;
425 }
426 
427 /* Assign a transport to a socket and call the .init transport callback.
428  *
429  * Note: for connection oriented socket this must be called when vsk->remote_addr
430  * is set (e.g. during the connect() or when a connection request on a listener
431  * socket is received).
432  * The vsk->remote_addr is used to decide which transport to use:
433  *  - remote CID == VMADDR_CID_LOCAL or g2h->local_cid or VMADDR_CID_HOST if
434  *    g2h is not loaded, will use local transport;
435  *  - remote CID <= VMADDR_CID_HOST or h2g is not loaded or remote flags field
436  *    includes VMADDR_FLAG_TO_HOST flag value, will use guest->host transport;
437  *  - remote CID > VMADDR_CID_HOST will use host->guest transport;
438  */
439 int vsock_assign_transport(struct vsock_sock *vsk, struct vsock_sock *psk)
440 {
441 	const struct vsock_transport *new_transport;
442 	struct sock *sk = sk_vsock(vsk);
443 	unsigned int remote_cid = vsk->remote_addr.svm_cid;
444 	__u8 remote_flags;
445 	int ret;
446 
447 	/* If the packet is coming with the source and destination CIDs higher
448 	 * than VMADDR_CID_HOST, then a vsock channel where all the packets are
449 	 * forwarded to the host should be established. Then the host will
450 	 * need to forward the packets to the guest.
451 	 *
452 	 * The flag is set on the (listen) receive path (psk is not NULL). On
453 	 * the connect path the flag can be set by the user space application.
454 	 */
455 	if (psk && vsk->local_addr.svm_cid > VMADDR_CID_HOST &&
456 	    vsk->remote_addr.svm_cid > VMADDR_CID_HOST)
457 		vsk->remote_addr.svm_flags |= VMADDR_FLAG_TO_HOST;
458 
459 	remote_flags = vsk->remote_addr.svm_flags;
460 
461 	switch (sk->sk_type) {
462 	case SOCK_DGRAM:
463 		new_transport = transport_dgram;
464 		break;
465 	case SOCK_STREAM:
466 	case SOCK_SEQPACKET:
467 		if (vsock_use_local_transport(remote_cid))
468 			new_transport = transport_local;
469 		else if (remote_cid <= VMADDR_CID_HOST || !transport_h2g ||
470 			 (remote_flags & VMADDR_FLAG_TO_HOST))
471 			new_transport = transport_g2h;
472 		else
473 			new_transport = transport_h2g;
474 		break;
475 	default:
476 		return -ESOCKTNOSUPPORT;
477 	}
478 
479 	if (vsk->transport) {
480 		if (vsk->transport == new_transport)
481 			return 0;
482 
483 		/* transport->release() must be called with sock lock acquired.
484 		 * This path can only be taken during vsock_connect(), where we
485 		 * have already held the sock lock. In the other cases, this
486 		 * function is called on a new socket which is not assigned to
487 		 * any transport.
488 		 */
489 		vsk->transport->release(vsk);
490 		vsock_deassign_transport(vsk);
491 	}
492 
493 	/* We increase the module refcnt to prevent the transport unloading
494 	 * while there are open sockets assigned to it.
495 	 */
496 	if (!new_transport || !try_module_get(new_transport->module))
497 		return -ENODEV;
498 
499 	if (sk->sk_type == SOCK_SEQPACKET) {
500 		if (!new_transport->seqpacket_allow ||
501 		    !new_transport->seqpacket_allow(remote_cid)) {
502 			module_put(new_transport->module);
503 			return -ESOCKTNOSUPPORT;
504 		}
505 	}
506 
507 	ret = new_transport->init(vsk, psk);
508 	if (ret) {
509 		module_put(new_transport->module);
510 		return ret;
511 	}
512 
513 	vsk->transport = new_transport;
514 
515 	return 0;
516 }
517 EXPORT_SYMBOL_GPL(vsock_assign_transport);
518 
519 bool vsock_find_cid(unsigned int cid)
520 {
521 	if (transport_g2h && cid == transport_g2h->get_local_cid())
522 		return true;
523 
524 	if (transport_h2g && cid == VMADDR_CID_HOST)
525 		return true;
526 
527 	if (transport_local && cid == VMADDR_CID_LOCAL)
528 		return true;
529 
530 	return false;
531 }
532 EXPORT_SYMBOL_GPL(vsock_find_cid);
533 
534 static struct sock *vsock_dequeue_accept(struct sock *listener)
535 {
536 	struct vsock_sock *vlistener;
537 	struct vsock_sock *vconnected;
538 
539 	vlistener = vsock_sk(listener);
540 
541 	if (list_empty(&vlistener->accept_queue))
542 		return NULL;
543 
544 	vconnected = list_entry(vlistener->accept_queue.next,
545 				struct vsock_sock, accept_queue);
546 
547 	list_del_init(&vconnected->accept_queue);
548 	sock_put(listener);
549 	/* The caller will need a reference on the connected socket so we let
550 	 * it call sock_put().
551 	 */
552 
553 	return sk_vsock(vconnected);
554 }
555 
556 static bool vsock_is_accept_queue_empty(struct sock *sk)
557 {
558 	struct vsock_sock *vsk = vsock_sk(sk);
559 	return list_empty(&vsk->accept_queue);
560 }
561 
562 static bool vsock_is_pending(struct sock *sk)
563 {
564 	struct vsock_sock *vsk = vsock_sk(sk);
565 	return !list_empty(&vsk->pending_links);
566 }
567 
568 static int vsock_send_shutdown(struct sock *sk, int mode)
569 {
570 	struct vsock_sock *vsk = vsock_sk(sk);
571 
572 	if (!vsk->transport)
573 		return -ENODEV;
574 
575 	return vsk->transport->shutdown(vsk, mode);
576 }
577 
578 static void vsock_pending_work(struct work_struct *work)
579 {
580 	struct sock *sk;
581 	struct sock *listener;
582 	struct vsock_sock *vsk;
583 	bool cleanup;
584 
585 	vsk = container_of(work, struct vsock_sock, pending_work.work);
586 	sk = sk_vsock(vsk);
587 	listener = vsk->listener;
588 	cleanup = true;
589 
590 	lock_sock(listener);
591 	lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
592 
593 	if (vsock_is_pending(sk)) {
594 		vsock_remove_pending(listener, sk);
595 
596 		sk_acceptq_removed(listener);
597 	} else if (!vsk->rejected) {
598 		/* We are not on the pending list and accept() did not reject
599 		 * us, so we must have been accepted by our user process.  We
600 		 * just need to drop our references to the sockets and be on
601 		 * our way.
602 		 */
603 		cleanup = false;
604 		goto out;
605 	}
606 
607 	/* We need to remove ourself from the global connected sockets list so
608 	 * incoming packets can't find this socket, and to reduce the reference
609 	 * count.
610 	 */
611 	vsock_remove_connected(vsk);
612 
613 	sk->sk_state = TCP_CLOSE;
614 
615 out:
616 	release_sock(sk);
617 	release_sock(listener);
618 	if (cleanup)
619 		sock_put(sk);
620 
621 	sock_put(sk);
622 	sock_put(listener);
623 }
624 
625 /**** SOCKET OPERATIONS ****/
626 
627 static int __vsock_bind_connectible(struct vsock_sock *vsk,
628 				    struct sockaddr_vm *addr)
629 {
630 	static u32 port;
631 	struct sockaddr_vm new_addr;
632 
633 	if (!port)
634 		port = get_random_u32_above(LAST_RESERVED_PORT);
635 
636 	vsock_addr_init(&new_addr, addr->svm_cid, addr->svm_port);
637 
638 	if (addr->svm_port == VMADDR_PORT_ANY) {
639 		bool found = false;
640 		unsigned int i;
641 
642 		for (i = 0; i < MAX_PORT_RETRIES; i++) {
643 			if (port <= LAST_RESERVED_PORT)
644 				port = LAST_RESERVED_PORT + 1;
645 
646 			new_addr.svm_port = port++;
647 
648 			if (!__vsock_find_bound_socket(&new_addr)) {
649 				found = true;
650 				break;
651 			}
652 		}
653 
654 		if (!found)
655 			return -EADDRNOTAVAIL;
656 	} else {
657 		/* If port is in reserved range, ensure caller
658 		 * has necessary privileges.
659 		 */
660 		if (addr->svm_port <= LAST_RESERVED_PORT &&
661 		    !capable(CAP_NET_BIND_SERVICE)) {
662 			return -EACCES;
663 		}
664 
665 		if (__vsock_find_bound_socket(&new_addr))
666 			return -EADDRINUSE;
667 	}
668 
669 	vsock_addr_init(&vsk->local_addr, new_addr.svm_cid, new_addr.svm_port);
670 
671 	/* Remove connection oriented sockets from the unbound list and add them
672 	 * to the hash table for easy lookup by its address.  The unbound list
673 	 * is simply an extra entry at the end of the hash table, a trick used
674 	 * by AF_UNIX.
675 	 */
676 	__vsock_remove_bound(vsk);
677 	__vsock_insert_bound(vsock_bound_sockets(&vsk->local_addr), vsk);
678 
679 	return 0;
680 }
681 
682 static int __vsock_bind_dgram(struct vsock_sock *vsk,
683 			      struct sockaddr_vm *addr)
684 {
685 	return vsk->transport->dgram_bind(vsk, addr);
686 }
687 
688 static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr)
689 {
690 	struct vsock_sock *vsk = vsock_sk(sk);
691 	int retval;
692 
693 	/* First ensure this socket isn't already bound. */
694 	if (vsock_addr_bound(&vsk->local_addr))
695 		return -EINVAL;
696 
697 	/* Now bind to the provided address or select appropriate values if
698 	 * none are provided (VMADDR_CID_ANY and VMADDR_PORT_ANY).  Note that
699 	 * like AF_INET prevents binding to a non-local IP address (in most
700 	 * cases), we only allow binding to a local CID.
701 	 */
702 	if (addr->svm_cid != VMADDR_CID_ANY && !vsock_find_cid(addr->svm_cid))
703 		return -EADDRNOTAVAIL;
704 
705 	switch (sk->sk_socket->type) {
706 	case SOCK_STREAM:
707 	case SOCK_SEQPACKET:
708 		spin_lock_bh(&vsock_table_lock);
709 		retval = __vsock_bind_connectible(vsk, addr);
710 		spin_unlock_bh(&vsock_table_lock);
711 		break;
712 
713 	case SOCK_DGRAM:
714 		retval = __vsock_bind_dgram(vsk, addr);
715 		break;
716 
717 	default:
718 		retval = -EINVAL;
719 		break;
720 	}
721 
722 	return retval;
723 }
724 
725 static void vsock_connect_timeout(struct work_struct *work);
726 
727 static struct sock *__vsock_create(struct net *net,
728 				   struct socket *sock,
729 				   struct sock *parent,
730 				   gfp_t priority,
731 				   unsigned short type,
732 				   int kern)
733 {
734 	struct sock *sk;
735 	struct vsock_sock *psk;
736 	struct vsock_sock *vsk;
737 
738 	sk = sk_alloc(net, AF_VSOCK, priority, &vsock_proto, kern);
739 	if (!sk)
740 		return NULL;
741 
742 	sock_init_data(sock, sk);
743 
744 	/* sk->sk_type is normally set in sock_init_data, but only if sock is
745 	 * non-NULL. We make sure that our sockets always have a type by
746 	 * setting it here if needed.
747 	 */
748 	if (!sock)
749 		sk->sk_type = type;
750 
751 	vsk = vsock_sk(sk);
752 	vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
753 	vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
754 
755 	sk->sk_destruct = vsock_sk_destruct;
756 	sk->sk_backlog_rcv = vsock_queue_rcv_skb;
757 	sock_reset_flag(sk, SOCK_DONE);
758 
759 	INIT_LIST_HEAD(&vsk->bound_table);
760 	INIT_LIST_HEAD(&vsk->connected_table);
761 	vsk->listener = NULL;
762 	INIT_LIST_HEAD(&vsk->pending_links);
763 	INIT_LIST_HEAD(&vsk->accept_queue);
764 	vsk->rejected = false;
765 	vsk->sent_request = false;
766 	vsk->ignore_connecting_rst = false;
767 	vsk->peer_shutdown = 0;
768 	INIT_DELAYED_WORK(&vsk->connect_work, vsock_connect_timeout);
769 	INIT_DELAYED_WORK(&vsk->pending_work, vsock_pending_work);
770 
771 	psk = parent ? vsock_sk(parent) : NULL;
772 	if (parent) {
773 		vsk->trusted = psk->trusted;
774 		vsk->owner = get_cred(psk->owner);
775 		vsk->connect_timeout = psk->connect_timeout;
776 		vsk->buffer_size = psk->buffer_size;
777 		vsk->buffer_min_size = psk->buffer_min_size;
778 		vsk->buffer_max_size = psk->buffer_max_size;
779 		security_sk_clone(parent, sk);
780 	} else {
781 		vsk->trusted = ns_capable_noaudit(&init_user_ns, CAP_NET_ADMIN);
782 		vsk->owner = get_current_cred();
783 		vsk->connect_timeout = VSOCK_DEFAULT_CONNECT_TIMEOUT;
784 		vsk->buffer_size = VSOCK_DEFAULT_BUFFER_SIZE;
785 		vsk->buffer_min_size = VSOCK_DEFAULT_BUFFER_MIN_SIZE;
786 		vsk->buffer_max_size = VSOCK_DEFAULT_BUFFER_MAX_SIZE;
787 	}
788 
789 	return sk;
790 }
791 
792 static bool sock_type_connectible(u16 type)
793 {
794 	return (type == SOCK_STREAM) || (type == SOCK_SEQPACKET);
795 }
796 
797 static void __vsock_release(struct sock *sk, int level)
798 {
799 	if (sk) {
800 		struct sock *pending;
801 		struct vsock_sock *vsk;
802 
803 		vsk = vsock_sk(sk);
804 		pending = NULL;	/* Compiler warning. */
805 
806 		/* When "level" is SINGLE_DEPTH_NESTING, use the nested
807 		 * version to avoid the warning "possible recursive locking
808 		 * detected". When "level" is 0, lock_sock_nested(sk, level)
809 		 * is the same as lock_sock(sk).
810 		 */
811 		lock_sock_nested(sk, level);
812 
813 		if (vsk->transport)
814 			vsk->transport->release(vsk);
815 		else if (sock_type_connectible(sk->sk_type))
816 			vsock_remove_sock(vsk);
817 
818 		sock_orphan(sk);
819 		sk->sk_shutdown = SHUTDOWN_MASK;
820 
821 		skb_queue_purge(&sk->sk_receive_queue);
822 
823 		/* Clean up any sockets that never were accepted. */
824 		while ((pending = vsock_dequeue_accept(sk)) != NULL) {
825 			__vsock_release(pending, SINGLE_DEPTH_NESTING);
826 			sock_put(pending);
827 		}
828 
829 		release_sock(sk);
830 		sock_put(sk);
831 	}
832 }
833 
834 static void vsock_sk_destruct(struct sock *sk)
835 {
836 	struct vsock_sock *vsk = vsock_sk(sk);
837 
838 	vsock_deassign_transport(vsk);
839 
840 	/* When clearing these addresses, there's no need to set the family and
841 	 * possibly register the address family with the kernel.
842 	 */
843 	vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
844 	vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
845 
846 	put_cred(vsk->owner);
847 }
848 
849 static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
850 {
851 	int err;
852 
853 	err = sock_queue_rcv_skb(sk, skb);
854 	if (err)
855 		kfree_skb(skb);
856 
857 	return err;
858 }
859 
860 struct sock *vsock_create_connected(struct sock *parent)
861 {
862 	return __vsock_create(sock_net(parent), NULL, parent, GFP_KERNEL,
863 			      parent->sk_type, 0);
864 }
865 EXPORT_SYMBOL_GPL(vsock_create_connected);
866 
867 s64 vsock_stream_has_data(struct vsock_sock *vsk)
868 {
869 	return vsk->transport->stream_has_data(vsk);
870 }
871 EXPORT_SYMBOL_GPL(vsock_stream_has_data);
872 
873 s64 vsock_connectible_has_data(struct vsock_sock *vsk)
874 {
875 	struct sock *sk = sk_vsock(vsk);
876 
877 	if (sk->sk_type == SOCK_SEQPACKET)
878 		return vsk->transport->seqpacket_has_data(vsk);
879 	else
880 		return vsock_stream_has_data(vsk);
881 }
882 EXPORT_SYMBOL_GPL(vsock_connectible_has_data);
883 
884 s64 vsock_stream_has_space(struct vsock_sock *vsk)
885 {
886 	return vsk->transport->stream_has_space(vsk);
887 }
888 EXPORT_SYMBOL_GPL(vsock_stream_has_space);
889 
890 void vsock_data_ready(struct sock *sk)
891 {
892 	struct vsock_sock *vsk = vsock_sk(sk);
893 
894 	if (vsock_stream_has_data(vsk) >= sk->sk_rcvlowat ||
895 	    sock_flag(sk, SOCK_DONE))
896 		sk->sk_data_ready(sk);
897 }
898 EXPORT_SYMBOL_GPL(vsock_data_ready);
899 
900 static int vsock_release(struct socket *sock)
901 {
902 	__vsock_release(sock->sk, 0);
903 	sock->sk = NULL;
904 	sock->state = SS_FREE;
905 
906 	return 0;
907 }
908 
909 static int
910 vsock_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
911 {
912 	int err;
913 	struct sock *sk;
914 	struct sockaddr_vm *vm_addr;
915 
916 	sk = sock->sk;
917 
918 	if (vsock_addr_cast(addr, addr_len, &vm_addr) != 0)
919 		return -EINVAL;
920 
921 	lock_sock(sk);
922 	err = __vsock_bind(sk, vm_addr);
923 	release_sock(sk);
924 
925 	return err;
926 }
927 
928 static int vsock_getname(struct socket *sock,
929 			 struct sockaddr *addr, int peer)
930 {
931 	int err;
932 	struct sock *sk;
933 	struct vsock_sock *vsk;
934 	struct sockaddr_vm *vm_addr;
935 
936 	sk = sock->sk;
937 	vsk = vsock_sk(sk);
938 	err = 0;
939 
940 	lock_sock(sk);
941 
942 	if (peer) {
943 		if (sock->state != SS_CONNECTED) {
944 			err = -ENOTCONN;
945 			goto out;
946 		}
947 		vm_addr = &vsk->remote_addr;
948 	} else {
949 		vm_addr = &vsk->local_addr;
950 	}
951 
952 	if (!vm_addr) {
953 		err = -EINVAL;
954 		goto out;
955 	}
956 
957 	/* sys_getsockname() and sys_getpeername() pass us a
958 	 * MAX_SOCK_ADDR-sized buffer and don't set addr_len.  Unfortunately
959 	 * that macro is defined in socket.c instead of .h, so we hardcode its
960 	 * value here.
961 	 */
962 	BUILD_BUG_ON(sizeof(*vm_addr) > 128);
963 	memcpy(addr, vm_addr, sizeof(*vm_addr));
964 	err = sizeof(*vm_addr);
965 
966 out:
967 	release_sock(sk);
968 	return err;
969 }
970 
971 static int vsock_shutdown(struct socket *sock, int mode)
972 {
973 	int err;
974 	struct sock *sk;
975 
976 	/* User level uses SHUT_RD (0) and SHUT_WR (1), but the kernel uses
977 	 * RCV_SHUTDOWN (1) and SEND_SHUTDOWN (2), so we must increment mode
978 	 * here like the other address families do.  Note also that the
979 	 * increment makes SHUT_RDWR (2) into RCV_SHUTDOWN | SEND_SHUTDOWN (3),
980 	 * which is what we want.
981 	 */
982 	mode++;
983 
984 	if ((mode & ~SHUTDOWN_MASK) || !mode)
985 		return -EINVAL;
986 
987 	/* If this is a connection oriented socket and it is not connected then
988 	 * bail out immediately.  If it is a DGRAM socket then we must first
989 	 * kick the socket so that it wakes up from any sleeping calls, for
990 	 * example recv(), and then afterwards return the error.
991 	 */
992 
993 	sk = sock->sk;
994 
995 	lock_sock(sk);
996 	if (sock->state == SS_UNCONNECTED) {
997 		err = -ENOTCONN;
998 		if (sock_type_connectible(sk->sk_type))
999 			goto out;
1000 	} else {
1001 		sock->state = SS_DISCONNECTING;
1002 		err = 0;
1003 	}
1004 
1005 	/* Receive and send shutdowns are treated alike. */
1006 	mode = mode & (RCV_SHUTDOWN | SEND_SHUTDOWN);
1007 	if (mode) {
1008 		sk->sk_shutdown |= mode;
1009 		sk->sk_state_change(sk);
1010 
1011 		if (sock_type_connectible(sk->sk_type)) {
1012 			sock_reset_flag(sk, SOCK_DONE);
1013 			vsock_send_shutdown(sk, mode);
1014 		}
1015 	}
1016 
1017 out:
1018 	release_sock(sk);
1019 	return err;
1020 }
1021 
1022 static __poll_t vsock_poll(struct file *file, struct socket *sock,
1023 			       poll_table *wait)
1024 {
1025 	struct sock *sk;
1026 	__poll_t mask;
1027 	struct vsock_sock *vsk;
1028 
1029 	sk = sock->sk;
1030 	vsk = vsock_sk(sk);
1031 
1032 	poll_wait(file, sk_sleep(sk), wait);
1033 	mask = 0;
1034 
1035 	if (sk->sk_err || !skb_queue_empty_lockless(&sk->sk_error_queue))
1036 		/* Signify that there has been an error on this socket. */
1037 		mask |= EPOLLERR;
1038 
1039 	/* INET sockets treat local write shutdown and peer write shutdown as a
1040 	 * case of EPOLLHUP set.
1041 	 */
1042 	if ((sk->sk_shutdown == SHUTDOWN_MASK) ||
1043 	    ((sk->sk_shutdown & SEND_SHUTDOWN) &&
1044 	     (vsk->peer_shutdown & SEND_SHUTDOWN))) {
1045 		mask |= EPOLLHUP;
1046 	}
1047 
1048 	if (sk->sk_shutdown & RCV_SHUTDOWN ||
1049 	    vsk->peer_shutdown & SEND_SHUTDOWN) {
1050 		mask |= EPOLLRDHUP;
1051 	}
1052 
1053 	if (sock->type == SOCK_DGRAM) {
1054 		/* For datagram sockets we can read if there is something in
1055 		 * the queue and write as long as the socket isn't shutdown for
1056 		 * sending.
1057 		 */
1058 		if (!skb_queue_empty_lockless(&sk->sk_receive_queue) ||
1059 		    (sk->sk_shutdown & RCV_SHUTDOWN)) {
1060 			mask |= EPOLLIN | EPOLLRDNORM;
1061 		}
1062 
1063 		if (!(sk->sk_shutdown & SEND_SHUTDOWN))
1064 			mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND;
1065 
1066 	} else if (sock_type_connectible(sk->sk_type)) {
1067 		const struct vsock_transport *transport;
1068 
1069 		lock_sock(sk);
1070 
1071 		transport = vsk->transport;
1072 
1073 		/* Listening sockets that have connections in their accept
1074 		 * queue can be read.
1075 		 */
1076 		if (sk->sk_state == TCP_LISTEN
1077 		    && !vsock_is_accept_queue_empty(sk))
1078 			mask |= EPOLLIN | EPOLLRDNORM;
1079 
1080 		/* If there is something in the queue then we can read. */
1081 		if (transport && transport->stream_is_active(vsk) &&
1082 		    !(sk->sk_shutdown & RCV_SHUTDOWN)) {
1083 			bool data_ready_now = false;
1084 			int target = sock_rcvlowat(sk, 0, INT_MAX);
1085 			int ret = transport->notify_poll_in(
1086 					vsk, target, &data_ready_now);
1087 			if (ret < 0) {
1088 				mask |= EPOLLERR;
1089 			} else {
1090 				if (data_ready_now)
1091 					mask |= EPOLLIN | EPOLLRDNORM;
1092 
1093 			}
1094 		}
1095 
1096 		/* Sockets whose connections have been closed, reset, or
1097 		 * terminated should also be considered read, and we check the
1098 		 * shutdown flag for that.
1099 		 */
1100 		if (sk->sk_shutdown & RCV_SHUTDOWN ||
1101 		    vsk->peer_shutdown & SEND_SHUTDOWN) {
1102 			mask |= EPOLLIN | EPOLLRDNORM;
1103 		}
1104 
1105 		/* Connected sockets that can produce data can be written. */
1106 		if (transport && sk->sk_state == TCP_ESTABLISHED) {
1107 			if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1108 				bool space_avail_now = false;
1109 				int ret = transport->notify_poll_out(
1110 						vsk, 1, &space_avail_now);
1111 				if (ret < 0) {
1112 					mask |= EPOLLERR;
1113 				} else {
1114 					if (space_avail_now)
1115 						/* Remove EPOLLWRBAND since INET
1116 						 * sockets are not setting it.
1117 						 */
1118 						mask |= EPOLLOUT | EPOLLWRNORM;
1119 
1120 				}
1121 			}
1122 		}
1123 
1124 		/* Simulate INET socket poll behaviors, which sets
1125 		 * EPOLLOUT|EPOLLWRNORM when peer is closed and nothing to read,
1126 		 * but local send is not shutdown.
1127 		 */
1128 		if (sk->sk_state == TCP_CLOSE || sk->sk_state == TCP_CLOSING) {
1129 			if (!(sk->sk_shutdown & SEND_SHUTDOWN))
1130 				mask |= EPOLLOUT | EPOLLWRNORM;
1131 
1132 		}
1133 
1134 		release_sock(sk);
1135 	}
1136 
1137 	return mask;
1138 }
1139 
1140 static int vsock_read_skb(struct sock *sk, skb_read_actor_t read_actor)
1141 {
1142 	struct vsock_sock *vsk = vsock_sk(sk);
1143 
1144 	return vsk->transport->read_skb(vsk, read_actor);
1145 }
1146 
1147 static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg,
1148 			       size_t len)
1149 {
1150 	int err;
1151 	struct sock *sk;
1152 	struct vsock_sock *vsk;
1153 	struct sockaddr_vm *remote_addr;
1154 	const struct vsock_transport *transport;
1155 
1156 	if (msg->msg_flags & MSG_OOB)
1157 		return -EOPNOTSUPP;
1158 
1159 	/* For now, MSG_DONTWAIT is always assumed... */
1160 	err = 0;
1161 	sk = sock->sk;
1162 	vsk = vsock_sk(sk);
1163 
1164 	lock_sock(sk);
1165 
1166 	transport = vsk->transport;
1167 
1168 	err = vsock_auto_bind(vsk);
1169 	if (err)
1170 		goto out;
1171 
1172 
1173 	/* If the provided message contains an address, use that.  Otherwise
1174 	 * fall back on the socket's remote handle (if it has been connected).
1175 	 */
1176 	if (msg->msg_name &&
1177 	    vsock_addr_cast(msg->msg_name, msg->msg_namelen,
1178 			    &remote_addr) == 0) {
1179 		/* Ensure this address is of the right type and is a valid
1180 		 * destination.
1181 		 */
1182 
1183 		if (remote_addr->svm_cid == VMADDR_CID_ANY)
1184 			remote_addr->svm_cid = transport->get_local_cid();
1185 
1186 		if (!vsock_addr_bound(remote_addr)) {
1187 			err = -EINVAL;
1188 			goto out;
1189 		}
1190 	} else if (sock->state == SS_CONNECTED) {
1191 		remote_addr = &vsk->remote_addr;
1192 
1193 		if (remote_addr->svm_cid == VMADDR_CID_ANY)
1194 			remote_addr->svm_cid = transport->get_local_cid();
1195 
1196 		/* XXX Should connect() or this function ensure remote_addr is
1197 		 * bound?
1198 		 */
1199 		if (!vsock_addr_bound(&vsk->remote_addr)) {
1200 			err = -EINVAL;
1201 			goto out;
1202 		}
1203 	} else {
1204 		err = -EINVAL;
1205 		goto out;
1206 	}
1207 
1208 	if (!transport->dgram_allow(remote_addr->svm_cid,
1209 				    remote_addr->svm_port)) {
1210 		err = -EINVAL;
1211 		goto out;
1212 	}
1213 
1214 	err = transport->dgram_enqueue(vsk, remote_addr, msg, len);
1215 
1216 out:
1217 	release_sock(sk);
1218 	return err;
1219 }
1220 
1221 static int vsock_dgram_connect(struct socket *sock,
1222 			       struct sockaddr *addr, int addr_len, int flags)
1223 {
1224 	int err;
1225 	struct sock *sk;
1226 	struct vsock_sock *vsk;
1227 	struct sockaddr_vm *remote_addr;
1228 
1229 	sk = sock->sk;
1230 	vsk = vsock_sk(sk);
1231 
1232 	err = vsock_addr_cast(addr, addr_len, &remote_addr);
1233 	if (err == -EAFNOSUPPORT && remote_addr->svm_family == AF_UNSPEC) {
1234 		lock_sock(sk);
1235 		vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY,
1236 				VMADDR_PORT_ANY);
1237 		sock->state = SS_UNCONNECTED;
1238 		release_sock(sk);
1239 		return 0;
1240 	} else if (err != 0)
1241 		return -EINVAL;
1242 
1243 	lock_sock(sk);
1244 
1245 	err = vsock_auto_bind(vsk);
1246 	if (err)
1247 		goto out;
1248 
1249 	if (!vsk->transport->dgram_allow(remote_addr->svm_cid,
1250 					 remote_addr->svm_port)) {
1251 		err = -EINVAL;
1252 		goto out;
1253 	}
1254 
1255 	memcpy(&vsk->remote_addr, remote_addr, sizeof(vsk->remote_addr));
1256 	sock->state = SS_CONNECTED;
1257 
1258 	/* sock map disallows redirection of non-TCP sockets with sk_state !=
1259 	 * TCP_ESTABLISHED (see sock_map_redirect_allowed()), so we set
1260 	 * TCP_ESTABLISHED here to allow redirection of connected vsock dgrams.
1261 	 *
1262 	 * This doesn't seem to be abnormal state for datagram sockets, as the
1263 	 * same approach can be see in other datagram socket types as well
1264 	 * (such as unix sockets).
1265 	 */
1266 	sk->sk_state = TCP_ESTABLISHED;
1267 
1268 out:
1269 	release_sock(sk);
1270 	return err;
1271 }
1272 
1273 int __vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg,
1274 			  size_t len, int flags)
1275 {
1276 	struct sock *sk = sock->sk;
1277 	struct vsock_sock *vsk = vsock_sk(sk);
1278 
1279 	return vsk->transport->dgram_dequeue(vsk, msg, len, flags);
1280 }
1281 
1282 int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg,
1283 			size_t len, int flags)
1284 {
1285 #ifdef CONFIG_BPF_SYSCALL
1286 	struct sock *sk = sock->sk;
1287 	const struct proto *prot;
1288 
1289 	prot = READ_ONCE(sk->sk_prot);
1290 	if (prot != &vsock_proto)
1291 		return prot->recvmsg(sk, msg, len, flags, NULL);
1292 #endif
1293 
1294 	return __vsock_dgram_recvmsg(sock, msg, len, flags);
1295 }
1296 EXPORT_SYMBOL_GPL(vsock_dgram_recvmsg);
1297 
1298 static const struct proto_ops vsock_dgram_ops = {
1299 	.family = PF_VSOCK,
1300 	.owner = THIS_MODULE,
1301 	.release = vsock_release,
1302 	.bind = vsock_bind,
1303 	.connect = vsock_dgram_connect,
1304 	.socketpair = sock_no_socketpair,
1305 	.accept = sock_no_accept,
1306 	.getname = vsock_getname,
1307 	.poll = vsock_poll,
1308 	.ioctl = sock_no_ioctl,
1309 	.listen = sock_no_listen,
1310 	.shutdown = vsock_shutdown,
1311 	.sendmsg = vsock_dgram_sendmsg,
1312 	.recvmsg = vsock_dgram_recvmsg,
1313 	.mmap = sock_no_mmap,
1314 	.read_skb = vsock_read_skb,
1315 };
1316 
1317 static int vsock_transport_cancel_pkt(struct vsock_sock *vsk)
1318 {
1319 	const struct vsock_transport *transport = vsk->transport;
1320 
1321 	if (!transport || !transport->cancel_pkt)
1322 		return -EOPNOTSUPP;
1323 
1324 	return transport->cancel_pkt(vsk);
1325 }
1326 
1327 static void vsock_connect_timeout(struct work_struct *work)
1328 {
1329 	struct sock *sk;
1330 	struct vsock_sock *vsk;
1331 
1332 	vsk = container_of(work, struct vsock_sock, connect_work.work);
1333 	sk = sk_vsock(vsk);
1334 
1335 	lock_sock(sk);
1336 	if (sk->sk_state == TCP_SYN_SENT &&
1337 	    (sk->sk_shutdown != SHUTDOWN_MASK)) {
1338 		sk->sk_state = TCP_CLOSE;
1339 		sk->sk_socket->state = SS_UNCONNECTED;
1340 		sk->sk_err = ETIMEDOUT;
1341 		sk_error_report(sk);
1342 		vsock_transport_cancel_pkt(vsk);
1343 	}
1344 	release_sock(sk);
1345 
1346 	sock_put(sk);
1347 }
1348 
1349 static int vsock_connect(struct socket *sock, struct sockaddr *addr,
1350 			 int addr_len, int flags)
1351 {
1352 	int err;
1353 	struct sock *sk;
1354 	struct vsock_sock *vsk;
1355 	const struct vsock_transport *transport;
1356 	struct sockaddr_vm *remote_addr;
1357 	long timeout;
1358 	DEFINE_WAIT(wait);
1359 
1360 	err = 0;
1361 	sk = sock->sk;
1362 	vsk = vsock_sk(sk);
1363 
1364 	lock_sock(sk);
1365 
1366 	/* XXX AF_UNSPEC should make us disconnect like AF_INET. */
1367 	switch (sock->state) {
1368 	case SS_CONNECTED:
1369 		err = -EISCONN;
1370 		goto out;
1371 	case SS_DISCONNECTING:
1372 		err = -EINVAL;
1373 		goto out;
1374 	case SS_CONNECTING:
1375 		/* This continues on so we can move sock into the SS_CONNECTED
1376 		 * state once the connection has completed (at which point err
1377 		 * will be set to zero also).  Otherwise, we will either wait
1378 		 * for the connection or return -EALREADY should this be a
1379 		 * non-blocking call.
1380 		 */
1381 		err = -EALREADY;
1382 		if (flags & O_NONBLOCK)
1383 			goto out;
1384 		break;
1385 	default:
1386 		if ((sk->sk_state == TCP_LISTEN) ||
1387 		    vsock_addr_cast(addr, addr_len, &remote_addr) != 0) {
1388 			err = -EINVAL;
1389 			goto out;
1390 		}
1391 
1392 		/* Set the remote address that we are connecting to. */
1393 		memcpy(&vsk->remote_addr, remote_addr,
1394 		       sizeof(vsk->remote_addr));
1395 
1396 		err = vsock_assign_transport(vsk, NULL);
1397 		if (err)
1398 			goto out;
1399 
1400 		transport = vsk->transport;
1401 
1402 		/* The hypervisor and well-known contexts do not have socket
1403 		 * endpoints.
1404 		 */
1405 		if (!transport ||
1406 		    !transport->stream_allow(remote_addr->svm_cid,
1407 					     remote_addr->svm_port)) {
1408 			err = -ENETUNREACH;
1409 			goto out;
1410 		}
1411 
1412 		if (vsock_msgzerocopy_allow(transport)) {
1413 			set_bit(SOCK_SUPPORT_ZC, &sk->sk_socket->flags);
1414 		} else if (sock_flag(sk, SOCK_ZEROCOPY)) {
1415 			/* If this option was set before 'connect()',
1416 			 * when transport was unknown, check that this
1417 			 * feature is supported here.
1418 			 */
1419 			err = -EOPNOTSUPP;
1420 			goto out;
1421 		}
1422 
1423 		err = vsock_auto_bind(vsk);
1424 		if (err)
1425 			goto out;
1426 
1427 		sk->sk_state = TCP_SYN_SENT;
1428 
1429 		err = transport->connect(vsk);
1430 		if (err < 0)
1431 			goto out;
1432 
1433 		/* Mark sock as connecting and set the error code to in
1434 		 * progress in case this is a non-blocking connect.
1435 		 */
1436 		sock->state = SS_CONNECTING;
1437 		err = -EINPROGRESS;
1438 	}
1439 
1440 	/* The receive path will handle all communication until we are able to
1441 	 * enter the connected state.  Here we wait for the connection to be
1442 	 * completed or a notification of an error.
1443 	 */
1444 	timeout = vsk->connect_timeout;
1445 	prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1446 
1447 	while (sk->sk_state != TCP_ESTABLISHED && sk->sk_err == 0) {
1448 		if (flags & O_NONBLOCK) {
1449 			/* If we're not going to block, we schedule a timeout
1450 			 * function to generate a timeout on the connection
1451 			 * attempt, in case the peer doesn't respond in a
1452 			 * timely manner. We hold on to the socket until the
1453 			 * timeout fires.
1454 			 */
1455 			sock_hold(sk);
1456 
1457 			/* If the timeout function is already scheduled,
1458 			 * reschedule it, then ungrab the socket refcount to
1459 			 * keep it balanced.
1460 			 */
1461 			if (mod_delayed_work(system_wq, &vsk->connect_work,
1462 					     timeout))
1463 				sock_put(sk);
1464 
1465 			/* Skip ahead to preserve error code set above. */
1466 			goto out_wait;
1467 		}
1468 
1469 		release_sock(sk);
1470 		timeout = schedule_timeout(timeout);
1471 		lock_sock(sk);
1472 
1473 		if (signal_pending(current)) {
1474 			err = sock_intr_errno(timeout);
1475 			sk->sk_state = sk->sk_state == TCP_ESTABLISHED ? TCP_CLOSING : TCP_CLOSE;
1476 			sock->state = SS_UNCONNECTED;
1477 			vsock_transport_cancel_pkt(vsk);
1478 			vsock_remove_connected(vsk);
1479 			goto out_wait;
1480 		} else if ((sk->sk_state != TCP_ESTABLISHED) && (timeout == 0)) {
1481 			err = -ETIMEDOUT;
1482 			sk->sk_state = TCP_CLOSE;
1483 			sock->state = SS_UNCONNECTED;
1484 			vsock_transport_cancel_pkt(vsk);
1485 			goto out_wait;
1486 		}
1487 
1488 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1489 	}
1490 
1491 	if (sk->sk_err) {
1492 		err = -sk->sk_err;
1493 		sk->sk_state = TCP_CLOSE;
1494 		sock->state = SS_UNCONNECTED;
1495 	} else {
1496 		err = 0;
1497 	}
1498 
1499 out_wait:
1500 	finish_wait(sk_sleep(sk), &wait);
1501 out:
1502 	release_sock(sk);
1503 	return err;
1504 }
1505 
1506 static int vsock_accept(struct socket *sock, struct socket *newsock,
1507 			struct proto_accept_arg *arg)
1508 {
1509 	struct sock *listener;
1510 	int err;
1511 	struct sock *connected;
1512 	struct vsock_sock *vconnected;
1513 	long timeout;
1514 	DEFINE_WAIT(wait);
1515 
1516 	err = 0;
1517 	listener = sock->sk;
1518 
1519 	lock_sock(listener);
1520 
1521 	if (!sock_type_connectible(sock->type)) {
1522 		err = -EOPNOTSUPP;
1523 		goto out;
1524 	}
1525 
1526 	if (listener->sk_state != TCP_LISTEN) {
1527 		err = -EINVAL;
1528 		goto out;
1529 	}
1530 
1531 	/* Wait for children sockets to appear; these are the new sockets
1532 	 * created upon connection establishment.
1533 	 */
1534 	timeout = sock_rcvtimeo(listener, arg->flags & O_NONBLOCK);
1535 	prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1536 
1537 	while ((connected = vsock_dequeue_accept(listener)) == NULL &&
1538 	       listener->sk_err == 0) {
1539 		release_sock(listener);
1540 		timeout = schedule_timeout(timeout);
1541 		finish_wait(sk_sleep(listener), &wait);
1542 		lock_sock(listener);
1543 
1544 		if (signal_pending(current)) {
1545 			err = sock_intr_errno(timeout);
1546 			goto out;
1547 		} else if (timeout == 0) {
1548 			err = -EAGAIN;
1549 			goto out;
1550 		}
1551 
1552 		prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1553 	}
1554 	finish_wait(sk_sleep(listener), &wait);
1555 
1556 	if (listener->sk_err)
1557 		err = -listener->sk_err;
1558 
1559 	if (connected) {
1560 		sk_acceptq_removed(listener);
1561 
1562 		lock_sock_nested(connected, SINGLE_DEPTH_NESTING);
1563 		vconnected = vsock_sk(connected);
1564 
1565 		/* If the listener socket has received an error, then we should
1566 		 * reject this socket and return.  Note that we simply mark the
1567 		 * socket rejected, drop our reference, and let the cleanup
1568 		 * function handle the cleanup; the fact that we found it in
1569 		 * the listener's accept queue guarantees that the cleanup
1570 		 * function hasn't run yet.
1571 		 */
1572 		if (err) {
1573 			vconnected->rejected = true;
1574 		} else {
1575 			newsock->state = SS_CONNECTED;
1576 			sock_graft(connected, newsock);
1577 			if (vsock_msgzerocopy_allow(vconnected->transport))
1578 				set_bit(SOCK_SUPPORT_ZC,
1579 					&connected->sk_socket->flags);
1580 		}
1581 
1582 		release_sock(connected);
1583 		sock_put(connected);
1584 	}
1585 
1586 out:
1587 	release_sock(listener);
1588 	return err;
1589 }
1590 
1591 static int vsock_listen(struct socket *sock, int backlog)
1592 {
1593 	int err;
1594 	struct sock *sk;
1595 	struct vsock_sock *vsk;
1596 
1597 	sk = sock->sk;
1598 
1599 	lock_sock(sk);
1600 
1601 	if (!sock_type_connectible(sk->sk_type)) {
1602 		err = -EOPNOTSUPP;
1603 		goto out;
1604 	}
1605 
1606 	if (sock->state != SS_UNCONNECTED) {
1607 		err = -EINVAL;
1608 		goto out;
1609 	}
1610 
1611 	vsk = vsock_sk(sk);
1612 
1613 	if (!vsock_addr_bound(&vsk->local_addr)) {
1614 		err = -EINVAL;
1615 		goto out;
1616 	}
1617 
1618 	sk->sk_max_ack_backlog = backlog;
1619 	sk->sk_state = TCP_LISTEN;
1620 
1621 	err = 0;
1622 
1623 out:
1624 	release_sock(sk);
1625 	return err;
1626 }
1627 
1628 static void vsock_update_buffer_size(struct vsock_sock *vsk,
1629 				     const struct vsock_transport *transport,
1630 				     u64 val)
1631 {
1632 	if (val > vsk->buffer_max_size)
1633 		val = vsk->buffer_max_size;
1634 
1635 	if (val < vsk->buffer_min_size)
1636 		val = vsk->buffer_min_size;
1637 
1638 	if (val != vsk->buffer_size &&
1639 	    transport && transport->notify_buffer_size)
1640 		transport->notify_buffer_size(vsk, &val);
1641 
1642 	vsk->buffer_size = val;
1643 }
1644 
1645 static int vsock_connectible_setsockopt(struct socket *sock,
1646 					int level,
1647 					int optname,
1648 					sockptr_t optval,
1649 					unsigned int optlen)
1650 {
1651 	int err;
1652 	struct sock *sk;
1653 	struct vsock_sock *vsk;
1654 	const struct vsock_transport *transport;
1655 	u64 val;
1656 
1657 	if (level != AF_VSOCK && level != SOL_SOCKET)
1658 		return -ENOPROTOOPT;
1659 
1660 #define COPY_IN(_v)                                       \
1661 	do {						  \
1662 		if (optlen < sizeof(_v)) {		  \
1663 			err = -EINVAL;			  \
1664 			goto exit;			  \
1665 		}					  \
1666 		if (copy_from_sockptr(&_v, optval, sizeof(_v)) != 0) {	\
1667 			err = -EFAULT;					\
1668 			goto exit;					\
1669 		}							\
1670 	} while (0)
1671 
1672 	err = 0;
1673 	sk = sock->sk;
1674 	vsk = vsock_sk(sk);
1675 
1676 	lock_sock(sk);
1677 
1678 	transport = vsk->transport;
1679 
1680 	if (level == SOL_SOCKET) {
1681 		int zerocopy;
1682 
1683 		if (optname != SO_ZEROCOPY) {
1684 			release_sock(sk);
1685 			return sock_setsockopt(sock, level, optname, optval, optlen);
1686 		}
1687 
1688 		/* Use 'int' type here, because variable to
1689 		 * set this option usually has this type.
1690 		 */
1691 		COPY_IN(zerocopy);
1692 
1693 		if (zerocopy < 0 || zerocopy > 1) {
1694 			err = -EINVAL;
1695 			goto exit;
1696 		}
1697 
1698 		if (transport && !vsock_msgzerocopy_allow(transport)) {
1699 			err = -EOPNOTSUPP;
1700 			goto exit;
1701 		}
1702 
1703 		sock_valbool_flag(sk, SOCK_ZEROCOPY, zerocopy);
1704 		goto exit;
1705 	}
1706 
1707 	switch (optname) {
1708 	case SO_VM_SOCKETS_BUFFER_SIZE:
1709 		COPY_IN(val);
1710 		vsock_update_buffer_size(vsk, transport, val);
1711 		break;
1712 
1713 	case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1714 		COPY_IN(val);
1715 		vsk->buffer_max_size = val;
1716 		vsock_update_buffer_size(vsk, transport, vsk->buffer_size);
1717 		break;
1718 
1719 	case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1720 		COPY_IN(val);
1721 		vsk->buffer_min_size = val;
1722 		vsock_update_buffer_size(vsk, transport, vsk->buffer_size);
1723 		break;
1724 
1725 	case SO_VM_SOCKETS_CONNECT_TIMEOUT_NEW:
1726 	case SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD: {
1727 		struct __kernel_sock_timeval tv;
1728 
1729 		err = sock_copy_user_timeval(&tv, optval, optlen,
1730 					     optname == SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD);
1731 		if (err)
1732 			break;
1733 		if (tv.tv_sec >= 0 && tv.tv_usec < USEC_PER_SEC &&
1734 		    tv.tv_sec < (MAX_SCHEDULE_TIMEOUT / HZ - 1)) {
1735 			vsk->connect_timeout = tv.tv_sec * HZ +
1736 				DIV_ROUND_UP((unsigned long)tv.tv_usec, (USEC_PER_SEC / HZ));
1737 			if (vsk->connect_timeout == 0)
1738 				vsk->connect_timeout =
1739 				    VSOCK_DEFAULT_CONNECT_TIMEOUT;
1740 
1741 		} else {
1742 			err = -ERANGE;
1743 		}
1744 		break;
1745 	}
1746 
1747 	default:
1748 		err = -ENOPROTOOPT;
1749 		break;
1750 	}
1751 
1752 #undef COPY_IN
1753 
1754 exit:
1755 	release_sock(sk);
1756 	return err;
1757 }
1758 
1759 static int vsock_connectible_getsockopt(struct socket *sock,
1760 					int level, int optname,
1761 					char __user *optval,
1762 					int __user *optlen)
1763 {
1764 	struct sock *sk = sock->sk;
1765 	struct vsock_sock *vsk = vsock_sk(sk);
1766 
1767 	union {
1768 		u64 val64;
1769 		struct old_timeval32 tm32;
1770 		struct __kernel_old_timeval tm;
1771 		struct  __kernel_sock_timeval stm;
1772 	} v;
1773 
1774 	int lv = sizeof(v.val64);
1775 	int len;
1776 
1777 	if (level != AF_VSOCK)
1778 		return -ENOPROTOOPT;
1779 
1780 	if (get_user(len, optlen))
1781 		return -EFAULT;
1782 
1783 	memset(&v, 0, sizeof(v));
1784 
1785 	switch (optname) {
1786 	case SO_VM_SOCKETS_BUFFER_SIZE:
1787 		v.val64 = vsk->buffer_size;
1788 		break;
1789 
1790 	case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1791 		v.val64 = vsk->buffer_max_size;
1792 		break;
1793 
1794 	case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1795 		v.val64 = vsk->buffer_min_size;
1796 		break;
1797 
1798 	case SO_VM_SOCKETS_CONNECT_TIMEOUT_NEW:
1799 	case SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD:
1800 		lv = sock_get_timeout(vsk->connect_timeout, &v,
1801 				      optname == SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD);
1802 		break;
1803 
1804 	default:
1805 		return -ENOPROTOOPT;
1806 	}
1807 
1808 	if (len < lv)
1809 		return -EINVAL;
1810 	if (len > lv)
1811 		len = lv;
1812 	if (copy_to_user(optval, &v, len))
1813 		return -EFAULT;
1814 
1815 	if (put_user(len, optlen))
1816 		return -EFAULT;
1817 
1818 	return 0;
1819 }
1820 
1821 static int vsock_connectible_sendmsg(struct socket *sock, struct msghdr *msg,
1822 				     size_t len)
1823 {
1824 	struct sock *sk;
1825 	struct vsock_sock *vsk;
1826 	const struct vsock_transport *transport;
1827 	ssize_t total_written;
1828 	long timeout;
1829 	int err;
1830 	struct vsock_transport_send_notify_data send_data;
1831 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
1832 
1833 	sk = sock->sk;
1834 	vsk = vsock_sk(sk);
1835 	total_written = 0;
1836 	err = 0;
1837 
1838 	if (msg->msg_flags & MSG_OOB)
1839 		return -EOPNOTSUPP;
1840 
1841 	lock_sock(sk);
1842 
1843 	transport = vsk->transport;
1844 
1845 	/* Callers should not provide a destination with connection oriented
1846 	 * sockets.
1847 	 */
1848 	if (msg->msg_namelen) {
1849 		err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP;
1850 		goto out;
1851 	}
1852 
1853 	/* Send data only if both sides are not shutdown in the direction. */
1854 	if (sk->sk_shutdown & SEND_SHUTDOWN ||
1855 	    vsk->peer_shutdown & RCV_SHUTDOWN) {
1856 		err = -EPIPE;
1857 		goto out;
1858 	}
1859 
1860 	if (!transport || sk->sk_state != TCP_ESTABLISHED ||
1861 	    !vsock_addr_bound(&vsk->local_addr)) {
1862 		err = -ENOTCONN;
1863 		goto out;
1864 	}
1865 
1866 	if (!vsock_addr_bound(&vsk->remote_addr)) {
1867 		err = -EDESTADDRREQ;
1868 		goto out;
1869 	}
1870 
1871 	if (msg->msg_flags & MSG_ZEROCOPY &&
1872 	    !vsock_msgzerocopy_allow(transport)) {
1873 		err = -EOPNOTSUPP;
1874 		goto out;
1875 	}
1876 
1877 	/* Wait for room in the produce queue to enqueue our user's data. */
1878 	timeout = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1879 
1880 	err = transport->notify_send_init(vsk, &send_data);
1881 	if (err < 0)
1882 		goto out;
1883 
1884 	while (total_written < len) {
1885 		ssize_t written;
1886 
1887 		add_wait_queue(sk_sleep(sk), &wait);
1888 		while (vsock_stream_has_space(vsk) == 0 &&
1889 		       sk->sk_err == 0 &&
1890 		       !(sk->sk_shutdown & SEND_SHUTDOWN) &&
1891 		       !(vsk->peer_shutdown & RCV_SHUTDOWN)) {
1892 
1893 			/* Don't wait for non-blocking sockets. */
1894 			if (timeout == 0) {
1895 				err = -EAGAIN;
1896 				remove_wait_queue(sk_sleep(sk), &wait);
1897 				goto out_err;
1898 			}
1899 
1900 			err = transport->notify_send_pre_block(vsk, &send_data);
1901 			if (err < 0) {
1902 				remove_wait_queue(sk_sleep(sk), &wait);
1903 				goto out_err;
1904 			}
1905 
1906 			release_sock(sk);
1907 			timeout = wait_woken(&wait, TASK_INTERRUPTIBLE, timeout);
1908 			lock_sock(sk);
1909 			if (signal_pending(current)) {
1910 				err = sock_intr_errno(timeout);
1911 				remove_wait_queue(sk_sleep(sk), &wait);
1912 				goto out_err;
1913 			} else if (timeout == 0) {
1914 				err = -EAGAIN;
1915 				remove_wait_queue(sk_sleep(sk), &wait);
1916 				goto out_err;
1917 			}
1918 		}
1919 		remove_wait_queue(sk_sleep(sk), &wait);
1920 
1921 		/* These checks occur both as part of and after the loop
1922 		 * conditional since we need to check before and after
1923 		 * sleeping.
1924 		 */
1925 		if (sk->sk_err) {
1926 			err = -sk->sk_err;
1927 			goto out_err;
1928 		} else if ((sk->sk_shutdown & SEND_SHUTDOWN) ||
1929 			   (vsk->peer_shutdown & RCV_SHUTDOWN)) {
1930 			err = -EPIPE;
1931 			goto out_err;
1932 		}
1933 
1934 		err = transport->notify_send_pre_enqueue(vsk, &send_data);
1935 		if (err < 0)
1936 			goto out_err;
1937 
1938 		/* Note that enqueue will only write as many bytes as are free
1939 		 * in the produce queue, so we don't need to ensure len is
1940 		 * smaller than the queue size.  It is the caller's
1941 		 * responsibility to check how many bytes we were able to send.
1942 		 */
1943 
1944 		if (sk->sk_type == SOCK_SEQPACKET) {
1945 			written = transport->seqpacket_enqueue(vsk,
1946 						msg, len - total_written);
1947 		} else {
1948 			written = transport->stream_enqueue(vsk,
1949 					msg, len - total_written);
1950 		}
1951 
1952 		if (written < 0) {
1953 			err = written;
1954 			goto out_err;
1955 		}
1956 
1957 		total_written += written;
1958 
1959 		err = transport->notify_send_post_enqueue(
1960 				vsk, written, &send_data);
1961 		if (err < 0)
1962 			goto out_err;
1963 
1964 	}
1965 
1966 out_err:
1967 	if (total_written > 0) {
1968 		/* Return number of written bytes only if:
1969 		 * 1) SOCK_STREAM socket.
1970 		 * 2) SOCK_SEQPACKET socket when whole buffer is sent.
1971 		 */
1972 		if (sk->sk_type == SOCK_STREAM || total_written == len)
1973 			err = total_written;
1974 	}
1975 out:
1976 	if (sk->sk_type == SOCK_STREAM)
1977 		err = sk_stream_error(sk, msg->msg_flags, err);
1978 
1979 	release_sock(sk);
1980 	return err;
1981 }
1982 
1983 static int vsock_connectible_wait_data(struct sock *sk,
1984 				       struct wait_queue_entry *wait,
1985 				       long timeout,
1986 				       struct vsock_transport_recv_notify_data *recv_data,
1987 				       size_t target)
1988 {
1989 	const struct vsock_transport *transport;
1990 	struct vsock_sock *vsk;
1991 	s64 data;
1992 	int err;
1993 
1994 	vsk = vsock_sk(sk);
1995 	err = 0;
1996 	transport = vsk->transport;
1997 
1998 	while (1) {
1999 		prepare_to_wait(sk_sleep(sk), wait, TASK_INTERRUPTIBLE);
2000 		data = vsock_connectible_has_data(vsk);
2001 		if (data != 0)
2002 			break;
2003 
2004 		if (sk->sk_err != 0 ||
2005 		    (sk->sk_shutdown & RCV_SHUTDOWN) ||
2006 		    (vsk->peer_shutdown & SEND_SHUTDOWN)) {
2007 			break;
2008 		}
2009 
2010 		/* Don't wait for non-blocking sockets. */
2011 		if (timeout == 0) {
2012 			err = -EAGAIN;
2013 			break;
2014 		}
2015 
2016 		if (recv_data) {
2017 			err = transport->notify_recv_pre_block(vsk, target, recv_data);
2018 			if (err < 0)
2019 				break;
2020 		}
2021 
2022 		release_sock(sk);
2023 		timeout = schedule_timeout(timeout);
2024 		lock_sock(sk);
2025 
2026 		if (signal_pending(current)) {
2027 			err = sock_intr_errno(timeout);
2028 			break;
2029 		} else if (timeout == 0) {
2030 			err = -EAGAIN;
2031 			break;
2032 		}
2033 	}
2034 
2035 	finish_wait(sk_sleep(sk), wait);
2036 
2037 	if (err)
2038 		return err;
2039 
2040 	/* Internal transport error when checking for available
2041 	 * data. XXX This should be changed to a connection
2042 	 * reset in a later change.
2043 	 */
2044 	if (data < 0)
2045 		return -ENOMEM;
2046 
2047 	return data;
2048 }
2049 
2050 static int __vsock_stream_recvmsg(struct sock *sk, struct msghdr *msg,
2051 				  size_t len, int flags)
2052 {
2053 	struct vsock_transport_recv_notify_data recv_data;
2054 	const struct vsock_transport *transport;
2055 	struct vsock_sock *vsk;
2056 	ssize_t copied;
2057 	size_t target;
2058 	long timeout;
2059 	int err;
2060 
2061 	DEFINE_WAIT(wait);
2062 
2063 	vsk = vsock_sk(sk);
2064 	transport = vsk->transport;
2065 
2066 	/* We must not copy less than target bytes into the user's buffer
2067 	 * before returning successfully, so we wait for the consume queue to
2068 	 * have that much data to consume before dequeueing.  Note that this
2069 	 * makes it impossible to handle cases where target is greater than the
2070 	 * queue size.
2071 	 */
2072 	target = sock_rcvlowat(sk, flags & MSG_WAITALL, len);
2073 	if (target >= transport->stream_rcvhiwat(vsk)) {
2074 		err = -ENOMEM;
2075 		goto out;
2076 	}
2077 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2078 	copied = 0;
2079 
2080 	err = transport->notify_recv_init(vsk, target, &recv_data);
2081 	if (err < 0)
2082 		goto out;
2083 
2084 
2085 	while (1) {
2086 		ssize_t read;
2087 
2088 		err = vsock_connectible_wait_data(sk, &wait, timeout,
2089 						  &recv_data, target);
2090 		if (err <= 0)
2091 			break;
2092 
2093 		err = transport->notify_recv_pre_dequeue(vsk, target,
2094 							 &recv_data);
2095 		if (err < 0)
2096 			break;
2097 
2098 		read = transport->stream_dequeue(vsk, msg, len - copied, flags);
2099 		if (read < 0) {
2100 			err = read;
2101 			break;
2102 		}
2103 
2104 		copied += read;
2105 
2106 		err = transport->notify_recv_post_dequeue(vsk, target, read,
2107 						!(flags & MSG_PEEK), &recv_data);
2108 		if (err < 0)
2109 			goto out;
2110 
2111 		if (read >= target || flags & MSG_PEEK)
2112 			break;
2113 
2114 		target -= read;
2115 	}
2116 
2117 	if (sk->sk_err)
2118 		err = -sk->sk_err;
2119 	else if (sk->sk_shutdown & RCV_SHUTDOWN)
2120 		err = 0;
2121 
2122 	if (copied > 0)
2123 		err = copied;
2124 
2125 out:
2126 	return err;
2127 }
2128 
2129 static int __vsock_seqpacket_recvmsg(struct sock *sk, struct msghdr *msg,
2130 				     size_t len, int flags)
2131 {
2132 	const struct vsock_transport *transport;
2133 	struct vsock_sock *vsk;
2134 	ssize_t msg_len;
2135 	long timeout;
2136 	int err = 0;
2137 	DEFINE_WAIT(wait);
2138 
2139 	vsk = vsock_sk(sk);
2140 	transport = vsk->transport;
2141 
2142 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2143 
2144 	err = vsock_connectible_wait_data(sk, &wait, timeout, NULL, 0);
2145 	if (err <= 0)
2146 		goto out;
2147 
2148 	msg_len = transport->seqpacket_dequeue(vsk, msg, flags);
2149 
2150 	if (msg_len < 0) {
2151 		err = msg_len;
2152 		goto out;
2153 	}
2154 
2155 	if (sk->sk_err) {
2156 		err = -sk->sk_err;
2157 	} else if (sk->sk_shutdown & RCV_SHUTDOWN) {
2158 		err = 0;
2159 	} else {
2160 		/* User sets MSG_TRUNC, so return real length of
2161 		 * packet.
2162 		 */
2163 		if (flags & MSG_TRUNC)
2164 			err = msg_len;
2165 		else
2166 			err = len - msg_data_left(msg);
2167 
2168 		/* Always set MSG_TRUNC if real length of packet is
2169 		 * bigger than user's buffer.
2170 		 */
2171 		if (msg_len > len)
2172 			msg->msg_flags |= MSG_TRUNC;
2173 	}
2174 
2175 out:
2176 	return err;
2177 }
2178 
2179 int
2180 __vsock_connectible_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
2181 			    int flags)
2182 {
2183 	struct sock *sk;
2184 	struct vsock_sock *vsk;
2185 	const struct vsock_transport *transport;
2186 	int err;
2187 
2188 	sk = sock->sk;
2189 
2190 	if (unlikely(flags & MSG_ERRQUEUE))
2191 		return sock_recv_errqueue(sk, msg, len, SOL_VSOCK, VSOCK_RECVERR);
2192 
2193 	vsk = vsock_sk(sk);
2194 	err = 0;
2195 
2196 	lock_sock(sk);
2197 
2198 	transport = vsk->transport;
2199 
2200 	if (!transport || sk->sk_state != TCP_ESTABLISHED) {
2201 		/* Recvmsg is supposed to return 0 if a peer performs an
2202 		 * orderly shutdown. Differentiate between that case and when a
2203 		 * peer has not connected or a local shutdown occurred with the
2204 		 * SOCK_DONE flag.
2205 		 */
2206 		if (sock_flag(sk, SOCK_DONE))
2207 			err = 0;
2208 		else
2209 			err = -ENOTCONN;
2210 
2211 		goto out;
2212 	}
2213 
2214 	if (flags & MSG_OOB) {
2215 		err = -EOPNOTSUPP;
2216 		goto out;
2217 	}
2218 
2219 	/* We don't check peer_shutdown flag here since peer may actually shut
2220 	 * down, but there can be data in the queue that a local socket can
2221 	 * receive.
2222 	 */
2223 	if (sk->sk_shutdown & RCV_SHUTDOWN) {
2224 		err = 0;
2225 		goto out;
2226 	}
2227 
2228 	/* It is valid on Linux to pass in a zero-length receive buffer.  This
2229 	 * is not an error.  We may as well bail out now.
2230 	 */
2231 	if (!len) {
2232 		err = 0;
2233 		goto out;
2234 	}
2235 
2236 	if (sk->sk_type == SOCK_STREAM)
2237 		err = __vsock_stream_recvmsg(sk, msg, len, flags);
2238 	else
2239 		err = __vsock_seqpacket_recvmsg(sk, msg, len, flags);
2240 
2241 out:
2242 	release_sock(sk);
2243 	return err;
2244 }
2245 
2246 int
2247 vsock_connectible_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
2248 			  int flags)
2249 {
2250 #ifdef CONFIG_BPF_SYSCALL
2251 	struct sock *sk = sock->sk;
2252 	const struct proto *prot;
2253 
2254 	prot = READ_ONCE(sk->sk_prot);
2255 	if (prot != &vsock_proto)
2256 		return prot->recvmsg(sk, msg, len, flags, NULL);
2257 #endif
2258 
2259 	return __vsock_connectible_recvmsg(sock, msg, len, flags);
2260 }
2261 EXPORT_SYMBOL_GPL(vsock_connectible_recvmsg);
2262 
2263 static int vsock_set_rcvlowat(struct sock *sk, int val)
2264 {
2265 	const struct vsock_transport *transport;
2266 	struct vsock_sock *vsk;
2267 
2268 	vsk = vsock_sk(sk);
2269 
2270 	if (val > vsk->buffer_size)
2271 		return -EINVAL;
2272 
2273 	transport = vsk->transport;
2274 
2275 	if (transport && transport->notify_set_rcvlowat) {
2276 		int err;
2277 
2278 		err = transport->notify_set_rcvlowat(vsk, val);
2279 		if (err)
2280 			return err;
2281 	}
2282 
2283 	WRITE_ONCE(sk->sk_rcvlowat, val ? : 1);
2284 	return 0;
2285 }
2286 
2287 static const struct proto_ops vsock_stream_ops = {
2288 	.family = PF_VSOCK,
2289 	.owner = THIS_MODULE,
2290 	.release = vsock_release,
2291 	.bind = vsock_bind,
2292 	.connect = vsock_connect,
2293 	.socketpair = sock_no_socketpair,
2294 	.accept = vsock_accept,
2295 	.getname = vsock_getname,
2296 	.poll = vsock_poll,
2297 	.ioctl = sock_no_ioctl,
2298 	.listen = vsock_listen,
2299 	.shutdown = vsock_shutdown,
2300 	.setsockopt = vsock_connectible_setsockopt,
2301 	.getsockopt = vsock_connectible_getsockopt,
2302 	.sendmsg = vsock_connectible_sendmsg,
2303 	.recvmsg = vsock_connectible_recvmsg,
2304 	.mmap = sock_no_mmap,
2305 	.set_rcvlowat = vsock_set_rcvlowat,
2306 	.read_skb = vsock_read_skb,
2307 };
2308 
2309 static const struct proto_ops vsock_seqpacket_ops = {
2310 	.family = PF_VSOCK,
2311 	.owner = THIS_MODULE,
2312 	.release = vsock_release,
2313 	.bind = vsock_bind,
2314 	.connect = vsock_connect,
2315 	.socketpair = sock_no_socketpair,
2316 	.accept = vsock_accept,
2317 	.getname = vsock_getname,
2318 	.poll = vsock_poll,
2319 	.ioctl = sock_no_ioctl,
2320 	.listen = vsock_listen,
2321 	.shutdown = vsock_shutdown,
2322 	.setsockopt = vsock_connectible_setsockopt,
2323 	.getsockopt = vsock_connectible_getsockopt,
2324 	.sendmsg = vsock_connectible_sendmsg,
2325 	.recvmsg = vsock_connectible_recvmsg,
2326 	.mmap = sock_no_mmap,
2327 	.read_skb = vsock_read_skb,
2328 };
2329 
2330 static int vsock_create(struct net *net, struct socket *sock,
2331 			int protocol, int kern)
2332 {
2333 	struct vsock_sock *vsk;
2334 	struct sock *sk;
2335 	int ret;
2336 
2337 	if (!sock)
2338 		return -EINVAL;
2339 
2340 	if (protocol && protocol != PF_VSOCK)
2341 		return -EPROTONOSUPPORT;
2342 
2343 	switch (sock->type) {
2344 	case SOCK_DGRAM:
2345 		sock->ops = &vsock_dgram_ops;
2346 		break;
2347 	case SOCK_STREAM:
2348 		sock->ops = &vsock_stream_ops;
2349 		break;
2350 	case SOCK_SEQPACKET:
2351 		sock->ops = &vsock_seqpacket_ops;
2352 		break;
2353 	default:
2354 		return -ESOCKTNOSUPPORT;
2355 	}
2356 
2357 	sock->state = SS_UNCONNECTED;
2358 
2359 	sk = __vsock_create(net, sock, NULL, GFP_KERNEL, 0, kern);
2360 	if (!sk)
2361 		return -ENOMEM;
2362 
2363 	vsk = vsock_sk(sk);
2364 
2365 	if (sock->type == SOCK_DGRAM) {
2366 		ret = vsock_assign_transport(vsk, NULL);
2367 		if (ret < 0) {
2368 			sock_put(sk);
2369 			return ret;
2370 		}
2371 	}
2372 
2373 	/* SOCK_DGRAM doesn't have 'setsockopt' callback set in its
2374 	 * proto_ops, so there is no handler for custom logic.
2375 	 */
2376 	if (sock_type_connectible(sock->type))
2377 		set_bit(SOCK_CUSTOM_SOCKOPT, &sk->sk_socket->flags);
2378 
2379 	vsock_insert_unbound(vsk);
2380 
2381 	return 0;
2382 }
2383 
2384 static const struct net_proto_family vsock_family_ops = {
2385 	.family = AF_VSOCK,
2386 	.create = vsock_create,
2387 	.owner = THIS_MODULE,
2388 };
2389 
2390 static long vsock_dev_do_ioctl(struct file *filp,
2391 			       unsigned int cmd, void __user *ptr)
2392 {
2393 	u32 __user *p = ptr;
2394 	u32 cid = VMADDR_CID_ANY;
2395 	int retval = 0;
2396 
2397 	switch (cmd) {
2398 	case IOCTL_VM_SOCKETS_GET_LOCAL_CID:
2399 		/* To be compatible with the VMCI behavior, we prioritize the
2400 		 * guest CID instead of well-know host CID (VMADDR_CID_HOST).
2401 		 */
2402 		if (transport_g2h)
2403 			cid = transport_g2h->get_local_cid();
2404 		else if (transport_h2g)
2405 			cid = transport_h2g->get_local_cid();
2406 
2407 		if (put_user(cid, p) != 0)
2408 			retval = -EFAULT;
2409 		break;
2410 
2411 	default:
2412 		retval = -ENOIOCTLCMD;
2413 	}
2414 
2415 	return retval;
2416 }
2417 
2418 static long vsock_dev_ioctl(struct file *filp,
2419 			    unsigned int cmd, unsigned long arg)
2420 {
2421 	return vsock_dev_do_ioctl(filp, cmd, (void __user *)arg);
2422 }
2423 
2424 #ifdef CONFIG_COMPAT
2425 static long vsock_dev_compat_ioctl(struct file *filp,
2426 				   unsigned int cmd, unsigned long arg)
2427 {
2428 	return vsock_dev_do_ioctl(filp, cmd, compat_ptr(arg));
2429 }
2430 #endif
2431 
2432 static const struct file_operations vsock_device_ops = {
2433 	.owner		= THIS_MODULE,
2434 	.unlocked_ioctl	= vsock_dev_ioctl,
2435 #ifdef CONFIG_COMPAT
2436 	.compat_ioctl	= vsock_dev_compat_ioctl,
2437 #endif
2438 	.open		= nonseekable_open,
2439 };
2440 
2441 static struct miscdevice vsock_device = {
2442 	.name		= "vsock",
2443 	.fops		= &vsock_device_ops,
2444 };
2445 
2446 static int __init vsock_init(void)
2447 {
2448 	int err = 0;
2449 
2450 	vsock_init_tables();
2451 
2452 	vsock_proto.owner = THIS_MODULE;
2453 	vsock_device.minor = MISC_DYNAMIC_MINOR;
2454 	err = misc_register(&vsock_device);
2455 	if (err) {
2456 		pr_err("Failed to register misc device\n");
2457 		goto err_reset_transport;
2458 	}
2459 
2460 	err = proto_register(&vsock_proto, 1);	/* we want our slab */
2461 	if (err) {
2462 		pr_err("Cannot register vsock protocol\n");
2463 		goto err_deregister_misc;
2464 	}
2465 
2466 	err = sock_register(&vsock_family_ops);
2467 	if (err) {
2468 		pr_err("could not register af_vsock (%d) address family: %d\n",
2469 		       AF_VSOCK, err);
2470 		goto err_unregister_proto;
2471 	}
2472 
2473 	vsock_bpf_build_proto();
2474 
2475 	return 0;
2476 
2477 err_unregister_proto:
2478 	proto_unregister(&vsock_proto);
2479 err_deregister_misc:
2480 	misc_deregister(&vsock_device);
2481 err_reset_transport:
2482 	return err;
2483 }
2484 
2485 static void __exit vsock_exit(void)
2486 {
2487 	misc_deregister(&vsock_device);
2488 	sock_unregister(AF_VSOCK);
2489 	proto_unregister(&vsock_proto);
2490 }
2491 
2492 const struct vsock_transport *vsock_core_get_transport(struct vsock_sock *vsk)
2493 {
2494 	return vsk->transport;
2495 }
2496 EXPORT_SYMBOL_GPL(vsock_core_get_transport);
2497 
2498 int vsock_core_register(const struct vsock_transport *t, int features)
2499 {
2500 	const struct vsock_transport *t_h2g, *t_g2h, *t_dgram, *t_local;
2501 	int err = mutex_lock_interruptible(&vsock_register_mutex);
2502 
2503 	if (err)
2504 		return err;
2505 
2506 	t_h2g = transport_h2g;
2507 	t_g2h = transport_g2h;
2508 	t_dgram = transport_dgram;
2509 	t_local = transport_local;
2510 
2511 	if (features & VSOCK_TRANSPORT_F_H2G) {
2512 		if (t_h2g) {
2513 			err = -EBUSY;
2514 			goto err_busy;
2515 		}
2516 		t_h2g = t;
2517 	}
2518 
2519 	if (features & VSOCK_TRANSPORT_F_G2H) {
2520 		if (t_g2h) {
2521 			err = -EBUSY;
2522 			goto err_busy;
2523 		}
2524 		t_g2h = t;
2525 	}
2526 
2527 	if (features & VSOCK_TRANSPORT_F_DGRAM) {
2528 		if (t_dgram) {
2529 			err = -EBUSY;
2530 			goto err_busy;
2531 		}
2532 		t_dgram = t;
2533 	}
2534 
2535 	if (features & VSOCK_TRANSPORT_F_LOCAL) {
2536 		if (t_local) {
2537 			err = -EBUSY;
2538 			goto err_busy;
2539 		}
2540 		t_local = t;
2541 	}
2542 
2543 	transport_h2g = t_h2g;
2544 	transport_g2h = t_g2h;
2545 	transport_dgram = t_dgram;
2546 	transport_local = t_local;
2547 
2548 err_busy:
2549 	mutex_unlock(&vsock_register_mutex);
2550 	return err;
2551 }
2552 EXPORT_SYMBOL_GPL(vsock_core_register);
2553 
2554 void vsock_core_unregister(const struct vsock_transport *t)
2555 {
2556 	mutex_lock(&vsock_register_mutex);
2557 
2558 	if (transport_h2g == t)
2559 		transport_h2g = NULL;
2560 
2561 	if (transport_g2h == t)
2562 		transport_g2h = NULL;
2563 
2564 	if (transport_dgram == t)
2565 		transport_dgram = NULL;
2566 
2567 	if (transport_local == t)
2568 		transport_local = NULL;
2569 
2570 	mutex_unlock(&vsock_register_mutex);
2571 }
2572 EXPORT_SYMBOL_GPL(vsock_core_unregister);
2573 
2574 module_init(vsock_init);
2575 module_exit(vsock_exit);
2576 
2577 MODULE_AUTHOR("VMware, Inc.");
2578 MODULE_DESCRIPTION("VMware Virtual Socket Family");
2579 MODULE_VERSION("1.0.2.0-k");
2580 MODULE_LICENSE("GPL v2");
2581