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 VSOCK_SS_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 * - 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
79#include <linux/types.h>
80#include <linux/bitops.h>
81#include <linux/cred.h>
82#include <linux/init.h>
83#include <linux/io.h>
84#include <linux/kernel.h>
85#include <linux/kmod.h>
86#include <linux/list.h>
87#include <linux/miscdevice.h>
88#include <linux/module.h>
89#include <linux/mutex.h>
90#include <linux/net.h>
91#include <linux/poll.h>
92#include <linux/skbuff.h>
93#include <linux/smp.h>
94#include <linux/socket.h>
95#include <linux/stddef.h>
96#include <linux/unistd.h>
97#include <linux/wait.h>
98#include <linux/workqueue.h>
99#include <net/sock.h>
100#include <net/af_vsock.h>
101
102static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr);
103static void vsock_sk_destruct(struct sock *sk);
104static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
105
106/* Protocol family. */
107static struct proto vsock_proto = {
108	.name = "AF_VSOCK",
109	.owner = THIS_MODULE,
110	.obj_size = sizeof(struct vsock_sock),
111};
112
113/* The default peer timeout indicates how long we will wait for a peer response
114 * to a control message.
115 */
116#define VSOCK_DEFAULT_CONNECT_TIMEOUT (2 * HZ)
117
118static const struct vsock_transport *transport;
119static DEFINE_MUTEX(vsock_register_mutex);
120
121/**** EXPORTS ****/
122
123/* Get the ID of the local context.  This is transport dependent. */
124
125int vm_sockets_get_local_cid(void)
126{
127	return transport->get_local_cid();
128}
129EXPORT_SYMBOL_GPL(vm_sockets_get_local_cid);
130
131/**** UTILS ****/
132
133/* Each bound VSocket is stored in the bind hash table and each connected
134 * VSocket is stored in the connected hash table.
135 *
136 * Unbound sockets are all put on the same list attached to the end of the hash
137 * table (vsock_unbound_sockets).  Bound sockets are added to the hash table in
138 * the bucket that their local address hashes to (vsock_bound_sockets(addr)
139 * represents the list that addr hashes to).
140 *
141 * Specifically, we initialize the vsock_bind_table array to a size of
142 * VSOCK_HASH_SIZE + 1 so that vsock_bind_table[0] through
143 * vsock_bind_table[VSOCK_HASH_SIZE - 1] are for bound sockets and
144 * vsock_bind_table[VSOCK_HASH_SIZE] is for unbound sockets.  The hash function
145 * mods with VSOCK_HASH_SIZE to ensure this.
146 */
147#define VSOCK_HASH_SIZE         251
148#define MAX_PORT_RETRIES        24
149
150#define VSOCK_HASH(addr)        ((addr)->svm_port % VSOCK_HASH_SIZE)
151#define vsock_bound_sockets(addr) (&vsock_bind_table[VSOCK_HASH(addr)])
152#define vsock_unbound_sockets     (&vsock_bind_table[VSOCK_HASH_SIZE])
153
154/* XXX This can probably be implemented in a better way. */
155#define VSOCK_CONN_HASH(src, dst)				\
156	(((src)->svm_cid ^ (dst)->svm_port) % VSOCK_HASH_SIZE)
157#define vsock_connected_sockets(src, dst)		\
158	(&vsock_connected_table[VSOCK_CONN_HASH(src, dst)])
159#define vsock_connected_sockets_vsk(vsk)				\
160	vsock_connected_sockets(&(vsk)->remote_addr, &(vsk)->local_addr)
161
162static struct list_head vsock_bind_table[VSOCK_HASH_SIZE + 1];
163static struct list_head vsock_connected_table[VSOCK_HASH_SIZE];
164static DEFINE_SPINLOCK(vsock_table_lock);
165
166/* Autobind this socket to the local address if necessary. */
167static int vsock_auto_bind(struct vsock_sock *vsk)
168{
169	struct sock *sk = sk_vsock(vsk);
170	struct sockaddr_vm local_addr;
171
172	if (vsock_addr_bound(&vsk->local_addr))
173		return 0;
174	vsock_addr_init(&local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
175	return __vsock_bind(sk, &local_addr);
176}
177
178static void vsock_init_tables(void)
179{
180	int i;
181
182	for (i = 0; i < ARRAY_SIZE(vsock_bind_table); i++)
183		INIT_LIST_HEAD(&vsock_bind_table[i]);
184
185	for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++)
186		INIT_LIST_HEAD(&vsock_connected_table[i]);
187}
188
189static void __vsock_insert_bound(struct list_head *list,
190				 struct vsock_sock *vsk)
191{
192	sock_hold(&vsk->sk);
193	list_add(&vsk->bound_table, list);
194}
195
196static void __vsock_insert_connected(struct list_head *list,
197				     struct vsock_sock *vsk)
198{
199	sock_hold(&vsk->sk);
200	list_add(&vsk->connected_table, list);
201}
202
203static void __vsock_remove_bound(struct vsock_sock *vsk)
204{
205	list_del_init(&vsk->bound_table);
206	sock_put(&vsk->sk);
207}
208
209static void __vsock_remove_connected(struct vsock_sock *vsk)
210{
211	list_del_init(&vsk->connected_table);
212	sock_put(&vsk->sk);
213}
214
215static struct sock *__vsock_find_bound_socket(struct sockaddr_vm *addr)
216{
217	struct vsock_sock *vsk;
218
219	list_for_each_entry(vsk, vsock_bound_sockets(addr), bound_table)
220		if (addr->svm_port == vsk->local_addr.svm_port)
221			return sk_vsock(vsk);
222
223	return NULL;
224}
225
226static struct sock *__vsock_find_connected_socket(struct sockaddr_vm *src,
227						  struct sockaddr_vm *dst)
228{
229	struct vsock_sock *vsk;
230
231	list_for_each_entry(vsk, vsock_connected_sockets(src, dst),
232			    connected_table) {
233		if (vsock_addr_equals_addr(src, &vsk->remote_addr) &&
234		    dst->svm_port == vsk->local_addr.svm_port) {
235			return sk_vsock(vsk);
236		}
237	}
238
239	return NULL;
240}
241
242static bool __vsock_in_bound_table(struct vsock_sock *vsk)
243{
244	return !list_empty(&vsk->bound_table);
245}
246
247static bool __vsock_in_connected_table(struct vsock_sock *vsk)
248{
249	return !list_empty(&vsk->connected_table);
250}
251
252static void vsock_insert_unbound(struct vsock_sock *vsk)
253{
254	spin_lock_bh(&vsock_table_lock);
255	__vsock_insert_bound(vsock_unbound_sockets, vsk);
256	spin_unlock_bh(&vsock_table_lock);
257}
258
259void vsock_insert_connected(struct vsock_sock *vsk)
260{
261	struct list_head *list = vsock_connected_sockets(
262		&vsk->remote_addr, &vsk->local_addr);
263
264	spin_lock_bh(&vsock_table_lock);
265	__vsock_insert_connected(list, vsk);
266	spin_unlock_bh(&vsock_table_lock);
267}
268EXPORT_SYMBOL_GPL(vsock_insert_connected);
269
270void vsock_remove_bound(struct vsock_sock *vsk)
271{
272	spin_lock_bh(&vsock_table_lock);
273	__vsock_remove_bound(vsk);
274	spin_unlock_bh(&vsock_table_lock);
275}
276EXPORT_SYMBOL_GPL(vsock_remove_bound);
277
278void vsock_remove_connected(struct vsock_sock *vsk)
279{
280	spin_lock_bh(&vsock_table_lock);
281	__vsock_remove_connected(vsk);
282	spin_unlock_bh(&vsock_table_lock);
283}
284EXPORT_SYMBOL_GPL(vsock_remove_connected);
285
286struct sock *vsock_find_bound_socket(struct sockaddr_vm *addr)
287{
288	struct sock *sk;
289
290	spin_lock_bh(&vsock_table_lock);
291	sk = __vsock_find_bound_socket(addr);
292	if (sk)
293		sock_hold(sk);
294
295	spin_unlock_bh(&vsock_table_lock);
296
297	return sk;
298}
299EXPORT_SYMBOL_GPL(vsock_find_bound_socket);
300
301struct sock *vsock_find_connected_socket(struct sockaddr_vm *src,
302					 struct sockaddr_vm *dst)
303{
304	struct sock *sk;
305
306	spin_lock_bh(&vsock_table_lock);
307	sk = __vsock_find_connected_socket(src, dst);
308	if (sk)
309		sock_hold(sk);
310
311	spin_unlock_bh(&vsock_table_lock);
312
313	return sk;
314}
315EXPORT_SYMBOL_GPL(vsock_find_connected_socket);
316
317static bool vsock_in_bound_table(struct vsock_sock *vsk)
318{
319	bool ret;
320
321	spin_lock_bh(&vsock_table_lock);
322	ret = __vsock_in_bound_table(vsk);
323	spin_unlock_bh(&vsock_table_lock);
324
325	return ret;
326}
327
328static bool vsock_in_connected_table(struct vsock_sock *vsk)
329{
330	bool ret;
331
332	spin_lock_bh(&vsock_table_lock);
333	ret = __vsock_in_connected_table(vsk);
334	spin_unlock_bh(&vsock_table_lock);
335
336	return ret;
337}
338
339void vsock_for_each_connected_socket(void (*fn)(struct sock *sk))
340{
341	int i;
342
343	spin_lock_bh(&vsock_table_lock);
344
345	for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) {
346		struct vsock_sock *vsk;
347		list_for_each_entry(vsk, &vsock_connected_table[i],
348				    connected_table)
349			fn(sk_vsock(vsk));
350	}
351
352	spin_unlock_bh(&vsock_table_lock);
353}
354EXPORT_SYMBOL_GPL(vsock_for_each_connected_socket);
355
356void vsock_add_pending(struct sock *listener, struct sock *pending)
357{
358	struct vsock_sock *vlistener;
359	struct vsock_sock *vpending;
360
361	vlistener = vsock_sk(listener);
362	vpending = vsock_sk(pending);
363
364	sock_hold(pending);
365	sock_hold(listener);
366	list_add_tail(&vpending->pending_links, &vlistener->pending_links);
367}
368EXPORT_SYMBOL_GPL(vsock_add_pending);
369
370void vsock_remove_pending(struct sock *listener, struct sock *pending)
371{
372	struct vsock_sock *vpending = vsock_sk(pending);
373
374	list_del_init(&vpending->pending_links);
375	sock_put(listener);
376	sock_put(pending);
377}
378EXPORT_SYMBOL_GPL(vsock_remove_pending);
379
380void vsock_enqueue_accept(struct sock *listener, struct sock *connected)
381{
382	struct vsock_sock *vlistener;
383	struct vsock_sock *vconnected;
384
385	vlistener = vsock_sk(listener);
386	vconnected = vsock_sk(connected);
387
388	sock_hold(connected);
389	sock_hold(listener);
390	list_add_tail(&vconnected->accept_queue, &vlistener->accept_queue);
391}
392EXPORT_SYMBOL_GPL(vsock_enqueue_accept);
393
394static struct sock *vsock_dequeue_accept(struct sock *listener)
395{
396	struct vsock_sock *vlistener;
397	struct vsock_sock *vconnected;
398
399	vlistener = vsock_sk(listener);
400
401	if (list_empty(&vlistener->accept_queue))
402		return NULL;
403
404	vconnected = list_entry(vlistener->accept_queue.next,
405				struct vsock_sock, accept_queue);
406
407	list_del_init(&vconnected->accept_queue);
408	sock_put(listener);
409	/* The caller will need a reference on the connected socket so we let
410	 * it call sock_put().
411	 */
412
413	return sk_vsock(vconnected);
414}
415
416static bool vsock_is_accept_queue_empty(struct sock *sk)
417{
418	struct vsock_sock *vsk = vsock_sk(sk);
419	return list_empty(&vsk->accept_queue);
420}
421
422static bool vsock_is_pending(struct sock *sk)
423{
424	struct vsock_sock *vsk = vsock_sk(sk);
425	return !list_empty(&vsk->pending_links);
426}
427
428static int vsock_send_shutdown(struct sock *sk, int mode)
429{
430	return transport->shutdown(vsock_sk(sk), mode);
431}
432
433void vsock_pending_work(struct work_struct *work)
434{
435	struct sock *sk;
436	struct sock *listener;
437	struct vsock_sock *vsk;
438	bool cleanup;
439
440	vsk = container_of(work, struct vsock_sock, dwork.work);
441	sk = sk_vsock(vsk);
442	listener = vsk->listener;
443	cleanup = true;
444
445	lock_sock(listener);
446	lock_sock(sk);
447
448	if (vsock_is_pending(sk)) {
449		vsock_remove_pending(listener, sk);
450	} else if (!vsk->rejected) {
451		/* We are not on the pending list and accept() did not reject
452		 * us, so we must have been accepted by our user process.  We
453		 * just need to drop our references to the sockets and be on
454		 * our way.
455		 */
456		cleanup = false;
457		goto out;
458	}
459
460	listener->sk_ack_backlog--;
461
462	/* We need to remove ourself from the global connected sockets list so
463	 * incoming packets can't find this socket, and to reduce the reference
464	 * count.
465	 */
466	if (vsock_in_connected_table(vsk))
467		vsock_remove_connected(vsk);
468
469	sk->sk_state = SS_FREE;
470
471out:
472	release_sock(sk);
473	release_sock(listener);
474	if (cleanup)
475		sock_put(sk);
476
477	sock_put(sk);
478	sock_put(listener);
479}
480EXPORT_SYMBOL_GPL(vsock_pending_work);
481
482/**** SOCKET OPERATIONS ****/
483
484static int __vsock_bind_stream(struct vsock_sock *vsk,
485			       struct sockaddr_vm *addr)
486{
487	static u32 port = LAST_RESERVED_PORT + 1;
488	struct sockaddr_vm new_addr;
489
490	vsock_addr_init(&new_addr, addr->svm_cid, addr->svm_port);
491
492	if (addr->svm_port == VMADDR_PORT_ANY) {
493		bool found = false;
494		unsigned int i;
495
496		for (i = 0; i < MAX_PORT_RETRIES; i++) {
497			if (port <= LAST_RESERVED_PORT)
498				port = LAST_RESERVED_PORT + 1;
499
500			new_addr.svm_port = port++;
501
502			if (!__vsock_find_bound_socket(&new_addr)) {
503				found = true;
504				break;
505			}
506		}
507
508		if (!found)
509			return -EADDRNOTAVAIL;
510	} else {
511		/* If port is in reserved range, ensure caller
512		 * has necessary privileges.
513		 */
514		if (addr->svm_port <= LAST_RESERVED_PORT &&
515		    !capable(CAP_NET_BIND_SERVICE)) {
516			return -EACCES;
517		}
518
519		if (__vsock_find_bound_socket(&new_addr))
520			return -EADDRINUSE;
521	}
522
523	vsock_addr_init(&vsk->local_addr, new_addr.svm_cid, new_addr.svm_port);
524
525	/* Remove stream sockets from the unbound list and add them to the hash
526	 * table for easy lookup by its address.  The unbound list is simply an
527	 * extra entry at the end of the hash table, a trick used by AF_UNIX.
528	 */
529	__vsock_remove_bound(vsk);
530	__vsock_insert_bound(vsock_bound_sockets(&vsk->local_addr), vsk);
531
532	return 0;
533}
534
535static int __vsock_bind_dgram(struct vsock_sock *vsk,
536			      struct sockaddr_vm *addr)
537{
538	return transport->dgram_bind(vsk, addr);
539}
540
541static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr)
542{
543	struct vsock_sock *vsk = vsock_sk(sk);
544	u32 cid;
545	int retval;
546
547	/* First ensure this socket isn't already bound. */
548	if (vsock_addr_bound(&vsk->local_addr))
549		return -EINVAL;
550
551	/* Now bind to the provided address or select appropriate values if
552	 * none are provided (VMADDR_CID_ANY and VMADDR_PORT_ANY).  Note that
553	 * like AF_INET prevents binding to a non-local IP address (in most
554	 * cases), we only allow binding to the local CID.
555	 */
556	cid = transport->get_local_cid();
557	if (addr->svm_cid != cid && addr->svm_cid != VMADDR_CID_ANY)
558		return -EADDRNOTAVAIL;
559
560	switch (sk->sk_socket->type) {
561	case SOCK_STREAM:
562		spin_lock_bh(&vsock_table_lock);
563		retval = __vsock_bind_stream(vsk, addr);
564		spin_unlock_bh(&vsock_table_lock);
565		break;
566
567	case SOCK_DGRAM:
568		retval = __vsock_bind_dgram(vsk, addr);
569		break;
570
571	default:
572		retval = -EINVAL;
573		break;
574	}
575
576	return retval;
577}
578
579struct sock *__vsock_create(struct net *net,
580			    struct socket *sock,
581			    struct sock *parent,
582			    gfp_t priority,
583			    unsigned short type,
584			    int kern)
585{
586	struct sock *sk;
587	struct vsock_sock *psk;
588	struct vsock_sock *vsk;
589
590	sk = sk_alloc(net, AF_VSOCK, priority, &vsock_proto, kern);
591	if (!sk)
592		return NULL;
593
594	sock_init_data(sock, sk);
595
596	/* sk->sk_type is normally set in sock_init_data, but only if sock is
597	 * non-NULL. We make sure that our sockets always have a type by
598	 * setting it here if needed.
599	 */
600	if (!sock)
601		sk->sk_type = type;
602
603	vsk = vsock_sk(sk);
604	vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
605	vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
606
607	sk->sk_destruct = vsock_sk_destruct;
608	sk->sk_backlog_rcv = vsock_queue_rcv_skb;
609	sk->sk_state = 0;
610	sock_reset_flag(sk, SOCK_DONE);
611
612	INIT_LIST_HEAD(&vsk->bound_table);
613	INIT_LIST_HEAD(&vsk->connected_table);
614	vsk->listener = NULL;
615	INIT_LIST_HEAD(&vsk->pending_links);
616	INIT_LIST_HEAD(&vsk->accept_queue);
617	vsk->rejected = false;
618	vsk->sent_request = false;
619	vsk->ignore_connecting_rst = false;
620	vsk->peer_shutdown = 0;
621
622	psk = parent ? vsock_sk(parent) : NULL;
623	if (parent) {
624		vsk->trusted = psk->trusted;
625		vsk->owner = get_cred(psk->owner);
626		vsk->connect_timeout = psk->connect_timeout;
627	} else {
628		vsk->trusted = capable(CAP_NET_ADMIN);
629		vsk->owner = get_current_cred();
630		vsk->connect_timeout = VSOCK_DEFAULT_CONNECT_TIMEOUT;
631	}
632
633	if (transport->init(vsk, psk) < 0) {
634		sk_free(sk);
635		return NULL;
636	}
637
638	if (sock)
639		vsock_insert_unbound(vsk);
640
641	return sk;
642}
643EXPORT_SYMBOL_GPL(__vsock_create);
644
645static void __vsock_release(struct sock *sk)
646{
647	if (sk) {
648		struct sk_buff *skb;
649		struct sock *pending;
650		struct vsock_sock *vsk;
651
652		vsk = vsock_sk(sk);
653		pending = NULL;	/* Compiler warning. */
654
655		if (vsock_in_bound_table(vsk))
656			vsock_remove_bound(vsk);
657
658		if (vsock_in_connected_table(vsk))
659			vsock_remove_connected(vsk);
660
661		transport->release(vsk);
662
663		lock_sock(sk);
664		sock_orphan(sk);
665		sk->sk_shutdown = SHUTDOWN_MASK;
666
667		while ((skb = skb_dequeue(&sk->sk_receive_queue)))
668			kfree_skb(skb);
669
670		/* Clean up any sockets that never were accepted. */
671		while ((pending = vsock_dequeue_accept(sk)) != NULL) {
672			__vsock_release(pending);
673			sock_put(pending);
674		}
675
676		release_sock(sk);
677		sock_put(sk);
678	}
679}
680
681static void vsock_sk_destruct(struct sock *sk)
682{
683	struct vsock_sock *vsk = vsock_sk(sk);
684
685	transport->destruct(vsk);
686
687	/* When clearing these addresses, there's no need to set the family and
688	 * possibly register the address family with the kernel.
689	 */
690	vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
691	vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY);
692
693	put_cred(vsk->owner);
694}
695
696static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
697{
698	int err;
699
700	err = sock_queue_rcv_skb(sk, skb);
701	if (err)
702		kfree_skb(skb);
703
704	return err;
705}
706
707s64 vsock_stream_has_data(struct vsock_sock *vsk)
708{
709	return transport->stream_has_data(vsk);
710}
711EXPORT_SYMBOL_GPL(vsock_stream_has_data);
712
713s64 vsock_stream_has_space(struct vsock_sock *vsk)
714{
715	return transport->stream_has_space(vsk);
716}
717EXPORT_SYMBOL_GPL(vsock_stream_has_space);
718
719static int vsock_release(struct socket *sock)
720{
721	__vsock_release(sock->sk);
722	sock->sk = NULL;
723	sock->state = SS_FREE;
724
725	return 0;
726}
727
728static int
729vsock_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
730{
731	int err;
732	struct sock *sk;
733	struct sockaddr_vm *vm_addr;
734
735	sk = sock->sk;
736
737	if (vsock_addr_cast(addr, addr_len, &vm_addr) != 0)
738		return -EINVAL;
739
740	lock_sock(sk);
741	err = __vsock_bind(sk, vm_addr);
742	release_sock(sk);
743
744	return err;
745}
746
747static int vsock_getname(struct socket *sock,
748			 struct sockaddr *addr, int *addr_len, int peer)
749{
750	int err;
751	struct sock *sk;
752	struct vsock_sock *vsk;
753	struct sockaddr_vm *vm_addr;
754
755	sk = sock->sk;
756	vsk = vsock_sk(sk);
757	err = 0;
758
759	lock_sock(sk);
760
761	if (peer) {
762		if (sock->state != SS_CONNECTED) {
763			err = -ENOTCONN;
764			goto out;
765		}
766		vm_addr = &vsk->remote_addr;
767	} else {
768		vm_addr = &vsk->local_addr;
769	}
770
771	if (!vm_addr) {
772		err = -EINVAL;
773		goto out;
774	}
775
776	/* sys_getsockname() and sys_getpeername() pass us a
777	 * MAX_SOCK_ADDR-sized buffer and don't set addr_len.  Unfortunately
778	 * that macro is defined in socket.c instead of .h, so we hardcode its
779	 * value here.
780	 */
781	BUILD_BUG_ON(sizeof(*vm_addr) > 128);
782	memcpy(addr, vm_addr, sizeof(*vm_addr));
783	*addr_len = sizeof(*vm_addr);
784
785out:
786	release_sock(sk);
787	return err;
788}
789
790static int vsock_shutdown(struct socket *sock, int mode)
791{
792	int err;
793	struct sock *sk;
794
795	/* User level uses SHUT_RD (0) and SHUT_WR (1), but the kernel uses
796	 * RCV_SHUTDOWN (1) and SEND_SHUTDOWN (2), so we must increment mode
797	 * here like the other address families do.  Note also that the
798	 * increment makes SHUT_RDWR (2) into RCV_SHUTDOWN | SEND_SHUTDOWN (3),
799	 * which is what we want.
800	 */
801	mode++;
802
803	if ((mode & ~SHUTDOWN_MASK) || !mode)
804		return -EINVAL;
805
806	/* If this is a STREAM socket and it is not connected then bail out
807	 * immediately.  If it is a DGRAM socket then we must first kick the
808	 * socket so that it wakes up from any sleeping calls, for example
809	 * recv(), and then afterwards return the error.
810	 */
811
812	sk = sock->sk;
813	if (sock->state == SS_UNCONNECTED) {
814		err = -ENOTCONN;
815		if (sk->sk_type == SOCK_STREAM)
816			return err;
817	} else {
818		sock->state = SS_DISCONNECTING;
819		err = 0;
820	}
821
822	/* Receive and send shutdowns are treated alike. */
823	mode = mode & (RCV_SHUTDOWN | SEND_SHUTDOWN);
824	if (mode) {
825		lock_sock(sk);
826		sk->sk_shutdown |= mode;
827		sk->sk_state_change(sk);
828		release_sock(sk);
829
830		if (sk->sk_type == SOCK_STREAM) {
831			sock_reset_flag(sk, SOCK_DONE);
832			vsock_send_shutdown(sk, mode);
833		}
834	}
835
836	return err;
837}
838
839static unsigned int vsock_poll(struct file *file, struct socket *sock,
840			       poll_table *wait)
841{
842	struct sock *sk;
843	unsigned int mask;
844	struct vsock_sock *vsk;
845
846	sk = sock->sk;
847	vsk = vsock_sk(sk);
848
849	poll_wait(file, sk_sleep(sk), wait);
850	mask = 0;
851
852	if (sk->sk_err)
853		/* Signify that there has been an error on this socket. */
854		mask |= POLLERR;
855
856	/* INET sockets treat local write shutdown and peer write shutdown as a
857	 * case of POLLHUP set.
858	 */
859	if ((sk->sk_shutdown == SHUTDOWN_MASK) ||
860	    ((sk->sk_shutdown & SEND_SHUTDOWN) &&
861	     (vsk->peer_shutdown & SEND_SHUTDOWN))) {
862		mask |= POLLHUP;
863	}
864
865	if (sk->sk_shutdown & RCV_SHUTDOWN ||
866	    vsk->peer_shutdown & SEND_SHUTDOWN) {
867		mask |= POLLRDHUP;
868	}
869
870	if (sock->type == SOCK_DGRAM) {
871		/* For datagram sockets we can read if there is something in
872		 * the queue and write as long as the socket isn't shutdown for
873		 * sending.
874		 */
875		if (!skb_queue_empty(&sk->sk_receive_queue) ||
876		    (sk->sk_shutdown & RCV_SHUTDOWN)) {
877			mask |= POLLIN | POLLRDNORM;
878		}
879
880		if (!(sk->sk_shutdown & SEND_SHUTDOWN))
881			mask |= POLLOUT | POLLWRNORM | POLLWRBAND;
882
883	} else if (sock->type == SOCK_STREAM) {
884		lock_sock(sk);
885
886		/* Listening sockets that have connections in their accept
887		 * queue can be read.
888		 */
889		if (sk->sk_state == VSOCK_SS_LISTEN
890		    && !vsock_is_accept_queue_empty(sk))
891			mask |= POLLIN | POLLRDNORM;
892
893		/* If there is something in the queue then we can read. */
894		if (transport->stream_is_active(vsk) &&
895		    !(sk->sk_shutdown & RCV_SHUTDOWN)) {
896			bool data_ready_now = false;
897			int ret = transport->notify_poll_in(
898					vsk, 1, &data_ready_now);
899			if (ret < 0) {
900				mask |= POLLERR;
901			} else {
902				if (data_ready_now)
903					mask |= POLLIN | POLLRDNORM;
904
905			}
906		}
907
908		/* Sockets whose connections have been closed, reset, or
909		 * terminated should also be considered read, and we check the
910		 * shutdown flag for that.
911		 */
912		if (sk->sk_shutdown & RCV_SHUTDOWN ||
913		    vsk->peer_shutdown & SEND_SHUTDOWN) {
914			mask |= POLLIN | POLLRDNORM;
915		}
916
917		/* Connected sockets that can produce data can be written. */
918		if (sk->sk_state == SS_CONNECTED) {
919			if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
920				bool space_avail_now = false;
921				int ret = transport->notify_poll_out(
922						vsk, 1, &space_avail_now);
923				if (ret < 0) {
924					mask |= POLLERR;
925				} else {
926					if (space_avail_now)
927						/* Remove POLLWRBAND since INET
928						 * sockets are not setting it.
929						 */
930						mask |= POLLOUT | POLLWRNORM;
931
932				}
933			}
934		}
935
936		/* Simulate INET socket poll behaviors, which sets
937		 * POLLOUT|POLLWRNORM when peer is closed and nothing to read,
938		 * but local send is not shutdown.
939		 */
940		if (sk->sk_state == SS_UNCONNECTED) {
941			if (!(sk->sk_shutdown & SEND_SHUTDOWN))
942				mask |= POLLOUT | POLLWRNORM;
943
944		}
945
946		release_sock(sk);
947	}
948
949	return mask;
950}
951
952static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg,
953			       size_t len)
954{
955	int err;
956	struct sock *sk;
957	struct vsock_sock *vsk;
958	struct sockaddr_vm *remote_addr;
959
960	if (msg->msg_flags & MSG_OOB)
961		return -EOPNOTSUPP;
962
963	/* For now, MSG_DONTWAIT is always assumed... */
964	err = 0;
965	sk = sock->sk;
966	vsk = vsock_sk(sk);
967
968	lock_sock(sk);
969
970	err = vsock_auto_bind(vsk);
971	if (err)
972		goto out;
973
974
975	/* If the provided message contains an address, use that.  Otherwise
976	 * fall back on the socket's remote handle (if it has been connected).
977	 */
978	if (msg->msg_name &&
979	    vsock_addr_cast(msg->msg_name, msg->msg_namelen,
980			    &remote_addr) == 0) {
981		/* Ensure this address is of the right type and is a valid
982		 * destination.
983		 */
984
985		if (remote_addr->svm_cid == VMADDR_CID_ANY)
986			remote_addr->svm_cid = transport->get_local_cid();
987
988		if (!vsock_addr_bound(remote_addr)) {
989			err = -EINVAL;
990			goto out;
991		}
992	} else if (sock->state == SS_CONNECTED) {
993		remote_addr = &vsk->remote_addr;
994
995		if (remote_addr->svm_cid == VMADDR_CID_ANY)
996			remote_addr->svm_cid = transport->get_local_cid();
997
998		/* XXX Should connect() or this function ensure remote_addr is
999		 * bound?
1000		 */
1001		if (!vsock_addr_bound(&vsk->remote_addr)) {
1002			err = -EINVAL;
1003			goto out;
1004		}
1005	} else {
1006		err = -EINVAL;
1007		goto out;
1008	}
1009
1010	if (!transport->dgram_allow(remote_addr->svm_cid,
1011				    remote_addr->svm_port)) {
1012		err = -EINVAL;
1013		goto out;
1014	}
1015
1016	err = transport->dgram_enqueue(vsk, remote_addr, msg, len);
1017
1018out:
1019	release_sock(sk);
1020	return err;
1021}
1022
1023static int vsock_dgram_connect(struct socket *sock,
1024			       struct sockaddr *addr, int addr_len, int flags)
1025{
1026	int err;
1027	struct sock *sk;
1028	struct vsock_sock *vsk;
1029	struct sockaddr_vm *remote_addr;
1030
1031	sk = sock->sk;
1032	vsk = vsock_sk(sk);
1033
1034	err = vsock_addr_cast(addr, addr_len, &remote_addr);
1035	if (err == -EAFNOSUPPORT && remote_addr->svm_family == AF_UNSPEC) {
1036		lock_sock(sk);
1037		vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY,
1038				VMADDR_PORT_ANY);
1039		sock->state = SS_UNCONNECTED;
1040		release_sock(sk);
1041		return 0;
1042	} else if (err != 0)
1043		return -EINVAL;
1044
1045	lock_sock(sk);
1046
1047	err = vsock_auto_bind(vsk);
1048	if (err)
1049		goto out;
1050
1051	if (!transport->dgram_allow(remote_addr->svm_cid,
1052				    remote_addr->svm_port)) {
1053		err = -EINVAL;
1054		goto out;
1055	}
1056
1057	memcpy(&vsk->remote_addr, remote_addr, sizeof(vsk->remote_addr));
1058	sock->state = SS_CONNECTED;
1059
1060out:
1061	release_sock(sk);
1062	return err;
1063}
1064
1065static int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg,
1066			       size_t len, int flags)
1067{
1068	return transport->dgram_dequeue(vsock_sk(sock->sk), msg, len, flags);
1069}
1070
1071static const struct proto_ops vsock_dgram_ops = {
1072	.family = PF_VSOCK,
1073	.owner = THIS_MODULE,
1074	.release = vsock_release,
1075	.bind = vsock_bind,
1076	.connect = vsock_dgram_connect,
1077	.socketpair = sock_no_socketpair,
1078	.accept = sock_no_accept,
1079	.getname = vsock_getname,
1080	.poll = vsock_poll,
1081	.ioctl = sock_no_ioctl,
1082	.listen = sock_no_listen,
1083	.shutdown = vsock_shutdown,
1084	.setsockopt = sock_no_setsockopt,
1085	.getsockopt = sock_no_getsockopt,
1086	.sendmsg = vsock_dgram_sendmsg,
1087	.recvmsg = vsock_dgram_recvmsg,
1088	.mmap = sock_no_mmap,
1089	.sendpage = sock_no_sendpage,
1090};
1091
1092static void vsock_connect_timeout(struct work_struct *work)
1093{
1094	struct sock *sk;
1095	struct vsock_sock *vsk;
1096
1097	vsk = container_of(work, struct vsock_sock, dwork.work);
1098	sk = sk_vsock(vsk);
1099
1100	lock_sock(sk);
1101	if (sk->sk_state == SS_CONNECTING &&
1102	    (sk->sk_shutdown != SHUTDOWN_MASK)) {
1103		sk->sk_state = SS_UNCONNECTED;
1104		sk->sk_err = ETIMEDOUT;
1105		sk->sk_error_report(sk);
1106	}
1107	release_sock(sk);
1108
1109	sock_put(sk);
1110}
1111
1112static int vsock_stream_connect(struct socket *sock, struct sockaddr *addr,
1113				int addr_len, int flags)
1114{
1115	int err;
1116	struct sock *sk;
1117	struct vsock_sock *vsk;
1118	struct sockaddr_vm *remote_addr;
1119	long timeout;
1120	DEFINE_WAIT(wait);
1121
1122	err = 0;
1123	sk = sock->sk;
1124	vsk = vsock_sk(sk);
1125
1126	lock_sock(sk);
1127
1128	/* XXX AF_UNSPEC should make us disconnect like AF_INET. */
1129	switch (sock->state) {
1130	case SS_CONNECTED:
1131		err = -EISCONN;
1132		goto out;
1133	case SS_DISCONNECTING:
1134		err = -EINVAL;
1135		goto out;
1136	case SS_CONNECTING:
1137		/* This continues on so we can move sock into the SS_CONNECTED
1138		 * state once the connection has completed (at which point err
1139		 * will be set to zero also).  Otherwise, we will either wait
1140		 * for the connection or return -EALREADY should this be a
1141		 * non-blocking call.
1142		 */
1143		err = -EALREADY;
1144		break;
1145	default:
1146		if ((sk->sk_state == VSOCK_SS_LISTEN) ||
1147		    vsock_addr_cast(addr, addr_len, &remote_addr) != 0) {
1148			err = -EINVAL;
1149			goto out;
1150		}
1151
1152		/* The hypervisor and well-known contexts do not have socket
1153		 * endpoints.
1154		 */
1155		if (!transport->stream_allow(remote_addr->svm_cid,
1156					     remote_addr->svm_port)) {
1157			err = -ENETUNREACH;
1158			goto out;
1159		}
1160
1161		/* Set the remote address that we are connecting to. */
1162		memcpy(&vsk->remote_addr, remote_addr,
1163		       sizeof(vsk->remote_addr));
1164
1165		err = vsock_auto_bind(vsk);
1166		if (err)
1167			goto out;
1168
1169		sk->sk_state = SS_CONNECTING;
1170
1171		err = transport->connect(vsk);
1172		if (err < 0)
1173			goto out;
1174
1175		/* Mark sock as connecting and set the error code to in
1176		 * progress in case this is a non-blocking connect.
1177		 */
1178		sock->state = SS_CONNECTING;
1179		err = -EINPROGRESS;
1180	}
1181
1182	/* The receive path will handle all communication until we are able to
1183	 * enter the connected state.  Here we wait for the connection to be
1184	 * completed or a notification of an error.
1185	 */
1186	timeout = vsk->connect_timeout;
1187	prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1188
1189	while (sk->sk_state != SS_CONNECTED && sk->sk_err == 0) {
1190		if (flags & O_NONBLOCK) {
1191			/* If we're not going to block, we schedule a timeout
1192			 * function to generate a timeout on the connection
1193			 * attempt, in case the peer doesn't respond in a
1194			 * timely manner. We hold on to the socket until the
1195			 * timeout fires.
1196			 */
1197			sock_hold(sk);
1198			INIT_DELAYED_WORK(&vsk->dwork,
1199					  vsock_connect_timeout);
1200			schedule_delayed_work(&vsk->dwork, timeout);
1201
1202			/* Skip ahead to preserve error code set above. */
1203			goto out_wait;
1204		}
1205
1206		release_sock(sk);
1207		timeout = schedule_timeout(timeout);
1208		lock_sock(sk);
1209
1210		if (signal_pending(current)) {
1211			err = sock_intr_errno(timeout);
1212			goto out_wait_error;
1213		} else if (timeout == 0) {
1214			err = -ETIMEDOUT;
1215			goto out_wait_error;
1216		}
1217
1218		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1219	}
1220
1221	if (sk->sk_err) {
1222		err = -sk->sk_err;
1223		goto out_wait_error;
1224	} else
1225		err = 0;
1226
1227out_wait:
1228	finish_wait(sk_sleep(sk), &wait);
1229out:
1230	release_sock(sk);
1231	return err;
1232
1233out_wait_error:
1234	sk->sk_state = SS_UNCONNECTED;
1235	sock->state = SS_UNCONNECTED;
1236	goto out_wait;
1237}
1238
1239static int vsock_accept(struct socket *sock, struct socket *newsock, int flags)
1240{
1241	struct sock *listener;
1242	int err;
1243	struct sock *connected;
1244	struct vsock_sock *vconnected;
1245	long timeout;
1246	DEFINE_WAIT(wait);
1247
1248	err = 0;
1249	listener = sock->sk;
1250
1251	lock_sock(listener);
1252
1253	if (sock->type != SOCK_STREAM) {
1254		err = -EOPNOTSUPP;
1255		goto out;
1256	}
1257
1258	if (listener->sk_state != VSOCK_SS_LISTEN) {
1259		err = -EINVAL;
1260		goto out;
1261	}
1262
1263	/* Wait for children sockets to appear; these are the new sockets
1264	 * created upon connection establishment.
1265	 */
1266	timeout = sock_sndtimeo(listener, flags & O_NONBLOCK);
1267	prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1268
1269	while ((connected = vsock_dequeue_accept(listener)) == NULL &&
1270	       listener->sk_err == 0) {
1271		release_sock(listener);
1272		timeout = schedule_timeout(timeout);
1273		lock_sock(listener);
1274
1275		if (signal_pending(current)) {
1276			err = sock_intr_errno(timeout);
1277			goto out_wait;
1278		} else if (timeout == 0) {
1279			err = -EAGAIN;
1280			goto out_wait;
1281		}
1282
1283		prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE);
1284	}
1285
1286	if (listener->sk_err)
1287		err = -listener->sk_err;
1288
1289	if (connected) {
1290		listener->sk_ack_backlog--;
1291
1292		lock_sock(connected);
1293		vconnected = vsock_sk(connected);
1294
1295		/* If the listener socket has received an error, then we should
1296		 * reject this socket and return.  Note that we simply mark the
1297		 * socket rejected, drop our reference, and let the cleanup
1298		 * function handle the cleanup; the fact that we found it in
1299		 * the listener's accept queue guarantees that the cleanup
1300		 * function hasn't run yet.
1301		 */
1302		if (err) {
1303			vconnected->rejected = true;
1304			release_sock(connected);
1305			sock_put(connected);
1306			goto out_wait;
1307		}
1308
1309		newsock->state = SS_CONNECTED;
1310		sock_graft(connected, newsock);
1311		release_sock(connected);
1312		sock_put(connected);
1313	}
1314
1315out_wait:
1316	finish_wait(sk_sleep(listener), &wait);
1317out:
1318	release_sock(listener);
1319	return err;
1320}
1321
1322static int vsock_listen(struct socket *sock, int backlog)
1323{
1324	int err;
1325	struct sock *sk;
1326	struct vsock_sock *vsk;
1327
1328	sk = sock->sk;
1329
1330	lock_sock(sk);
1331
1332	if (sock->type != SOCK_STREAM) {
1333		err = -EOPNOTSUPP;
1334		goto out;
1335	}
1336
1337	if (sock->state != SS_UNCONNECTED) {
1338		err = -EINVAL;
1339		goto out;
1340	}
1341
1342	vsk = vsock_sk(sk);
1343
1344	if (!vsock_addr_bound(&vsk->local_addr)) {
1345		err = -EINVAL;
1346		goto out;
1347	}
1348
1349	sk->sk_max_ack_backlog = backlog;
1350	sk->sk_state = VSOCK_SS_LISTEN;
1351
1352	err = 0;
1353
1354out:
1355	release_sock(sk);
1356	return err;
1357}
1358
1359static int vsock_stream_setsockopt(struct socket *sock,
1360				   int level,
1361				   int optname,
1362				   char __user *optval,
1363				   unsigned int optlen)
1364{
1365	int err;
1366	struct sock *sk;
1367	struct vsock_sock *vsk;
1368	u64 val;
1369
1370	if (level != AF_VSOCK)
1371		return -ENOPROTOOPT;
1372
1373#define COPY_IN(_v)                                       \
1374	do {						  \
1375		if (optlen < sizeof(_v)) {		  \
1376			err = -EINVAL;			  \
1377			goto exit;			  \
1378		}					  \
1379		if (copy_from_user(&_v, optval, sizeof(_v)) != 0) {	\
1380			err = -EFAULT;					\
1381			goto exit;					\
1382		}							\
1383	} while (0)
1384
1385	err = 0;
1386	sk = sock->sk;
1387	vsk = vsock_sk(sk);
1388
1389	lock_sock(sk);
1390
1391	switch (optname) {
1392	case SO_VM_SOCKETS_BUFFER_SIZE:
1393		COPY_IN(val);
1394		transport->set_buffer_size(vsk, val);
1395		break;
1396
1397	case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1398		COPY_IN(val);
1399		transport->set_max_buffer_size(vsk, val);
1400		break;
1401
1402	case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1403		COPY_IN(val);
1404		transport->set_min_buffer_size(vsk, val);
1405		break;
1406
1407	case SO_VM_SOCKETS_CONNECT_TIMEOUT: {
1408		struct timeval tv;
1409		COPY_IN(tv);
1410		if (tv.tv_sec >= 0 && tv.tv_usec < USEC_PER_SEC &&
1411		    tv.tv_sec < (MAX_SCHEDULE_TIMEOUT / HZ - 1)) {
1412			vsk->connect_timeout = tv.tv_sec * HZ +
1413			    DIV_ROUND_UP(tv.tv_usec, (1000000 / HZ));
1414			if (vsk->connect_timeout == 0)
1415				vsk->connect_timeout =
1416				    VSOCK_DEFAULT_CONNECT_TIMEOUT;
1417
1418		} else {
1419			err = -ERANGE;
1420		}
1421		break;
1422	}
1423
1424	default:
1425		err = -ENOPROTOOPT;
1426		break;
1427	}
1428
1429#undef COPY_IN
1430
1431exit:
1432	release_sock(sk);
1433	return err;
1434}
1435
1436static int vsock_stream_getsockopt(struct socket *sock,
1437				   int level, int optname,
1438				   char __user *optval,
1439				   int __user *optlen)
1440{
1441	int err;
1442	int len;
1443	struct sock *sk;
1444	struct vsock_sock *vsk;
1445	u64 val;
1446
1447	if (level != AF_VSOCK)
1448		return -ENOPROTOOPT;
1449
1450	err = get_user(len, optlen);
1451	if (err != 0)
1452		return err;
1453
1454#define COPY_OUT(_v)                            \
1455	do {					\
1456		if (len < sizeof(_v))		\
1457			return -EINVAL;		\
1458						\
1459		len = sizeof(_v);		\
1460		if (copy_to_user(optval, &_v, len) != 0)	\
1461			return -EFAULT;				\
1462								\
1463	} while (0)
1464
1465	err = 0;
1466	sk = sock->sk;
1467	vsk = vsock_sk(sk);
1468
1469	switch (optname) {
1470	case SO_VM_SOCKETS_BUFFER_SIZE:
1471		val = transport->get_buffer_size(vsk);
1472		COPY_OUT(val);
1473		break;
1474
1475	case SO_VM_SOCKETS_BUFFER_MAX_SIZE:
1476		val = transport->get_max_buffer_size(vsk);
1477		COPY_OUT(val);
1478		break;
1479
1480	case SO_VM_SOCKETS_BUFFER_MIN_SIZE:
1481		val = transport->get_min_buffer_size(vsk);
1482		COPY_OUT(val);
1483		break;
1484
1485	case SO_VM_SOCKETS_CONNECT_TIMEOUT: {
1486		struct timeval tv;
1487		tv.tv_sec = vsk->connect_timeout / HZ;
1488		tv.tv_usec =
1489		    (vsk->connect_timeout -
1490		     tv.tv_sec * HZ) * (1000000 / HZ);
1491		COPY_OUT(tv);
1492		break;
1493	}
1494	default:
1495		return -ENOPROTOOPT;
1496	}
1497
1498	err = put_user(len, optlen);
1499	if (err != 0)
1500		return -EFAULT;
1501
1502#undef COPY_OUT
1503
1504	return 0;
1505}
1506
1507static int vsock_stream_sendmsg(struct socket *sock, struct msghdr *msg,
1508				size_t len)
1509{
1510	struct sock *sk;
1511	struct vsock_sock *vsk;
1512	ssize_t total_written;
1513	long timeout;
1514	int err;
1515	struct vsock_transport_send_notify_data send_data;
1516
1517	DEFINE_WAIT(wait);
1518
1519	sk = sock->sk;
1520	vsk = vsock_sk(sk);
1521	total_written = 0;
1522	err = 0;
1523
1524	if (msg->msg_flags & MSG_OOB)
1525		return -EOPNOTSUPP;
1526
1527	lock_sock(sk);
1528
1529	/* Callers should not provide a destination with stream sockets. */
1530	if (msg->msg_namelen) {
1531		err = sk->sk_state == SS_CONNECTED ? -EISCONN : -EOPNOTSUPP;
1532		goto out;
1533	}
1534
1535	/* Send data only if both sides are not shutdown in the direction. */
1536	if (sk->sk_shutdown & SEND_SHUTDOWN ||
1537	    vsk->peer_shutdown & RCV_SHUTDOWN) {
1538		err = -EPIPE;
1539		goto out;
1540	}
1541
1542	if (sk->sk_state != SS_CONNECTED ||
1543	    !vsock_addr_bound(&vsk->local_addr)) {
1544		err = -ENOTCONN;
1545		goto out;
1546	}
1547
1548	if (!vsock_addr_bound(&vsk->remote_addr)) {
1549		err = -EDESTADDRREQ;
1550		goto out;
1551	}
1552
1553	/* Wait for room in the produce queue to enqueue our user's data. */
1554	timeout = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1555
1556	err = transport->notify_send_init(vsk, &send_data);
1557	if (err < 0)
1558		goto out;
1559
1560	prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1561
1562	while (total_written < len) {
1563		ssize_t written;
1564
1565		while (vsock_stream_has_space(vsk) == 0 &&
1566		       sk->sk_err == 0 &&
1567		       !(sk->sk_shutdown & SEND_SHUTDOWN) &&
1568		       !(vsk->peer_shutdown & RCV_SHUTDOWN)) {
1569
1570			/* Don't wait for non-blocking sockets. */
1571			if (timeout == 0) {
1572				err = -EAGAIN;
1573				goto out_wait;
1574			}
1575
1576			err = transport->notify_send_pre_block(vsk, &send_data);
1577			if (err < 0)
1578				goto out_wait;
1579
1580			release_sock(sk);
1581			timeout = schedule_timeout(timeout);
1582			lock_sock(sk);
1583			if (signal_pending(current)) {
1584				err = sock_intr_errno(timeout);
1585				goto out_wait;
1586			} else if (timeout == 0) {
1587				err = -EAGAIN;
1588				goto out_wait;
1589			}
1590
1591			prepare_to_wait(sk_sleep(sk), &wait,
1592					TASK_INTERRUPTIBLE);
1593		}
1594
1595		/* These checks occur both as part of and after the loop
1596		 * conditional since we need to check before and after
1597		 * sleeping.
1598		 */
1599		if (sk->sk_err) {
1600			err = -sk->sk_err;
1601			goto out_wait;
1602		} else if ((sk->sk_shutdown & SEND_SHUTDOWN) ||
1603			   (vsk->peer_shutdown & RCV_SHUTDOWN)) {
1604			err = -EPIPE;
1605			goto out_wait;
1606		}
1607
1608		err = transport->notify_send_pre_enqueue(vsk, &send_data);
1609		if (err < 0)
1610			goto out_wait;
1611
1612		/* Note that enqueue will only write as many bytes as are free
1613		 * in the produce queue, so we don't need to ensure len is
1614		 * smaller than the queue size.  It is the caller's
1615		 * responsibility to check how many bytes we were able to send.
1616		 */
1617
1618		written = transport->stream_enqueue(
1619				vsk, msg,
1620				len - total_written);
1621		if (written < 0) {
1622			err = -ENOMEM;
1623			goto out_wait;
1624		}
1625
1626		total_written += written;
1627
1628		err = transport->notify_send_post_enqueue(
1629				vsk, written, &send_data);
1630		if (err < 0)
1631			goto out_wait;
1632
1633	}
1634
1635out_wait:
1636	if (total_written > 0)
1637		err = total_written;
1638	finish_wait(sk_sleep(sk), &wait);
1639out:
1640	release_sock(sk);
1641	return err;
1642}
1643
1644
1645static int
1646vsock_stream_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
1647		     int flags)
1648{
1649	struct sock *sk;
1650	struct vsock_sock *vsk;
1651	int err;
1652	size_t target;
1653	ssize_t copied;
1654	long timeout;
1655	struct vsock_transport_recv_notify_data recv_data;
1656
1657	DEFINE_WAIT(wait);
1658
1659	sk = sock->sk;
1660	vsk = vsock_sk(sk);
1661	err = 0;
1662
1663	lock_sock(sk);
1664
1665	if (sk->sk_state != SS_CONNECTED) {
1666		/* Recvmsg is supposed to return 0 if a peer performs an
1667		 * orderly shutdown. Differentiate between that case and when a
1668		 * peer has not connected or a local shutdown occured with the
1669		 * SOCK_DONE flag.
1670		 */
1671		if (sock_flag(sk, SOCK_DONE))
1672			err = 0;
1673		else
1674			err = -ENOTCONN;
1675
1676		goto out;
1677	}
1678
1679	if (flags & MSG_OOB) {
1680		err = -EOPNOTSUPP;
1681		goto out;
1682	}
1683
1684	/* We don't check peer_shutdown flag here since peer may actually shut
1685	 * down, but there can be data in the queue that a local socket can
1686	 * receive.
1687	 */
1688	if (sk->sk_shutdown & RCV_SHUTDOWN) {
1689		err = 0;
1690		goto out;
1691	}
1692
1693	/* It is valid on Linux to pass in a zero-length receive buffer.  This
1694	 * is not an error.  We may as well bail out now.
1695	 */
1696	if (!len) {
1697		err = 0;
1698		goto out;
1699	}
1700
1701	/* We must not copy less than target bytes into the user's buffer
1702	 * before returning successfully, so we wait for the consume queue to
1703	 * have that much data to consume before dequeueing.  Note that this
1704	 * makes it impossible to handle cases where target is greater than the
1705	 * queue size.
1706	 */
1707	target = sock_rcvlowat(sk, flags & MSG_WAITALL, len);
1708	if (target >= transport->stream_rcvhiwat(vsk)) {
1709		err = -ENOMEM;
1710		goto out;
1711	}
1712	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1713	copied = 0;
1714
1715	err = transport->notify_recv_init(vsk, target, &recv_data);
1716	if (err < 0)
1717		goto out;
1718
1719	prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1720
1721	while (1) {
1722		s64 ready = vsock_stream_has_data(vsk);
1723
1724		if (ready < 0) {
1725			/* Invalid queue pair content. XXX This should be
1726			 * changed to a connection reset in a later change.
1727			 */
1728
1729			err = -ENOMEM;
1730			goto out_wait;
1731		} else if (ready > 0) {
1732			ssize_t read;
1733
1734			err = transport->notify_recv_pre_dequeue(
1735					vsk, target, &recv_data);
1736			if (err < 0)
1737				break;
1738
1739			read = transport->stream_dequeue(
1740					vsk, msg,
1741					len - copied, flags);
1742			if (read < 0) {
1743				err = -ENOMEM;
1744				break;
1745			}
1746
1747			copied += read;
1748
1749			err = transport->notify_recv_post_dequeue(
1750					vsk, target, read,
1751					!(flags & MSG_PEEK), &recv_data);
1752			if (err < 0)
1753				goto out_wait;
1754
1755			if (read >= target || flags & MSG_PEEK)
1756				break;
1757
1758			target -= read;
1759		} else {
1760			if (sk->sk_err != 0 || (sk->sk_shutdown & RCV_SHUTDOWN)
1761			    || (vsk->peer_shutdown & SEND_SHUTDOWN)) {
1762				break;
1763			}
1764			/* Don't wait for non-blocking sockets. */
1765			if (timeout == 0) {
1766				err = -EAGAIN;
1767				break;
1768			}
1769
1770			err = transport->notify_recv_pre_block(
1771					vsk, target, &recv_data);
1772			if (err < 0)
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				break;
1782			} else if (timeout == 0) {
1783				err = -EAGAIN;
1784				break;
1785			}
1786
1787			prepare_to_wait(sk_sleep(sk), &wait,
1788					TASK_INTERRUPTIBLE);
1789		}
1790	}
1791
1792	if (sk->sk_err)
1793		err = -sk->sk_err;
1794	else if (sk->sk_shutdown & RCV_SHUTDOWN)
1795		err = 0;
1796
1797	if (copied > 0)
1798		err = copied;
1799
1800out_wait:
1801	finish_wait(sk_sleep(sk), &wait);
1802out:
1803	release_sock(sk);
1804	return err;
1805}
1806
1807static const struct proto_ops vsock_stream_ops = {
1808	.family = PF_VSOCK,
1809	.owner = THIS_MODULE,
1810	.release = vsock_release,
1811	.bind = vsock_bind,
1812	.connect = vsock_stream_connect,
1813	.socketpair = sock_no_socketpair,
1814	.accept = vsock_accept,
1815	.getname = vsock_getname,
1816	.poll = vsock_poll,
1817	.ioctl = sock_no_ioctl,
1818	.listen = vsock_listen,
1819	.shutdown = vsock_shutdown,
1820	.setsockopt = vsock_stream_setsockopt,
1821	.getsockopt = vsock_stream_getsockopt,
1822	.sendmsg = vsock_stream_sendmsg,
1823	.recvmsg = vsock_stream_recvmsg,
1824	.mmap = sock_no_mmap,
1825	.sendpage = sock_no_sendpage,
1826};
1827
1828static int vsock_create(struct net *net, struct socket *sock,
1829			int protocol, int kern)
1830{
1831	if (!sock)
1832		return -EINVAL;
1833
1834	if (protocol && protocol != PF_VSOCK)
1835		return -EPROTONOSUPPORT;
1836
1837	switch (sock->type) {
1838	case SOCK_DGRAM:
1839		sock->ops = &vsock_dgram_ops;
1840		break;
1841	case SOCK_STREAM:
1842		sock->ops = &vsock_stream_ops;
1843		break;
1844	default:
1845		return -ESOCKTNOSUPPORT;
1846	}
1847
1848	sock->state = SS_UNCONNECTED;
1849
1850	return __vsock_create(net, sock, NULL, GFP_KERNEL, 0, kern) ? 0 : -ENOMEM;
1851}
1852
1853static const struct net_proto_family vsock_family_ops = {
1854	.family = AF_VSOCK,
1855	.create = vsock_create,
1856	.owner = THIS_MODULE,
1857};
1858
1859static long vsock_dev_do_ioctl(struct file *filp,
1860			       unsigned int cmd, void __user *ptr)
1861{
1862	u32 __user *p = ptr;
1863	int retval = 0;
1864
1865	switch (cmd) {
1866	case IOCTL_VM_SOCKETS_GET_LOCAL_CID:
1867		if (put_user(transport->get_local_cid(), p) != 0)
1868			retval = -EFAULT;
1869		break;
1870
1871	default:
1872		pr_err("Unknown ioctl %d\n", cmd);
1873		retval = -EINVAL;
1874	}
1875
1876	return retval;
1877}
1878
1879static long vsock_dev_ioctl(struct file *filp,
1880			    unsigned int cmd, unsigned long arg)
1881{
1882	return vsock_dev_do_ioctl(filp, cmd, (void __user *)arg);
1883}
1884
1885#ifdef CONFIG_COMPAT
1886static long vsock_dev_compat_ioctl(struct file *filp,
1887				   unsigned int cmd, unsigned long arg)
1888{
1889	return vsock_dev_do_ioctl(filp, cmd, compat_ptr(arg));
1890}
1891#endif
1892
1893static const struct file_operations vsock_device_ops = {
1894	.owner		= THIS_MODULE,
1895	.unlocked_ioctl	= vsock_dev_ioctl,
1896#ifdef CONFIG_COMPAT
1897	.compat_ioctl	= vsock_dev_compat_ioctl,
1898#endif
1899	.open		= nonseekable_open,
1900};
1901
1902static struct miscdevice vsock_device = {
1903	.name		= "vsock",
1904	.fops		= &vsock_device_ops,
1905};
1906
1907int __vsock_core_init(const struct vsock_transport *t, struct module *owner)
1908{
1909	int err = mutex_lock_interruptible(&vsock_register_mutex);
1910
1911	if (err)
1912		return err;
1913
1914	if (transport) {
1915		err = -EBUSY;
1916		goto err_busy;
1917	}
1918
1919	/* Transport must be the owner of the protocol so that it can't
1920	 * unload while there are open sockets.
1921	 */
1922	vsock_proto.owner = owner;
1923	transport = t;
1924
1925	vsock_init_tables();
1926
1927	vsock_device.minor = MISC_DYNAMIC_MINOR;
1928	err = misc_register(&vsock_device);
1929	if (err) {
1930		pr_err("Failed to register misc device\n");
1931		goto err_reset_transport;
1932	}
1933
1934	err = proto_register(&vsock_proto, 1);	/* we want our slab */
1935	if (err) {
1936		pr_err("Cannot register vsock protocol\n");
1937		goto err_deregister_misc;
1938	}
1939
1940	err = sock_register(&vsock_family_ops);
1941	if (err) {
1942		pr_err("could not register af_vsock (%d) address family: %d\n",
1943		       AF_VSOCK, err);
1944		goto err_unregister_proto;
1945	}
1946
1947	mutex_unlock(&vsock_register_mutex);
1948	return 0;
1949
1950err_unregister_proto:
1951	proto_unregister(&vsock_proto);
1952err_deregister_misc:
1953	misc_deregister(&vsock_device);
1954err_reset_transport:
1955	transport = NULL;
1956err_busy:
1957	mutex_unlock(&vsock_register_mutex);
1958	return err;
1959}
1960EXPORT_SYMBOL_GPL(__vsock_core_init);
1961
1962void vsock_core_exit(void)
1963{
1964	mutex_lock(&vsock_register_mutex);
1965
1966	misc_deregister(&vsock_device);
1967	sock_unregister(AF_VSOCK);
1968	proto_unregister(&vsock_proto);
1969
1970	/* We do not want the assignment below re-ordered. */
1971	mb();
1972	transport = NULL;
1973
1974	mutex_unlock(&vsock_register_mutex);
1975}
1976EXPORT_SYMBOL_GPL(vsock_core_exit);
1977
1978MODULE_AUTHOR("VMware, Inc.");
1979MODULE_DESCRIPTION("VMware Virtual Socket Family");
1980MODULE_VERSION("1.0.1.0-k");
1981MODULE_LICENSE("GPL v2");
1982