1685a6bf8SThomas Gleixner // SPDX-License-Identifier: GPL-2.0-only 2d021c344SAndy King /* 3d021c344SAndy King * VMware vSockets Driver 4d021c344SAndy King * 5d021c344SAndy King * Copyright (C) 2007-2013 VMware, Inc. All rights reserved. 6d021c344SAndy King */ 7d021c344SAndy King 8d021c344SAndy King /* Implementation notes: 9d021c344SAndy King * 10d021c344SAndy King * - There are two kinds of sockets: those created by user action (such as 11d021c344SAndy King * calling socket(2)) and those created by incoming connection request packets. 12d021c344SAndy King * 13d021c344SAndy King * - There are two "global" tables, one for bound sockets (sockets that have 14d021c344SAndy King * specified an address that they are responsible for) and one for connected 15d021c344SAndy King * sockets (sockets that have established a connection with another socket). 16d021c344SAndy King * These tables are "global" in that all sockets on the system are placed 17d021c344SAndy King * within them. - Note, though, that the bound table contains an extra entry 18d021c344SAndy King * for a list of unbound sockets and SOCK_DGRAM sockets will always remain in 19d021c344SAndy King * that list. The bound table is used solely for lookup of sockets when packets 20d021c344SAndy King * are received and that's not necessary for SOCK_DGRAM sockets since we create 21d021c344SAndy King * a datagram handle for each and need not perform a lookup. Keeping SOCK_DGRAM 22d021c344SAndy King * sockets out of the bound hash buckets will reduce the chance of collisions 23d021c344SAndy King * when looking for SOCK_STREAM sockets and prevents us from having to check the 24d021c344SAndy King * socket type in the hash table lookups. 25d021c344SAndy King * 26d021c344SAndy King * - Sockets created by user action will either be "client" sockets that 27d021c344SAndy King * initiate a connection or "server" sockets that listen for connections; we do 28d021c344SAndy King * not support simultaneous connects (two "client" sockets connecting). 29d021c344SAndy King * 30d021c344SAndy King * - "Server" sockets are referred to as listener sockets throughout this 313b4477d2SStefan Hajnoczi * implementation because they are in the TCP_LISTEN state. When a 32ea3803c1SStefan Hajnoczi * connection request is received (the second kind of socket mentioned above), 33ea3803c1SStefan Hajnoczi * we create a new socket and refer to it as a pending socket. These pending 34ea3803c1SStefan Hajnoczi * sockets are placed on the pending connection list of the listener socket. 35ea3803c1SStefan Hajnoczi * When future packets are received for the address the listener socket is 36ea3803c1SStefan Hajnoczi * bound to, we check if the source of the packet is from one that has an 37ea3803c1SStefan Hajnoczi * existing pending connection. If it does, we process the packet for the 38ea3803c1SStefan Hajnoczi * pending socket. When that socket reaches the connected state, it is removed 39ea3803c1SStefan Hajnoczi * from the listener socket's pending list and enqueued in the listener 40ea3803c1SStefan Hajnoczi * socket's accept queue. Callers of accept(2) will accept connected sockets 41ea3803c1SStefan Hajnoczi * from the listener socket's accept queue. If the socket cannot be accepted 42ea3803c1SStefan Hajnoczi * for some reason then it is marked rejected. Once the connection is 43ea3803c1SStefan Hajnoczi * accepted, it is owned by the user process and the responsibility for cleanup 44ea3803c1SStefan Hajnoczi * falls with that user process. 45d021c344SAndy King * 46d021c344SAndy King * - It is possible that these pending sockets will never reach the connected 47d021c344SAndy King * state; in fact, we may never receive another packet after the connection 48d021c344SAndy King * request. Because of this, we must schedule a cleanup function to run in the 49d021c344SAndy King * future, after some amount of time passes where a connection should have been 50d021c344SAndy King * established. This function ensures that the socket is off all lists so it 51d021c344SAndy King * cannot be retrieved, then drops all references to the socket so it is cleaned 52d021c344SAndy King * up (sock_put() -> sk_free() -> our sk_destruct implementation). Note this 53d021c344SAndy King * function will also cleanup rejected sockets, those that reach the connected 54d021c344SAndy King * state but leave it before they have been accepted. 55d021c344SAndy King * 564192f672SStefan Hajnoczi * - Lock ordering for pending or accept queue sockets is: 574192f672SStefan Hajnoczi * 584192f672SStefan Hajnoczi * lock_sock(listener); 594192f672SStefan Hajnoczi * lock_sock_nested(pending, SINGLE_DEPTH_NESTING); 604192f672SStefan Hajnoczi * 614192f672SStefan Hajnoczi * Using explicit nested locking keeps lockdep happy since normally only one 624192f672SStefan Hajnoczi * lock of a given class may be taken at a time. 634192f672SStefan Hajnoczi * 64d021c344SAndy King * - Sockets created by user action will be cleaned up when the user process 65d021c344SAndy King * calls close(2), causing our release implementation to be called. Our release 66d021c344SAndy King * implementation will perform some cleanup then drop the last reference so our 67d021c344SAndy King * sk_destruct implementation is invoked. Our sk_destruct implementation will 68d021c344SAndy King * perform additional cleanup that's common for both types of sockets. 69d021c344SAndy King * 70d021c344SAndy King * - A socket's reference count is what ensures that the structure won't be 71d021c344SAndy King * freed. Each entry in a list (such as the "global" bound and connected tables 72d021c344SAndy King * and the listener socket's pending list and connected queue) ensures a 73d021c344SAndy King * reference. When we defer work until process context and pass a socket as our 74d021c344SAndy King * argument, we must ensure the reference count is increased to ensure the 75d021c344SAndy King * socket isn't freed before the function is run; the deferred function will 76d021c344SAndy King * then drop the reference. 773b4477d2SStefan Hajnoczi * 783b4477d2SStefan Hajnoczi * - sk->sk_state uses the TCP state constants because they are widely used by 793b4477d2SStefan Hajnoczi * other address families and exposed to userspace tools like ss(8): 803b4477d2SStefan Hajnoczi * 813b4477d2SStefan Hajnoczi * TCP_CLOSE - unconnected 823b4477d2SStefan Hajnoczi * TCP_SYN_SENT - connecting 833b4477d2SStefan Hajnoczi * TCP_ESTABLISHED - connected 843b4477d2SStefan Hajnoczi * TCP_CLOSING - disconnecting 853b4477d2SStefan Hajnoczi * TCP_LISTEN - listening 86d021c344SAndy King */ 87d021c344SAndy King 88d021c344SAndy King #include <linux/types.h> 89d021c344SAndy King #include <linux/bitops.h> 90d021c344SAndy King #include <linux/cred.h> 91d021c344SAndy King #include <linux/init.h> 92d021c344SAndy King #include <linux/io.h> 93d021c344SAndy King #include <linux/kernel.h> 94174cd4b1SIngo Molnar #include <linux/sched/signal.h> 95d021c344SAndy King #include <linux/kmod.h> 96d021c344SAndy King #include <linux/list.h> 97d021c344SAndy King #include <linux/miscdevice.h> 98d021c344SAndy King #include <linux/module.h> 99d021c344SAndy King #include <linux/mutex.h> 100d021c344SAndy King #include <linux/net.h> 101d021c344SAndy King #include <linux/poll.h> 1028236b08cSLepton Wu #include <linux/random.h> 103d021c344SAndy King #include <linux/skbuff.h> 104d021c344SAndy King #include <linux/smp.h> 105d021c344SAndy King #include <linux/socket.h> 106d021c344SAndy King #include <linux/stddef.h> 107d021c344SAndy King #include <linux/unistd.h> 108d021c344SAndy King #include <linux/wait.h> 109d021c344SAndy King #include <linux/workqueue.h> 110d021c344SAndy King #include <net/sock.h> 11182a54d0eSAsias He #include <net/af_vsock.h> 112d021c344SAndy King 113d021c344SAndy King static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr); 114d021c344SAndy King static void vsock_sk_destruct(struct sock *sk); 115d021c344SAndy King static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb); 116d021c344SAndy King 117d021c344SAndy King /* Protocol family. */ 118d021c344SAndy King static struct proto vsock_proto = { 119d021c344SAndy King .name = "AF_VSOCK", 120d021c344SAndy King .owner = THIS_MODULE, 121d021c344SAndy King .obj_size = sizeof(struct vsock_sock), 122d021c344SAndy King }; 123d021c344SAndy King 124d021c344SAndy King /* The default peer timeout indicates how long we will wait for a peer response 125d021c344SAndy King * to a control message. 126d021c344SAndy King */ 127d021c344SAndy King #define VSOCK_DEFAULT_CONNECT_TIMEOUT (2 * HZ) 128d021c344SAndy King 129b9f2b0ffSStefano Garzarella #define VSOCK_DEFAULT_BUFFER_SIZE (1024 * 256) 130b9f2b0ffSStefano Garzarella #define VSOCK_DEFAULT_BUFFER_MAX_SIZE (1024 * 256) 131b9f2b0ffSStefano Garzarella #define VSOCK_DEFAULT_BUFFER_MIN_SIZE 128 132b9f2b0ffSStefano Garzarella 133c0cfa2d8SStefano Garzarella /* Transport used for host->guest communication */ 134c0cfa2d8SStefano Garzarella static const struct vsock_transport *transport_h2g; 135c0cfa2d8SStefano Garzarella /* Transport used for guest->host communication */ 136c0cfa2d8SStefano Garzarella static const struct vsock_transport *transport_g2h; 137c0cfa2d8SStefano Garzarella /* Transport used for DGRAM communication */ 138c0cfa2d8SStefano Garzarella static const struct vsock_transport *transport_dgram; 139d021c344SAndy King static DEFINE_MUTEX(vsock_register_mutex); 140d021c344SAndy King 141d021c344SAndy King /**** UTILS ****/ 142d021c344SAndy King 143d021c344SAndy King /* Each bound VSocket is stored in the bind hash table and each connected 144d021c344SAndy King * VSocket is stored in the connected hash table. 145d021c344SAndy King * 146d021c344SAndy King * Unbound sockets are all put on the same list attached to the end of the hash 147d021c344SAndy King * table (vsock_unbound_sockets). Bound sockets are added to the hash table in 148d021c344SAndy King * the bucket that their local address hashes to (vsock_bound_sockets(addr) 149d021c344SAndy King * represents the list that addr hashes to). 150d021c344SAndy King * 151d021c344SAndy King * Specifically, we initialize the vsock_bind_table array to a size of 152d021c344SAndy King * VSOCK_HASH_SIZE + 1 so that vsock_bind_table[0] through 153d021c344SAndy King * vsock_bind_table[VSOCK_HASH_SIZE - 1] are for bound sockets and 154d021c344SAndy King * vsock_bind_table[VSOCK_HASH_SIZE] is for unbound sockets. The hash function 155a49dd9dcSAsias He * mods with VSOCK_HASH_SIZE to ensure this. 156d021c344SAndy King */ 157d021c344SAndy King #define MAX_PORT_RETRIES 24 158d021c344SAndy King 159a49dd9dcSAsias He #define VSOCK_HASH(addr) ((addr)->svm_port % VSOCK_HASH_SIZE) 160d021c344SAndy King #define vsock_bound_sockets(addr) (&vsock_bind_table[VSOCK_HASH(addr)]) 161d021c344SAndy King #define vsock_unbound_sockets (&vsock_bind_table[VSOCK_HASH_SIZE]) 162d021c344SAndy King 163d021c344SAndy King /* XXX This can probably be implemented in a better way. */ 164d021c344SAndy King #define VSOCK_CONN_HASH(src, dst) \ 165a49dd9dcSAsias He (((src)->svm_cid ^ (dst)->svm_port) % VSOCK_HASH_SIZE) 166d021c344SAndy King #define vsock_connected_sockets(src, dst) \ 167d021c344SAndy King (&vsock_connected_table[VSOCK_CONN_HASH(src, dst)]) 168d021c344SAndy King #define vsock_connected_sockets_vsk(vsk) \ 169d021c344SAndy King vsock_connected_sockets(&(vsk)->remote_addr, &(vsk)->local_addr) 170d021c344SAndy King 17144f20980SStefan Hajnoczi struct list_head vsock_bind_table[VSOCK_HASH_SIZE + 1]; 17244f20980SStefan Hajnoczi EXPORT_SYMBOL_GPL(vsock_bind_table); 17344f20980SStefan Hajnoczi struct list_head vsock_connected_table[VSOCK_HASH_SIZE]; 17444f20980SStefan Hajnoczi EXPORT_SYMBOL_GPL(vsock_connected_table); 17544f20980SStefan Hajnoczi DEFINE_SPINLOCK(vsock_table_lock); 17644f20980SStefan Hajnoczi EXPORT_SYMBOL_GPL(vsock_table_lock); 177d021c344SAndy King 178b3a6dfe8SAsias He /* Autobind this socket to the local address if necessary. */ 179b3a6dfe8SAsias He static int vsock_auto_bind(struct vsock_sock *vsk) 180b3a6dfe8SAsias He { 181b3a6dfe8SAsias He struct sock *sk = sk_vsock(vsk); 182b3a6dfe8SAsias He struct sockaddr_vm local_addr; 183b3a6dfe8SAsias He 184b3a6dfe8SAsias He if (vsock_addr_bound(&vsk->local_addr)) 185b3a6dfe8SAsias He return 0; 186b3a6dfe8SAsias He vsock_addr_init(&local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 187b3a6dfe8SAsias He return __vsock_bind(sk, &local_addr); 188b3a6dfe8SAsias He } 189b3a6dfe8SAsias He 190c0cfa2d8SStefano Garzarella static void vsock_init_tables(void) 191d021c344SAndy King { 192d021c344SAndy King int i; 193d021c344SAndy King 194d021c344SAndy King for (i = 0; i < ARRAY_SIZE(vsock_bind_table); i++) 195d021c344SAndy King INIT_LIST_HEAD(&vsock_bind_table[i]); 196d021c344SAndy King 197d021c344SAndy King for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) 198d021c344SAndy King INIT_LIST_HEAD(&vsock_connected_table[i]); 199d021c344SAndy King } 200d021c344SAndy King 201d021c344SAndy King static void __vsock_insert_bound(struct list_head *list, 202d021c344SAndy King struct vsock_sock *vsk) 203d021c344SAndy King { 204d021c344SAndy King sock_hold(&vsk->sk); 205d021c344SAndy King list_add(&vsk->bound_table, list); 206d021c344SAndy King } 207d021c344SAndy King 208d021c344SAndy King static void __vsock_insert_connected(struct list_head *list, 209d021c344SAndy King struct vsock_sock *vsk) 210d021c344SAndy King { 211d021c344SAndy King sock_hold(&vsk->sk); 212d021c344SAndy King list_add(&vsk->connected_table, list); 213d021c344SAndy King } 214d021c344SAndy King 215d021c344SAndy King static void __vsock_remove_bound(struct vsock_sock *vsk) 216d021c344SAndy King { 217d021c344SAndy King list_del_init(&vsk->bound_table); 218d021c344SAndy King sock_put(&vsk->sk); 219d021c344SAndy King } 220d021c344SAndy King 221d021c344SAndy King static void __vsock_remove_connected(struct vsock_sock *vsk) 222d021c344SAndy King { 223d021c344SAndy King list_del_init(&vsk->connected_table); 224d021c344SAndy King sock_put(&vsk->sk); 225d021c344SAndy King } 226d021c344SAndy King 227d021c344SAndy King static struct sock *__vsock_find_bound_socket(struct sockaddr_vm *addr) 228d021c344SAndy King { 229d021c344SAndy King struct vsock_sock *vsk; 230d021c344SAndy King 231d021c344SAndy King list_for_each_entry(vsk, vsock_bound_sockets(addr), bound_table) 232990454b5SReilly Grant if (addr->svm_port == vsk->local_addr.svm_port) 233d021c344SAndy King return sk_vsock(vsk); 234d021c344SAndy King 235d021c344SAndy King return NULL; 236d021c344SAndy King } 237d021c344SAndy King 238d021c344SAndy King static struct sock *__vsock_find_connected_socket(struct sockaddr_vm *src, 239d021c344SAndy King struct sockaddr_vm *dst) 240d021c344SAndy King { 241d021c344SAndy King struct vsock_sock *vsk; 242d021c344SAndy King 243d021c344SAndy King list_for_each_entry(vsk, vsock_connected_sockets(src, dst), 244d021c344SAndy King connected_table) { 245990454b5SReilly Grant if (vsock_addr_equals_addr(src, &vsk->remote_addr) && 246990454b5SReilly Grant dst->svm_port == vsk->local_addr.svm_port) { 247d021c344SAndy King return sk_vsock(vsk); 248d021c344SAndy King } 249d021c344SAndy King } 250d021c344SAndy King 251d021c344SAndy King return NULL; 252d021c344SAndy King } 253d021c344SAndy King 254d021c344SAndy King static void vsock_insert_unbound(struct vsock_sock *vsk) 255d021c344SAndy King { 256d021c344SAndy King spin_lock_bh(&vsock_table_lock); 257d021c344SAndy King __vsock_insert_bound(vsock_unbound_sockets, vsk); 258d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 259d021c344SAndy King } 260d021c344SAndy King 261d021c344SAndy King void vsock_insert_connected(struct vsock_sock *vsk) 262d021c344SAndy King { 263d021c344SAndy King struct list_head *list = vsock_connected_sockets( 264d021c344SAndy King &vsk->remote_addr, &vsk->local_addr); 265d021c344SAndy King 266d021c344SAndy King spin_lock_bh(&vsock_table_lock); 267d021c344SAndy King __vsock_insert_connected(list, vsk); 268d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 269d021c344SAndy King } 270d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_insert_connected); 271d021c344SAndy King 272d021c344SAndy King void vsock_remove_bound(struct vsock_sock *vsk) 273d021c344SAndy King { 274d021c344SAndy King spin_lock_bh(&vsock_table_lock); 275d5afa82cSSunil Muthuswamy if (__vsock_in_bound_table(vsk)) 276d021c344SAndy King __vsock_remove_bound(vsk); 277d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 278d021c344SAndy King } 279d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_remove_bound); 280d021c344SAndy King 281d021c344SAndy King void vsock_remove_connected(struct vsock_sock *vsk) 282d021c344SAndy King { 283d021c344SAndy King spin_lock_bh(&vsock_table_lock); 284d5afa82cSSunil Muthuswamy if (__vsock_in_connected_table(vsk)) 285d021c344SAndy King __vsock_remove_connected(vsk); 286d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 287d021c344SAndy King } 288d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_remove_connected); 289d021c344SAndy King 290d021c344SAndy King struct sock *vsock_find_bound_socket(struct sockaddr_vm *addr) 291d021c344SAndy King { 292d021c344SAndy King struct sock *sk; 293d021c344SAndy King 294d021c344SAndy King spin_lock_bh(&vsock_table_lock); 295d021c344SAndy King sk = __vsock_find_bound_socket(addr); 296d021c344SAndy King if (sk) 297d021c344SAndy King sock_hold(sk); 298d021c344SAndy King 299d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 300d021c344SAndy King 301d021c344SAndy King return sk; 302d021c344SAndy King } 303d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_find_bound_socket); 304d021c344SAndy King 305d021c344SAndy King struct sock *vsock_find_connected_socket(struct sockaddr_vm *src, 306d021c344SAndy King struct sockaddr_vm *dst) 307d021c344SAndy King { 308d021c344SAndy King struct sock *sk; 309d021c344SAndy King 310d021c344SAndy King spin_lock_bh(&vsock_table_lock); 311d021c344SAndy King sk = __vsock_find_connected_socket(src, dst); 312d021c344SAndy King if (sk) 313d021c344SAndy King sock_hold(sk); 314d021c344SAndy King 315d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 316d021c344SAndy King 317d021c344SAndy King return sk; 318d021c344SAndy King } 319d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_find_connected_socket); 320d021c344SAndy King 3216773b7dcSStefan Hajnoczi void vsock_remove_sock(struct vsock_sock *vsk) 3226773b7dcSStefan Hajnoczi { 3236773b7dcSStefan Hajnoczi vsock_remove_bound(vsk); 3246773b7dcSStefan Hajnoczi vsock_remove_connected(vsk); 3256773b7dcSStefan Hajnoczi } 3266773b7dcSStefan Hajnoczi EXPORT_SYMBOL_GPL(vsock_remove_sock); 3276773b7dcSStefan Hajnoczi 328d021c344SAndy King void vsock_for_each_connected_socket(void (*fn)(struct sock *sk)) 329d021c344SAndy King { 330d021c344SAndy King int i; 331d021c344SAndy King 332d021c344SAndy King spin_lock_bh(&vsock_table_lock); 333d021c344SAndy King 334d021c344SAndy King for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) { 335d021c344SAndy King struct vsock_sock *vsk; 336d021c344SAndy King list_for_each_entry(vsk, &vsock_connected_table[i], 337d9af2d67SJulia Lawall connected_table) 338d021c344SAndy King fn(sk_vsock(vsk)); 339d021c344SAndy King } 340d021c344SAndy King 341d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 342d021c344SAndy King } 343d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_for_each_connected_socket); 344d021c344SAndy King 345d021c344SAndy King void vsock_add_pending(struct sock *listener, struct sock *pending) 346d021c344SAndy King { 347d021c344SAndy King struct vsock_sock *vlistener; 348d021c344SAndy King struct vsock_sock *vpending; 349d021c344SAndy King 350d021c344SAndy King vlistener = vsock_sk(listener); 351d021c344SAndy King vpending = vsock_sk(pending); 352d021c344SAndy King 353d021c344SAndy King sock_hold(pending); 354d021c344SAndy King sock_hold(listener); 355d021c344SAndy King list_add_tail(&vpending->pending_links, &vlistener->pending_links); 356d021c344SAndy King } 357d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_add_pending); 358d021c344SAndy King 359d021c344SAndy King void vsock_remove_pending(struct sock *listener, struct sock *pending) 360d021c344SAndy King { 361d021c344SAndy King struct vsock_sock *vpending = vsock_sk(pending); 362d021c344SAndy King 363d021c344SAndy King list_del_init(&vpending->pending_links); 364d021c344SAndy King sock_put(listener); 365d021c344SAndy King sock_put(pending); 366d021c344SAndy King } 367d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_remove_pending); 368d021c344SAndy King 369d021c344SAndy King void vsock_enqueue_accept(struct sock *listener, struct sock *connected) 370d021c344SAndy King { 371d021c344SAndy King struct vsock_sock *vlistener; 372d021c344SAndy King struct vsock_sock *vconnected; 373d021c344SAndy King 374d021c344SAndy King vlistener = vsock_sk(listener); 375d021c344SAndy King vconnected = vsock_sk(connected); 376d021c344SAndy King 377d021c344SAndy King sock_hold(connected); 378d021c344SAndy King sock_hold(listener); 379d021c344SAndy King list_add_tail(&vconnected->accept_queue, &vlistener->accept_queue); 380d021c344SAndy King } 381d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_enqueue_accept); 382d021c344SAndy King 383*6a2c0962SStefano Garzarella static void vsock_deassign_transport(struct vsock_sock *vsk) 384*6a2c0962SStefano Garzarella { 385*6a2c0962SStefano Garzarella if (!vsk->transport) 386*6a2c0962SStefano Garzarella return; 387*6a2c0962SStefano Garzarella 388*6a2c0962SStefano Garzarella vsk->transport->destruct(vsk); 389*6a2c0962SStefano Garzarella module_put(vsk->transport->module); 390*6a2c0962SStefano Garzarella vsk->transport = NULL; 391*6a2c0962SStefano Garzarella } 392*6a2c0962SStefano Garzarella 393c0cfa2d8SStefano Garzarella /* Assign a transport to a socket and call the .init transport callback. 394c0cfa2d8SStefano Garzarella * 395c0cfa2d8SStefano Garzarella * Note: for stream socket this must be called when vsk->remote_addr is set 396c0cfa2d8SStefano Garzarella * (e.g. during the connect() or when a connection request on a listener 397c0cfa2d8SStefano Garzarella * socket is received). 398c0cfa2d8SStefano Garzarella * The vsk->remote_addr is used to decide which transport to use: 399c0cfa2d8SStefano Garzarella * - remote CID <= VMADDR_CID_HOST will use guest->host transport; 400c0cfa2d8SStefano Garzarella * - remote CID == local_cid (guest->host transport) will use guest->host 401c0cfa2d8SStefano Garzarella * transport for loopback (host->guest transports don't support loopback); 402c0cfa2d8SStefano Garzarella * - remote CID > VMADDR_CID_HOST will use host->guest transport; 403c0cfa2d8SStefano Garzarella */ 404c0cfa2d8SStefano Garzarella int vsock_assign_transport(struct vsock_sock *vsk, struct vsock_sock *psk) 405c0cfa2d8SStefano Garzarella { 406c0cfa2d8SStefano Garzarella const struct vsock_transport *new_transport; 407c0cfa2d8SStefano Garzarella struct sock *sk = sk_vsock(vsk); 408c0cfa2d8SStefano Garzarella unsigned int remote_cid = vsk->remote_addr.svm_cid; 409c0cfa2d8SStefano Garzarella 410c0cfa2d8SStefano Garzarella switch (sk->sk_type) { 411c0cfa2d8SStefano Garzarella case SOCK_DGRAM: 412c0cfa2d8SStefano Garzarella new_transport = transport_dgram; 413c0cfa2d8SStefano Garzarella break; 414c0cfa2d8SStefano Garzarella case SOCK_STREAM: 415c0cfa2d8SStefano Garzarella if (remote_cid <= VMADDR_CID_HOST || 416c0cfa2d8SStefano Garzarella (transport_g2h && 417c0cfa2d8SStefano Garzarella remote_cid == transport_g2h->get_local_cid())) 418c0cfa2d8SStefano Garzarella new_transport = transport_g2h; 419c0cfa2d8SStefano Garzarella else 420c0cfa2d8SStefano Garzarella new_transport = transport_h2g; 421c0cfa2d8SStefano Garzarella break; 422c0cfa2d8SStefano Garzarella default: 423c0cfa2d8SStefano Garzarella return -ESOCKTNOSUPPORT; 424c0cfa2d8SStefano Garzarella } 425c0cfa2d8SStefano Garzarella 426c0cfa2d8SStefano Garzarella if (vsk->transport) { 427c0cfa2d8SStefano Garzarella if (vsk->transport == new_transport) 428c0cfa2d8SStefano Garzarella return 0; 429c0cfa2d8SStefano Garzarella 430c0cfa2d8SStefano Garzarella vsk->transport->release(vsk); 431*6a2c0962SStefano Garzarella vsock_deassign_transport(vsk); 432c0cfa2d8SStefano Garzarella } 433c0cfa2d8SStefano Garzarella 434*6a2c0962SStefano Garzarella /* We increase the module refcnt to prevent the transport unloading 435*6a2c0962SStefano Garzarella * while there are open sockets assigned to it. 436*6a2c0962SStefano Garzarella */ 437*6a2c0962SStefano Garzarella if (!new_transport || !try_module_get(new_transport->module)) 438c0cfa2d8SStefano Garzarella return -ENODEV; 439c0cfa2d8SStefano Garzarella 440c0cfa2d8SStefano Garzarella vsk->transport = new_transport; 441c0cfa2d8SStefano Garzarella 442c0cfa2d8SStefano Garzarella return vsk->transport->init(vsk, psk); 443c0cfa2d8SStefano Garzarella } 444c0cfa2d8SStefano Garzarella EXPORT_SYMBOL_GPL(vsock_assign_transport); 445c0cfa2d8SStefano Garzarella 446c0cfa2d8SStefano Garzarella bool vsock_find_cid(unsigned int cid) 447c0cfa2d8SStefano Garzarella { 448c0cfa2d8SStefano Garzarella if (transport_g2h && cid == transport_g2h->get_local_cid()) 449c0cfa2d8SStefano Garzarella return true; 450c0cfa2d8SStefano Garzarella 451c0cfa2d8SStefano Garzarella if (transport_h2g && cid == VMADDR_CID_HOST) 452c0cfa2d8SStefano Garzarella return true; 453c0cfa2d8SStefano Garzarella 454c0cfa2d8SStefano Garzarella return false; 455c0cfa2d8SStefano Garzarella } 456c0cfa2d8SStefano Garzarella EXPORT_SYMBOL_GPL(vsock_find_cid); 457c0cfa2d8SStefano Garzarella 458d021c344SAndy King static struct sock *vsock_dequeue_accept(struct sock *listener) 459d021c344SAndy King { 460d021c344SAndy King struct vsock_sock *vlistener; 461d021c344SAndy King struct vsock_sock *vconnected; 462d021c344SAndy King 463d021c344SAndy King vlistener = vsock_sk(listener); 464d021c344SAndy King 465d021c344SAndy King if (list_empty(&vlistener->accept_queue)) 466d021c344SAndy King return NULL; 467d021c344SAndy King 468d021c344SAndy King vconnected = list_entry(vlistener->accept_queue.next, 469d021c344SAndy King struct vsock_sock, accept_queue); 470d021c344SAndy King 471d021c344SAndy King list_del_init(&vconnected->accept_queue); 472d021c344SAndy King sock_put(listener); 473d021c344SAndy King /* The caller will need a reference on the connected socket so we let 474d021c344SAndy King * it call sock_put(). 475d021c344SAndy King */ 476d021c344SAndy King 477d021c344SAndy King return sk_vsock(vconnected); 478d021c344SAndy King } 479d021c344SAndy King 480d021c344SAndy King static bool vsock_is_accept_queue_empty(struct sock *sk) 481d021c344SAndy King { 482d021c344SAndy King struct vsock_sock *vsk = vsock_sk(sk); 483d021c344SAndy King return list_empty(&vsk->accept_queue); 484d021c344SAndy King } 485d021c344SAndy King 486d021c344SAndy King static bool vsock_is_pending(struct sock *sk) 487d021c344SAndy King { 488d021c344SAndy King struct vsock_sock *vsk = vsock_sk(sk); 489d021c344SAndy King return !list_empty(&vsk->pending_links); 490d021c344SAndy King } 491d021c344SAndy King 492d021c344SAndy King static int vsock_send_shutdown(struct sock *sk, int mode) 493d021c344SAndy King { 494fe502c4aSStefano Garzarella struct vsock_sock *vsk = vsock_sk(sk); 495fe502c4aSStefano Garzarella 496c0cfa2d8SStefano Garzarella if (!vsk->transport) 497c0cfa2d8SStefano Garzarella return -ENODEV; 498c0cfa2d8SStefano Garzarella 499fe502c4aSStefano Garzarella return vsk->transport->shutdown(vsk, mode); 500d021c344SAndy King } 501d021c344SAndy King 502455f05ecSCong Wang static void vsock_pending_work(struct work_struct *work) 503d021c344SAndy King { 504d021c344SAndy King struct sock *sk; 505d021c344SAndy King struct sock *listener; 506d021c344SAndy King struct vsock_sock *vsk; 507d021c344SAndy King bool cleanup; 508d021c344SAndy King 509455f05ecSCong Wang vsk = container_of(work, struct vsock_sock, pending_work.work); 510d021c344SAndy King sk = sk_vsock(vsk); 511d021c344SAndy King listener = vsk->listener; 512d021c344SAndy King cleanup = true; 513d021c344SAndy King 514d021c344SAndy King lock_sock(listener); 5154192f672SStefan Hajnoczi lock_sock_nested(sk, SINGLE_DEPTH_NESTING); 516d021c344SAndy King 517d021c344SAndy King if (vsock_is_pending(sk)) { 518d021c344SAndy King vsock_remove_pending(listener, sk); 5191190cfdbSJorgen Hansen 5207976a11bSEric Dumazet sk_acceptq_removed(listener); 521d021c344SAndy King } else if (!vsk->rejected) { 522d021c344SAndy King /* We are not on the pending list and accept() did not reject 523d021c344SAndy King * us, so we must have been accepted by our user process. We 524d021c344SAndy King * just need to drop our references to the sockets and be on 525d021c344SAndy King * our way. 526d021c344SAndy King */ 527d021c344SAndy King cleanup = false; 528d021c344SAndy King goto out; 529d021c344SAndy King } 530d021c344SAndy King 531d021c344SAndy King /* We need to remove ourself from the global connected sockets list so 532d021c344SAndy King * incoming packets can't find this socket, and to reduce the reference 533d021c344SAndy King * count. 534d021c344SAndy King */ 535d021c344SAndy King vsock_remove_connected(vsk); 536d021c344SAndy King 5373b4477d2SStefan Hajnoczi sk->sk_state = TCP_CLOSE; 538d021c344SAndy King 539d021c344SAndy King out: 540d021c344SAndy King release_sock(sk); 541d021c344SAndy King release_sock(listener); 542d021c344SAndy King if (cleanup) 543d021c344SAndy King sock_put(sk); 544d021c344SAndy King 545d021c344SAndy King sock_put(sk); 546d021c344SAndy King sock_put(listener); 547d021c344SAndy King } 548d021c344SAndy King 549d021c344SAndy King /**** SOCKET OPERATIONS ****/ 550d021c344SAndy King 551d021c344SAndy King static int __vsock_bind_stream(struct vsock_sock *vsk, 552d021c344SAndy King struct sockaddr_vm *addr) 553d021c344SAndy King { 554a22d3251SLepton Wu static u32 port; 555d021c344SAndy King struct sockaddr_vm new_addr; 556d021c344SAndy King 5578236b08cSLepton Wu if (!port) 5588236b08cSLepton Wu port = LAST_RESERVED_PORT + 1 + 5598236b08cSLepton Wu prandom_u32_max(U32_MAX - LAST_RESERVED_PORT); 5608236b08cSLepton Wu 561d021c344SAndy King vsock_addr_init(&new_addr, addr->svm_cid, addr->svm_port); 562d021c344SAndy King 563d021c344SAndy King if (addr->svm_port == VMADDR_PORT_ANY) { 564d021c344SAndy King bool found = false; 565d021c344SAndy King unsigned int i; 566d021c344SAndy King 567d021c344SAndy King for (i = 0; i < MAX_PORT_RETRIES; i++) { 568d021c344SAndy King if (port <= LAST_RESERVED_PORT) 569d021c344SAndy King port = LAST_RESERVED_PORT + 1; 570d021c344SAndy King 571d021c344SAndy King new_addr.svm_port = port++; 572d021c344SAndy King 573d021c344SAndy King if (!__vsock_find_bound_socket(&new_addr)) { 574d021c344SAndy King found = true; 575d021c344SAndy King break; 576d021c344SAndy King } 577d021c344SAndy King } 578d021c344SAndy King 579d021c344SAndy King if (!found) 580d021c344SAndy King return -EADDRNOTAVAIL; 581d021c344SAndy King } else { 582d021c344SAndy King /* If port is in reserved range, ensure caller 583d021c344SAndy King * has necessary privileges. 584d021c344SAndy King */ 585d021c344SAndy King if (addr->svm_port <= LAST_RESERVED_PORT && 586d021c344SAndy King !capable(CAP_NET_BIND_SERVICE)) { 587d021c344SAndy King return -EACCES; 588d021c344SAndy King } 589d021c344SAndy King 590d021c344SAndy King if (__vsock_find_bound_socket(&new_addr)) 591d021c344SAndy King return -EADDRINUSE; 592d021c344SAndy King } 593d021c344SAndy King 594d021c344SAndy King vsock_addr_init(&vsk->local_addr, new_addr.svm_cid, new_addr.svm_port); 595d021c344SAndy King 596d021c344SAndy King /* Remove stream sockets from the unbound list and add them to the hash 597d021c344SAndy King * table for easy lookup by its address. The unbound list is simply an 598d021c344SAndy King * extra entry at the end of the hash table, a trick used by AF_UNIX. 599d021c344SAndy King */ 600d021c344SAndy King __vsock_remove_bound(vsk); 601d021c344SAndy King __vsock_insert_bound(vsock_bound_sockets(&vsk->local_addr), vsk); 602d021c344SAndy King 603d021c344SAndy King return 0; 604d021c344SAndy King } 605d021c344SAndy King 606d021c344SAndy King static int __vsock_bind_dgram(struct vsock_sock *vsk, 607d021c344SAndy King struct sockaddr_vm *addr) 608d021c344SAndy King { 609fe502c4aSStefano Garzarella return vsk->transport->dgram_bind(vsk, addr); 610d021c344SAndy King } 611d021c344SAndy King 612d021c344SAndy King static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr) 613d021c344SAndy King { 614d021c344SAndy King struct vsock_sock *vsk = vsock_sk(sk); 615d021c344SAndy King int retval; 616d021c344SAndy King 617d021c344SAndy King /* First ensure this socket isn't already bound. */ 618d021c344SAndy King if (vsock_addr_bound(&vsk->local_addr)) 619d021c344SAndy King return -EINVAL; 620d021c344SAndy King 621d021c344SAndy King /* Now bind to the provided address or select appropriate values if 622d021c344SAndy King * none are provided (VMADDR_CID_ANY and VMADDR_PORT_ANY). Note that 623d021c344SAndy King * like AF_INET prevents binding to a non-local IP address (in most 624c0cfa2d8SStefano Garzarella * cases), we only allow binding to a local CID. 625d021c344SAndy King */ 626c0cfa2d8SStefano Garzarella if (addr->svm_cid != VMADDR_CID_ANY && !vsock_find_cid(addr->svm_cid)) 627d021c344SAndy King return -EADDRNOTAVAIL; 628d021c344SAndy King 629d021c344SAndy King switch (sk->sk_socket->type) { 630d021c344SAndy King case SOCK_STREAM: 631d021c344SAndy King spin_lock_bh(&vsock_table_lock); 632d021c344SAndy King retval = __vsock_bind_stream(vsk, addr); 633d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 634d021c344SAndy King break; 635d021c344SAndy King 636d021c344SAndy King case SOCK_DGRAM: 637d021c344SAndy King retval = __vsock_bind_dgram(vsk, addr); 638d021c344SAndy King break; 639d021c344SAndy King 640d021c344SAndy King default: 641d021c344SAndy King retval = -EINVAL; 642d021c344SAndy King break; 643d021c344SAndy King } 644d021c344SAndy King 645d021c344SAndy King return retval; 646d021c344SAndy King } 647d021c344SAndy King 648455f05ecSCong Wang static void vsock_connect_timeout(struct work_struct *work); 649455f05ecSCong Wang 650b9ca2f5fSStefano Garzarella static struct sock *__vsock_create(struct net *net, 651d021c344SAndy King struct socket *sock, 652d021c344SAndy King struct sock *parent, 653d021c344SAndy King gfp_t priority, 65411aa9c28SEric W. Biederman unsigned short type, 65511aa9c28SEric W. Biederman int kern) 656d021c344SAndy King { 657d021c344SAndy King struct sock *sk; 658d021c344SAndy King struct vsock_sock *psk; 659d021c344SAndy King struct vsock_sock *vsk; 660d021c344SAndy King 66111aa9c28SEric W. Biederman sk = sk_alloc(net, AF_VSOCK, priority, &vsock_proto, kern); 662d021c344SAndy King if (!sk) 663d021c344SAndy King return NULL; 664d021c344SAndy King 665d021c344SAndy King sock_init_data(sock, sk); 666d021c344SAndy King 667d021c344SAndy King /* sk->sk_type is normally set in sock_init_data, but only if sock is 668d021c344SAndy King * non-NULL. We make sure that our sockets always have a type by 669d021c344SAndy King * setting it here if needed. 670d021c344SAndy King */ 671d021c344SAndy King if (!sock) 672d021c344SAndy King sk->sk_type = type; 673d021c344SAndy King 674d021c344SAndy King vsk = vsock_sk(sk); 675d021c344SAndy King vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 676d021c344SAndy King vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 677d021c344SAndy King 678d021c344SAndy King sk->sk_destruct = vsock_sk_destruct; 679d021c344SAndy King sk->sk_backlog_rcv = vsock_queue_rcv_skb; 680d021c344SAndy King sock_reset_flag(sk, SOCK_DONE); 681d021c344SAndy King 682d021c344SAndy King INIT_LIST_HEAD(&vsk->bound_table); 683d021c344SAndy King INIT_LIST_HEAD(&vsk->connected_table); 684d021c344SAndy King vsk->listener = NULL; 685d021c344SAndy King INIT_LIST_HEAD(&vsk->pending_links); 686d021c344SAndy King INIT_LIST_HEAD(&vsk->accept_queue); 687d021c344SAndy King vsk->rejected = false; 688d021c344SAndy King vsk->sent_request = false; 689d021c344SAndy King vsk->ignore_connecting_rst = false; 690d021c344SAndy King vsk->peer_shutdown = 0; 691455f05ecSCong Wang INIT_DELAYED_WORK(&vsk->connect_work, vsock_connect_timeout); 692455f05ecSCong Wang INIT_DELAYED_WORK(&vsk->pending_work, vsock_pending_work); 693d021c344SAndy King 694d021c344SAndy King psk = parent ? vsock_sk(parent) : NULL; 695d021c344SAndy King if (parent) { 696d021c344SAndy King vsk->trusted = psk->trusted; 697d021c344SAndy King vsk->owner = get_cred(psk->owner); 698d021c344SAndy King vsk->connect_timeout = psk->connect_timeout; 699b9f2b0ffSStefano Garzarella vsk->buffer_size = psk->buffer_size; 700b9f2b0ffSStefano Garzarella vsk->buffer_min_size = psk->buffer_min_size; 701b9f2b0ffSStefano Garzarella vsk->buffer_max_size = psk->buffer_max_size; 702d021c344SAndy King } else { 703d021c344SAndy King vsk->trusted = capable(CAP_NET_ADMIN); 704d021c344SAndy King vsk->owner = get_current_cred(); 705d021c344SAndy King vsk->connect_timeout = VSOCK_DEFAULT_CONNECT_TIMEOUT; 706b9f2b0ffSStefano Garzarella vsk->buffer_size = VSOCK_DEFAULT_BUFFER_SIZE; 707b9f2b0ffSStefano Garzarella vsk->buffer_min_size = VSOCK_DEFAULT_BUFFER_MIN_SIZE; 708b9f2b0ffSStefano Garzarella vsk->buffer_max_size = VSOCK_DEFAULT_BUFFER_MAX_SIZE; 709d021c344SAndy King } 710d021c344SAndy King 711d021c344SAndy King return sk; 712d021c344SAndy King } 713d021c344SAndy King 7140d9138ffSDexuan Cui static void __vsock_release(struct sock *sk, int level) 715d021c344SAndy King { 716d021c344SAndy King if (sk) { 717d021c344SAndy King struct sock *pending; 718d021c344SAndy King struct vsock_sock *vsk; 719d021c344SAndy King 720d021c344SAndy King vsk = vsock_sk(sk); 721d021c344SAndy King pending = NULL; /* Compiler warning. */ 722d021c344SAndy King 7230d9138ffSDexuan Cui /* The release call is supposed to use lock_sock_nested() 7240d9138ffSDexuan Cui * rather than lock_sock(), if a sock lock should be acquired. 7250d9138ffSDexuan Cui */ 726c0cfa2d8SStefano Garzarella if (vsk->transport) 727fe502c4aSStefano Garzarella vsk->transport->release(vsk); 728c0cfa2d8SStefano Garzarella else if (sk->sk_type == SOCK_STREAM) 729c0cfa2d8SStefano Garzarella vsock_remove_sock(vsk); 730d021c344SAndy King 7310d9138ffSDexuan Cui /* When "level" is SINGLE_DEPTH_NESTING, use the nested 7320d9138ffSDexuan Cui * version to avoid the warning "possible recursive locking 7330d9138ffSDexuan Cui * detected". When "level" is 0, lock_sock_nested(sk, level) 7340d9138ffSDexuan Cui * is the same as lock_sock(sk). 7350d9138ffSDexuan Cui */ 7360d9138ffSDexuan Cui lock_sock_nested(sk, level); 737d021c344SAndy King sock_orphan(sk); 738d021c344SAndy King sk->sk_shutdown = SHUTDOWN_MASK; 739d021c344SAndy King 7403b7ad08bSChristophe JAILLET skb_queue_purge(&sk->sk_receive_queue); 741d021c344SAndy King 742d021c344SAndy King /* Clean up any sockets that never were accepted. */ 743d021c344SAndy King while ((pending = vsock_dequeue_accept(sk)) != NULL) { 7440d9138ffSDexuan Cui __vsock_release(pending, SINGLE_DEPTH_NESTING); 745d021c344SAndy King sock_put(pending); 746d021c344SAndy King } 747d021c344SAndy King 748d021c344SAndy King release_sock(sk); 749d021c344SAndy King sock_put(sk); 750d021c344SAndy King } 751d021c344SAndy King } 752d021c344SAndy King 753d021c344SAndy King static void vsock_sk_destruct(struct sock *sk) 754d021c344SAndy King { 755d021c344SAndy King struct vsock_sock *vsk = vsock_sk(sk); 756d021c344SAndy King 757*6a2c0962SStefano Garzarella vsock_deassign_transport(vsk); 758d021c344SAndy King 759d021c344SAndy King /* When clearing these addresses, there's no need to set the family and 760d021c344SAndy King * possibly register the address family with the kernel. 761d021c344SAndy King */ 762d021c344SAndy King vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 763d021c344SAndy King vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 764d021c344SAndy King 765d021c344SAndy King put_cred(vsk->owner); 766d021c344SAndy King } 767d021c344SAndy King 768d021c344SAndy King static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) 769d021c344SAndy King { 770d021c344SAndy King int err; 771d021c344SAndy King 772d021c344SAndy King err = sock_queue_rcv_skb(sk, skb); 773d021c344SAndy King if (err) 774d021c344SAndy King kfree_skb(skb); 775d021c344SAndy King 776d021c344SAndy King return err; 777d021c344SAndy King } 778d021c344SAndy King 779b9ca2f5fSStefano Garzarella struct sock *vsock_create_connected(struct sock *parent) 780b9ca2f5fSStefano Garzarella { 781b9ca2f5fSStefano Garzarella return __vsock_create(sock_net(parent), NULL, parent, GFP_KERNEL, 782b9ca2f5fSStefano Garzarella parent->sk_type, 0); 783b9ca2f5fSStefano Garzarella } 784b9ca2f5fSStefano Garzarella EXPORT_SYMBOL_GPL(vsock_create_connected); 785b9ca2f5fSStefano Garzarella 786d021c344SAndy King s64 vsock_stream_has_data(struct vsock_sock *vsk) 787d021c344SAndy King { 788fe502c4aSStefano Garzarella return vsk->transport->stream_has_data(vsk); 789d021c344SAndy King } 790d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_stream_has_data); 791d021c344SAndy King 792d021c344SAndy King s64 vsock_stream_has_space(struct vsock_sock *vsk) 793d021c344SAndy King { 794fe502c4aSStefano Garzarella return vsk->transport->stream_has_space(vsk); 795d021c344SAndy King } 796d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_stream_has_space); 797d021c344SAndy King 798d021c344SAndy King static int vsock_release(struct socket *sock) 799d021c344SAndy King { 8000d9138ffSDexuan Cui __vsock_release(sock->sk, 0); 801d021c344SAndy King sock->sk = NULL; 802d021c344SAndy King sock->state = SS_FREE; 803d021c344SAndy King 804d021c344SAndy King return 0; 805d021c344SAndy King } 806d021c344SAndy King 807d021c344SAndy King static int 808d021c344SAndy King vsock_bind(struct socket *sock, struct sockaddr *addr, int addr_len) 809d021c344SAndy King { 810d021c344SAndy King int err; 811d021c344SAndy King struct sock *sk; 812d021c344SAndy King struct sockaddr_vm *vm_addr; 813d021c344SAndy King 814d021c344SAndy King sk = sock->sk; 815d021c344SAndy King 816d021c344SAndy King if (vsock_addr_cast(addr, addr_len, &vm_addr) != 0) 817d021c344SAndy King return -EINVAL; 818d021c344SAndy King 819d021c344SAndy King lock_sock(sk); 820d021c344SAndy King err = __vsock_bind(sk, vm_addr); 821d021c344SAndy King release_sock(sk); 822d021c344SAndy King 823d021c344SAndy King return err; 824d021c344SAndy King } 825d021c344SAndy King 826d021c344SAndy King static int vsock_getname(struct socket *sock, 8279b2c45d4SDenys Vlasenko struct sockaddr *addr, int peer) 828d021c344SAndy King { 829d021c344SAndy King int err; 830d021c344SAndy King struct sock *sk; 831d021c344SAndy King struct vsock_sock *vsk; 832d021c344SAndy King struct sockaddr_vm *vm_addr; 833d021c344SAndy King 834d021c344SAndy King sk = sock->sk; 835d021c344SAndy King vsk = vsock_sk(sk); 836d021c344SAndy King err = 0; 837d021c344SAndy King 838d021c344SAndy King lock_sock(sk); 839d021c344SAndy King 840d021c344SAndy King if (peer) { 841d021c344SAndy King if (sock->state != SS_CONNECTED) { 842d021c344SAndy King err = -ENOTCONN; 843d021c344SAndy King goto out; 844d021c344SAndy King } 845d021c344SAndy King vm_addr = &vsk->remote_addr; 846d021c344SAndy King } else { 847d021c344SAndy King vm_addr = &vsk->local_addr; 848d021c344SAndy King } 849d021c344SAndy King 850d021c344SAndy King if (!vm_addr) { 851d021c344SAndy King err = -EINVAL; 852d021c344SAndy King goto out; 853d021c344SAndy King } 854d021c344SAndy King 855d021c344SAndy King /* sys_getsockname() and sys_getpeername() pass us a 856d021c344SAndy King * MAX_SOCK_ADDR-sized buffer and don't set addr_len. Unfortunately 857d021c344SAndy King * that macro is defined in socket.c instead of .h, so we hardcode its 858d021c344SAndy King * value here. 859d021c344SAndy King */ 860d021c344SAndy King BUILD_BUG_ON(sizeof(*vm_addr) > 128); 861d021c344SAndy King memcpy(addr, vm_addr, sizeof(*vm_addr)); 8629b2c45d4SDenys Vlasenko err = sizeof(*vm_addr); 863d021c344SAndy King 864d021c344SAndy King out: 865d021c344SAndy King release_sock(sk); 866d021c344SAndy King return err; 867d021c344SAndy King } 868d021c344SAndy King 869d021c344SAndy King static int vsock_shutdown(struct socket *sock, int mode) 870d021c344SAndy King { 871d021c344SAndy King int err; 872d021c344SAndy King struct sock *sk; 873d021c344SAndy King 874d021c344SAndy King /* User level uses SHUT_RD (0) and SHUT_WR (1), but the kernel uses 875d021c344SAndy King * RCV_SHUTDOWN (1) and SEND_SHUTDOWN (2), so we must increment mode 876d021c344SAndy King * here like the other address families do. Note also that the 877d021c344SAndy King * increment makes SHUT_RDWR (2) into RCV_SHUTDOWN | SEND_SHUTDOWN (3), 878d021c344SAndy King * which is what we want. 879d021c344SAndy King */ 880d021c344SAndy King mode++; 881d021c344SAndy King 882d021c344SAndy King if ((mode & ~SHUTDOWN_MASK) || !mode) 883d021c344SAndy King return -EINVAL; 884d021c344SAndy King 885d021c344SAndy King /* If this is a STREAM socket and it is not connected then bail out 886d021c344SAndy King * immediately. If it is a DGRAM socket then we must first kick the 887d021c344SAndy King * socket so that it wakes up from any sleeping calls, for example 888d021c344SAndy King * recv(), and then afterwards return the error. 889d021c344SAndy King */ 890d021c344SAndy King 891d021c344SAndy King sk = sock->sk; 892d021c344SAndy King if (sock->state == SS_UNCONNECTED) { 893d021c344SAndy King err = -ENOTCONN; 894d021c344SAndy King if (sk->sk_type == SOCK_STREAM) 895d021c344SAndy King return err; 896d021c344SAndy King } else { 897d021c344SAndy King sock->state = SS_DISCONNECTING; 898d021c344SAndy King err = 0; 899d021c344SAndy King } 900d021c344SAndy King 901d021c344SAndy King /* Receive and send shutdowns are treated alike. */ 902d021c344SAndy King mode = mode & (RCV_SHUTDOWN | SEND_SHUTDOWN); 903d021c344SAndy King if (mode) { 904d021c344SAndy King lock_sock(sk); 905d021c344SAndy King sk->sk_shutdown |= mode; 906d021c344SAndy King sk->sk_state_change(sk); 907d021c344SAndy King release_sock(sk); 908d021c344SAndy King 909d021c344SAndy King if (sk->sk_type == SOCK_STREAM) { 910d021c344SAndy King sock_reset_flag(sk, SOCK_DONE); 911d021c344SAndy King vsock_send_shutdown(sk, mode); 912d021c344SAndy King } 913d021c344SAndy King } 914d021c344SAndy King 915d021c344SAndy King return err; 916d021c344SAndy King } 917d021c344SAndy King 918a11e1d43SLinus Torvalds static __poll_t vsock_poll(struct file *file, struct socket *sock, 919a11e1d43SLinus Torvalds poll_table *wait) 920d021c344SAndy King { 921a11e1d43SLinus Torvalds struct sock *sk; 922a11e1d43SLinus Torvalds __poll_t mask; 923a11e1d43SLinus Torvalds struct vsock_sock *vsk; 924a11e1d43SLinus Torvalds 925a11e1d43SLinus Torvalds sk = sock->sk; 926a11e1d43SLinus Torvalds vsk = vsock_sk(sk); 927a11e1d43SLinus Torvalds 928a11e1d43SLinus Torvalds poll_wait(file, sk_sleep(sk), wait); 929a11e1d43SLinus Torvalds mask = 0; 930d021c344SAndy King 931d021c344SAndy King if (sk->sk_err) 932d021c344SAndy King /* Signify that there has been an error on this socket. */ 933a9a08845SLinus Torvalds mask |= EPOLLERR; 934d021c344SAndy King 935d021c344SAndy King /* INET sockets treat local write shutdown and peer write shutdown as a 936a9a08845SLinus Torvalds * case of EPOLLHUP set. 937d021c344SAndy King */ 938d021c344SAndy King if ((sk->sk_shutdown == SHUTDOWN_MASK) || 939d021c344SAndy King ((sk->sk_shutdown & SEND_SHUTDOWN) && 940d021c344SAndy King (vsk->peer_shutdown & SEND_SHUTDOWN))) { 941a9a08845SLinus Torvalds mask |= EPOLLHUP; 942d021c344SAndy King } 943d021c344SAndy King 944d021c344SAndy King if (sk->sk_shutdown & RCV_SHUTDOWN || 945d021c344SAndy King vsk->peer_shutdown & SEND_SHUTDOWN) { 946a9a08845SLinus Torvalds mask |= EPOLLRDHUP; 947d021c344SAndy King } 948d021c344SAndy King 949d021c344SAndy King if (sock->type == SOCK_DGRAM) { 950d021c344SAndy King /* For datagram sockets we can read if there is something in 951d021c344SAndy King * the queue and write as long as the socket isn't shutdown for 952d021c344SAndy King * sending. 953d021c344SAndy King */ 9543ef7cf57SEric Dumazet if (!skb_queue_empty_lockless(&sk->sk_receive_queue) || 955d021c344SAndy King (sk->sk_shutdown & RCV_SHUTDOWN)) { 956a9a08845SLinus Torvalds mask |= EPOLLIN | EPOLLRDNORM; 957d021c344SAndy King } 958d021c344SAndy King 959d021c344SAndy King if (!(sk->sk_shutdown & SEND_SHUTDOWN)) 960a9a08845SLinus Torvalds mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND; 961d021c344SAndy King 962d021c344SAndy King } else if (sock->type == SOCK_STREAM) { 963fe502c4aSStefano Garzarella const struct vsock_transport *transport = vsk->transport; 964d021c344SAndy King lock_sock(sk); 965d021c344SAndy King 966d021c344SAndy King /* Listening sockets that have connections in their accept 967d021c344SAndy King * queue can be read. 968d021c344SAndy King */ 9693b4477d2SStefan Hajnoczi if (sk->sk_state == TCP_LISTEN 970d021c344SAndy King && !vsock_is_accept_queue_empty(sk)) 971a9a08845SLinus Torvalds mask |= EPOLLIN | EPOLLRDNORM; 972d021c344SAndy King 973d021c344SAndy King /* If there is something in the queue then we can read. */ 974c0cfa2d8SStefano Garzarella if (transport && transport->stream_is_active(vsk) && 975d021c344SAndy King !(sk->sk_shutdown & RCV_SHUTDOWN)) { 976d021c344SAndy King bool data_ready_now = false; 977d021c344SAndy King int ret = transport->notify_poll_in( 978d021c344SAndy King vsk, 1, &data_ready_now); 979d021c344SAndy King if (ret < 0) { 980a9a08845SLinus Torvalds mask |= EPOLLERR; 981d021c344SAndy King } else { 982d021c344SAndy King if (data_ready_now) 983a9a08845SLinus Torvalds mask |= EPOLLIN | EPOLLRDNORM; 984d021c344SAndy King 985d021c344SAndy King } 986d021c344SAndy King } 987d021c344SAndy King 988d021c344SAndy King /* Sockets whose connections have been closed, reset, or 989d021c344SAndy King * terminated should also be considered read, and we check the 990d021c344SAndy King * shutdown flag for that. 991d021c344SAndy King */ 992d021c344SAndy King if (sk->sk_shutdown & RCV_SHUTDOWN || 993d021c344SAndy King vsk->peer_shutdown & SEND_SHUTDOWN) { 994a9a08845SLinus Torvalds mask |= EPOLLIN | EPOLLRDNORM; 995d021c344SAndy King } 996d021c344SAndy King 997d021c344SAndy King /* Connected sockets that can produce data can be written. */ 9983b4477d2SStefan Hajnoczi if (sk->sk_state == TCP_ESTABLISHED) { 999d021c344SAndy King if (!(sk->sk_shutdown & SEND_SHUTDOWN)) { 1000d021c344SAndy King bool space_avail_now = false; 1001d021c344SAndy King int ret = transport->notify_poll_out( 1002d021c344SAndy King vsk, 1, &space_avail_now); 1003d021c344SAndy King if (ret < 0) { 1004a9a08845SLinus Torvalds mask |= EPOLLERR; 1005d021c344SAndy King } else { 1006d021c344SAndy King if (space_avail_now) 1007a9a08845SLinus Torvalds /* Remove EPOLLWRBAND since INET 1008d021c344SAndy King * sockets are not setting it. 1009d021c344SAndy King */ 1010a9a08845SLinus Torvalds mask |= EPOLLOUT | EPOLLWRNORM; 1011d021c344SAndy King 1012d021c344SAndy King } 1013d021c344SAndy King } 1014d021c344SAndy King } 1015d021c344SAndy King 1016d021c344SAndy King /* Simulate INET socket poll behaviors, which sets 1017a9a08845SLinus Torvalds * EPOLLOUT|EPOLLWRNORM when peer is closed and nothing to read, 1018d021c344SAndy King * but local send is not shutdown. 1019d021c344SAndy King */ 1020ba3169fcSStefan Hajnoczi if (sk->sk_state == TCP_CLOSE || sk->sk_state == TCP_CLOSING) { 1021d021c344SAndy King if (!(sk->sk_shutdown & SEND_SHUTDOWN)) 1022a9a08845SLinus Torvalds mask |= EPOLLOUT | EPOLLWRNORM; 1023d021c344SAndy King 1024d021c344SAndy King } 1025d021c344SAndy King 1026d021c344SAndy King release_sock(sk); 1027d021c344SAndy King } 1028d021c344SAndy King 1029d021c344SAndy King return mask; 1030d021c344SAndy King } 1031d021c344SAndy King 10321b784140SYing Xue static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg, 10331b784140SYing Xue size_t len) 1034d021c344SAndy King { 1035d021c344SAndy King int err; 1036d021c344SAndy King struct sock *sk; 1037d021c344SAndy King struct vsock_sock *vsk; 1038d021c344SAndy King struct sockaddr_vm *remote_addr; 1039fe502c4aSStefano Garzarella const struct vsock_transport *transport; 1040d021c344SAndy King 1041d021c344SAndy King if (msg->msg_flags & MSG_OOB) 1042d021c344SAndy King return -EOPNOTSUPP; 1043d021c344SAndy King 1044d021c344SAndy King /* For now, MSG_DONTWAIT is always assumed... */ 1045d021c344SAndy King err = 0; 1046d021c344SAndy King sk = sock->sk; 1047d021c344SAndy King vsk = vsock_sk(sk); 1048fe502c4aSStefano Garzarella transport = vsk->transport; 1049d021c344SAndy King 1050d021c344SAndy King lock_sock(sk); 1051d021c344SAndy King 1052b3a6dfe8SAsias He err = vsock_auto_bind(vsk); 1053b3a6dfe8SAsias He if (err) 1054d021c344SAndy King goto out; 1055d021c344SAndy King 1056d021c344SAndy King 1057d021c344SAndy King /* If the provided message contains an address, use that. Otherwise 1058d021c344SAndy King * fall back on the socket's remote handle (if it has been connected). 1059d021c344SAndy King */ 1060d021c344SAndy King if (msg->msg_name && 1061d021c344SAndy King vsock_addr_cast(msg->msg_name, msg->msg_namelen, 1062d021c344SAndy King &remote_addr) == 0) { 1063d021c344SAndy King /* Ensure this address is of the right type and is a valid 1064d021c344SAndy King * destination. 1065d021c344SAndy King */ 1066d021c344SAndy King 1067d021c344SAndy King if (remote_addr->svm_cid == VMADDR_CID_ANY) 1068d021c344SAndy King remote_addr->svm_cid = transport->get_local_cid(); 1069d021c344SAndy King 1070d021c344SAndy King if (!vsock_addr_bound(remote_addr)) { 1071d021c344SAndy King err = -EINVAL; 1072d021c344SAndy King goto out; 1073d021c344SAndy King } 1074d021c344SAndy King } else if (sock->state == SS_CONNECTED) { 1075d021c344SAndy King remote_addr = &vsk->remote_addr; 1076d021c344SAndy King 1077d021c344SAndy King if (remote_addr->svm_cid == VMADDR_CID_ANY) 1078d021c344SAndy King remote_addr->svm_cid = transport->get_local_cid(); 1079d021c344SAndy King 1080d021c344SAndy King /* XXX Should connect() or this function ensure remote_addr is 1081d021c344SAndy King * bound? 1082d021c344SAndy King */ 1083d021c344SAndy King if (!vsock_addr_bound(&vsk->remote_addr)) { 1084d021c344SAndy King err = -EINVAL; 1085d021c344SAndy King goto out; 1086d021c344SAndy King } 1087d021c344SAndy King } else { 1088d021c344SAndy King err = -EINVAL; 1089d021c344SAndy King goto out; 1090d021c344SAndy King } 1091d021c344SAndy King 1092d021c344SAndy King if (!transport->dgram_allow(remote_addr->svm_cid, 1093d021c344SAndy King remote_addr->svm_port)) { 1094d021c344SAndy King err = -EINVAL; 1095d021c344SAndy King goto out; 1096d021c344SAndy King } 1097d021c344SAndy King 10980f7db23aSAl Viro err = transport->dgram_enqueue(vsk, remote_addr, msg, len); 1099d021c344SAndy King 1100d021c344SAndy King out: 1101d021c344SAndy King release_sock(sk); 1102d021c344SAndy King return err; 1103d021c344SAndy King } 1104d021c344SAndy King 1105d021c344SAndy King static int vsock_dgram_connect(struct socket *sock, 1106d021c344SAndy King struct sockaddr *addr, int addr_len, int flags) 1107d021c344SAndy King { 1108d021c344SAndy King int err; 1109d021c344SAndy King struct sock *sk; 1110d021c344SAndy King struct vsock_sock *vsk; 1111d021c344SAndy King struct sockaddr_vm *remote_addr; 1112d021c344SAndy King 1113d021c344SAndy King sk = sock->sk; 1114d021c344SAndy King vsk = vsock_sk(sk); 1115d021c344SAndy King 1116d021c344SAndy King err = vsock_addr_cast(addr, addr_len, &remote_addr); 1117d021c344SAndy King if (err == -EAFNOSUPPORT && remote_addr->svm_family == AF_UNSPEC) { 1118d021c344SAndy King lock_sock(sk); 1119d021c344SAndy King vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, 1120d021c344SAndy King VMADDR_PORT_ANY); 1121d021c344SAndy King sock->state = SS_UNCONNECTED; 1122d021c344SAndy King release_sock(sk); 1123d021c344SAndy King return 0; 1124d021c344SAndy King } else if (err != 0) 1125d021c344SAndy King return -EINVAL; 1126d021c344SAndy King 1127d021c344SAndy King lock_sock(sk); 1128d021c344SAndy King 1129b3a6dfe8SAsias He err = vsock_auto_bind(vsk); 1130b3a6dfe8SAsias He if (err) 1131d021c344SAndy King goto out; 1132d021c344SAndy King 1133fe502c4aSStefano Garzarella if (!vsk->transport->dgram_allow(remote_addr->svm_cid, 1134d021c344SAndy King remote_addr->svm_port)) { 1135d021c344SAndy King err = -EINVAL; 1136d021c344SAndy King goto out; 1137d021c344SAndy King } 1138d021c344SAndy King 1139d021c344SAndy King memcpy(&vsk->remote_addr, remote_addr, sizeof(vsk->remote_addr)); 1140d021c344SAndy King sock->state = SS_CONNECTED; 1141d021c344SAndy King 1142d021c344SAndy King out: 1143d021c344SAndy King release_sock(sk); 1144d021c344SAndy King return err; 1145d021c344SAndy King } 1146d021c344SAndy King 11471b784140SYing Xue static int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg, 11481b784140SYing Xue size_t len, int flags) 1149d021c344SAndy King { 1150fe502c4aSStefano Garzarella struct vsock_sock *vsk = vsock_sk(sock->sk); 1151fe502c4aSStefano Garzarella 1152fe502c4aSStefano Garzarella return vsk->transport->dgram_dequeue(vsk, msg, len, flags); 1153d021c344SAndy King } 1154d021c344SAndy King 1155d021c344SAndy King static const struct proto_ops vsock_dgram_ops = { 1156d021c344SAndy King .family = PF_VSOCK, 1157d021c344SAndy King .owner = THIS_MODULE, 1158d021c344SAndy King .release = vsock_release, 1159d021c344SAndy King .bind = vsock_bind, 1160d021c344SAndy King .connect = vsock_dgram_connect, 1161d021c344SAndy King .socketpair = sock_no_socketpair, 1162d021c344SAndy King .accept = sock_no_accept, 1163d021c344SAndy King .getname = vsock_getname, 1164a11e1d43SLinus Torvalds .poll = vsock_poll, 1165d021c344SAndy King .ioctl = sock_no_ioctl, 1166d021c344SAndy King .listen = sock_no_listen, 1167d021c344SAndy King .shutdown = vsock_shutdown, 1168d021c344SAndy King .setsockopt = sock_no_setsockopt, 1169d021c344SAndy King .getsockopt = sock_no_getsockopt, 1170d021c344SAndy King .sendmsg = vsock_dgram_sendmsg, 1171d021c344SAndy King .recvmsg = vsock_dgram_recvmsg, 1172d021c344SAndy King .mmap = sock_no_mmap, 1173d021c344SAndy King .sendpage = sock_no_sendpage, 1174d021c344SAndy King }; 1175d021c344SAndy King 1176380feae0SPeng Tao static int vsock_transport_cancel_pkt(struct vsock_sock *vsk) 1177380feae0SPeng Tao { 1178fe502c4aSStefano Garzarella const struct vsock_transport *transport = vsk->transport; 1179fe502c4aSStefano Garzarella 1180380feae0SPeng Tao if (!transport->cancel_pkt) 1181380feae0SPeng Tao return -EOPNOTSUPP; 1182380feae0SPeng Tao 1183380feae0SPeng Tao return transport->cancel_pkt(vsk); 1184380feae0SPeng Tao } 1185380feae0SPeng Tao 1186d021c344SAndy King static void vsock_connect_timeout(struct work_struct *work) 1187d021c344SAndy King { 1188d021c344SAndy King struct sock *sk; 1189d021c344SAndy King struct vsock_sock *vsk; 1190380feae0SPeng Tao int cancel = 0; 1191d021c344SAndy King 1192455f05ecSCong Wang vsk = container_of(work, struct vsock_sock, connect_work.work); 1193d021c344SAndy King sk = sk_vsock(vsk); 1194d021c344SAndy King 1195d021c344SAndy King lock_sock(sk); 11963b4477d2SStefan Hajnoczi if (sk->sk_state == TCP_SYN_SENT && 1197d021c344SAndy King (sk->sk_shutdown != SHUTDOWN_MASK)) { 11983b4477d2SStefan Hajnoczi sk->sk_state = TCP_CLOSE; 1199d021c344SAndy King sk->sk_err = ETIMEDOUT; 1200d021c344SAndy King sk->sk_error_report(sk); 1201380feae0SPeng Tao cancel = 1; 1202d021c344SAndy King } 1203d021c344SAndy King release_sock(sk); 1204380feae0SPeng Tao if (cancel) 1205380feae0SPeng Tao vsock_transport_cancel_pkt(vsk); 1206d021c344SAndy King 1207d021c344SAndy King sock_put(sk); 1208d021c344SAndy King } 1209d021c344SAndy King 1210d021c344SAndy King static int vsock_stream_connect(struct socket *sock, struct sockaddr *addr, 1211d021c344SAndy King int addr_len, int flags) 1212d021c344SAndy King { 1213d021c344SAndy King int err; 1214d021c344SAndy King struct sock *sk; 1215d021c344SAndy King struct vsock_sock *vsk; 1216fe502c4aSStefano Garzarella const struct vsock_transport *transport; 1217d021c344SAndy King struct sockaddr_vm *remote_addr; 1218d021c344SAndy King long timeout; 1219d021c344SAndy King DEFINE_WAIT(wait); 1220d021c344SAndy King 1221d021c344SAndy King err = 0; 1222d021c344SAndy King sk = sock->sk; 1223d021c344SAndy King vsk = vsock_sk(sk); 1224d021c344SAndy King 1225d021c344SAndy King lock_sock(sk); 1226d021c344SAndy King 1227d021c344SAndy King /* XXX AF_UNSPEC should make us disconnect like AF_INET. */ 1228d021c344SAndy King switch (sock->state) { 1229d021c344SAndy King case SS_CONNECTED: 1230d021c344SAndy King err = -EISCONN; 1231d021c344SAndy King goto out; 1232d021c344SAndy King case SS_DISCONNECTING: 1233d021c344SAndy King err = -EINVAL; 1234d021c344SAndy King goto out; 1235d021c344SAndy King case SS_CONNECTING: 1236d021c344SAndy King /* This continues on so we can move sock into the SS_CONNECTED 1237d021c344SAndy King * state once the connection has completed (at which point err 1238d021c344SAndy King * will be set to zero also). Otherwise, we will either wait 1239d021c344SAndy King * for the connection or return -EALREADY should this be a 1240d021c344SAndy King * non-blocking call. 1241d021c344SAndy King */ 1242d021c344SAndy King err = -EALREADY; 1243d021c344SAndy King break; 1244d021c344SAndy King default: 12453b4477d2SStefan Hajnoczi if ((sk->sk_state == TCP_LISTEN) || 1246d021c344SAndy King vsock_addr_cast(addr, addr_len, &remote_addr) != 0) { 1247d021c344SAndy King err = -EINVAL; 1248d021c344SAndy King goto out; 1249d021c344SAndy King } 1250d021c344SAndy King 1251c0cfa2d8SStefano Garzarella /* Set the remote address that we are connecting to. */ 1252c0cfa2d8SStefano Garzarella memcpy(&vsk->remote_addr, remote_addr, 1253c0cfa2d8SStefano Garzarella sizeof(vsk->remote_addr)); 1254c0cfa2d8SStefano Garzarella 1255c0cfa2d8SStefano Garzarella err = vsock_assign_transport(vsk, NULL); 1256c0cfa2d8SStefano Garzarella if (err) 1257c0cfa2d8SStefano Garzarella goto out; 1258c0cfa2d8SStefano Garzarella 1259c0cfa2d8SStefano Garzarella transport = vsk->transport; 1260c0cfa2d8SStefano Garzarella 1261d021c344SAndy King /* The hypervisor and well-known contexts do not have socket 1262d021c344SAndy King * endpoints. 1263d021c344SAndy King */ 1264c0cfa2d8SStefano Garzarella if (!transport || 1265c0cfa2d8SStefano Garzarella !transport->stream_allow(remote_addr->svm_cid, 1266d021c344SAndy King remote_addr->svm_port)) { 1267d021c344SAndy King err = -ENETUNREACH; 1268d021c344SAndy King goto out; 1269d021c344SAndy King } 1270d021c344SAndy King 1271b3a6dfe8SAsias He err = vsock_auto_bind(vsk); 1272b3a6dfe8SAsias He if (err) 1273d021c344SAndy King goto out; 1274d021c344SAndy King 12753b4477d2SStefan Hajnoczi sk->sk_state = TCP_SYN_SENT; 1276d021c344SAndy King 1277d021c344SAndy King err = transport->connect(vsk); 1278d021c344SAndy King if (err < 0) 1279d021c344SAndy King goto out; 1280d021c344SAndy King 1281d021c344SAndy King /* Mark sock as connecting and set the error code to in 1282d021c344SAndy King * progress in case this is a non-blocking connect. 1283d021c344SAndy King */ 1284d021c344SAndy King sock->state = SS_CONNECTING; 1285d021c344SAndy King err = -EINPROGRESS; 1286d021c344SAndy King } 1287d021c344SAndy King 1288d021c344SAndy King /* The receive path will handle all communication until we are able to 1289d021c344SAndy King * enter the connected state. Here we wait for the connection to be 1290d021c344SAndy King * completed or a notification of an error. 1291d021c344SAndy King */ 1292d021c344SAndy King timeout = vsk->connect_timeout; 1293d021c344SAndy King prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); 1294d021c344SAndy King 12953b4477d2SStefan Hajnoczi while (sk->sk_state != TCP_ESTABLISHED && sk->sk_err == 0) { 1296d021c344SAndy King if (flags & O_NONBLOCK) { 1297d021c344SAndy King /* If we're not going to block, we schedule a timeout 1298d021c344SAndy King * function to generate a timeout on the connection 1299d021c344SAndy King * attempt, in case the peer doesn't respond in a 1300d021c344SAndy King * timely manner. We hold on to the socket until the 1301d021c344SAndy King * timeout fires. 1302d021c344SAndy King */ 1303d021c344SAndy King sock_hold(sk); 1304455f05ecSCong Wang schedule_delayed_work(&vsk->connect_work, timeout); 1305d021c344SAndy King 1306d021c344SAndy King /* Skip ahead to preserve error code set above. */ 1307d021c344SAndy King goto out_wait; 1308d021c344SAndy King } 1309d021c344SAndy King 1310d021c344SAndy King release_sock(sk); 1311d021c344SAndy King timeout = schedule_timeout(timeout); 1312d021c344SAndy King lock_sock(sk); 1313d021c344SAndy King 1314d021c344SAndy King if (signal_pending(current)) { 1315d021c344SAndy King err = sock_intr_errno(timeout); 13163b4477d2SStefan Hajnoczi sk->sk_state = TCP_CLOSE; 1317f7f9b5e7SClaudio Imbrenda sock->state = SS_UNCONNECTED; 1318380feae0SPeng Tao vsock_transport_cancel_pkt(vsk); 1319f7f9b5e7SClaudio Imbrenda goto out_wait; 1320d021c344SAndy King } else if (timeout == 0) { 1321d021c344SAndy King err = -ETIMEDOUT; 13223b4477d2SStefan Hajnoczi sk->sk_state = TCP_CLOSE; 1323f7f9b5e7SClaudio Imbrenda sock->state = SS_UNCONNECTED; 1324380feae0SPeng Tao vsock_transport_cancel_pkt(vsk); 1325f7f9b5e7SClaudio Imbrenda goto out_wait; 1326d021c344SAndy King } 1327d021c344SAndy King 1328d021c344SAndy King prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); 1329d021c344SAndy King } 1330d021c344SAndy King 1331d021c344SAndy King if (sk->sk_err) { 1332d021c344SAndy King err = -sk->sk_err; 13333b4477d2SStefan Hajnoczi sk->sk_state = TCP_CLOSE; 1334f7f9b5e7SClaudio Imbrenda sock->state = SS_UNCONNECTED; 1335f7f9b5e7SClaudio Imbrenda } else { 1336d021c344SAndy King err = 0; 1337f7f9b5e7SClaudio Imbrenda } 1338d021c344SAndy King 1339d021c344SAndy King out_wait: 1340d021c344SAndy King finish_wait(sk_sleep(sk), &wait); 1341d021c344SAndy King out: 1342d021c344SAndy King release_sock(sk); 1343d021c344SAndy King return err; 1344d021c344SAndy King } 1345d021c344SAndy King 1346cdfbabfbSDavid Howells static int vsock_accept(struct socket *sock, struct socket *newsock, int flags, 1347cdfbabfbSDavid Howells bool kern) 1348d021c344SAndy King { 1349d021c344SAndy King struct sock *listener; 1350d021c344SAndy King int err; 1351d021c344SAndy King struct sock *connected; 1352d021c344SAndy King struct vsock_sock *vconnected; 1353d021c344SAndy King long timeout; 1354d021c344SAndy King DEFINE_WAIT(wait); 1355d021c344SAndy King 1356d021c344SAndy King err = 0; 1357d021c344SAndy King listener = sock->sk; 1358d021c344SAndy King 1359d021c344SAndy King lock_sock(listener); 1360d021c344SAndy King 1361d021c344SAndy King if (sock->type != SOCK_STREAM) { 1362d021c344SAndy King err = -EOPNOTSUPP; 1363d021c344SAndy King goto out; 1364d021c344SAndy King } 1365d021c344SAndy King 13663b4477d2SStefan Hajnoczi if (listener->sk_state != TCP_LISTEN) { 1367d021c344SAndy King err = -EINVAL; 1368d021c344SAndy King goto out; 1369d021c344SAndy King } 1370d021c344SAndy King 1371d021c344SAndy King /* Wait for children sockets to appear; these are the new sockets 1372d021c344SAndy King * created upon connection establishment. 1373d021c344SAndy King */ 1374d021c344SAndy King timeout = sock_sndtimeo(listener, flags & O_NONBLOCK); 1375d021c344SAndy King prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE); 1376d021c344SAndy King 1377d021c344SAndy King while ((connected = vsock_dequeue_accept(listener)) == NULL && 1378d021c344SAndy King listener->sk_err == 0) { 1379d021c344SAndy King release_sock(listener); 1380d021c344SAndy King timeout = schedule_timeout(timeout); 1381f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(listener), &wait); 1382d021c344SAndy King lock_sock(listener); 1383d021c344SAndy King 1384d021c344SAndy King if (signal_pending(current)) { 1385d021c344SAndy King err = sock_intr_errno(timeout); 1386f7f9b5e7SClaudio Imbrenda goto out; 1387d021c344SAndy King } else if (timeout == 0) { 1388d021c344SAndy King err = -EAGAIN; 1389f7f9b5e7SClaudio Imbrenda goto out; 1390d021c344SAndy King } 1391d021c344SAndy King 1392d021c344SAndy King prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE); 1393d021c344SAndy King } 1394f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(listener), &wait); 1395d021c344SAndy King 1396d021c344SAndy King if (listener->sk_err) 1397d021c344SAndy King err = -listener->sk_err; 1398d021c344SAndy King 1399d021c344SAndy King if (connected) { 14007976a11bSEric Dumazet sk_acceptq_removed(listener); 1401d021c344SAndy King 14024192f672SStefan Hajnoczi lock_sock_nested(connected, SINGLE_DEPTH_NESTING); 1403d021c344SAndy King vconnected = vsock_sk(connected); 1404d021c344SAndy King 1405d021c344SAndy King /* If the listener socket has received an error, then we should 1406d021c344SAndy King * reject this socket and return. Note that we simply mark the 1407d021c344SAndy King * socket rejected, drop our reference, and let the cleanup 1408d021c344SAndy King * function handle the cleanup; the fact that we found it in 1409d021c344SAndy King * the listener's accept queue guarantees that the cleanup 1410d021c344SAndy King * function hasn't run yet. 1411d021c344SAndy King */ 1412d021c344SAndy King if (err) { 1413d021c344SAndy King vconnected->rejected = true; 1414f7f9b5e7SClaudio Imbrenda } else { 1415d021c344SAndy King newsock->state = SS_CONNECTED; 1416d021c344SAndy King sock_graft(connected, newsock); 1417f7f9b5e7SClaudio Imbrenda } 1418f7f9b5e7SClaudio Imbrenda 1419d021c344SAndy King release_sock(connected); 1420d021c344SAndy King sock_put(connected); 1421d021c344SAndy King } 1422d021c344SAndy King 1423d021c344SAndy King out: 1424d021c344SAndy King release_sock(listener); 1425d021c344SAndy King return err; 1426d021c344SAndy King } 1427d021c344SAndy King 1428d021c344SAndy King static int vsock_listen(struct socket *sock, int backlog) 1429d021c344SAndy King { 1430d021c344SAndy King int err; 1431d021c344SAndy King struct sock *sk; 1432d021c344SAndy King struct vsock_sock *vsk; 1433d021c344SAndy King 1434d021c344SAndy King sk = sock->sk; 1435d021c344SAndy King 1436d021c344SAndy King lock_sock(sk); 1437d021c344SAndy King 1438d021c344SAndy King if (sock->type != SOCK_STREAM) { 1439d021c344SAndy King err = -EOPNOTSUPP; 1440d021c344SAndy King goto out; 1441d021c344SAndy King } 1442d021c344SAndy King 1443d021c344SAndy King if (sock->state != SS_UNCONNECTED) { 1444d021c344SAndy King err = -EINVAL; 1445d021c344SAndy King goto out; 1446d021c344SAndy King } 1447d021c344SAndy King 1448d021c344SAndy King vsk = vsock_sk(sk); 1449d021c344SAndy King 1450d021c344SAndy King if (!vsock_addr_bound(&vsk->local_addr)) { 1451d021c344SAndy King err = -EINVAL; 1452d021c344SAndy King goto out; 1453d021c344SAndy King } 1454d021c344SAndy King 1455d021c344SAndy King sk->sk_max_ack_backlog = backlog; 14563b4477d2SStefan Hajnoczi sk->sk_state = TCP_LISTEN; 1457d021c344SAndy King 1458d021c344SAndy King err = 0; 1459d021c344SAndy King 1460d021c344SAndy King out: 1461d021c344SAndy King release_sock(sk); 1462d021c344SAndy King return err; 1463d021c344SAndy King } 1464d021c344SAndy King 1465b9f2b0ffSStefano Garzarella static void vsock_update_buffer_size(struct vsock_sock *vsk, 1466b9f2b0ffSStefano Garzarella const struct vsock_transport *transport, 1467b9f2b0ffSStefano Garzarella u64 val) 1468b9f2b0ffSStefano Garzarella { 1469b9f2b0ffSStefano Garzarella if (val > vsk->buffer_max_size) 1470b9f2b0ffSStefano Garzarella val = vsk->buffer_max_size; 1471b9f2b0ffSStefano Garzarella 1472b9f2b0ffSStefano Garzarella if (val < vsk->buffer_min_size) 1473b9f2b0ffSStefano Garzarella val = vsk->buffer_min_size; 1474b9f2b0ffSStefano Garzarella 1475b9f2b0ffSStefano Garzarella if (val != vsk->buffer_size && 1476b9f2b0ffSStefano Garzarella transport && transport->notify_buffer_size) 1477b9f2b0ffSStefano Garzarella transport->notify_buffer_size(vsk, &val); 1478b9f2b0ffSStefano Garzarella 1479b9f2b0ffSStefano Garzarella vsk->buffer_size = val; 1480b9f2b0ffSStefano Garzarella } 1481b9f2b0ffSStefano Garzarella 1482d021c344SAndy King static int vsock_stream_setsockopt(struct socket *sock, 1483d021c344SAndy King int level, 1484d021c344SAndy King int optname, 1485d021c344SAndy King char __user *optval, 1486d021c344SAndy King unsigned int optlen) 1487d021c344SAndy King { 1488d021c344SAndy King int err; 1489d021c344SAndy King struct sock *sk; 1490d021c344SAndy King struct vsock_sock *vsk; 1491fe502c4aSStefano Garzarella const struct vsock_transport *transport; 1492d021c344SAndy King u64 val; 1493d021c344SAndy King 1494d021c344SAndy King if (level != AF_VSOCK) 1495d021c344SAndy King return -ENOPROTOOPT; 1496d021c344SAndy King 1497d021c344SAndy King #define COPY_IN(_v) \ 1498d021c344SAndy King do { \ 1499d021c344SAndy King if (optlen < sizeof(_v)) { \ 1500d021c344SAndy King err = -EINVAL; \ 1501d021c344SAndy King goto exit; \ 1502d021c344SAndy King } \ 1503d021c344SAndy King if (copy_from_user(&_v, optval, sizeof(_v)) != 0) { \ 1504d021c344SAndy King err = -EFAULT; \ 1505d021c344SAndy King goto exit; \ 1506d021c344SAndy King } \ 1507d021c344SAndy King } while (0) 1508d021c344SAndy King 1509d021c344SAndy King err = 0; 1510d021c344SAndy King sk = sock->sk; 1511d021c344SAndy King vsk = vsock_sk(sk); 1512fe502c4aSStefano Garzarella transport = vsk->transport; 1513d021c344SAndy King 1514d021c344SAndy King lock_sock(sk); 1515d021c344SAndy King 1516d021c344SAndy King switch (optname) { 1517d021c344SAndy King case SO_VM_SOCKETS_BUFFER_SIZE: 1518d021c344SAndy King COPY_IN(val); 1519b9f2b0ffSStefano Garzarella vsock_update_buffer_size(vsk, transport, val); 1520d021c344SAndy King break; 1521d021c344SAndy King 1522d021c344SAndy King case SO_VM_SOCKETS_BUFFER_MAX_SIZE: 1523d021c344SAndy King COPY_IN(val); 1524b9f2b0ffSStefano Garzarella vsk->buffer_max_size = val; 1525b9f2b0ffSStefano Garzarella vsock_update_buffer_size(vsk, transport, vsk->buffer_size); 1526d021c344SAndy King break; 1527d021c344SAndy King 1528d021c344SAndy King case SO_VM_SOCKETS_BUFFER_MIN_SIZE: 1529d021c344SAndy King COPY_IN(val); 1530b9f2b0ffSStefano Garzarella vsk->buffer_min_size = val; 1531b9f2b0ffSStefano Garzarella vsock_update_buffer_size(vsk, transport, vsk->buffer_size); 1532d021c344SAndy King break; 1533d021c344SAndy King 1534d021c344SAndy King case SO_VM_SOCKETS_CONNECT_TIMEOUT: { 1535fe0c72f3SArnd Bergmann struct __kernel_old_timeval tv; 1536d021c344SAndy King COPY_IN(tv); 1537d021c344SAndy King if (tv.tv_sec >= 0 && tv.tv_usec < USEC_PER_SEC && 1538d021c344SAndy King tv.tv_sec < (MAX_SCHEDULE_TIMEOUT / HZ - 1)) { 1539d021c344SAndy King vsk->connect_timeout = tv.tv_sec * HZ + 1540d021c344SAndy King DIV_ROUND_UP(tv.tv_usec, (1000000 / HZ)); 1541d021c344SAndy King if (vsk->connect_timeout == 0) 1542d021c344SAndy King vsk->connect_timeout = 1543d021c344SAndy King VSOCK_DEFAULT_CONNECT_TIMEOUT; 1544d021c344SAndy King 1545d021c344SAndy King } else { 1546d021c344SAndy King err = -ERANGE; 1547d021c344SAndy King } 1548d021c344SAndy King break; 1549d021c344SAndy King } 1550d021c344SAndy King 1551d021c344SAndy King default: 1552d021c344SAndy King err = -ENOPROTOOPT; 1553d021c344SAndy King break; 1554d021c344SAndy King } 1555d021c344SAndy King 1556d021c344SAndy King #undef COPY_IN 1557d021c344SAndy King 1558d021c344SAndy King exit: 1559d021c344SAndy King release_sock(sk); 1560d021c344SAndy King return err; 1561d021c344SAndy King } 1562d021c344SAndy King 1563d021c344SAndy King static int vsock_stream_getsockopt(struct socket *sock, 1564d021c344SAndy King int level, int optname, 1565d021c344SAndy King char __user *optval, 1566d021c344SAndy King int __user *optlen) 1567d021c344SAndy King { 1568d021c344SAndy King int err; 1569d021c344SAndy King int len; 1570d021c344SAndy King struct sock *sk; 1571d021c344SAndy King struct vsock_sock *vsk; 1572d021c344SAndy King u64 val; 1573d021c344SAndy King 1574d021c344SAndy King if (level != AF_VSOCK) 1575d021c344SAndy King return -ENOPROTOOPT; 1576d021c344SAndy King 1577d021c344SAndy King err = get_user(len, optlen); 1578d021c344SAndy King if (err != 0) 1579d021c344SAndy King return err; 1580d021c344SAndy King 1581d021c344SAndy King #define COPY_OUT(_v) \ 1582d021c344SAndy King do { \ 1583d021c344SAndy King if (len < sizeof(_v)) \ 1584d021c344SAndy King return -EINVAL; \ 1585d021c344SAndy King \ 1586d021c344SAndy King len = sizeof(_v); \ 1587d021c344SAndy King if (copy_to_user(optval, &_v, len) != 0) \ 1588d021c344SAndy King return -EFAULT; \ 1589d021c344SAndy King \ 1590d021c344SAndy King } while (0) 1591d021c344SAndy King 1592d021c344SAndy King err = 0; 1593d021c344SAndy King sk = sock->sk; 1594d021c344SAndy King vsk = vsock_sk(sk); 1595d021c344SAndy King 1596d021c344SAndy King switch (optname) { 1597d021c344SAndy King case SO_VM_SOCKETS_BUFFER_SIZE: 1598b9f2b0ffSStefano Garzarella val = vsk->buffer_size; 1599d021c344SAndy King COPY_OUT(val); 1600d021c344SAndy King break; 1601d021c344SAndy King 1602d021c344SAndy King case SO_VM_SOCKETS_BUFFER_MAX_SIZE: 1603b9f2b0ffSStefano Garzarella val = vsk->buffer_max_size; 1604d021c344SAndy King COPY_OUT(val); 1605d021c344SAndy King break; 1606d021c344SAndy King 1607d021c344SAndy King case SO_VM_SOCKETS_BUFFER_MIN_SIZE: 1608b9f2b0ffSStefano Garzarella val = vsk->buffer_min_size; 1609d021c344SAndy King COPY_OUT(val); 1610d021c344SAndy King break; 1611d021c344SAndy King 1612d021c344SAndy King case SO_VM_SOCKETS_CONNECT_TIMEOUT: { 1613fe0c72f3SArnd Bergmann struct __kernel_old_timeval tv; 1614d021c344SAndy King tv.tv_sec = vsk->connect_timeout / HZ; 1615d021c344SAndy King tv.tv_usec = 1616d021c344SAndy King (vsk->connect_timeout - 1617d021c344SAndy King tv.tv_sec * HZ) * (1000000 / HZ); 1618d021c344SAndy King COPY_OUT(tv); 1619d021c344SAndy King break; 1620d021c344SAndy King } 1621d021c344SAndy King default: 1622d021c344SAndy King return -ENOPROTOOPT; 1623d021c344SAndy King } 1624d021c344SAndy King 1625d021c344SAndy King err = put_user(len, optlen); 1626d021c344SAndy King if (err != 0) 1627d021c344SAndy King return -EFAULT; 1628d021c344SAndy King 1629d021c344SAndy King #undef COPY_OUT 1630d021c344SAndy King 1631d021c344SAndy King return 0; 1632d021c344SAndy King } 1633d021c344SAndy King 16341b784140SYing Xue static int vsock_stream_sendmsg(struct socket *sock, struct msghdr *msg, 16351b784140SYing Xue size_t len) 1636d021c344SAndy King { 1637d021c344SAndy King struct sock *sk; 1638d021c344SAndy King struct vsock_sock *vsk; 1639fe502c4aSStefano Garzarella const struct vsock_transport *transport; 1640d021c344SAndy King ssize_t total_written; 1641d021c344SAndy King long timeout; 1642d021c344SAndy King int err; 1643d021c344SAndy King struct vsock_transport_send_notify_data send_data; 1644499fde66SWANG Cong DEFINE_WAIT_FUNC(wait, woken_wake_function); 1645d021c344SAndy King 1646d021c344SAndy King sk = sock->sk; 1647d021c344SAndy King vsk = vsock_sk(sk); 1648fe502c4aSStefano Garzarella transport = vsk->transport; 1649d021c344SAndy King total_written = 0; 1650d021c344SAndy King err = 0; 1651d021c344SAndy King 1652d021c344SAndy King if (msg->msg_flags & MSG_OOB) 1653d021c344SAndy King return -EOPNOTSUPP; 1654d021c344SAndy King 1655d021c344SAndy King lock_sock(sk); 1656d021c344SAndy King 1657d021c344SAndy King /* Callers should not provide a destination with stream sockets. */ 1658d021c344SAndy King if (msg->msg_namelen) { 16593b4477d2SStefan Hajnoczi err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP; 1660d021c344SAndy King goto out; 1661d021c344SAndy King } 1662d021c344SAndy King 1663d021c344SAndy King /* Send data only if both sides are not shutdown in the direction. */ 1664d021c344SAndy King if (sk->sk_shutdown & SEND_SHUTDOWN || 1665d021c344SAndy King vsk->peer_shutdown & RCV_SHUTDOWN) { 1666d021c344SAndy King err = -EPIPE; 1667d021c344SAndy King goto out; 1668d021c344SAndy King } 1669d021c344SAndy King 1670c0cfa2d8SStefano Garzarella if (!transport || sk->sk_state != TCP_ESTABLISHED || 1671d021c344SAndy King !vsock_addr_bound(&vsk->local_addr)) { 1672d021c344SAndy King err = -ENOTCONN; 1673d021c344SAndy King goto out; 1674d021c344SAndy King } 1675d021c344SAndy King 1676d021c344SAndy King if (!vsock_addr_bound(&vsk->remote_addr)) { 1677d021c344SAndy King err = -EDESTADDRREQ; 1678d021c344SAndy King goto out; 1679d021c344SAndy King } 1680d021c344SAndy King 1681d021c344SAndy King /* Wait for room in the produce queue to enqueue our user's data. */ 1682d021c344SAndy King timeout = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT); 1683d021c344SAndy King 1684d021c344SAndy King err = transport->notify_send_init(vsk, &send_data); 1685d021c344SAndy King if (err < 0) 1686d021c344SAndy King goto out; 1687d021c344SAndy King 1688d021c344SAndy King while (total_written < len) { 1689d021c344SAndy King ssize_t written; 1690d021c344SAndy King 1691499fde66SWANG Cong add_wait_queue(sk_sleep(sk), &wait); 1692d021c344SAndy King while (vsock_stream_has_space(vsk) == 0 && 1693d021c344SAndy King sk->sk_err == 0 && 1694d021c344SAndy King !(sk->sk_shutdown & SEND_SHUTDOWN) && 1695d021c344SAndy King !(vsk->peer_shutdown & RCV_SHUTDOWN)) { 1696d021c344SAndy King 1697d021c344SAndy King /* Don't wait for non-blocking sockets. */ 1698d021c344SAndy King if (timeout == 0) { 1699d021c344SAndy King err = -EAGAIN; 1700499fde66SWANG Cong remove_wait_queue(sk_sleep(sk), &wait); 1701f7f9b5e7SClaudio Imbrenda goto out_err; 1702d021c344SAndy King } 1703d021c344SAndy King 1704d021c344SAndy King err = transport->notify_send_pre_block(vsk, &send_data); 1705f7f9b5e7SClaudio Imbrenda if (err < 0) { 1706499fde66SWANG Cong remove_wait_queue(sk_sleep(sk), &wait); 1707f7f9b5e7SClaudio Imbrenda goto out_err; 1708f7f9b5e7SClaudio Imbrenda } 1709d021c344SAndy King 1710d021c344SAndy King release_sock(sk); 1711499fde66SWANG Cong timeout = wait_woken(&wait, TASK_INTERRUPTIBLE, timeout); 1712d021c344SAndy King lock_sock(sk); 1713d021c344SAndy King if (signal_pending(current)) { 1714d021c344SAndy King err = sock_intr_errno(timeout); 1715499fde66SWANG Cong remove_wait_queue(sk_sleep(sk), &wait); 1716f7f9b5e7SClaudio Imbrenda goto out_err; 1717d021c344SAndy King } else if (timeout == 0) { 1718d021c344SAndy King err = -EAGAIN; 1719499fde66SWANG Cong remove_wait_queue(sk_sleep(sk), &wait); 1720f7f9b5e7SClaudio Imbrenda goto out_err; 1721d021c344SAndy King } 1722d021c344SAndy King } 1723499fde66SWANG Cong remove_wait_queue(sk_sleep(sk), &wait); 1724d021c344SAndy King 1725d021c344SAndy King /* These checks occur both as part of and after the loop 1726d021c344SAndy King * conditional since we need to check before and after 1727d021c344SAndy King * sleeping. 1728d021c344SAndy King */ 1729d021c344SAndy King if (sk->sk_err) { 1730d021c344SAndy King err = -sk->sk_err; 1731f7f9b5e7SClaudio Imbrenda goto out_err; 1732d021c344SAndy King } else if ((sk->sk_shutdown & SEND_SHUTDOWN) || 1733d021c344SAndy King (vsk->peer_shutdown & RCV_SHUTDOWN)) { 1734d021c344SAndy King err = -EPIPE; 1735f7f9b5e7SClaudio Imbrenda goto out_err; 1736d021c344SAndy King } 1737d021c344SAndy King 1738d021c344SAndy King err = transport->notify_send_pre_enqueue(vsk, &send_data); 1739d021c344SAndy King if (err < 0) 1740f7f9b5e7SClaudio Imbrenda goto out_err; 1741d021c344SAndy King 1742d021c344SAndy King /* Note that enqueue will only write as many bytes as are free 1743d021c344SAndy King * in the produce queue, so we don't need to ensure len is 1744d021c344SAndy King * smaller than the queue size. It is the caller's 1745d021c344SAndy King * responsibility to check how many bytes we were able to send. 1746d021c344SAndy King */ 1747d021c344SAndy King 1748d021c344SAndy King written = transport->stream_enqueue( 17490f7db23aSAl Viro vsk, msg, 1750d021c344SAndy King len - total_written); 1751d021c344SAndy King if (written < 0) { 1752d021c344SAndy King err = -ENOMEM; 1753f7f9b5e7SClaudio Imbrenda goto out_err; 1754d021c344SAndy King } 1755d021c344SAndy King 1756d021c344SAndy King total_written += written; 1757d021c344SAndy King 1758d021c344SAndy King err = transport->notify_send_post_enqueue( 1759d021c344SAndy King vsk, written, &send_data); 1760d021c344SAndy King if (err < 0) 1761f7f9b5e7SClaudio Imbrenda goto out_err; 1762d021c344SAndy King 1763d021c344SAndy King } 1764d021c344SAndy King 1765f7f9b5e7SClaudio Imbrenda out_err: 1766d021c344SAndy King if (total_written > 0) 1767d021c344SAndy King err = total_written; 1768d021c344SAndy King out: 1769d021c344SAndy King release_sock(sk); 1770d021c344SAndy King return err; 1771d021c344SAndy King } 1772d021c344SAndy King 1773d021c344SAndy King 1774d021c344SAndy King static int 17751b784140SYing Xue vsock_stream_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, 17761b784140SYing Xue int flags) 1777d021c344SAndy King { 1778d021c344SAndy King struct sock *sk; 1779d021c344SAndy King struct vsock_sock *vsk; 1780fe502c4aSStefano Garzarella const struct vsock_transport *transport; 1781d021c344SAndy King int err; 1782d021c344SAndy King size_t target; 1783d021c344SAndy King ssize_t copied; 1784d021c344SAndy King long timeout; 1785d021c344SAndy King struct vsock_transport_recv_notify_data recv_data; 1786d021c344SAndy King 1787d021c344SAndy King DEFINE_WAIT(wait); 1788d021c344SAndy King 1789d021c344SAndy King sk = sock->sk; 1790d021c344SAndy King vsk = vsock_sk(sk); 1791fe502c4aSStefano Garzarella transport = vsk->transport; 1792d021c344SAndy King err = 0; 1793d021c344SAndy King 1794d021c344SAndy King lock_sock(sk); 1795d021c344SAndy King 1796c0cfa2d8SStefano Garzarella if (!transport || sk->sk_state != TCP_ESTABLISHED) { 1797d021c344SAndy King /* Recvmsg is supposed to return 0 if a peer performs an 1798d021c344SAndy King * orderly shutdown. Differentiate between that case and when a 1799d021c344SAndy King * peer has not connected or a local shutdown occured with the 1800d021c344SAndy King * SOCK_DONE flag. 1801d021c344SAndy King */ 1802d021c344SAndy King if (sock_flag(sk, SOCK_DONE)) 1803d021c344SAndy King err = 0; 1804d021c344SAndy King else 1805d021c344SAndy King err = -ENOTCONN; 1806d021c344SAndy King 1807d021c344SAndy King goto out; 1808d021c344SAndy King } 1809d021c344SAndy King 1810d021c344SAndy King if (flags & MSG_OOB) { 1811d021c344SAndy King err = -EOPNOTSUPP; 1812d021c344SAndy King goto out; 1813d021c344SAndy King } 1814d021c344SAndy King 1815d021c344SAndy King /* We don't check peer_shutdown flag here since peer may actually shut 1816d021c344SAndy King * down, but there can be data in the queue that a local socket can 1817d021c344SAndy King * receive. 1818d021c344SAndy King */ 1819d021c344SAndy King if (sk->sk_shutdown & RCV_SHUTDOWN) { 1820d021c344SAndy King err = 0; 1821d021c344SAndy King goto out; 1822d021c344SAndy King } 1823d021c344SAndy King 1824d021c344SAndy King /* It is valid on Linux to pass in a zero-length receive buffer. This 1825d021c344SAndy King * is not an error. We may as well bail out now. 1826d021c344SAndy King */ 1827d021c344SAndy King if (!len) { 1828d021c344SAndy King err = 0; 1829d021c344SAndy King goto out; 1830d021c344SAndy King } 1831d021c344SAndy King 1832d021c344SAndy King /* We must not copy less than target bytes into the user's buffer 1833d021c344SAndy King * before returning successfully, so we wait for the consume queue to 1834d021c344SAndy King * have that much data to consume before dequeueing. Note that this 1835d021c344SAndy King * makes it impossible to handle cases where target is greater than the 1836d021c344SAndy King * queue size. 1837d021c344SAndy King */ 1838d021c344SAndy King target = sock_rcvlowat(sk, flags & MSG_WAITALL, len); 1839d021c344SAndy King if (target >= transport->stream_rcvhiwat(vsk)) { 1840d021c344SAndy King err = -ENOMEM; 1841d021c344SAndy King goto out; 1842d021c344SAndy King } 1843d021c344SAndy King timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT); 1844d021c344SAndy King copied = 0; 1845d021c344SAndy King 1846d021c344SAndy King err = transport->notify_recv_init(vsk, target, &recv_data); 1847d021c344SAndy King if (err < 0) 1848d021c344SAndy King goto out; 1849d021c344SAndy King 1850d021c344SAndy King 1851d021c344SAndy King while (1) { 1852f7f9b5e7SClaudio Imbrenda s64 ready; 1853f7f9b5e7SClaudio Imbrenda 1854f7f9b5e7SClaudio Imbrenda prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); 1855f7f9b5e7SClaudio Imbrenda ready = vsock_stream_has_data(vsk); 1856f7f9b5e7SClaudio Imbrenda 1857f7f9b5e7SClaudio Imbrenda if (ready == 0) { 1858f7f9b5e7SClaudio Imbrenda if (sk->sk_err != 0 || 1859f7f9b5e7SClaudio Imbrenda (sk->sk_shutdown & RCV_SHUTDOWN) || 1860f7f9b5e7SClaudio Imbrenda (vsk->peer_shutdown & SEND_SHUTDOWN)) { 1861f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(sk), &wait); 1862f7f9b5e7SClaudio Imbrenda break; 1863f7f9b5e7SClaudio Imbrenda } 1864f7f9b5e7SClaudio Imbrenda /* Don't wait for non-blocking sockets. */ 1865f7f9b5e7SClaudio Imbrenda if (timeout == 0) { 1866f7f9b5e7SClaudio Imbrenda err = -EAGAIN; 1867f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(sk), &wait); 1868f7f9b5e7SClaudio Imbrenda break; 1869f7f9b5e7SClaudio Imbrenda } 1870f7f9b5e7SClaudio Imbrenda 1871f7f9b5e7SClaudio Imbrenda err = transport->notify_recv_pre_block( 1872f7f9b5e7SClaudio Imbrenda vsk, target, &recv_data); 1873f7f9b5e7SClaudio Imbrenda if (err < 0) { 1874f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(sk), &wait); 1875f7f9b5e7SClaudio Imbrenda break; 1876f7f9b5e7SClaudio Imbrenda } 1877f7f9b5e7SClaudio Imbrenda release_sock(sk); 1878f7f9b5e7SClaudio Imbrenda timeout = schedule_timeout(timeout); 1879f7f9b5e7SClaudio Imbrenda lock_sock(sk); 1880f7f9b5e7SClaudio Imbrenda 1881f7f9b5e7SClaudio Imbrenda if (signal_pending(current)) { 1882f7f9b5e7SClaudio Imbrenda err = sock_intr_errno(timeout); 1883f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(sk), &wait); 1884f7f9b5e7SClaudio Imbrenda break; 1885f7f9b5e7SClaudio Imbrenda } else if (timeout == 0) { 1886f7f9b5e7SClaudio Imbrenda err = -EAGAIN; 1887f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(sk), &wait); 1888f7f9b5e7SClaudio Imbrenda break; 1889f7f9b5e7SClaudio Imbrenda } 1890f7f9b5e7SClaudio Imbrenda } else { 1891f7f9b5e7SClaudio Imbrenda ssize_t read; 1892f7f9b5e7SClaudio Imbrenda 1893f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(sk), &wait); 1894d021c344SAndy King 1895d021c344SAndy King if (ready < 0) { 1896f7f9b5e7SClaudio Imbrenda /* Invalid queue pair content. XXX This should 1897f7f9b5e7SClaudio Imbrenda * be changed to a connection reset in a later 1898f7f9b5e7SClaudio Imbrenda * change. 1899d021c344SAndy King */ 1900d021c344SAndy King 1901d021c344SAndy King err = -ENOMEM; 1902f7f9b5e7SClaudio Imbrenda goto out; 1903f7f9b5e7SClaudio Imbrenda } 1904d021c344SAndy King 1905d021c344SAndy King err = transport->notify_recv_pre_dequeue( 1906d021c344SAndy King vsk, target, &recv_data); 1907d021c344SAndy King if (err < 0) 1908d021c344SAndy King break; 1909d021c344SAndy King 1910d021c344SAndy King read = transport->stream_dequeue( 19110f7db23aSAl Viro vsk, msg, 1912d021c344SAndy King len - copied, flags); 1913d021c344SAndy King if (read < 0) { 1914d021c344SAndy King err = -ENOMEM; 1915d021c344SAndy King break; 1916d021c344SAndy King } 1917d021c344SAndy King 1918d021c344SAndy King copied += read; 1919d021c344SAndy King 1920d021c344SAndy King err = transport->notify_recv_post_dequeue( 1921d021c344SAndy King vsk, target, read, 1922d021c344SAndy King !(flags & MSG_PEEK), &recv_data); 1923d021c344SAndy King if (err < 0) 1924f7f9b5e7SClaudio Imbrenda goto out; 1925d021c344SAndy King 1926d021c344SAndy King if (read >= target || flags & MSG_PEEK) 1927d021c344SAndy King break; 1928d021c344SAndy King 1929d021c344SAndy King target -= read; 1930d021c344SAndy King } 1931d021c344SAndy King } 1932d021c344SAndy King 1933d021c344SAndy King if (sk->sk_err) 1934d021c344SAndy King err = -sk->sk_err; 1935d021c344SAndy King else if (sk->sk_shutdown & RCV_SHUTDOWN) 1936d021c344SAndy King err = 0; 1937d021c344SAndy King 1938dedc58e0SIan Campbell if (copied > 0) 1939d021c344SAndy King err = copied; 1940d021c344SAndy King 1941d021c344SAndy King out: 1942d021c344SAndy King release_sock(sk); 1943d021c344SAndy King return err; 1944d021c344SAndy King } 1945d021c344SAndy King 1946d021c344SAndy King static const struct proto_ops vsock_stream_ops = { 1947d021c344SAndy King .family = PF_VSOCK, 1948d021c344SAndy King .owner = THIS_MODULE, 1949d021c344SAndy King .release = vsock_release, 1950d021c344SAndy King .bind = vsock_bind, 1951d021c344SAndy King .connect = vsock_stream_connect, 1952d021c344SAndy King .socketpair = sock_no_socketpair, 1953d021c344SAndy King .accept = vsock_accept, 1954d021c344SAndy King .getname = vsock_getname, 1955a11e1d43SLinus Torvalds .poll = vsock_poll, 1956d021c344SAndy King .ioctl = sock_no_ioctl, 1957d021c344SAndy King .listen = vsock_listen, 1958d021c344SAndy King .shutdown = vsock_shutdown, 1959d021c344SAndy King .setsockopt = vsock_stream_setsockopt, 1960d021c344SAndy King .getsockopt = vsock_stream_getsockopt, 1961d021c344SAndy King .sendmsg = vsock_stream_sendmsg, 1962d021c344SAndy King .recvmsg = vsock_stream_recvmsg, 1963d021c344SAndy King .mmap = sock_no_mmap, 1964d021c344SAndy King .sendpage = sock_no_sendpage, 1965d021c344SAndy King }; 1966d021c344SAndy King 1967d021c344SAndy King static int vsock_create(struct net *net, struct socket *sock, 1968d021c344SAndy King int protocol, int kern) 1969d021c344SAndy King { 1970c0cfa2d8SStefano Garzarella struct vsock_sock *vsk; 197155f3e149SStefano Garzarella struct sock *sk; 1972c0cfa2d8SStefano Garzarella int ret; 197355f3e149SStefano Garzarella 1974d021c344SAndy King if (!sock) 1975d021c344SAndy King return -EINVAL; 1976d021c344SAndy King 19776cf1c5fcSAndy King if (protocol && protocol != PF_VSOCK) 1978d021c344SAndy King return -EPROTONOSUPPORT; 1979d021c344SAndy King 1980d021c344SAndy King switch (sock->type) { 1981d021c344SAndy King case SOCK_DGRAM: 1982d021c344SAndy King sock->ops = &vsock_dgram_ops; 1983d021c344SAndy King break; 1984d021c344SAndy King case SOCK_STREAM: 1985d021c344SAndy King sock->ops = &vsock_stream_ops; 1986d021c344SAndy King break; 1987d021c344SAndy King default: 1988d021c344SAndy King return -ESOCKTNOSUPPORT; 1989d021c344SAndy King } 1990d021c344SAndy King 1991d021c344SAndy King sock->state = SS_UNCONNECTED; 1992d021c344SAndy King 199355f3e149SStefano Garzarella sk = __vsock_create(net, sock, NULL, GFP_KERNEL, 0, kern); 199455f3e149SStefano Garzarella if (!sk) 199555f3e149SStefano Garzarella return -ENOMEM; 199655f3e149SStefano Garzarella 1997c0cfa2d8SStefano Garzarella vsk = vsock_sk(sk); 1998c0cfa2d8SStefano Garzarella 1999c0cfa2d8SStefano Garzarella if (sock->type == SOCK_DGRAM) { 2000c0cfa2d8SStefano Garzarella ret = vsock_assign_transport(vsk, NULL); 2001c0cfa2d8SStefano Garzarella if (ret < 0) { 2002c0cfa2d8SStefano Garzarella sock_put(sk); 2003c0cfa2d8SStefano Garzarella return ret; 2004c0cfa2d8SStefano Garzarella } 2005c0cfa2d8SStefano Garzarella } 2006c0cfa2d8SStefano Garzarella 2007c0cfa2d8SStefano Garzarella vsock_insert_unbound(vsk); 200855f3e149SStefano Garzarella 200955f3e149SStefano Garzarella return 0; 2010d021c344SAndy King } 2011d021c344SAndy King 2012d021c344SAndy King static const struct net_proto_family vsock_family_ops = { 2013d021c344SAndy King .family = AF_VSOCK, 2014d021c344SAndy King .create = vsock_create, 2015d021c344SAndy King .owner = THIS_MODULE, 2016d021c344SAndy King }; 2017d021c344SAndy King 2018d021c344SAndy King static long vsock_dev_do_ioctl(struct file *filp, 2019d021c344SAndy King unsigned int cmd, void __user *ptr) 2020d021c344SAndy King { 2021d021c344SAndy King u32 __user *p = ptr; 2022c0cfa2d8SStefano Garzarella u32 cid = VMADDR_CID_ANY; 2023d021c344SAndy King int retval = 0; 2024d021c344SAndy King 2025d021c344SAndy King switch (cmd) { 2026d021c344SAndy King case IOCTL_VM_SOCKETS_GET_LOCAL_CID: 2027c0cfa2d8SStefano Garzarella /* To be compatible with the VMCI behavior, we prioritize the 2028c0cfa2d8SStefano Garzarella * guest CID instead of well-know host CID (VMADDR_CID_HOST). 2029c0cfa2d8SStefano Garzarella */ 2030c0cfa2d8SStefano Garzarella if (transport_g2h) 2031c0cfa2d8SStefano Garzarella cid = transport_g2h->get_local_cid(); 2032c0cfa2d8SStefano Garzarella else if (transport_h2g) 2033c0cfa2d8SStefano Garzarella cid = transport_h2g->get_local_cid(); 2034c0cfa2d8SStefano Garzarella 2035c0cfa2d8SStefano Garzarella if (put_user(cid, p) != 0) 2036d021c344SAndy King retval = -EFAULT; 2037d021c344SAndy King break; 2038d021c344SAndy King 2039d021c344SAndy King default: 2040d021c344SAndy King pr_err("Unknown ioctl %d\n", cmd); 2041d021c344SAndy King retval = -EINVAL; 2042d021c344SAndy King } 2043d021c344SAndy King 2044d021c344SAndy King return retval; 2045d021c344SAndy King } 2046d021c344SAndy King 2047d021c344SAndy King static long vsock_dev_ioctl(struct file *filp, 2048d021c344SAndy King unsigned int cmd, unsigned long arg) 2049d021c344SAndy King { 2050d021c344SAndy King return vsock_dev_do_ioctl(filp, cmd, (void __user *)arg); 2051d021c344SAndy King } 2052d021c344SAndy King 2053d021c344SAndy King #ifdef CONFIG_COMPAT 2054d021c344SAndy King static long vsock_dev_compat_ioctl(struct file *filp, 2055d021c344SAndy King unsigned int cmd, unsigned long arg) 2056d021c344SAndy King { 2057d021c344SAndy King return vsock_dev_do_ioctl(filp, cmd, compat_ptr(arg)); 2058d021c344SAndy King } 2059d021c344SAndy King #endif 2060d021c344SAndy King 2061d021c344SAndy King static const struct file_operations vsock_device_ops = { 2062d021c344SAndy King .owner = THIS_MODULE, 2063d021c344SAndy King .unlocked_ioctl = vsock_dev_ioctl, 2064d021c344SAndy King #ifdef CONFIG_COMPAT 2065d021c344SAndy King .compat_ioctl = vsock_dev_compat_ioctl, 2066d021c344SAndy King #endif 2067d021c344SAndy King .open = nonseekable_open, 2068d021c344SAndy King }; 2069d021c344SAndy King 2070d021c344SAndy King static struct miscdevice vsock_device = { 2071d021c344SAndy King .name = "vsock", 2072d021c344SAndy King .fops = &vsock_device_ops, 2073d021c344SAndy King }; 2074d021c344SAndy King 2075c0cfa2d8SStefano Garzarella static int __init vsock_init(void) 2076d021c344SAndy King { 2077c0cfa2d8SStefano Garzarella int err = 0; 20782c4a336eSAndy King 2079c0cfa2d8SStefano Garzarella vsock_init_tables(); 20802c4a336eSAndy King 2081c0cfa2d8SStefano Garzarella vsock_proto.owner = THIS_MODULE; 20826ad0b2f7SAsias He vsock_device.minor = MISC_DYNAMIC_MINOR; 2083d021c344SAndy King err = misc_register(&vsock_device); 2084d021c344SAndy King if (err) { 2085d021c344SAndy King pr_err("Failed to register misc device\n"); 2086f6a835bbSGao feng goto err_reset_transport; 2087d021c344SAndy King } 2088d021c344SAndy King 2089d021c344SAndy King err = proto_register(&vsock_proto, 1); /* we want our slab */ 2090d021c344SAndy King if (err) { 2091d021c344SAndy King pr_err("Cannot register vsock protocol\n"); 2092f6a835bbSGao feng goto err_deregister_misc; 2093d021c344SAndy King } 2094d021c344SAndy King 2095d021c344SAndy King err = sock_register(&vsock_family_ops); 2096d021c344SAndy King if (err) { 2097d021c344SAndy King pr_err("could not register af_vsock (%d) address family: %d\n", 2098d021c344SAndy King AF_VSOCK, err); 2099d021c344SAndy King goto err_unregister_proto; 2100d021c344SAndy King } 2101d021c344SAndy King 2102d021c344SAndy King return 0; 2103d021c344SAndy King 2104d021c344SAndy King err_unregister_proto: 2105d021c344SAndy King proto_unregister(&vsock_proto); 2106f6a835bbSGao feng err_deregister_misc: 2107d021c344SAndy King misc_deregister(&vsock_device); 2108f6a835bbSGao feng err_reset_transport: 2109d021c344SAndy King return err; 2110d021c344SAndy King } 2111d021c344SAndy King 2112c0cfa2d8SStefano Garzarella static void __exit vsock_exit(void) 2113d021c344SAndy King { 2114d021c344SAndy King misc_deregister(&vsock_device); 2115d021c344SAndy King sock_unregister(AF_VSOCK); 2116d021c344SAndy King proto_unregister(&vsock_proto); 2117d021c344SAndy King } 2118d021c344SAndy King 2119daabfbcaSStefano Garzarella const struct vsock_transport *vsock_core_get_transport(struct vsock_sock *vsk) 21200b01aeb3SStefan Hajnoczi { 2121daabfbcaSStefano Garzarella return vsk->transport; 21220b01aeb3SStefan Hajnoczi } 21230b01aeb3SStefan Hajnoczi EXPORT_SYMBOL_GPL(vsock_core_get_transport); 21240b01aeb3SStefan Hajnoczi 2125c0cfa2d8SStefano Garzarella int vsock_core_register(const struct vsock_transport *t, int features) 212605e489b1SStefan Hajnoczi { 2127c0cfa2d8SStefano Garzarella const struct vsock_transport *t_h2g, *t_g2h, *t_dgram; 2128c0cfa2d8SStefano Garzarella int err = mutex_lock_interruptible(&vsock_register_mutex); 2129c0cfa2d8SStefano Garzarella 2130c0cfa2d8SStefano Garzarella if (err) 2131c0cfa2d8SStefano Garzarella return err; 2132c0cfa2d8SStefano Garzarella 2133c0cfa2d8SStefano Garzarella t_h2g = transport_h2g; 2134c0cfa2d8SStefano Garzarella t_g2h = transport_g2h; 2135c0cfa2d8SStefano Garzarella t_dgram = transport_dgram; 2136c0cfa2d8SStefano Garzarella 2137c0cfa2d8SStefano Garzarella if (features & VSOCK_TRANSPORT_F_H2G) { 2138c0cfa2d8SStefano Garzarella if (t_h2g) { 2139c0cfa2d8SStefano Garzarella err = -EBUSY; 2140c0cfa2d8SStefano Garzarella goto err_busy; 2141c0cfa2d8SStefano Garzarella } 2142c0cfa2d8SStefano Garzarella t_h2g = t; 214305e489b1SStefan Hajnoczi } 214405e489b1SStefan Hajnoczi 2145c0cfa2d8SStefano Garzarella if (features & VSOCK_TRANSPORT_F_G2H) { 2146c0cfa2d8SStefano Garzarella if (t_g2h) { 2147c0cfa2d8SStefano Garzarella err = -EBUSY; 2148c0cfa2d8SStefano Garzarella goto err_busy; 2149c0cfa2d8SStefano Garzarella } 2150c0cfa2d8SStefano Garzarella t_g2h = t; 2151c0cfa2d8SStefano Garzarella } 2152c0cfa2d8SStefano Garzarella 2153c0cfa2d8SStefano Garzarella if (features & VSOCK_TRANSPORT_F_DGRAM) { 2154c0cfa2d8SStefano Garzarella if (t_dgram) { 2155c0cfa2d8SStefano Garzarella err = -EBUSY; 2156c0cfa2d8SStefano Garzarella goto err_busy; 2157c0cfa2d8SStefano Garzarella } 2158c0cfa2d8SStefano Garzarella t_dgram = t; 2159c0cfa2d8SStefano Garzarella } 2160c0cfa2d8SStefano Garzarella 2161c0cfa2d8SStefano Garzarella transport_h2g = t_h2g; 2162c0cfa2d8SStefano Garzarella transport_g2h = t_g2h; 2163c0cfa2d8SStefano Garzarella transport_dgram = t_dgram; 2164c0cfa2d8SStefano Garzarella 2165c0cfa2d8SStefano Garzarella err_busy: 2166c0cfa2d8SStefano Garzarella mutex_unlock(&vsock_register_mutex); 2167c0cfa2d8SStefano Garzarella return err; 2168c0cfa2d8SStefano Garzarella } 2169c0cfa2d8SStefano Garzarella EXPORT_SYMBOL_GPL(vsock_core_register); 2170c0cfa2d8SStefano Garzarella 2171c0cfa2d8SStefano Garzarella void vsock_core_unregister(const struct vsock_transport *t) 2172c0cfa2d8SStefano Garzarella { 2173c0cfa2d8SStefano Garzarella mutex_lock(&vsock_register_mutex); 2174c0cfa2d8SStefano Garzarella 2175c0cfa2d8SStefano Garzarella if (transport_h2g == t) 2176c0cfa2d8SStefano Garzarella transport_h2g = NULL; 2177c0cfa2d8SStefano Garzarella 2178c0cfa2d8SStefano Garzarella if (transport_g2h == t) 2179c0cfa2d8SStefano Garzarella transport_g2h = NULL; 2180c0cfa2d8SStefano Garzarella 2181c0cfa2d8SStefano Garzarella if (transport_dgram == t) 2182c0cfa2d8SStefano Garzarella transport_dgram = NULL; 2183c0cfa2d8SStefano Garzarella 2184c0cfa2d8SStefano Garzarella mutex_unlock(&vsock_register_mutex); 2185c0cfa2d8SStefano Garzarella } 2186c0cfa2d8SStefano Garzarella EXPORT_SYMBOL_GPL(vsock_core_unregister); 2187c0cfa2d8SStefano Garzarella 2188c0cfa2d8SStefano Garzarella module_init(vsock_init); 218905e489b1SStefan Hajnoczi module_exit(vsock_exit); 2190c1eef220SCong Wang 2191d021c344SAndy King MODULE_AUTHOR("VMware, Inc."); 2192d021c344SAndy King MODULE_DESCRIPTION("VMware Virtual Socket Family"); 21931190cfdbSJorgen Hansen MODULE_VERSION("1.0.2.0-k"); 2194d021c344SAndy King MODULE_LICENSE("GPL v2"); 2195