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; 1390e121905SStefano Garzarella /* Transport used for local communication */ 1400e121905SStefano Garzarella static const struct vsock_transport *transport_local; 141d021c344SAndy King static DEFINE_MUTEX(vsock_register_mutex); 142d021c344SAndy King 143d021c344SAndy King /**** UTILS ****/ 144d021c344SAndy King 145d021c344SAndy King /* Each bound VSocket is stored in the bind hash table and each connected 146d021c344SAndy King * VSocket is stored in the connected hash table. 147d021c344SAndy King * 148d021c344SAndy King * Unbound sockets are all put on the same list attached to the end of the hash 149d021c344SAndy King * table (vsock_unbound_sockets). Bound sockets are added to the hash table in 150d021c344SAndy King * the bucket that their local address hashes to (vsock_bound_sockets(addr) 151d021c344SAndy King * represents the list that addr hashes to). 152d021c344SAndy King * 153d021c344SAndy King * Specifically, we initialize the vsock_bind_table array to a size of 154d021c344SAndy King * VSOCK_HASH_SIZE + 1 so that vsock_bind_table[0] through 155d021c344SAndy King * vsock_bind_table[VSOCK_HASH_SIZE - 1] are for bound sockets and 156d021c344SAndy King * vsock_bind_table[VSOCK_HASH_SIZE] is for unbound sockets. The hash function 157a49dd9dcSAsias He * mods with VSOCK_HASH_SIZE to ensure this. 158d021c344SAndy King */ 159d021c344SAndy King #define MAX_PORT_RETRIES 24 160d021c344SAndy King 161a49dd9dcSAsias He #define VSOCK_HASH(addr) ((addr)->svm_port % VSOCK_HASH_SIZE) 162d021c344SAndy King #define vsock_bound_sockets(addr) (&vsock_bind_table[VSOCK_HASH(addr)]) 163d021c344SAndy King #define vsock_unbound_sockets (&vsock_bind_table[VSOCK_HASH_SIZE]) 164d021c344SAndy King 165d021c344SAndy King /* XXX This can probably be implemented in a better way. */ 166d021c344SAndy King #define VSOCK_CONN_HASH(src, dst) \ 167a49dd9dcSAsias He (((src)->svm_cid ^ (dst)->svm_port) % VSOCK_HASH_SIZE) 168d021c344SAndy King #define vsock_connected_sockets(src, dst) \ 169d021c344SAndy King (&vsock_connected_table[VSOCK_CONN_HASH(src, dst)]) 170d021c344SAndy King #define vsock_connected_sockets_vsk(vsk) \ 171d021c344SAndy King vsock_connected_sockets(&(vsk)->remote_addr, &(vsk)->local_addr) 172d021c344SAndy King 17344f20980SStefan Hajnoczi struct list_head vsock_bind_table[VSOCK_HASH_SIZE + 1]; 17444f20980SStefan Hajnoczi EXPORT_SYMBOL_GPL(vsock_bind_table); 17544f20980SStefan Hajnoczi struct list_head vsock_connected_table[VSOCK_HASH_SIZE]; 17644f20980SStefan Hajnoczi EXPORT_SYMBOL_GPL(vsock_connected_table); 17744f20980SStefan Hajnoczi DEFINE_SPINLOCK(vsock_table_lock); 17844f20980SStefan Hajnoczi EXPORT_SYMBOL_GPL(vsock_table_lock); 179d021c344SAndy King 180b3a6dfe8SAsias He /* Autobind this socket to the local address if necessary. */ 181b3a6dfe8SAsias He static int vsock_auto_bind(struct vsock_sock *vsk) 182b3a6dfe8SAsias He { 183b3a6dfe8SAsias He struct sock *sk = sk_vsock(vsk); 184b3a6dfe8SAsias He struct sockaddr_vm local_addr; 185b3a6dfe8SAsias He 186b3a6dfe8SAsias He if (vsock_addr_bound(&vsk->local_addr)) 187b3a6dfe8SAsias He return 0; 188b3a6dfe8SAsias He vsock_addr_init(&local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 189b3a6dfe8SAsias He return __vsock_bind(sk, &local_addr); 190b3a6dfe8SAsias He } 191b3a6dfe8SAsias He 192c0cfa2d8SStefano Garzarella static void vsock_init_tables(void) 193d021c344SAndy King { 194d021c344SAndy King int i; 195d021c344SAndy King 196d021c344SAndy King for (i = 0; i < ARRAY_SIZE(vsock_bind_table); i++) 197d021c344SAndy King INIT_LIST_HEAD(&vsock_bind_table[i]); 198d021c344SAndy King 199d021c344SAndy King for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) 200d021c344SAndy King INIT_LIST_HEAD(&vsock_connected_table[i]); 201d021c344SAndy King } 202d021c344SAndy King 203d021c344SAndy King static void __vsock_insert_bound(struct list_head *list, 204d021c344SAndy King struct vsock_sock *vsk) 205d021c344SAndy King { 206d021c344SAndy King sock_hold(&vsk->sk); 207d021c344SAndy King list_add(&vsk->bound_table, list); 208d021c344SAndy King } 209d021c344SAndy King 210d021c344SAndy King static void __vsock_insert_connected(struct list_head *list, 211d021c344SAndy King struct vsock_sock *vsk) 212d021c344SAndy King { 213d021c344SAndy King sock_hold(&vsk->sk); 214d021c344SAndy King list_add(&vsk->connected_table, list); 215d021c344SAndy King } 216d021c344SAndy King 217d021c344SAndy King static void __vsock_remove_bound(struct vsock_sock *vsk) 218d021c344SAndy King { 219d021c344SAndy King list_del_init(&vsk->bound_table); 220d021c344SAndy King sock_put(&vsk->sk); 221d021c344SAndy King } 222d021c344SAndy King 223d021c344SAndy King static void __vsock_remove_connected(struct vsock_sock *vsk) 224d021c344SAndy King { 225d021c344SAndy King list_del_init(&vsk->connected_table); 226d021c344SAndy King sock_put(&vsk->sk); 227d021c344SAndy King } 228d021c344SAndy King 229d021c344SAndy King static struct sock *__vsock_find_bound_socket(struct sockaddr_vm *addr) 230d021c344SAndy King { 231d021c344SAndy King struct vsock_sock *vsk; 232d021c344SAndy King 23336c5b48bSStefano Garzarella list_for_each_entry(vsk, vsock_bound_sockets(addr), bound_table) { 23436c5b48bSStefano Garzarella if (vsock_addr_equals_addr(addr, &vsk->local_addr)) 235d021c344SAndy King return sk_vsock(vsk); 236d021c344SAndy King 23736c5b48bSStefano Garzarella if (addr->svm_port == vsk->local_addr.svm_port && 23836c5b48bSStefano Garzarella (vsk->local_addr.svm_cid == VMADDR_CID_ANY || 23936c5b48bSStefano Garzarella addr->svm_cid == VMADDR_CID_ANY)) 24036c5b48bSStefano Garzarella return sk_vsock(vsk); 24136c5b48bSStefano Garzarella } 24236c5b48bSStefano Garzarella 243d021c344SAndy King return NULL; 244d021c344SAndy King } 245d021c344SAndy King 246d021c344SAndy King static struct sock *__vsock_find_connected_socket(struct sockaddr_vm *src, 247d021c344SAndy King struct sockaddr_vm *dst) 248d021c344SAndy King { 249d021c344SAndy King struct vsock_sock *vsk; 250d021c344SAndy King 251d021c344SAndy King list_for_each_entry(vsk, vsock_connected_sockets(src, dst), 252d021c344SAndy King connected_table) { 253990454b5SReilly Grant if (vsock_addr_equals_addr(src, &vsk->remote_addr) && 254990454b5SReilly Grant dst->svm_port == vsk->local_addr.svm_port) { 255d021c344SAndy King return sk_vsock(vsk); 256d021c344SAndy King } 257d021c344SAndy King } 258d021c344SAndy King 259d021c344SAndy King return NULL; 260d021c344SAndy King } 261d021c344SAndy King 262d021c344SAndy King static void vsock_insert_unbound(struct vsock_sock *vsk) 263d021c344SAndy King { 264d021c344SAndy King spin_lock_bh(&vsock_table_lock); 265d021c344SAndy King __vsock_insert_bound(vsock_unbound_sockets, vsk); 266d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 267d021c344SAndy King } 268d021c344SAndy King 269d021c344SAndy King void vsock_insert_connected(struct vsock_sock *vsk) 270d021c344SAndy King { 271d021c344SAndy King struct list_head *list = vsock_connected_sockets( 272d021c344SAndy King &vsk->remote_addr, &vsk->local_addr); 273d021c344SAndy King 274d021c344SAndy King spin_lock_bh(&vsock_table_lock); 275d021c344SAndy King __vsock_insert_connected(list, vsk); 276d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 277d021c344SAndy King } 278d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_insert_connected); 279d021c344SAndy King 280d021c344SAndy King void vsock_remove_bound(struct vsock_sock *vsk) 281d021c344SAndy King { 282d021c344SAndy King spin_lock_bh(&vsock_table_lock); 283d5afa82cSSunil Muthuswamy if (__vsock_in_bound_table(vsk)) 284d021c344SAndy King __vsock_remove_bound(vsk); 285d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 286d021c344SAndy King } 287d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_remove_bound); 288d021c344SAndy King 289d021c344SAndy King void vsock_remove_connected(struct vsock_sock *vsk) 290d021c344SAndy King { 291d021c344SAndy King spin_lock_bh(&vsock_table_lock); 292d5afa82cSSunil Muthuswamy if (__vsock_in_connected_table(vsk)) 293d021c344SAndy King __vsock_remove_connected(vsk); 294d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 295d021c344SAndy King } 296d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_remove_connected); 297d021c344SAndy King 298d021c344SAndy King struct sock *vsock_find_bound_socket(struct sockaddr_vm *addr) 299d021c344SAndy King { 300d021c344SAndy King struct sock *sk; 301d021c344SAndy King 302d021c344SAndy King spin_lock_bh(&vsock_table_lock); 303d021c344SAndy King sk = __vsock_find_bound_socket(addr); 304d021c344SAndy King if (sk) 305d021c344SAndy King sock_hold(sk); 306d021c344SAndy King 307d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 308d021c344SAndy King 309d021c344SAndy King return sk; 310d021c344SAndy King } 311d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_find_bound_socket); 312d021c344SAndy King 313d021c344SAndy King struct sock *vsock_find_connected_socket(struct sockaddr_vm *src, 314d021c344SAndy King struct sockaddr_vm *dst) 315d021c344SAndy King { 316d021c344SAndy King struct sock *sk; 317d021c344SAndy King 318d021c344SAndy King spin_lock_bh(&vsock_table_lock); 319d021c344SAndy King sk = __vsock_find_connected_socket(src, dst); 320d021c344SAndy King if (sk) 321d021c344SAndy King sock_hold(sk); 322d021c344SAndy King 323d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 324d021c344SAndy King 325d021c344SAndy King return sk; 326d021c344SAndy King } 327d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_find_connected_socket); 328d021c344SAndy King 3296773b7dcSStefan Hajnoczi void vsock_remove_sock(struct vsock_sock *vsk) 3306773b7dcSStefan Hajnoczi { 3316773b7dcSStefan Hajnoczi vsock_remove_bound(vsk); 3326773b7dcSStefan Hajnoczi vsock_remove_connected(vsk); 3336773b7dcSStefan Hajnoczi } 3346773b7dcSStefan Hajnoczi EXPORT_SYMBOL_GPL(vsock_remove_sock); 3356773b7dcSStefan Hajnoczi 336d021c344SAndy King void vsock_for_each_connected_socket(void (*fn)(struct sock *sk)) 337d021c344SAndy King { 338d021c344SAndy King int i; 339d021c344SAndy King 340d021c344SAndy King spin_lock_bh(&vsock_table_lock); 341d021c344SAndy King 342d021c344SAndy King for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) { 343d021c344SAndy King struct vsock_sock *vsk; 344d021c344SAndy King list_for_each_entry(vsk, &vsock_connected_table[i], 345d9af2d67SJulia Lawall connected_table) 346d021c344SAndy King fn(sk_vsock(vsk)); 347d021c344SAndy King } 348d021c344SAndy King 349d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 350d021c344SAndy King } 351d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_for_each_connected_socket); 352d021c344SAndy King 353d021c344SAndy King void vsock_add_pending(struct sock *listener, struct sock *pending) 354d021c344SAndy King { 355d021c344SAndy King struct vsock_sock *vlistener; 356d021c344SAndy King struct vsock_sock *vpending; 357d021c344SAndy King 358d021c344SAndy King vlistener = vsock_sk(listener); 359d021c344SAndy King vpending = vsock_sk(pending); 360d021c344SAndy King 361d021c344SAndy King sock_hold(pending); 362d021c344SAndy King sock_hold(listener); 363d021c344SAndy King list_add_tail(&vpending->pending_links, &vlistener->pending_links); 364d021c344SAndy King } 365d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_add_pending); 366d021c344SAndy King 367d021c344SAndy King void vsock_remove_pending(struct sock *listener, struct sock *pending) 368d021c344SAndy King { 369d021c344SAndy King struct vsock_sock *vpending = vsock_sk(pending); 370d021c344SAndy King 371d021c344SAndy King list_del_init(&vpending->pending_links); 372d021c344SAndy King sock_put(listener); 373d021c344SAndy King sock_put(pending); 374d021c344SAndy King } 375d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_remove_pending); 376d021c344SAndy King 377d021c344SAndy King void vsock_enqueue_accept(struct sock *listener, struct sock *connected) 378d021c344SAndy King { 379d021c344SAndy King struct vsock_sock *vlistener; 380d021c344SAndy King struct vsock_sock *vconnected; 381d021c344SAndy King 382d021c344SAndy King vlistener = vsock_sk(listener); 383d021c344SAndy King vconnected = vsock_sk(connected); 384d021c344SAndy King 385d021c344SAndy King sock_hold(connected); 386d021c344SAndy King sock_hold(listener); 387d021c344SAndy King list_add_tail(&vconnected->accept_queue, &vlistener->accept_queue); 388d021c344SAndy King } 389d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_enqueue_accept); 390d021c344SAndy King 391408624afSStefano Garzarella static bool vsock_use_local_transport(unsigned int remote_cid) 392408624afSStefano Garzarella { 393408624afSStefano Garzarella if (!transport_local) 394408624afSStefano Garzarella return false; 395408624afSStefano Garzarella 396408624afSStefano Garzarella if (remote_cid == VMADDR_CID_LOCAL) 397408624afSStefano Garzarella return true; 398408624afSStefano Garzarella 399408624afSStefano Garzarella if (transport_g2h) { 400408624afSStefano Garzarella return remote_cid == transport_g2h->get_local_cid(); 401408624afSStefano Garzarella } else { 402408624afSStefano Garzarella return remote_cid == VMADDR_CID_HOST; 403408624afSStefano Garzarella } 404408624afSStefano Garzarella } 405408624afSStefano Garzarella 4066a2c0962SStefano Garzarella static void vsock_deassign_transport(struct vsock_sock *vsk) 4076a2c0962SStefano Garzarella { 4086a2c0962SStefano Garzarella if (!vsk->transport) 4096a2c0962SStefano Garzarella return; 4106a2c0962SStefano Garzarella 4116a2c0962SStefano Garzarella vsk->transport->destruct(vsk); 4126a2c0962SStefano Garzarella module_put(vsk->transport->module); 4136a2c0962SStefano Garzarella vsk->transport = NULL; 4146a2c0962SStefano Garzarella } 4156a2c0962SStefano Garzarella 416c0cfa2d8SStefano Garzarella /* Assign a transport to a socket and call the .init transport callback. 417c0cfa2d8SStefano Garzarella * 418c0cfa2d8SStefano Garzarella * Note: for stream socket this must be called when vsk->remote_addr is set 419c0cfa2d8SStefano Garzarella * (e.g. during the connect() or when a connection request on a listener 420c0cfa2d8SStefano Garzarella * socket is received). 421c0cfa2d8SStefano Garzarella * The vsk->remote_addr is used to decide which transport to use: 422408624afSStefano Garzarella * - remote CID == VMADDR_CID_LOCAL or g2h->local_cid or VMADDR_CID_HOST if 423408624afSStefano Garzarella * g2h is not loaded, will use local transport; 4247f816984SAndra Paraschiv * - remote CID <= VMADDR_CID_HOST or h2g is not loaded or remote flags field 4257f816984SAndra Paraschiv * includes VMADDR_FLAG_TO_HOST flag value, will use guest->host transport; 426c0cfa2d8SStefano Garzarella * - remote CID > VMADDR_CID_HOST will use host->guest transport; 427c0cfa2d8SStefano Garzarella */ 428c0cfa2d8SStefano Garzarella int vsock_assign_transport(struct vsock_sock *vsk, struct vsock_sock *psk) 429c0cfa2d8SStefano Garzarella { 430c0cfa2d8SStefano Garzarella const struct vsock_transport *new_transport; 431c0cfa2d8SStefano Garzarella struct sock *sk = sk_vsock(vsk); 432c0cfa2d8SStefano Garzarella unsigned int remote_cid = vsk->remote_addr.svm_cid; 4337f816984SAndra Paraschiv __u8 remote_flags; 434039fcccaSStefano Garzarella int ret; 435c0cfa2d8SStefano Garzarella 4361b5f2ab9SAndra Paraschiv /* If the packet is coming with the source and destination CIDs higher 4371b5f2ab9SAndra Paraschiv * than VMADDR_CID_HOST, then a vsock channel where all the packets are 4381b5f2ab9SAndra Paraschiv * forwarded to the host should be established. Then the host will 4391b5f2ab9SAndra Paraschiv * need to forward the packets to the guest. 4401b5f2ab9SAndra Paraschiv * 4411b5f2ab9SAndra Paraschiv * The flag is set on the (listen) receive path (psk is not NULL). On 4421b5f2ab9SAndra Paraschiv * the connect path the flag can be set by the user space application. 4431b5f2ab9SAndra Paraschiv */ 4441b5f2ab9SAndra Paraschiv if (psk && vsk->local_addr.svm_cid > VMADDR_CID_HOST && 4451b5f2ab9SAndra Paraschiv vsk->remote_addr.svm_cid > VMADDR_CID_HOST) 4461b5f2ab9SAndra Paraschiv vsk->remote_addr.svm_flags |= VMADDR_FLAG_TO_HOST; 4471b5f2ab9SAndra Paraschiv 4487f816984SAndra Paraschiv remote_flags = vsk->remote_addr.svm_flags; 4497f816984SAndra Paraschiv 450c0cfa2d8SStefano Garzarella switch (sk->sk_type) { 451c0cfa2d8SStefano Garzarella case SOCK_DGRAM: 452c0cfa2d8SStefano Garzarella new_transport = transport_dgram; 453c0cfa2d8SStefano Garzarella break; 454c0cfa2d8SStefano Garzarella case SOCK_STREAM: 455408624afSStefano Garzarella if (vsock_use_local_transport(remote_cid)) 456408624afSStefano Garzarella new_transport = transport_local; 4577f816984SAndra Paraschiv else if (remote_cid <= VMADDR_CID_HOST || !transport_h2g || 4587f816984SAndra Paraschiv (remote_flags & VMADDR_FLAG_TO_HOST)) 459c0cfa2d8SStefano Garzarella new_transport = transport_g2h; 460c0cfa2d8SStefano Garzarella else 461c0cfa2d8SStefano Garzarella new_transport = transport_h2g; 462c0cfa2d8SStefano Garzarella break; 463c0cfa2d8SStefano Garzarella default: 464c0cfa2d8SStefano Garzarella return -ESOCKTNOSUPPORT; 465c0cfa2d8SStefano Garzarella } 466c0cfa2d8SStefano Garzarella 467c0cfa2d8SStefano Garzarella if (vsk->transport) { 468c0cfa2d8SStefano Garzarella if (vsk->transport == new_transport) 469c0cfa2d8SStefano Garzarella return 0; 470c0cfa2d8SStefano Garzarella 4713f74957fSStefano Garzarella /* transport->release() must be called with sock lock acquired. 4723f74957fSStefano Garzarella * This path can only be taken during vsock_stream_connect(), 4733f74957fSStefano Garzarella * where we have already held the sock lock. 4743f74957fSStefano Garzarella * In the other cases, this function is called on a new socket 4753f74957fSStefano Garzarella * which is not assigned to any transport. 4763f74957fSStefano Garzarella */ 477c0cfa2d8SStefano Garzarella vsk->transport->release(vsk); 4786a2c0962SStefano Garzarella vsock_deassign_transport(vsk); 479c0cfa2d8SStefano Garzarella } 480c0cfa2d8SStefano Garzarella 4816a2c0962SStefano Garzarella /* We increase the module refcnt to prevent the transport unloading 4826a2c0962SStefano Garzarella * while there are open sockets assigned to it. 4836a2c0962SStefano Garzarella */ 4846a2c0962SStefano Garzarella if (!new_transport || !try_module_get(new_transport->module)) 485c0cfa2d8SStefano Garzarella return -ENODEV; 486c0cfa2d8SStefano Garzarella 487039fcccaSStefano Garzarella ret = new_transport->init(vsk, psk); 488039fcccaSStefano Garzarella if (ret) { 489039fcccaSStefano Garzarella module_put(new_transport->module); 490039fcccaSStefano Garzarella return ret; 491039fcccaSStefano Garzarella } 492039fcccaSStefano Garzarella 493c0cfa2d8SStefano Garzarella vsk->transport = new_transport; 494c0cfa2d8SStefano Garzarella 495039fcccaSStefano Garzarella return 0; 496c0cfa2d8SStefano Garzarella } 497c0cfa2d8SStefano Garzarella EXPORT_SYMBOL_GPL(vsock_assign_transport); 498c0cfa2d8SStefano Garzarella 499c0cfa2d8SStefano Garzarella bool vsock_find_cid(unsigned int cid) 500c0cfa2d8SStefano Garzarella { 501c0cfa2d8SStefano Garzarella if (transport_g2h && cid == transport_g2h->get_local_cid()) 502c0cfa2d8SStefano Garzarella return true; 503c0cfa2d8SStefano Garzarella 504c0cfa2d8SStefano Garzarella if (transport_h2g && cid == VMADDR_CID_HOST) 505c0cfa2d8SStefano Garzarella return true; 506c0cfa2d8SStefano Garzarella 507408624afSStefano Garzarella if (transport_local && cid == VMADDR_CID_LOCAL) 508408624afSStefano Garzarella return true; 509408624afSStefano Garzarella 510c0cfa2d8SStefano Garzarella return false; 511c0cfa2d8SStefano Garzarella } 512c0cfa2d8SStefano Garzarella EXPORT_SYMBOL_GPL(vsock_find_cid); 513c0cfa2d8SStefano Garzarella 514d021c344SAndy King static struct sock *vsock_dequeue_accept(struct sock *listener) 515d021c344SAndy King { 516d021c344SAndy King struct vsock_sock *vlistener; 517d021c344SAndy King struct vsock_sock *vconnected; 518d021c344SAndy King 519d021c344SAndy King vlistener = vsock_sk(listener); 520d021c344SAndy King 521d021c344SAndy King if (list_empty(&vlistener->accept_queue)) 522d021c344SAndy King return NULL; 523d021c344SAndy King 524d021c344SAndy King vconnected = list_entry(vlistener->accept_queue.next, 525d021c344SAndy King struct vsock_sock, accept_queue); 526d021c344SAndy King 527d021c344SAndy King list_del_init(&vconnected->accept_queue); 528d021c344SAndy King sock_put(listener); 529d021c344SAndy King /* The caller will need a reference on the connected socket so we let 530d021c344SAndy King * it call sock_put(). 531d021c344SAndy King */ 532d021c344SAndy King 533d021c344SAndy King return sk_vsock(vconnected); 534d021c344SAndy King } 535d021c344SAndy King 536d021c344SAndy King static bool vsock_is_accept_queue_empty(struct sock *sk) 537d021c344SAndy King { 538d021c344SAndy King struct vsock_sock *vsk = vsock_sk(sk); 539d021c344SAndy King return list_empty(&vsk->accept_queue); 540d021c344SAndy King } 541d021c344SAndy King 542d021c344SAndy King static bool vsock_is_pending(struct sock *sk) 543d021c344SAndy King { 544d021c344SAndy King struct vsock_sock *vsk = vsock_sk(sk); 545d021c344SAndy King return !list_empty(&vsk->pending_links); 546d021c344SAndy King } 547d021c344SAndy King 548d021c344SAndy King static int vsock_send_shutdown(struct sock *sk, int mode) 549d021c344SAndy King { 550fe502c4aSStefano Garzarella struct vsock_sock *vsk = vsock_sk(sk); 551fe502c4aSStefano Garzarella 552c0cfa2d8SStefano Garzarella if (!vsk->transport) 553c0cfa2d8SStefano Garzarella return -ENODEV; 554c0cfa2d8SStefano Garzarella 555fe502c4aSStefano Garzarella return vsk->transport->shutdown(vsk, mode); 556d021c344SAndy King } 557d021c344SAndy King 558455f05ecSCong Wang static void vsock_pending_work(struct work_struct *work) 559d021c344SAndy King { 560d021c344SAndy King struct sock *sk; 561d021c344SAndy King struct sock *listener; 562d021c344SAndy King struct vsock_sock *vsk; 563d021c344SAndy King bool cleanup; 564d021c344SAndy King 565455f05ecSCong Wang vsk = container_of(work, struct vsock_sock, pending_work.work); 566d021c344SAndy King sk = sk_vsock(vsk); 567d021c344SAndy King listener = vsk->listener; 568d021c344SAndy King cleanup = true; 569d021c344SAndy King 570d021c344SAndy King lock_sock(listener); 5714192f672SStefan Hajnoczi lock_sock_nested(sk, SINGLE_DEPTH_NESTING); 572d021c344SAndy King 573d021c344SAndy King if (vsock_is_pending(sk)) { 574d021c344SAndy King vsock_remove_pending(listener, sk); 5751190cfdbSJorgen Hansen 5767976a11bSEric Dumazet sk_acceptq_removed(listener); 577d021c344SAndy King } else if (!vsk->rejected) { 578d021c344SAndy King /* We are not on the pending list and accept() did not reject 579d021c344SAndy King * us, so we must have been accepted by our user process. We 580d021c344SAndy King * just need to drop our references to the sockets and be on 581d021c344SAndy King * our way. 582d021c344SAndy King */ 583d021c344SAndy King cleanup = false; 584d021c344SAndy King goto out; 585d021c344SAndy King } 586d021c344SAndy King 587d021c344SAndy King /* We need to remove ourself from the global connected sockets list so 588d021c344SAndy King * incoming packets can't find this socket, and to reduce the reference 589d021c344SAndy King * count. 590d021c344SAndy King */ 591d021c344SAndy King vsock_remove_connected(vsk); 592d021c344SAndy King 5933b4477d2SStefan Hajnoczi sk->sk_state = TCP_CLOSE; 594d021c344SAndy King 595d021c344SAndy King out: 596d021c344SAndy King release_sock(sk); 597d021c344SAndy King release_sock(listener); 598d021c344SAndy King if (cleanup) 599d021c344SAndy King sock_put(sk); 600d021c344SAndy King 601d021c344SAndy King sock_put(sk); 602d021c344SAndy King sock_put(listener); 603d021c344SAndy King } 604d021c344SAndy King 605d021c344SAndy King /**** SOCKET OPERATIONS ****/ 606d021c344SAndy King 607a9e29e55SArseny Krasnov static int __vsock_bind_connectible(struct vsock_sock *vsk, 608d021c344SAndy King struct sockaddr_vm *addr) 609d021c344SAndy King { 610a22d3251SLepton Wu static u32 port; 611d021c344SAndy King struct sockaddr_vm new_addr; 612d021c344SAndy King 6138236b08cSLepton Wu if (!port) 6148236b08cSLepton Wu port = LAST_RESERVED_PORT + 1 + 6158236b08cSLepton Wu prandom_u32_max(U32_MAX - LAST_RESERVED_PORT); 6168236b08cSLepton Wu 617d021c344SAndy King vsock_addr_init(&new_addr, addr->svm_cid, addr->svm_port); 618d021c344SAndy King 619d021c344SAndy King if (addr->svm_port == VMADDR_PORT_ANY) { 620d021c344SAndy King bool found = false; 621d021c344SAndy King unsigned int i; 622d021c344SAndy King 623d021c344SAndy King for (i = 0; i < MAX_PORT_RETRIES; i++) { 624d021c344SAndy King if (port <= LAST_RESERVED_PORT) 625d021c344SAndy King port = LAST_RESERVED_PORT + 1; 626d021c344SAndy King 627d021c344SAndy King new_addr.svm_port = port++; 628d021c344SAndy King 629d021c344SAndy King if (!__vsock_find_bound_socket(&new_addr)) { 630d021c344SAndy King found = true; 631d021c344SAndy King break; 632d021c344SAndy King } 633d021c344SAndy King } 634d021c344SAndy King 635d021c344SAndy King if (!found) 636d021c344SAndy King return -EADDRNOTAVAIL; 637d021c344SAndy King } else { 638d021c344SAndy King /* If port is in reserved range, ensure caller 639d021c344SAndy King * has necessary privileges. 640d021c344SAndy King */ 641d021c344SAndy King if (addr->svm_port <= LAST_RESERVED_PORT && 642d021c344SAndy King !capable(CAP_NET_BIND_SERVICE)) { 643d021c344SAndy King return -EACCES; 644d021c344SAndy King } 645d021c344SAndy King 646d021c344SAndy King if (__vsock_find_bound_socket(&new_addr)) 647d021c344SAndy King return -EADDRINUSE; 648d021c344SAndy King } 649d021c344SAndy King 650d021c344SAndy King vsock_addr_init(&vsk->local_addr, new_addr.svm_cid, new_addr.svm_port); 651d021c344SAndy King 652d021c344SAndy King /* Remove stream sockets from the unbound list and add them to the hash 653d021c344SAndy King * table for easy lookup by its address. The unbound list is simply an 654d021c344SAndy King * extra entry at the end of the hash table, a trick used by AF_UNIX. 655d021c344SAndy King */ 656d021c344SAndy King __vsock_remove_bound(vsk); 657d021c344SAndy King __vsock_insert_bound(vsock_bound_sockets(&vsk->local_addr), vsk); 658d021c344SAndy King 659d021c344SAndy King return 0; 660d021c344SAndy King } 661d021c344SAndy King 662d021c344SAndy King static int __vsock_bind_dgram(struct vsock_sock *vsk, 663d021c344SAndy King struct sockaddr_vm *addr) 664d021c344SAndy King { 665fe502c4aSStefano Garzarella return vsk->transport->dgram_bind(vsk, addr); 666d021c344SAndy King } 667d021c344SAndy King 668d021c344SAndy King static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr) 669d021c344SAndy King { 670d021c344SAndy King struct vsock_sock *vsk = vsock_sk(sk); 671d021c344SAndy King int retval; 672d021c344SAndy King 673d021c344SAndy King /* First ensure this socket isn't already bound. */ 674d021c344SAndy King if (vsock_addr_bound(&vsk->local_addr)) 675d021c344SAndy King return -EINVAL; 676d021c344SAndy King 677d021c344SAndy King /* Now bind to the provided address or select appropriate values if 678d021c344SAndy King * none are provided (VMADDR_CID_ANY and VMADDR_PORT_ANY). Note that 679d021c344SAndy King * like AF_INET prevents binding to a non-local IP address (in most 680c0cfa2d8SStefano Garzarella * cases), we only allow binding to a local CID. 681d021c344SAndy King */ 682c0cfa2d8SStefano Garzarella if (addr->svm_cid != VMADDR_CID_ANY && !vsock_find_cid(addr->svm_cid)) 683d021c344SAndy King return -EADDRNOTAVAIL; 684d021c344SAndy King 685d021c344SAndy King switch (sk->sk_socket->type) { 686d021c344SAndy King case SOCK_STREAM: 687d021c344SAndy King spin_lock_bh(&vsock_table_lock); 688a9e29e55SArseny Krasnov retval = __vsock_bind_connectible(vsk, addr); 689d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 690d021c344SAndy King break; 691d021c344SAndy King 692d021c344SAndy King case SOCK_DGRAM: 693d021c344SAndy King retval = __vsock_bind_dgram(vsk, addr); 694d021c344SAndy King break; 695d021c344SAndy King 696d021c344SAndy King default: 697d021c344SAndy King retval = -EINVAL; 698d021c344SAndy King break; 699d021c344SAndy King } 700d021c344SAndy King 701d021c344SAndy King return retval; 702d021c344SAndy King } 703d021c344SAndy King 704455f05ecSCong Wang static void vsock_connect_timeout(struct work_struct *work); 705455f05ecSCong Wang 706b9ca2f5fSStefano Garzarella static struct sock *__vsock_create(struct net *net, 707d021c344SAndy King struct socket *sock, 708d021c344SAndy King struct sock *parent, 709d021c344SAndy King gfp_t priority, 71011aa9c28SEric W. Biederman unsigned short type, 71111aa9c28SEric W. Biederman int kern) 712d021c344SAndy King { 713d021c344SAndy King struct sock *sk; 714d021c344SAndy King struct vsock_sock *psk; 715d021c344SAndy King struct vsock_sock *vsk; 716d021c344SAndy King 71711aa9c28SEric W. Biederman sk = sk_alloc(net, AF_VSOCK, priority, &vsock_proto, kern); 718d021c344SAndy King if (!sk) 719d021c344SAndy King return NULL; 720d021c344SAndy King 721d021c344SAndy King sock_init_data(sock, sk); 722d021c344SAndy King 723d021c344SAndy King /* sk->sk_type is normally set in sock_init_data, but only if sock is 724d021c344SAndy King * non-NULL. We make sure that our sockets always have a type by 725d021c344SAndy King * setting it here if needed. 726d021c344SAndy King */ 727d021c344SAndy King if (!sock) 728d021c344SAndy King sk->sk_type = type; 729d021c344SAndy King 730d021c344SAndy King vsk = vsock_sk(sk); 731d021c344SAndy King vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 732d021c344SAndy King vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 733d021c344SAndy King 734d021c344SAndy King sk->sk_destruct = vsock_sk_destruct; 735d021c344SAndy King sk->sk_backlog_rcv = vsock_queue_rcv_skb; 736d021c344SAndy King sock_reset_flag(sk, SOCK_DONE); 737d021c344SAndy King 738d021c344SAndy King INIT_LIST_HEAD(&vsk->bound_table); 739d021c344SAndy King INIT_LIST_HEAD(&vsk->connected_table); 740d021c344SAndy King vsk->listener = NULL; 741d021c344SAndy King INIT_LIST_HEAD(&vsk->pending_links); 742d021c344SAndy King INIT_LIST_HEAD(&vsk->accept_queue); 743d021c344SAndy King vsk->rejected = false; 744d021c344SAndy King vsk->sent_request = false; 745d021c344SAndy King vsk->ignore_connecting_rst = false; 746d021c344SAndy King vsk->peer_shutdown = 0; 747455f05ecSCong Wang INIT_DELAYED_WORK(&vsk->connect_work, vsock_connect_timeout); 748455f05ecSCong Wang INIT_DELAYED_WORK(&vsk->pending_work, vsock_pending_work); 749d021c344SAndy King 750d021c344SAndy King psk = parent ? vsock_sk(parent) : NULL; 751d021c344SAndy King if (parent) { 752d021c344SAndy King vsk->trusted = psk->trusted; 753d021c344SAndy King vsk->owner = get_cred(psk->owner); 754d021c344SAndy King vsk->connect_timeout = psk->connect_timeout; 755b9f2b0ffSStefano Garzarella vsk->buffer_size = psk->buffer_size; 756b9f2b0ffSStefano Garzarella vsk->buffer_min_size = psk->buffer_min_size; 757b9f2b0ffSStefano Garzarella vsk->buffer_max_size = psk->buffer_max_size; 7581f935e8eSDavid Brazdil security_sk_clone(parent, sk); 759d021c344SAndy King } else { 760af545bb5SJeff Vander Stoep vsk->trusted = ns_capable_noaudit(&init_user_ns, CAP_NET_ADMIN); 761d021c344SAndy King vsk->owner = get_current_cred(); 762d021c344SAndy King vsk->connect_timeout = VSOCK_DEFAULT_CONNECT_TIMEOUT; 763b9f2b0ffSStefano Garzarella vsk->buffer_size = VSOCK_DEFAULT_BUFFER_SIZE; 764b9f2b0ffSStefano Garzarella vsk->buffer_min_size = VSOCK_DEFAULT_BUFFER_MIN_SIZE; 765b9f2b0ffSStefano Garzarella vsk->buffer_max_size = VSOCK_DEFAULT_BUFFER_MAX_SIZE; 766d021c344SAndy King } 767d021c344SAndy King 768d021c344SAndy King return sk; 769d021c344SAndy King } 770d021c344SAndy King 771a9e29e55SArseny Krasnov static bool sock_type_connectible(u16 type) 772a9e29e55SArseny Krasnov { 773a9e29e55SArseny Krasnov return type == SOCK_STREAM; 774a9e29e55SArseny Krasnov } 775a9e29e55SArseny Krasnov 7760d9138ffSDexuan Cui static void __vsock_release(struct sock *sk, int level) 777d021c344SAndy King { 778d021c344SAndy King if (sk) { 779d021c344SAndy King struct sock *pending; 780d021c344SAndy King struct vsock_sock *vsk; 781d021c344SAndy King 782d021c344SAndy King vsk = vsock_sk(sk); 783d021c344SAndy King pending = NULL; /* Compiler warning. */ 784d021c344SAndy King 7850d9138ffSDexuan Cui /* When "level" is SINGLE_DEPTH_NESTING, use the nested 7860d9138ffSDexuan Cui * version to avoid the warning "possible recursive locking 7870d9138ffSDexuan Cui * detected". When "level" is 0, lock_sock_nested(sk, level) 7880d9138ffSDexuan Cui * is the same as lock_sock(sk). 7890d9138ffSDexuan Cui */ 7900d9138ffSDexuan Cui lock_sock_nested(sk, level); 7913f74957fSStefano Garzarella 7923f74957fSStefano Garzarella if (vsk->transport) 7933f74957fSStefano Garzarella vsk->transport->release(vsk); 794a9e29e55SArseny Krasnov else if (sock_type_connectible(sk->sk_type)) 7953f74957fSStefano Garzarella vsock_remove_sock(vsk); 7963f74957fSStefano Garzarella 797d021c344SAndy King sock_orphan(sk); 798d021c344SAndy King sk->sk_shutdown = SHUTDOWN_MASK; 799d021c344SAndy King 8003b7ad08bSChristophe JAILLET skb_queue_purge(&sk->sk_receive_queue); 801d021c344SAndy King 802d021c344SAndy King /* Clean up any sockets that never were accepted. */ 803d021c344SAndy King while ((pending = vsock_dequeue_accept(sk)) != NULL) { 8040d9138ffSDexuan Cui __vsock_release(pending, SINGLE_DEPTH_NESTING); 805d021c344SAndy King sock_put(pending); 806d021c344SAndy King } 807d021c344SAndy King 808d021c344SAndy King release_sock(sk); 809d021c344SAndy King sock_put(sk); 810d021c344SAndy King } 811d021c344SAndy King } 812d021c344SAndy King 813d021c344SAndy King static void vsock_sk_destruct(struct sock *sk) 814d021c344SAndy King { 815d021c344SAndy King struct vsock_sock *vsk = vsock_sk(sk); 816d021c344SAndy King 8176a2c0962SStefano Garzarella vsock_deassign_transport(vsk); 818d021c344SAndy King 819d021c344SAndy King /* When clearing these addresses, there's no need to set the family and 820d021c344SAndy King * possibly register the address family with the kernel. 821d021c344SAndy King */ 822d021c344SAndy King vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 823d021c344SAndy King vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 824d021c344SAndy King 825d021c344SAndy King put_cred(vsk->owner); 826d021c344SAndy King } 827d021c344SAndy King 828d021c344SAndy King static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) 829d021c344SAndy King { 830d021c344SAndy King int err; 831d021c344SAndy King 832d021c344SAndy King err = sock_queue_rcv_skb(sk, skb); 833d021c344SAndy King if (err) 834d021c344SAndy King kfree_skb(skb); 835d021c344SAndy King 836d021c344SAndy King return err; 837d021c344SAndy King } 838d021c344SAndy King 839b9ca2f5fSStefano Garzarella struct sock *vsock_create_connected(struct sock *parent) 840b9ca2f5fSStefano Garzarella { 841b9ca2f5fSStefano Garzarella return __vsock_create(sock_net(parent), NULL, parent, GFP_KERNEL, 842b9ca2f5fSStefano Garzarella parent->sk_type, 0); 843b9ca2f5fSStefano Garzarella } 844b9ca2f5fSStefano Garzarella EXPORT_SYMBOL_GPL(vsock_create_connected); 845b9ca2f5fSStefano Garzarella 846d021c344SAndy King s64 vsock_stream_has_data(struct vsock_sock *vsk) 847d021c344SAndy King { 848fe502c4aSStefano Garzarella return vsk->transport->stream_has_data(vsk); 849d021c344SAndy King } 850d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_stream_has_data); 851d021c344SAndy King 852d021c344SAndy King s64 vsock_stream_has_space(struct vsock_sock *vsk) 853d021c344SAndy King { 854fe502c4aSStefano Garzarella return vsk->transport->stream_has_space(vsk); 855d021c344SAndy King } 856d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_stream_has_space); 857d021c344SAndy King 858d021c344SAndy King static int vsock_release(struct socket *sock) 859d021c344SAndy King { 8600d9138ffSDexuan Cui __vsock_release(sock->sk, 0); 861d021c344SAndy King sock->sk = NULL; 862d021c344SAndy King sock->state = SS_FREE; 863d021c344SAndy King 864d021c344SAndy King return 0; 865d021c344SAndy King } 866d021c344SAndy King 867d021c344SAndy King static int 868d021c344SAndy King vsock_bind(struct socket *sock, struct sockaddr *addr, int addr_len) 869d021c344SAndy King { 870d021c344SAndy King int err; 871d021c344SAndy King struct sock *sk; 872d021c344SAndy King struct sockaddr_vm *vm_addr; 873d021c344SAndy King 874d021c344SAndy King sk = sock->sk; 875d021c344SAndy King 876d021c344SAndy King if (vsock_addr_cast(addr, addr_len, &vm_addr) != 0) 877d021c344SAndy King return -EINVAL; 878d021c344SAndy King 879d021c344SAndy King lock_sock(sk); 880d021c344SAndy King err = __vsock_bind(sk, vm_addr); 881d021c344SAndy King release_sock(sk); 882d021c344SAndy King 883d021c344SAndy King return err; 884d021c344SAndy King } 885d021c344SAndy King 886d021c344SAndy King static int vsock_getname(struct socket *sock, 8879b2c45d4SDenys Vlasenko struct sockaddr *addr, int peer) 888d021c344SAndy King { 889d021c344SAndy King int err; 890d021c344SAndy King struct sock *sk; 891d021c344SAndy King struct vsock_sock *vsk; 892d021c344SAndy King struct sockaddr_vm *vm_addr; 893d021c344SAndy King 894d021c344SAndy King sk = sock->sk; 895d021c344SAndy King vsk = vsock_sk(sk); 896d021c344SAndy King err = 0; 897d021c344SAndy King 898d021c344SAndy King lock_sock(sk); 899d021c344SAndy King 900d021c344SAndy King if (peer) { 901d021c344SAndy King if (sock->state != SS_CONNECTED) { 902d021c344SAndy King err = -ENOTCONN; 903d021c344SAndy King goto out; 904d021c344SAndy King } 905d021c344SAndy King vm_addr = &vsk->remote_addr; 906d021c344SAndy King } else { 907d021c344SAndy King vm_addr = &vsk->local_addr; 908d021c344SAndy King } 909d021c344SAndy King 910d021c344SAndy King if (!vm_addr) { 911d021c344SAndy King err = -EINVAL; 912d021c344SAndy King goto out; 913d021c344SAndy King } 914d021c344SAndy King 915d021c344SAndy King /* sys_getsockname() and sys_getpeername() pass us a 916d021c344SAndy King * MAX_SOCK_ADDR-sized buffer and don't set addr_len. Unfortunately 917d021c344SAndy King * that macro is defined in socket.c instead of .h, so we hardcode its 918d021c344SAndy King * value here. 919d021c344SAndy King */ 920d021c344SAndy King BUILD_BUG_ON(sizeof(*vm_addr) > 128); 921d021c344SAndy King memcpy(addr, vm_addr, sizeof(*vm_addr)); 9229b2c45d4SDenys Vlasenko err = sizeof(*vm_addr); 923d021c344SAndy King 924d021c344SAndy King out: 925d021c344SAndy King release_sock(sk); 926d021c344SAndy King return err; 927d021c344SAndy King } 928d021c344SAndy King 929d021c344SAndy King static int vsock_shutdown(struct socket *sock, int mode) 930d021c344SAndy King { 931d021c344SAndy King int err; 932d021c344SAndy King struct sock *sk; 933d021c344SAndy King 934d021c344SAndy King /* User level uses SHUT_RD (0) and SHUT_WR (1), but the kernel uses 935d021c344SAndy King * RCV_SHUTDOWN (1) and SEND_SHUTDOWN (2), so we must increment mode 936d021c344SAndy King * here like the other address families do. Note also that the 937d021c344SAndy King * increment makes SHUT_RDWR (2) into RCV_SHUTDOWN | SEND_SHUTDOWN (3), 938d021c344SAndy King * which is what we want. 939d021c344SAndy King */ 940d021c344SAndy King mode++; 941d021c344SAndy King 942d021c344SAndy King if ((mode & ~SHUTDOWN_MASK) || !mode) 943d021c344SAndy King return -EINVAL; 944d021c344SAndy King 945d021c344SAndy King /* If this is a STREAM socket and it is not connected then bail out 946d021c344SAndy King * immediately. If it is a DGRAM socket then we must first kick the 947d021c344SAndy King * socket so that it wakes up from any sleeping calls, for example 948d021c344SAndy King * recv(), and then afterwards return the error. 949d021c344SAndy King */ 950d021c344SAndy King 951d021c344SAndy King sk = sock->sk; 9521c5fae9cSStefano Garzarella 9531c5fae9cSStefano Garzarella lock_sock(sk); 954d021c344SAndy King if (sock->state == SS_UNCONNECTED) { 955d021c344SAndy King err = -ENOTCONN; 956a9e29e55SArseny Krasnov if (sock_type_connectible(sk->sk_type)) 9571c5fae9cSStefano Garzarella goto out; 958d021c344SAndy King } else { 959d021c344SAndy King sock->state = SS_DISCONNECTING; 960d021c344SAndy King err = 0; 961d021c344SAndy King } 962d021c344SAndy King 963d021c344SAndy King /* Receive and send shutdowns are treated alike. */ 964d021c344SAndy King mode = mode & (RCV_SHUTDOWN | SEND_SHUTDOWN); 965d021c344SAndy King if (mode) { 966d021c344SAndy King sk->sk_shutdown |= mode; 967d021c344SAndy King sk->sk_state_change(sk); 968d021c344SAndy King 969a9e29e55SArseny Krasnov if (sock_type_connectible(sk->sk_type)) { 970d021c344SAndy King sock_reset_flag(sk, SOCK_DONE); 971d021c344SAndy King vsock_send_shutdown(sk, mode); 972d021c344SAndy King } 973d021c344SAndy King } 974d021c344SAndy King 9751c5fae9cSStefano Garzarella out: 9761c5fae9cSStefano Garzarella release_sock(sk); 977d021c344SAndy King return err; 978d021c344SAndy King } 979d021c344SAndy King 980a11e1d43SLinus Torvalds static __poll_t vsock_poll(struct file *file, struct socket *sock, 981a11e1d43SLinus Torvalds poll_table *wait) 982d021c344SAndy King { 983a11e1d43SLinus Torvalds struct sock *sk; 984a11e1d43SLinus Torvalds __poll_t mask; 985a11e1d43SLinus Torvalds struct vsock_sock *vsk; 986a11e1d43SLinus Torvalds 987a11e1d43SLinus Torvalds sk = sock->sk; 988a11e1d43SLinus Torvalds vsk = vsock_sk(sk); 989a11e1d43SLinus Torvalds 990a11e1d43SLinus Torvalds poll_wait(file, sk_sleep(sk), wait); 991a11e1d43SLinus Torvalds mask = 0; 992d021c344SAndy King 993d021c344SAndy King if (sk->sk_err) 994d021c344SAndy King /* Signify that there has been an error on this socket. */ 995a9a08845SLinus Torvalds mask |= EPOLLERR; 996d021c344SAndy King 997d021c344SAndy King /* INET sockets treat local write shutdown and peer write shutdown as a 998a9a08845SLinus Torvalds * case of EPOLLHUP set. 999d021c344SAndy King */ 1000d021c344SAndy King if ((sk->sk_shutdown == SHUTDOWN_MASK) || 1001d021c344SAndy King ((sk->sk_shutdown & SEND_SHUTDOWN) && 1002d021c344SAndy King (vsk->peer_shutdown & SEND_SHUTDOWN))) { 1003a9a08845SLinus Torvalds mask |= EPOLLHUP; 1004d021c344SAndy King } 1005d021c344SAndy King 1006d021c344SAndy King if (sk->sk_shutdown & RCV_SHUTDOWN || 1007d021c344SAndy King vsk->peer_shutdown & SEND_SHUTDOWN) { 1008a9a08845SLinus Torvalds mask |= EPOLLRDHUP; 1009d021c344SAndy King } 1010d021c344SAndy King 1011d021c344SAndy King if (sock->type == SOCK_DGRAM) { 1012d021c344SAndy King /* For datagram sockets we can read if there is something in 1013d021c344SAndy King * the queue and write as long as the socket isn't shutdown for 1014d021c344SAndy King * sending. 1015d021c344SAndy King */ 10163ef7cf57SEric Dumazet if (!skb_queue_empty_lockless(&sk->sk_receive_queue) || 1017d021c344SAndy King (sk->sk_shutdown & RCV_SHUTDOWN)) { 1018a9a08845SLinus Torvalds mask |= EPOLLIN | EPOLLRDNORM; 1019d021c344SAndy King } 1020d021c344SAndy King 1021d021c344SAndy King if (!(sk->sk_shutdown & SEND_SHUTDOWN)) 1022a9a08845SLinus Torvalds mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND; 1023d021c344SAndy King 1024a9e29e55SArseny Krasnov } else if (sock_type_connectible(sk->sk_type)) { 1025c518adafSAlexander Popov const struct vsock_transport *transport; 1026c518adafSAlexander Popov 1027d021c344SAndy King lock_sock(sk); 1028d021c344SAndy King 1029c518adafSAlexander Popov transport = vsk->transport; 1030c518adafSAlexander Popov 1031d021c344SAndy King /* Listening sockets that have connections in their accept 1032d021c344SAndy King * queue can be read. 1033d021c344SAndy King */ 10343b4477d2SStefan Hajnoczi if (sk->sk_state == TCP_LISTEN 1035d021c344SAndy King && !vsock_is_accept_queue_empty(sk)) 1036a9a08845SLinus Torvalds mask |= EPOLLIN | EPOLLRDNORM; 1037d021c344SAndy King 1038d021c344SAndy King /* If there is something in the queue then we can read. */ 1039c0cfa2d8SStefano Garzarella if (transport && transport->stream_is_active(vsk) && 1040d021c344SAndy King !(sk->sk_shutdown & RCV_SHUTDOWN)) { 1041d021c344SAndy King bool data_ready_now = false; 1042d021c344SAndy King int ret = transport->notify_poll_in( 1043d021c344SAndy King vsk, 1, &data_ready_now); 1044d021c344SAndy King if (ret < 0) { 1045a9a08845SLinus Torvalds mask |= EPOLLERR; 1046d021c344SAndy King } else { 1047d021c344SAndy King if (data_ready_now) 1048a9a08845SLinus Torvalds mask |= EPOLLIN | EPOLLRDNORM; 1049d021c344SAndy King 1050d021c344SAndy King } 1051d021c344SAndy King } 1052d021c344SAndy King 1053d021c344SAndy King /* Sockets whose connections have been closed, reset, or 1054d021c344SAndy King * terminated should also be considered read, and we check the 1055d021c344SAndy King * shutdown flag for that. 1056d021c344SAndy King */ 1057d021c344SAndy King if (sk->sk_shutdown & RCV_SHUTDOWN || 1058d021c344SAndy King vsk->peer_shutdown & SEND_SHUTDOWN) { 1059a9a08845SLinus Torvalds mask |= EPOLLIN | EPOLLRDNORM; 1060d021c344SAndy King } 1061d021c344SAndy King 1062d021c344SAndy King /* Connected sockets that can produce data can be written. */ 10631980c058SStefano Garzarella if (transport && sk->sk_state == TCP_ESTABLISHED) { 1064d021c344SAndy King if (!(sk->sk_shutdown & SEND_SHUTDOWN)) { 1065d021c344SAndy King bool space_avail_now = false; 1066d021c344SAndy King int ret = transport->notify_poll_out( 1067d021c344SAndy King vsk, 1, &space_avail_now); 1068d021c344SAndy King if (ret < 0) { 1069a9a08845SLinus Torvalds mask |= EPOLLERR; 1070d021c344SAndy King } else { 1071d021c344SAndy King if (space_avail_now) 1072a9a08845SLinus Torvalds /* Remove EPOLLWRBAND since INET 1073d021c344SAndy King * sockets are not setting it. 1074d021c344SAndy King */ 1075a9a08845SLinus Torvalds mask |= EPOLLOUT | EPOLLWRNORM; 1076d021c344SAndy King 1077d021c344SAndy King } 1078d021c344SAndy King } 1079d021c344SAndy King } 1080d021c344SAndy King 1081d021c344SAndy King /* Simulate INET socket poll behaviors, which sets 1082a9a08845SLinus Torvalds * EPOLLOUT|EPOLLWRNORM when peer is closed and nothing to read, 1083d021c344SAndy King * but local send is not shutdown. 1084d021c344SAndy King */ 1085ba3169fcSStefan Hajnoczi if (sk->sk_state == TCP_CLOSE || sk->sk_state == TCP_CLOSING) { 1086d021c344SAndy King if (!(sk->sk_shutdown & SEND_SHUTDOWN)) 1087a9a08845SLinus Torvalds mask |= EPOLLOUT | EPOLLWRNORM; 1088d021c344SAndy King 1089d021c344SAndy King } 1090d021c344SAndy King 1091d021c344SAndy King release_sock(sk); 1092d021c344SAndy King } 1093d021c344SAndy King 1094d021c344SAndy King return mask; 1095d021c344SAndy King } 1096d021c344SAndy King 10971b784140SYing Xue static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg, 10981b784140SYing Xue size_t len) 1099d021c344SAndy King { 1100d021c344SAndy King int err; 1101d021c344SAndy King struct sock *sk; 1102d021c344SAndy King struct vsock_sock *vsk; 1103d021c344SAndy King struct sockaddr_vm *remote_addr; 1104fe502c4aSStefano Garzarella const struct vsock_transport *transport; 1105d021c344SAndy King 1106d021c344SAndy King if (msg->msg_flags & MSG_OOB) 1107d021c344SAndy King return -EOPNOTSUPP; 1108d021c344SAndy King 1109d021c344SAndy King /* For now, MSG_DONTWAIT is always assumed... */ 1110d021c344SAndy King err = 0; 1111d021c344SAndy King sk = sock->sk; 1112d021c344SAndy King vsk = vsock_sk(sk); 1113d021c344SAndy King 1114d021c344SAndy King lock_sock(sk); 1115d021c344SAndy King 1116c518adafSAlexander Popov transport = vsk->transport; 1117c518adafSAlexander Popov 1118b3a6dfe8SAsias He err = vsock_auto_bind(vsk); 1119b3a6dfe8SAsias He if (err) 1120d021c344SAndy King goto out; 1121d021c344SAndy King 1122d021c344SAndy King 1123d021c344SAndy King /* If the provided message contains an address, use that. Otherwise 1124d021c344SAndy King * fall back on the socket's remote handle (if it has been connected). 1125d021c344SAndy King */ 1126d021c344SAndy King if (msg->msg_name && 1127d021c344SAndy King vsock_addr_cast(msg->msg_name, msg->msg_namelen, 1128d021c344SAndy King &remote_addr) == 0) { 1129d021c344SAndy King /* Ensure this address is of the right type and is a valid 1130d021c344SAndy King * destination. 1131d021c344SAndy King */ 1132d021c344SAndy King 1133d021c344SAndy King if (remote_addr->svm_cid == VMADDR_CID_ANY) 1134d021c344SAndy King remote_addr->svm_cid = transport->get_local_cid(); 1135d021c344SAndy King 1136d021c344SAndy King if (!vsock_addr_bound(remote_addr)) { 1137d021c344SAndy King err = -EINVAL; 1138d021c344SAndy King goto out; 1139d021c344SAndy King } 1140d021c344SAndy King } else if (sock->state == SS_CONNECTED) { 1141d021c344SAndy King remote_addr = &vsk->remote_addr; 1142d021c344SAndy King 1143d021c344SAndy King if (remote_addr->svm_cid == VMADDR_CID_ANY) 1144d021c344SAndy King remote_addr->svm_cid = transport->get_local_cid(); 1145d021c344SAndy King 1146d021c344SAndy King /* XXX Should connect() or this function ensure remote_addr is 1147d021c344SAndy King * bound? 1148d021c344SAndy King */ 1149d021c344SAndy King if (!vsock_addr_bound(&vsk->remote_addr)) { 1150d021c344SAndy King err = -EINVAL; 1151d021c344SAndy King goto out; 1152d021c344SAndy King } 1153d021c344SAndy King } else { 1154d021c344SAndy King err = -EINVAL; 1155d021c344SAndy King goto out; 1156d021c344SAndy King } 1157d021c344SAndy King 1158d021c344SAndy King if (!transport->dgram_allow(remote_addr->svm_cid, 1159d021c344SAndy King remote_addr->svm_port)) { 1160d021c344SAndy King err = -EINVAL; 1161d021c344SAndy King goto out; 1162d021c344SAndy King } 1163d021c344SAndy King 11640f7db23aSAl Viro err = transport->dgram_enqueue(vsk, remote_addr, msg, len); 1165d021c344SAndy King 1166d021c344SAndy King out: 1167d021c344SAndy King release_sock(sk); 1168d021c344SAndy King return err; 1169d021c344SAndy King } 1170d021c344SAndy King 1171d021c344SAndy King static int vsock_dgram_connect(struct socket *sock, 1172d021c344SAndy King struct sockaddr *addr, int addr_len, int flags) 1173d021c344SAndy King { 1174d021c344SAndy King int err; 1175d021c344SAndy King struct sock *sk; 1176d021c344SAndy King struct vsock_sock *vsk; 1177d021c344SAndy King struct sockaddr_vm *remote_addr; 1178d021c344SAndy King 1179d021c344SAndy King sk = sock->sk; 1180d021c344SAndy King vsk = vsock_sk(sk); 1181d021c344SAndy King 1182d021c344SAndy King err = vsock_addr_cast(addr, addr_len, &remote_addr); 1183d021c344SAndy King if (err == -EAFNOSUPPORT && remote_addr->svm_family == AF_UNSPEC) { 1184d021c344SAndy King lock_sock(sk); 1185d021c344SAndy King vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, 1186d021c344SAndy King VMADDR_PORT_ANY); 1187d021c344SAndy King sock->state = SS_UNCONNECTED; 1188d021c344SAndy King release_sock(sk); 1189d021c344SAndy King return 0; 1190d021c344SAndy King } else if (err != 0) 1191d021c344SAndy King return -EINVAL; 1192d021c344SAndy King 1193d021c344SAndy King lock_sock(sk); 1194d021c344SAndy King 1195b3a6dfe8SAsias He err = vsock_auto_bind(vsk); 1196b3a6dfe8SAsias He if (err) 1197d021c344SAndy King goto out; 1198d021c344SAndy King 1199fe502c4aSStefano Garzarella if (!vsk->transport->dgram_allow(remote_addr->svm_cid, 1200d021c344SAndy King remote_addr->svm_port)) { 1201d021c344SAndy King err = -EINVAL; 1202d021c344SAndy King goto out; 1203d021c344SAndy King } 1204d021c344SAndy King 1205d021c344SAndy King memcpy(&vsk->remote_addr, remote_addr, sizeof(vsk->remote_addr)); 1206d021c344SAndy King sock->state = SS_CONNECTED; 1207d021c344SAndy King 1208d021c344SAndy King out: 1209d021c344SAndy King release_sock(sk); 1210d021c344SAndy King return err; 1211d021c344SAndy King } 1212d021c344SAndy King 12131b784140SYing Xue static int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg, 12141b784140SYing Xue size_t len, int flags) 1215d021c344SAndy King { 1216fe502c4aSStefano Garzarella struct vsock_sock *vsk = vsock_sk(sock->sk); 1217fe502c4aSStefano Garzarella 1218fe502c4aSStefano Garzarella return vsk->transport->dgram_dequeue(vsk, msg, len, flags); 1219d021c344SAndy King } 1220d021c344SAndy King 1221d021c344SAndy King static const struct proto_ops vsock_dgram_ops = { 1222d021c344SAndy King .family = PF_VSOCK, 1223d021c344SAndy King .owner = THIS_MODULE, 1224d021c344SAndy King .release = vsock_release, 1225d021c344SAndy King .bind = vsock_bind, 1226d021c344SAndy King .connect = vsock_dgram_connect, 1227d021c344SAndy King .socketpair = sock_no_socketpair, 1228d021c344SAndy King .accept = sock_no_accept, 1229d021c344SAndy King .getname = vsock_getname, 1230a11e1d43SLinus Torvalds .poll = vsock_poll, 1231d021c344SAndy King .ioctl = sock_no_ioctl, 1232d021c344SAndy King .listen = sock_no_listen, 1233d021c344SAndy King .shutdown = vsock_shutdown, 1234d021c344SAndy King .sendmsg = vsock_dgram_sendmsg, 1235d021c344SAndy King .recvmsg = vsock_dgram_recvmsg, 1236d021c344SAndy King .mmap = sock_no_mmap, 1237d021c344SAndy King .sendpage = sock_no_sendpage, 1238d021c344SAndy King }; 1239d021c344SAndy King 1240380feae0SPeng Tao static int vsock_transport_cancel_pkt(struct vsock_sock *vsk) 1241380feae0SPeng Tao { 1242fe502c4aSStefano Garzarella const struct vsock_transport *transport = vsk->transport; 1243fe502c4aSStefano Garzarella 12445d1cbcc9SNorbert Slusarek if (!transport || !transport->cancel_pkt) 1245380feae0SPeng Tao return -EOPNOTSUPP; 1246380feae0SPeng Tao 1247380feae0SPeng Tao return transport->cancel_pkt(vsk); 1248380feae0SPeng Tao } 1249380feae0SPeng Tao 1250d021c344SAndy King static void vsock_connect_timeout(struct work_struct *work) 1251d021c344SAndy King { 1252d021c344SAndy King struct sock *sk; 1253d021c344SAndy King struct vsock_sock *vsk; 1254d021c344SAndy King 1255455f05ecSCong Wang vsk = container_of(work, struct vsock_sock, connect_work.work); 1256d021c344SAndy King sk = sk_vsock(vsk); 1257d021c344SAndy King 1258d021c344SAndy King lock_sock(sk); 12593b4477d2SStefan Hajnoczi if (sk->sk_state == TCP_SYN_SENT && 1260d021c344SAndy King (sk->sk_shutdown != SHUTDOWN_MASK)) { 12613b4477d2SStefan Hajnoczi sk->sk_state = TCP_CLOSE; 1262d021c344SAndy King sk->sk_err = ETIMEDOUT; 1263d021c344SAndy King sk->sk_error_report(sk); 12643d0bc44dSNorbert Slusarek vsock_transport_cancel_pkt(vsk); 1265d021c344SAndy King } 1266d021c344SAndy King release_sock(sk); 1267d021c344SAndy King 1268d021c344SAndy King sock_put(sk); 1269d021c344SAndy King } 1270d021c344SAndy King 1271a9e29e55SArseny Krasnov static int vsock_connect(struct socket *sock, struct sockaddr *addr, 1272d021c344SAndy King int addr_len, int flags) 1273d021c344SAndy King { 1274d021c344SAndy King int err; 1275d021c344SAndy King struct sock *sk; 1276d021c344SAndy King struct vsock_sock *vsk; 1277fe502c4aSStefano Garzarella const struct vsock_transport *transport; 1278d021c344SAndy King struct sockaddr_vm *remote_addr; 1279d021c344SAndy King long timeout; 1280d021c344SAndy King DEFINE_WAIT(wait); 1281d021c344SAndy King 1282d021c344SAndy King err = 0; 1283d021c344SAndy King sk = sock->sk; 1284d021c344SAndy King vsk = vsock_sk(sk); 1285d021c344SAndy King 1286d021c344SAndy King lock_sock(sk); 1287d021c344SAndy King 1288d021c344SAndy King /* XXX AF_UNSPEC should make us disconnect like AF_INET. */ 1289d021c344SAndy King switch (sock->state) { 1290d021c344SAndy King case SS_CONNECTED: 1291d021c344SAndy King err = -EISCONN; 1292d021c344SAndy King goto out; 1293d021c344SAndy King case SS_DISCONNECTING: 1294d021c344SAndy King err = -EINVAL; 1295d021c344SAndy King goto out; 1296d021c344SAndy King case SS_CONNECTING: 1297d021c344SAndy King /* This continues on so we can move sock into the SS_CONNECTED 1298d021c344SAndy King * state once the connection has completed (at which point err 1299d021c344SAndy King * will be set to zero also). Otherwise, we will either wait 1300d021c344SAndy King * for the connection or return -EALREADY should this be a 1301d021c344SAndy King * non-blocking call. 1302d021c344SAndy King */ 1303d021c344SAndy King err = -EALREADY; 1304d021c344SAndy King break; 1305d021c344SAndy King default: 13063b4477d2SStefan Hajnoczi if ((sk->sk_state == TCP_LISTEN) || 1307d021c344SAndy King vsock_addr_cast(addr, addr_len, &remote_addr) != 0) { 1308d021c344SAndy King err = -EINVAL; 1309d021c344SAndy King goto out; 1310d021c344SAndy King } 1311d021c344SAndy King 1312c0cfa2d8SStefano Garzarella /* Set the remote address that we are connecting to. */ 1313c0cfa2d8SStefano Garzarella memcpy(&vsk->remote_addr, remote_addr, 1314c0cfa2d8SStefano Garzarella sizeof(vsk->remote_addr)); 1315c0cfa2d8SStefano Garzarella 1316c0cfa2d8SStefano Garzarella err = vsock_assign_transport(vsk, NULL); 1317c0cfa2d8SStefano Garzarella if (err) 1318c0cfa2d8SStefano Garzarella goto out; 1319c0cfa2d8SStefano Garzarella 1320c0cfa2d8SStefano Garzarella transport = vsk->transport; 1321c0cfa2d8SStefano Garzarella 1322d021c344SAndy King /* The hypervisor and well-known contexts do not have socket 1323d021c344SAndy King * endpoints. 1324d021c344SAndy King */ 1325c0cfa2d8SStefano Garzarella if (!transport || 1326c0cfa2d8SStefano Garzarella !transport->stream_allow(remote_addr->svm_cid, 1327d021c344SAndy King remote_addr->svm_port)) { 1328d021c344SAndy King err = -ENETUNREACH; 1329d021c344SAndy King goto out; 1330d021c344SAndy King } 1331d021c344SAndy King 1332b3a6dfe8SAsias He err = vsock_auto_bind(vsk); 1333b3a6dfe8SAsias He if (err) 1334d021c344SAndy King goto out; 1335d021c344SAndy King 13363b4477d2SStefan Hajnoczi sk->sk_state = TCP_SYN_SENT; 1337d021c344SAndy King 1338d021c344SAndy King err = transport->connect(vsk); 1339d021c344SAndy King if (err < 0) 1340d021c344SAndy King goto out; 1341d021c344SAndy King 1342d021c344SAndy King /* Mark sock as connecting and set the error code to in 1343d021c344SAndy King * progress in case this is a non-blocking connect. 1344d021c344SAndy King */ 1345d021c344SAndy King sock->state = SS_CONNECTING; 1346d021c344SAndy King err = -EINPROGRESS; 1347d021c344SAndy King } 1348d021c344SAndy King 1349d021c344SAndy King /* The receive path will handle all communication until we are able to 1350d021c344SAndy King * enter the connected state. Here we wait for the connection to be 1351d021c344SAndy King * completed or a notification of an error. 1352d021c344SAndy King */ 1353d021c344SAndy King timeout = vsk->connect_timeout; 1354d021c344SAndy King prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); 1355d021c344SAndy King 13563b4477d2SStefan Hajnoczi while (sk->sk_state != TCP_ESTABLISHED && sk->sk_err == 0) { 1357d021c344SAndy King if (flags & O_NONBLOCK) { 1358d021c344SAndy King /* If we're not going to block, we schedule a timeout 1359d021c344SAndy King * function to generate a timeout on the connection 1360d021c344SAndy King * attempt, in case the peer doesn't respond in a 1361d021c344SAndy King * timely manner. We hold on to the socket until the 1362d021c344SAndy King * timeout fires. 1363d021c344SAndy King */ 1364d021c344SAndy King sock_hold(sk); 1365455f05ecSCong Wang schedule_delayed_work(&vsk->connect_work, timeout); 1366d021c344SAndy King 1367d021c344SAndy King /* Skip ahead to preserve error code set above. */ 1368d021c344SAndy King goto out_wait; 1369d021c344SAndy King } 1370d021c344SAndy King 1371d021c344SAndy King release_sock(sk); 1372d021c344SAndy King timeout = schedule_timeout(timeout); 1373d021c344SAndy King lock_sock(sk); 1374d021c344SAndy King 1375d021c344SAndy King if (signal_pending(current)) { 1376d021c344SAndy King err = sock_intr_errno(timeout); 13773b4477d2SStefan Hajnoczi sk->sk_state = TCP_CLOSE; 1378f7f9b5e7SClaudio Imbrenda sock->state = SS_UNCONNECTED; 1379380feae0SPeng Tao vsock_transport_cancel_pkt(vsk); 1380f7f9b5e7SClaudio Imbrenda goto out_wait; 1381d021c344SAndy King } else if (timeout == 0) { 1382d021c344SAndy King err = -ETIMEDOUT; 13833b4477d2SStefan Hajnoczi sk->sk_state = TCP_CLOSE; 1384f7f9b5e7SClaudio Imbrenda sock->state = SS_UNCONNECTED; 1385380feae0SPeng Tao vsock_transport_cancel_pkt(vsk); 1386f7f9b5e7SClaudio Imbrenda goto out_wait; 1387d021c344SAndy King } 1388d021c344SAndy King 1389d021c344SAndy King prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); 1390d021c344SAndy King } 1391d021c344SAndy King 1392d021c344SAndy King if (sk->sk_err) { 1393d021c344SAndy King err = -sk->sk_err; 13943b4477d2SStefan Hajnoczi sk->sk_state = TCP_CLOSE; 1395f7f9b5e7SClaudio Imbrenda sock->state = SS_UNCONNECTED; 1396f7f9b5e7SClaudio Imbrenda } else { 1397d021c344SAndy King err = 0; 1398f7f9b5e7SClaudio Imbrenda } 1399d021c344SAndy King 1400d021c344SAndy King out_wait: 1401d021c344SAndy King finish_wait(sk_sleep(sk), &wait); 1402d021c344SAndy King out: 1403d021c344SAndy King release_sock(sk); 1404d021c344SAndy King return err; 1405d021c344SAndy King } 1406d021c344SAndy King 1407cdfbabfbSDavid Howells static int vsock_accept(struct socket *sock, struct socket *newsock, int flags, 1408cdfbabfbSDavid Howells bool kern) 1409d021c344SAndy King { 1410d021c344SAndy King struct sock *listener; 1411d021c344SAndy King int err; 1412d021c344SAndy King struct sock *connected; 1413d021c344SAndy King struct vsock_sock *vconnected; 1414d021c344SAndy King long timeout; 1415d021c344SAndy King DEFINE_WAIT(wait); 1416d021c344SAndy King 1417d021c344SAndy King err = 0; 1418d021c344SAndy King listener = sock->sk; 1419d021c344SAndy King 1420d021c344SAndy King lock_sock(listener); 1421d021c344SAndy King 1422a9e29e55SArseny Krasnov if (!sock_type_connectible(sock->type)) { 1423d021c344SAndy King err = -EOPNOTSUPP; 1424d021c344SAndy King goto out; 1425d021c344SAndy King } 1426d021c344SAndy King 14273b4477d2SStefan Hajnoczi if (listener->sk_state != TCP_LISTEN) { 1428d021c344SAndy King err = -EINVAL; 1429d021c344SAndy King goto out; 1430d021c344SAndy King } 1431d021c344SAndy King 1432d021c344SAndy King /* Wait for children sockets to appear; these are the new sockets 1433d021c344SAndy King * created upon connection establishment. 1434d021c344SAndy King */ 14357e0afbdfSStefano Garzarella timeout = sock_rcvtimeo(listener, flags & O_NONBLOCK); 1436d021c344SAndy King prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE); 1437d021c344SAndy King 1438d021c344SAndy King while ((connected = vsock_dequeue_accept(listener)) == NULL && 1439d021c344SAndy King listener->sk_err == 0) { 1440d021c344SAndy King release_sock(listener); 1441d021c344SAndy King timeout = schedule_timeout(timeout); 1442f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(listener), &wait); 1443d021c344SAndy King lock_sock(listener); 1444d021c344SAndy King 1445d021c344SAndy King if (signal_pending(current)) { 1446d021c344SAndy King err = sock_intr_errno(timeout); 1447f7f9b5e7SClaudio Imbrenda goto out; 1448d021c344SAndy King } else if (timeout == 0) { 1449d021c344SAndy King err = -EAGAIN; 1450f7f9b5e7SClaudio Imbrenda goto out; 1451d021c344SAndy King } 1452d021c344SAndy King 1453d021c344SAndy King prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE); 1454d021c344SAndy King } 1455f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(listener), &wait); 1456d021c344SAndy King 1457d021c344SAndy King if (listener->sk_err) 1458d021c344SAndy King err = -listener->sk_err; 1459d021c344SAndy King 1460d021c344SAndy King if (connected) { 14617976a11bSEric Dumazet sk_acceptq_removed(listener); 1462d021c344SAndy King 14634192f672SStefan Hajnoczi lock_sock_nested(connected, SINGLE_DEPTH_NESTING); 1464d021c344SAndy King vconnected = vsock_sk(connected); 1465d021c344SAndy King 1466d021c344SAndy King /* If the listener socket has received an error, then we should 1467d021c344SAndy King * reject this socket and return. Note that we simply mark the 1468d021c344SAndy King * socket rejected, drop our reference, and let the cleanup 1469d021c344SAndy King * function handle the cleanup; the fact that we found it in 1470d021c344SAndy King * the listener's accept queue guarantees that the cleanup 1471d021c344SAndy King * function hasn't run yet. 1472d021c344SAndy King */ 1473d021c344SAndy King if (err) { 1474d021c344SAndy King vconnected->rejected = true; 1475f7f9b5e7SClaudio Imbrenda } else { 1476d021c344SAndy King newsock->state = SS_CONNECTED; 1477d021c344SAndy King sock_graft(connected, newsock); 1478f7f9b5e7SClaudio Imbrenda } 1479f7f9b5e7SClaudio Imbrenda 1480d021c344SAndy King release_sock(connected); 1481d021c344SAndy King sock_put(connected); 1482d021c344SAndy King } 1483d021c344SAndy King 1484d021c344SAndy King out: 1485d021c344SAndy King release_sock(listener); 1486d021c344SAndy King return err; 1487d021c344SAndy King } 1488d021c344SAndy King 1489d021c344SAndy King static int vsock_listen(struct socket *sock, int backlog) 1490d021c344SAndy King { 1491d021c344SAndy King int err; 1492d021c344SAndy King struct sock *sk; 1493d021c344SAndy King struct vsock_sock *vsk; 1494d021c344SAndy King 1495d021c344SAndy King sk = sock->sk; 1496d021c344SAndy King 1497d021c344SAndy King lock_sock(sk); 1498d021c344SAndy King 1499a9e29e55SArseny Krasnov if (!sock_type_connectible(sk->sk_type)) { 1500d021c344SAndy King err = -EOPNOTSUPP; 1501d021c344SAndy King goto out; 1502d021c344SAndy King } 1503d021c344SAndy King 1504d021c344SAndy King if (sock->state != SS_UNCONNECTED) { 1505d021c344SAndy King err = -EINVAL; 1506d021c344SAndy King goto out; 1507d021c344SAndy King } 1508d021c344SAndy King 1509d021c344SAndy King vsk = vsock_sk(sk); 1510d021c344SAndy King 1511d021c344SAndy King if (!vsock_addr_bound(&vsk->local_addr)) { 1512d021c344SAndy King err = -EINVAL; 1513d021c344SAndy King goto out; 1514d021c344SAndy King } 1515d021c344SAndy King 1516d021c344SAndy King sk->sk_max_ack_backlog = backlog; 15173b4477d2SStefan Hajnoczi sk->sk_state = TCP_LISTEN; 1518d021c344SAndy King 1519d021c344SAndy King err = 0; 1520d021c344SAndy King 1521d021c344SAndy King out: 1522d021c344SAndy King release_sock(sk); 1523d021c344SAndy King return err; 1524d021c344SAndy King } 1525d021c344SAndy King 1526b9f2b0ffSStefano Garzarella static void vsock_update_buffer_size(struct vsock_sock *vsk, 1527b9f2b0ffSStefano Garzarella const struct vsock_transport *transport, 1528b9f2b0ffSStefano Garzarella u64 val) 1529b9f2b0ffSStefano Garzarella { 1530b9f2b0ffSStefano Garzarella if (val > vsk->buffer_max_size) 1531b9f2b0ffSStefano Garzarella val = vsk->buffer_max_size; 1532b9f2b0ffSStefano Garzarella 1533b9f2b0ffSStefano Garzarella if (val < vsk->buffer_min_size) 1534b9f2b0ffSStefano Garzarella val = vsk->buffer_min_size; 1535b9f2b0ffSStefano Garzarella 1536b9f2b0ffSStefano Garzarella if (val != vsk->buffer_size && 1537b9f2b0ffSStefano Garzarella transport && transport->notify_buffer_size) 1538b9f2b0ffSStefano Garzarella transport->notify_buffer_size(vsk, &val); 1539b9f2b0ffSStefano Garzarella 1540b9f2b0ffSStefano Garzarella vsk->buffer_size = val; 1541b9f2b0ffSStefano Garzarella } 1542b9f2b0ffSStefano Garzarella 1543a9e29e55SArseny Krasnov static int vsock_connectible_setsockopt(struct socket *sock, 1544d021c344SAndy King int level, 1545d021c344SAndy King int optname, 1546a7b75c5aSChristoph Hellwig sockptr_t optval, 1547d021c344SAndy King unsigned int optlen) 1548d021c344SAndy King { 1549d021c344SAndy King int err; 1550d021c344SAndy King struct sock *sk; 1551d021c344SAndy King struct vsock_sock *vsk; 1552fe502c4aSStefano Garzarella const struct vsock_transport *transport; 1553d021c344SAndy King u64 val; 1554d021c344SAndy King 1555d021c344SAndy King if (level != AF_VSOCK) 1556d021c344SAndy King return -ENOPROTOOPT; 1557d021c344SAndy King 1558d021c344SAndy King #define COPY_IN(_v) \ 1559d021c344SAndy King do { \ 1560d021c344SAndy King if (optlen < sizeof(_v)) { \ 1561d021c344SAndy King err = -EINVAL; \ 1562d021c344SAndy King goto exit; \ 1563d021c344SAndy King } \ 1564a7b75c5aSChristoph Hellwig if (copy_from_sockptr(&_v, optval, sizeof(_v)) != 0) { \ 1565d021c344SAndy King err = -EFAULT; \ 1566d021c344SAndy King goto exit; \ 1567d021c344SAndy King } \ 1568d021c344SAndy King } while (0) 1569d021c344SAndy King 1570d021c344SAndy King err = 0; 1571d021c344SAndy King sk = sock->sk; 1572d021c344SAndy King vsk = vsock_sk(sk); 1573d021c344SAndy King 1574d021c344SAndy King lock_sock(sk); 1575d021c344SAndy King 1576c518adafSAlexander Popov transport = vsk->transport; 1577c518adafSAlexander Popov 1578d021c344SAndy King switch (optname) { 1579d021c344SAndy King case SO_VM_SOCKETS_BUFFER_SIZE: 1580d021c344SAndy King COPY_IN(val); 1581b9f2b0ffSStefano Garzarella vsock_update_buffer_size(vsk, transport, val); 1582d021c344SAndy King break; 1583d021c344SAndy King 1584d021c344SAndy King case SO_VM_SOCKETS_BUFFER_MAX_SIZE: 1585d021c344SAndy King COPY_IN(val); 1586b9f2b0ffSStefano Garzarella vsk->buffer_max_size = val; 1587b9f2b0ffSStefano Garzarella vsock_update_buffer_size(vsk, transport, vsk->buffer_size); 1588d021c344SAndy King break; 1589d021c344SAndy King 1590d021c344SAndy King case SO_VM_SOCKETS_BUFFER_MIN_SIZE: 1591d021c344SAndy King COPY_IN(val); 1592b9f2b0ffSStefano Garzarella vsk->buffer_min_size = val; 1593b9f2b0ffSStefano Garzarella vsock_update_buffer_size(vsk, transport, vsk->buffer_size); 1594d021c344SAndy King break; 1595d021c344SAndy King 1596d021c344SAndy King case SO_VM_SOCKETS_CONNECT_TIMEOUT: { 1597fe0c72f3SArnd Bergmann struct __kernel_old_timeval tv; 1598d021c344SAndy King COPY_IN(tv); 1599d021c344SAndy King if (tv.tv_sec >= 0 && tv.tv_usec < USEC_PER_SEC && 1600d021c344SAndy King tv.tv_sec < (MAX_SCHEDULE_TIMEOUT / HZ - 1)) { 1601d021c344SAndy King vsk->connect_timeout = tv.tv_sec * HZ + 1602d021c344SAndy King DIV_ROUND_UP(tv.tv_usec, (1000000 / HZ)); 1603d021c344SAndy King if (vsk->connect_timeout == 0) 1604d021c344SAndy King vsk->connect_timeout = 1605d021c344SAndy King VSOCK_DEFAULT_CONNECT_TIMEOUT; 1606d021c344SAndy King 1607d021c344SAndy King } else { 1608d021c344SAndy King err = -ERANGE; 1609d021c344SAndy King } 1610d021c344SAndy King break; 1611d021c344SAndy King } 1612d021c344SAndy King 1613d021c344SAndy King default: 1614d021c344SAndy King err = -ENOPROTOOPT; 1615d021c344SAndy King break; 1616d021c344SAndy King } 1617d021c344SAndy King 1618d021c344SAndy King #undef COPY_IN 1619d021c344SAndy King 1620d021c344SAndy King exit: 1621d021c344SAndy King release_sock(sk); 1622d021c344SAndy King return err; 1623d021c344SAndy King } 1624d021c344SAndy King 1625a9e29e55SArseny Krasnov static int vsock_connectible_getsockopt(struct socket *sock, 1626d021c344SAndy King int level, int optname, 1627d021c344SAndy King char __user *optval, 1628d021c344SAndy King int __user *optlen) 1629d021c344SAndy King { 1630d021c344SAndy King int err; 1631d021c344SAndy King int len; 1632d021c344SAndy King struct sock *sk; 1633d021c344SAndy King struct vsock_sock *vsk; 1634d021c344SAndy King u64 val; 1635d021c344SAndy King 1636d021c344SAndy King if (level != AF_VSOCK) 1637d021c344SAndy King return -ENOPROTOOPT; 1638d021c344SAndy King 1639d021c344SAndy King err = get_user(len, optlen); 1640d021c344SAndy King if (err != 0) 1641d021c344SAndy King return err; 1642d021c344SAndy King 1643d021c344SAndy King #define COPY_OUT(_v) \ 1644d021c344SAndy King do { \ 1645d021c344SAndy King if (len < sizeof(_v)) \ 1646d021c344SAndy King return -EINVAL; \ 1647d021c344SAndy King \ 1648d021c344SAndy King len = sizeof(_v); \ 1649d021c344SAndy King if (copy_to_user(optval, &_v, len) != 0) \ 1650d021c344SAndy King return -EFAULT; \ 1651d021c344SAndy King \ 1652d021c344SAndy King } while (0) 1653d021c344SAndy King 1654d021c344SAndy King err = 0; 1655d021c344SAndy King sk = sock->sk; 1656d021c344SAndy King vsk = vsock_sk(sk); 1657d021c344SAndy King 1658d021c344SAndy King switch (optname) { 1659d021c344SAndy King case SO_VM_SOCKETS_BUFFER_SIZE: 1660b9f2b0ffSStefano Garzarella val = vsk->buffer_size; 1661d021c344SAndy King COPY_OUT(val); 1662d021c344SAndy King break; 1663d021c344SAndy King 1664d021c344SAndy King case SO_VM_SOCKETS_BUFFER_MAX_SIZE: 1665b9f2b0ffSStefano Garzarella val = vsk->buffer_max_size; 1666d021c344SAndy King COPY_OUT(val); 1667d021c344SAndy King break; 1668d021c344SAndy King 1669d021c344SAndy King case SO_VM_SOCKETS_BUFFER_MIN_SIZE: 1670b9f2b0ffSStefano Garzarella val = vsk->buffer_min_size; 1671d021c344SAndy King COPY_OUT(val); 1672d021c344SAndy King break; 1673d021c344SAndy King 1674d021c344SAndy King case SO_VM_SOCKETS_CONNECT_TIMEOUT: { 1675fe0c72f3SArnd Bergmann struct __kernel_old_timeval tv; 1676d021c344SAndy King tv.tv_sec = vsk->connect_timeout / HZ; 1677d021c344SAndy King tv.tv_usec = 1678d021c344SAndy King (vsk->connect_timeout - 1679d021c344SAndy King tv.tv_sec * HZ) * (1000000 / HZ); 1680d021c344SAndy King COPY_OUT(tv); 1681d021c344SAndy King break; 1682d021c344SAndy King } 1683d021c344SAndy King default: 1684d021c344SAndy King return -ENOPROTOOPT; 1685d021c344SAndy King } 1686d021c344SAndy King 1687d021c344SAndy King err = put_user(len, optlen); 1688d021c344SAndy King if (err != 0) 1689d021c344SAndy King return -EFAULT; 1690d021c344SAndy King 1691d021c344SAndy King #undef COPY_OUT 1692d021c344SAndy King 1693d021c344SAndy King return 0; 1694d021c344SAndy King } 1695d021c344SAndy King 1696a9e29e55SArseny Krasnov static int vsock_connectible_sendmsg(struct socket *sock, struct msghdr *msg, 16971b784140SYing Xue size_t len) 1698d021c344SAndy King { 1699d021c344SAndy King struct sock *sk; 1700d021c344SAndy King struct vsock_sock *vsk; 1701fe502c4aSStefano Garzarella const struct vsock_transport *transport; 1702d021c344SAndy King ssize_t total_written; 1703d021c344SAndy King long timeout; 1704d021c344SAndy King int err; 1705d021c344SAndy King struct vsock_transport_send_notify_data send_data; 1706499fde66SWANG Cong DEFINE_WAIT_FUNC(wait, woken_wake_function); 1707d021c344SAndy King 1708d021c344SAndy King sk = sock->sk; 1709d021c344SAndy King vsk = vsock_sk(sk); 1710d021c344SAndy King total_written = 0; 1711d021c344SAndy King err = 0; 1712d021c344SAndy King 1713d021c344SAndy King if (msg->msg_flags & MSG_OOB) 1714d021c344SAndy King return -EOPNOTSUPP; 1715d021c344SAndy King 1716d021c344SAndy King lock_sock(sk); 1717d021c344SAndy King 1718c518adafSAlexander Popov transport = vsk->transport; 1719c518adafSAlexander Popov 1720d021c344SAndy King /* Callers should not provide a destination with stream sockets. */ 1721d021c344SAndy King if (msg->msg_namelen) { 17223b4477d2SStefan Hajnoczi err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP; 1723d021c344SAndy King goto out; 1724d021c344SAndy King } 1725d021c344SAndy King 1726d021c344SAndy King /* Send data only if both sides are not shutdown in the direction. */ 1727d021c344SAndy King if (sk->sk_shutdown & SEND_SHUTDOWN || 1728d021c344SAndy King vsk->peer_shutdown & RCV_SHUTDOWN) { 1729d021c344SAndy King err = -EPIPE; 1730d021c344SAndy King goto out; 1731d021c344SAndy King } 1732d021c344SAndy King 1733c0cfa2d8SStefano Garzarella if (!transport || sk->sk_state != TCP_ESTABLISHED || 1734d021c344SAndy King !vsock_addr_bound(&vsk->local_addr)) { 1735d021c344SAndy King err = -ENOTCONN; 1736d021c344SAndy King goto out; 1737d021c344SAndy King } 1738d021c344SAndy King 1739d021c344SAndy King if (!vsock_addr_bound(&vsk->remote_addr)) { 1740d021c344SAndy King err = -EDESTADDRREQ; 1741d021c344SAndy King goto out; 1742d021c344SAndy King } 1743d021c344SAndy King 1744d021c344SAndy King /* Wait for room in the produce queue to enqueue our user's data. */ 1745d021c344SAndy King timeout = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT); 1746d021c344SAndy King 1747d021c344SAndy King err = transport->notify_send_init(vsk, &send_data); 1748d021c344SAndy King if (err < 0) 1749d021c344SAndy King goto out; 1750d021c344SAndy King 1751d021c344SAndy King while (total_written < len) { 1752d021c344SAndy King ssize_t written; 1753d021c344SAndy King 1754499fde66SWANG Cong add_wait_queue(sk_sleep(sk), &wait); 1755d021c344SAndy King while (vsock_stream_has_space(vsk) == 0 && 1756d021c344SAndy King sk->sk_err == 0 && 1757d021c344SAndy King !(sk->sk_shutdown & SEND_SHUTDOWN) && 1758d021c344SAndy King !(vsk->peer_shutdown & RCV_SHUTDOWN)) { 1759d021c344SAndy King 1760d021c344SAndy King /* Don't wait for non-blocking sockets. */ 1761d021c344SAndy King if (timeout == 0) { 1762d021c344SAndy King err = -EAGAIN; 1763499fde66SWANG Cong remove_wait_queue(sk_sleep(sk), &wait); 1764f7f9b5e7SClaudio Imbrenda goto out_err; 1765d021c344SAndy King } 1766d021c344SAndy King 1767d021c344SAndy King err = transport->notify_send_pre_block(vsk, &send_data); 1768f7f9b5e7SClaudio Imbrenda if (err < 0) { 1769499fde66SWANG Cong remove_wait_queue(sk_sleep(sk), &wait); 1770f7f9b5e7SClaudio Imbrenda goto out_err; 1771f7f9b5e7SClaudio Imbrenda } 1772d021c344SAndy King 1773d021c344SAndy King release_sock(sk); 1774499fde66SWANG Cong timeout = wait_woken(&wait, TASK_INTERRUPTIBLE, timeout); 1775d021c344SAndy King lock_sock(sk); 1776d021c344SAndy King if (signal_pending(current)) { 1777d021c344SAndy King err = sock_intr_errno(timeout); 1778499fde66SWANG Cong remove_wait_queue(sk_sleep(sk), &wait); 1779f7f9b5e7SClaudio Imbrenda goto out_err; 1780d021c344SAndy King } else if (timeout == 0) { 1781d021c344SAndy King err = -EAGAIN; 1782499fde66SWANG Cong remove_wait_queue(sk_sleep(sk), &wait); 1783f7f9b5e7SClaudio Imbrenda goto out_err; 1784d021c344SAndy King } 1785d021c344SAndy King } 1786499fde66SWANG Cong remove_wait_queue(sk_sleep(sk), &wait); 1787d021c344SAndy King 1788d021c344SAndy King /* These checks occur both as part of and after the loop 1789d021c344SAndy King * conditional since we need to check before and after 1790d021c344SAndy King * sleeping. 1791d021c344SAndy King */ 1792d021c344SAndy King if (sk->sk_err) { 1793d021c344SAndy King err = -sk->sk_err; 1794f7f9b5e7SClaudio Imbrenda goto out_err; 1795d021c344SAndy King } else if ((sk->sk_shutdown & SEND_SHUTDOWN) || 1796d021c344SAndy King (vsk->peer_shutdown & RCV_SHUTDOWN)) { 1797d021c344SAndy King err = -EPIPE; 1798f7f9b5e7SClaudio Imbrenda goto out_err; 1799d021c344SAndy King } 1800d021c344SAndy King 1801d021c344SAndy King err = transport->notify_send_pre_enqueue(vsk, &send_data); 1802d021c344SAndy King if (err < 0) 1803f7f9b5e7SClaudio Imbrenda goto out_err; 1804d021c344SAndy King 1805d021c344SAndy King /* Note that enqueue will only write as many bytes as are free 1806d021c344SAndy King * in the produce queue, so we don't need to ensure len is 1807d021c344SAndy King * smaller than the queue size. It is the caller's 1808d021c344SAndy King * responsibility to check how many bytes we were able to send. 1809d021c344SAndy King */ 1810d021c344SAndy King 1811d021c344SAndy King written = transport->stream_enqueue( 18120f7db23aSAl Viro vsk, msg, 1813d021c344SAndy King len - total_written); 1814d021c344SAndy King if (written < 0) { 1815d021c344SAndy King err = -ENOMEM; 1816f7f9b5e7SClaudio Imbrenda goto out_err; 1817d021c344SAndy King } 1818d021c344SAndy King 1819d021c344SAndy King total_written += written; 1820d021c344SAndy King 1821d021c344SAndy King err = transport->notify_send_post_enqueue( 1822d021c344SAndy King vsk, written, &send_data); 1823d021c344SAndy King if (err < 0) 1824f7f9b5e7SClaudio Imbrenda goto out_err; 1825d021c344SAndy King 1826d021c344SAndy King } 1827d021c344SAndy King 1828f7f9b5e7SClaudio Imbrenda out_err: 1829d021c344SAndy King if (total_written > 0) 1830d021c344SAndy King err = total_written; 1831d021c344SAndy King out: 1832d021c344SAndy King release_sock(sk); 1833d021c344SAndy King return err; 1834d021c344SAndy King } 1835d021c344SAndy King 1836b3f7fd54SArseny Krasnov static int vsock_wait_data(struct sock *sk, struct wait_queue_entry *wait, 1837b3f7fd54SArseny Krasnov long timeout, 1838b3f7fd54SArseny Krasnov struct vsock_transport_recv_notify_data *recv_data, 1839b3f7fd54SArseny Krasnov size_t target) 1840b3f7fd54SArseny Krasnov { 1841b3f7fd54SArseny Krasnov const struct vsock_transport *transport; 1842b3f7fd54SArseny Krasnov struct vsock_sock *vsk; 1843b3f7fd54SArseny Krasnov s64 data; 1844b3f7fd54SArseny Krasnov int err; 1845b3f7fd54SArseny Krasnov 1846b3f7fd54SArseny Krasnov vsk = vsock_sk(sk); 1847b3f7fd54SArseny Krasnov err = 0; 1848b3f7fd54SArseny Krasnov transport = vsk->transport; 1849b3f7fd54SArseny Krasnov 1850b3f7fd54SArseny Krasnov while ((data = vsock_stream_has_data(vsk)) == 0) { 1851b3f7fd54SArseny Krasnov prepare_to_wait(sk_sleep(sk), wait, TASK_INTERRUPTIBLE); 1852b3f7fd54SArseny Krasnov 1853b3f7fd54SArseny Krasnov if (sk->sk_err != 0 || 1854b3f7fd54SArseny Krasnov (sk->sk_shutdown & RCV_SHUTDOWN) || 1855b3f7fd54SArseny Krasnov (vsk->peer_shutdown & SEND_SHUTDOWN)) { 1856b3f7fd54SArseny Krasnov break; 1857b3f7fd54SArseny Krasnov } 1858b3f7fd54SArseny Krasnov 1859b3f7fd54SArseny Krasnov /* Don't wait for non-blocking sockets. */ 1860b3f7fd54SArseny Krasnov if (timeout == 0) { 1861b3f7fd54SArseny Krasnov err = -EAGAIN; 1862b3f7fd54SArseny Krasnov break; 1863b3f7fd54SArseny Krasnov } 1864b3f7fd54SArseny Krasnov 1865b3f7fd54SArseny Krasnov if (recv_data) { 1866b3f7fd54SArseny Krasnov err = transport->notify_recv_pre_block(vsk, target, recv_data); 1867b3f7fd54SArseny Krasnov if (err < 0) 1868b3f7fd54SArseny Krasnov break; 1869b3f7fd54SArseny Krasnov } 1870b3f7fd54SArseny Krasnov 1871b3f7fd54SArseny Krasnov release_sock(sk); 1872b3f7fd54SArseny Krasnov timeout = schedule_timeout(timeout); 1873b3f7fd54SArseny Krasnov lock_sock(sk); 1874b3f7fd54SArseny Krasnov 1875b3f7fd54SArseny Krasnov if (signal_pending(current)) { 1876b3f7fd54SArseny Krasnov err = sock_intr_errno(timeout); 1877b3f7fd54SArseny Krasnov break; 1878b3f7fd54SArseny Krasnov } else if (timeout == 0) { 1879b3f7fd54SArseny Krasnov err = -EAGAIN; 1880b3f7fd54SArseny Krasnov break; 1881b3f7fd54SArseny Krasnov } 1882b3f7fd54SArseny Krasnov } 1883b3f7fd54SArseny Krasnov 1884b3f7fd54SArseny Krasnov finish_wait(sk_sleep(sk), wait); 1885b3f7fd54SArseny Krasnov 1886b3f7fd54SArseny Krasnov if (err) 1887b3f7fd54SArseny Krasnov return err; 1888b3f7fd54SArseny Krasnov 1889b3f7fd54SArseny Krasnov /* Internal transport error when checking for available 1890b3f7fd54SArseny Krasnov * data. XXX This should be changed to a connection 1891b3f7fd54SArseny Krasnov * reset in a later change. 1892b3f7fd54SArseny Krasnov */ 1893b3f7fd54SArseny Krasnov if (data < 0) 1894b3f7fd54SArseny Krasnov return -ENOMEM; 1895b3f7fd54SArseny Krasnov 1896b3f7fd54SArseny Krasnov return data; 1897b3f7fd54SArseny Krasnov } 1898b3f7fd54SArseny Krasnov 1899*19c1b90eSArseny Krasnov static int __vsock_stream_recvmsg(struct sock *sk, struct msghdr *msg, 1900*19c1b90eSArseny Krasnov size_t len, int flags) 1901d021c344SAndy King { 1902d021c344SAndy King struct vsock_transport_recv_notify_data recv_data; 1903*19c1b90eSArseny Krasnov const struct vsock_transport *transport; 1904*19c1b90eSArseny Krasnov struct vsock_sock *vsk; 1905*19c1b90eSArseny Krasnov ssize_t copied; 1906*19c1b90eSArseny Krasnov size_t target; 1907*19c1b90eSArseny Krasnov long timeout; 1908*19c1b90eSArseny Krasnov int err; 1909d021c344SAndy King 1910d021c344SAndy King DEFINE_WAIT(wait); 1911d021c344SAndy King 1912d021c344SAndy King vsk = vsock_sk(sk); 1913c518adafSAlexander Popov transport = vsk->transport; 1914c518adafSAlexander Popov 1915d021c344SAndy King /* We must not copy less than target bytes into the user's buffer 1916d021c344SAndy King * before returning successfully, so we wait for the consume queue to 1917d021c344SAndy King * have that much data to consume before dequeueing. Note that this 1918d021c344SAndy King * makes it impossible to handle cases where target is greater than the 1919d021c344SAndy King * queue size. 1920d021c344SAndy King */ 1921d021c344SAndy King target = sock_rcvlowat(sk, flags & MSG_WAITALL, len); 1922d021c344SAndy King if (target >= transport->stream_rcvhiwat(vsk)) { 1923d021c344SAndy King err = -ENOMEM; 1924d021c344SAndy King goto out; 1925d021c344SAndy King } 1926d021c344SAndy King timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT); 1927d021c344SAndy King copied = 0; 1928d021c344SAndy King 1929d021c344SAndy King err = transport->notify_recv_init(vsk, target, &recv_data); 1930d021c344SAndy King if (err < 0) 1931d021c344SAndy King goto out; 1932d021c344SAndy King 1933d021c344SAndy King 1934d021c344SAndy King while (1) { 1935f7f9b5e7SClaudio Imbrenda ssize_t read; 1936f7f9b5e7SClaudio Imbrenda 1937b3f7fd54SArseny Krasnov err = vsock_wait_data(sk, &wait, timeout, &recv_data, target); 1938b3f7fd54SArseny Krasnov if (err <= 0) 1939b3f7fd54SArseny Krasnov break; 1940d021c344SAndy King 1941b3f7fd54SArseny Krasnov err = transport->notify_recv_pre_dequeue(vsk, target, 1942b3f7fd54SArseny Krasnov &recv_data); 1943d021c344SAndy King if (err < 0) 1944d021c344SAndy King break; 1945d021c344SAndy King 1946b3f7fd54SArseny Krasnov read = transport->stream_dequeue(vsk, msg, len - copied, flags); 1947d021c344SAndy King if (read < 0) { 1948d021c344SAndy King err = -ENOMEM; 1949d021c344SAndy King break; 1950d021c344SAndy King } 1951d021c344SAndy King 1952d021c344SAndy King copied += read; 1953d021c344SAndy King 1954b3f7fd54SArseny Krasnov err = transport->notify_recv_post_dequeue(vsk, target, read, 1955d021c344SAndy King !(flags & MSG_PEEK), &recv_data); 1956d021c344SAndy King if (err < 0) 1957f7f9b5e7SClaudio Imbrenda goto out; 1958d021c344SAndy King 1959d021c344SAndy King if (read >= target || flags & MSG_PEEK) 1960d021c344SAndy King break; 1961d021c344SAndy King 1962d021c344SAndy King target -= read; 1963d021c344SAndy King } 1964d021c344SAndy King 1965d021c344SAndy King if (sk->sk_err) 1966d021c344SAndy King err = -sk->sk_err; 1967d021c344SAndy King else if (sk->sk_shutdown & RCV_SHUTDOWN) 1968d021c344SAndy King err = 0; 1969d021c344SAndy King 1970dedc58e0SIan Campbell if (copied > 0) 1971d021c344SAndy King err = copied; 1972d021c344SAndy King 1973d021c344SAndy King out: 1974*19c1b90eSArseny Krasnov return err; 1975*19c1b90eSArseny Krasnov } 1976*19c1b90eSArseny Krasnov 1977*19c1b90eSArseny Krasnov static int 1978*19c1b90eSArseny Krasnov vsock_connectible_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, 1979*19c1b90eSArseny Krasnov int flags) 1980*19c1b90eSArseny Krasnov { 1981*19c1b90eSArseny Krasnov struct sock *sk; 1982*19c1b90eSArseny Krasnov struct vsock_sock *vsk; 1983*19c1b90eSArseny Krasnov const struct vsock_transport *transport; 1984*19c1b90eSArseny Krasnov int err; 1985*19c1b90eSArseny Krasnov 1986*19c1b90eSArseny Krasnov DEFINE_WAIT(wait); 1987*19c1b90eSArseny Krasnov 1988*19c1b90eSArseny Krasnov sk = sock->sk; 1989*19c1b90eSArseny Krasnov vsk = vsock_sk(sk); 1990*19c1b90eSArseny Krasnov err = 0; 1991*19c1b90eSArseny Krasnov 1992*19c1b90eSArseny Krasnov lock_sock(sk); 1993*19c1b90eSArseny Krasnov 1994*19c1b90eSArseny Krasnov transport = vsk->transport; 1995*19c1b90eSArseny Krasnov 1996*19c1b90eSArseny Krasnov if (!transport || sk->sk_state != TCP_ESTABLISHED) { 1997*19c1b90eSArseny Krasnov /* Recvmsg is supposed to return 0 if a peer performs an 1998*19c1b90eSArseny Krasnov * orderly shutdown. Differentiate between that case and when a 1999*19c1b90eSArseny Krasnov * peer has not connected or a local shutdown occurred with the 2000*19c1b90eSArseny Krasnov * SOCK_DONE flag. 2001*19c1b90eSArseny Krasnov */ 2002*19c1b90eSArseny Krasnov if (sock_flag(sk, SOCK_DONE)) 2003*19c1b90eSArseny Krasnov err = 0; 2004*19c1b90eSArseny Krasnov else 2005*19c1b90eSArseny Krasnov err = -ENOTCONN; 2006*19c1b90eSArseny Krasnov 2007*19c1b90eSArseny Krasnov goto out; 2008*19c1b90eSArseny Krasnov } 2009*19c1b90eSArseny Krasnov 2010*19c1b90eSArseny Krasnov if (flags & MSG_OOB) { 2011*19c1b90eSArseny Krasnov err = -EOPNOTSUPP; 2012*19c1b90eSArseny Krasnov goto out; 2013*19c1b90eSArseny Krasnov } 2014*19c1b90eSArseny Krasnov 2015*19c1b90eSArseny Krasnov /* We don't check peer_shutdown flag here since peer may actually shut 2016*19c1b90eSArseny Krasnov * down, but there can be data in the queue that a local socket can 2017*19c1b90eSArseny Krasnov * receive. 2018*19c1b90eSArseny Krasnov */ 2019*19c1b90eSArseny Krasnov if (sk->sk_shutdown & RCV_SHUTDOWN) { 2020*19c1b90eSArseny Krasnov err = 0; 2021*19c1b90eSArseny Krasnov goto out; 2022*19c1b90eSArseny Krasnov } 2023*19c1b90eSArseny Krasnov 2024*19c1b90eSArseny Krasnov /* It is valid on Linux to pass in a zero-length receive buffer. This 2025*19c1b90eSArseny Krasnov * is not an error. We may as well bail out now. 2026*19c1b90eSArseny Krasnov */ 2027*19c1b90eSArseny Krasnov if (!len) { 2028*19c1b90eSArseny Krasnov err = 0; 2029*19c1b90eSArseny Krasnov goto out; 2030*19c1b90eSArseny Krasnov } 2031*19c1b90eSArseny Krasnov 2032*19c1b90eSArseny Krasnov err = __vsock_stream_recvmsg(sk, msg, len, flags); 2033*19c1b90eSArseny Krasnov 2034*19c1b90eSArseny Krasnov out: 2035d021c344SAndy King release_sock(sk); 2036d021c344SAndy King return err; 2037d021c344SAndy King } 2038d021c344SAndy King 2039d021c344SAndy King static const struct proto_ops vsock_stream_ops = { 2040d021c344SAndy King .family = PF_VSOCK, 2041d021c344SAndy King .owner = THIS_MODULE, 2042d021c344SAndy King .release = vsock_release, 2043d021c344SAndy King .bind = vsock_bind, 2044a9e29e55SArseny Krasnov .connect = vsock_connect, 2045d021c344SAndy King .socketpair = sock_no_socketpair, 2046d021c344SAndy King .accept = vsock_accept, 2047d021c344SAndy King .getname = vsock_getname, 2048a11e1d43SLinus Torvalds .poll = vsock_poll, 2049d021c344SAndy King .ioctl = sock_no_ioctl, 2050d021c344SAndy King .listen = vsock_listen, 2051d021c344SAndy King .shutdown = vsock_shutdown, 2052a9e29e55SArseny Krasnov .setsockopt = vsock_connectible_setsockopt, 2053a9e29e55SArseny Krasnov .getsockopt = vsock_connectible_getsockopt, 2054a9e29e55SArseny Krasnov .sendmsg = vsock_connectible_sendmsg, 2055a9e29e55SArseny Krasnov .recvmsg = vsock_connectible_recvmsg, 2056d021c344SAndy King .mmap = sock_no_mmap, 2057d021c344SAndy King .sendpage = sock_no_sendpage, 2058d021c344SAndy King }; 2059d021c344SAndy King 2060d021c344SAndy King static int vsock_create(struct net *net, struct socket *sock, 2061d021c344SAndy King int protocol, int kern) 2062d021c344SAndy King { 2063c0cfa2d8SStefano Garzarella struct vsock_sock *vsk; 206455f3e149SStefano Garzarella struct sock *sk; 2065c0cfa2d8SStefano Garzarella int ret; 206655f3e149SStefano Garzarella 2067d021c344SAndy King if (!sock) 2068d021c344SAndy King return -EINVAL; 2069d021c344SAndy King 20706cf1c5fcSAndy King if (protocol && protocol != PF_VSOCK) 2071d021c344SAndy King return -EPROTONOSUPPORT; 2072d021c344SAndy King 2073d021c344SAndy King switch (sock->type) { 2074d021c344SAndy King case SOCK_DGRAM: 2075d021c344SAndy King sock->ops = &vsock_dgram_ops; 2076d021c344SAndy King break; 2077d021c344SAndy King case SOCK_STREAM: 2078d021c344SAndy King sock->ops = &vsock_stream_ops; 2079d021c344SAndy King break; 2080d021c344SAndy King default: 2081d021c344SAndy King return -ESOCKTNOSUPPORT; 2082d021c344SAndy King } 2083d021c344SAndy King 2084d021c344SAndy King sock->state = SS_UNCONNECTED; 2085d021c344SAndy King 208655f3e149SStefano Garzarella sk = __vsock_create(net, sock, NULL, GFP_KERNEL, 0, kern); 208755f3e149SStefano Garzarella if (!sk) 208855f3e149SStefano Garzarella return -ENOMEM; 208955f3e149SStefano Garzarella 2090c0cfa2d8SStefano Garzarella vsk = vsock_sk(sk); 2091c0cfa2d8SStefano Garzarella 2092c0cfa2d8SStefano Garzarella if (sock->type == SOCK_DGRAM) { 2093c0cfa2d8SStefano Garzarella ret = vsock_assign_transport(vsk, NULL); 2094c0cfa2d8SStefano Garzarella if (ret < 0) { 2095c0cfa2d8SStefano Garzarella sock_put(sk); 2096c0cfa2d8SStefano Garzarella return ret; 2097c0cfa2d8SStefano Garzarella } 2098c0cfa2d8SStefano Garzarella } 2099c0cfa2d8SStefano Garzarella 2100c0cfa2d8SStefano Garzarella vsock_insert_unbound(vsk); 210155f3e149SStefano Garzarella 210255f3e149SStefano Garzarella return 0; 2103d021c344SAndy King } 2104d021c344SAndy King 2105d021c344SAndy King static const struct net_proto_family vsock_family_ops = { 2106d021c344SAndy King .family = AF_VSOCK, 2107d021c344SAndy King .create = vsock_create, 2108d021c344SAndy King .owner = THIS_MODULE, 2109d021c344SAndy King }; 2110d021c344SAndy King 2111d021c344SAndy King static long vsock_dev_do_ioctl(struct file *filp, 2112d021c344SAndy King unsigned int cmd, void __user *ptr) 2113d021c344SAndy King { 2114d021c344SAndy King u32 __user *p = ptr; 2115c0cfa2d8SStefano Garzarella u32 cid = VMADDR_CID_ANY; 2116d021c344SAndy King int retval = 0; 2117d021c344SAndy King 2118d021c344SAndy King switch (cmd) { 2119d021c344SAndy King case IOCTL_VM_SOCKETS_GET_LOCAL_CID: 2120c0cfa2d8SStefano Garzarella /* To be compatible with the VMCI behavior, we prioritize the 2121c0cfa2d8SStefano Garzarella * guest CID instead of well-know host CID (VMADDR_CID_HOST). 2122c0cfa2d8SStefano Garzarella */ 2123c0cfa2d8SStefano Garzarella if (transport_g2h) 2124c0cfa2d8SStefano Garzarella cid = transport_g2h->get_local_cid(); 2125c0cfa2d8SStefano Garzarella else if (transport_h2g) 2126c0cfa2d8SStefano Garzarella cid = transport_h2g->get_local_cid(); 2127c0cfa2d8SStefano Garzarella 2128c0cfa2d8SStefano Garzarella if (put_user(cid, p) != 0) 2129d021c344SAndy King retval = -EFAULT; 2130d021c344SAndy King break; 2131d021c344SAndy King 2132d021c344SAndy King default: 2133c3e448cdSColin Ian King retval = -ENOIOCTLCMD; 2134d021c344SAndy King } 2135d021c344SAndy King 2136d021c344SAndy King return retval; 2137d021c344SAndy King } 2138d021c344SAndy King 2139d021c344SAndy King static long vsock_dev_ioctl(struct file *filp, 2140d021c344SAndy King unsigned int cmd, unsigned long arg) 2141d021c344SAndy King { 2142d021c344SAndy King return vsock_dev_do_ioctl(filp, cmd, (void __user *)arg); 2143d021c344SAndy King } 2144d021c344SAndy King 2145d021c344SAndy King #ifdef CONFIG_COMPAT 2146d021c344SAndy King static long vsock_dev_compat_ioctl(struct file *filp, 2147d021c344SAndy King unsigned int cmd, unsigned long arg) 2148d021c344SAndy King { 2149d021c344SAndy King return vsock_dev_do_ioctl(filp, cmd, compat_ptr(arg)); 2150d021c344SAndy King } 2151d021c344SAndy King #endif 2152d021c344SAndy King 2153d021c344SAndy King static const struct file_operations vsock_device_ops = { 2154d021c344SAndy King .owner = THIS_MODULE, 2155d021c344SAndy King .unlocked_ioctl = vsock_dev_ioctl, 2156d021c344SAndy King #ifdef CONFIG_COMPAT 2157d021c344SAndy King .compat_ioctl = vsock_dev_compat_ioctl, 2158d021c344SAndy King #endif 2159d021c344SAndy King .open = nonseekable_open, 2160d021c344SAndy King }; 2161d021c344SAndy King 2162d021c344SAndy King static struct miscdevice vsock_device = { 2163d021c344SAndy King .name = "vsock", 2164d021c344SAndy King .fops = &vsock_device_ops, 2165d021c344SAndy King }; 2166d021c344SAndy King 2167c0cfa2d8SStefano Garzarella static int __init vsock_init(void) 2168d021c344SAndy King { 2169c0cfa2d8SStefano Garzarella int err = 0; 21702c4a336eSAndy King 2171c0cfa2d8SStefano Garzarella vsock_init_tables(); 21722c4a336eSAndy King 2173c0cfa2d8SStefano Garzarella vsock_proto.owner = THIS_MODULE; 21746ad0b2f7SAsias He vsock_device.minor = MISC_DYNAMIC_MINOR; 2175d021c344SAndy King err = misc_register(&vsock_device); 2176d021c344SAndy King if (err) { 2177d021c344SAndy King pr_err("Failed to register misc device\n"); 2178f6a835bbSGao feng goto err_reset_transport; 2179d021c344SAndy King } 2180d021c344SAndy King 2181d021c344SAndy King err = proto_register(&vsock_proto, 1); /* we want our slab */ 2182d021c344SAndy King if (err) { 2183d021c344SAndy King pr_err("Cannot register vsock protocol\n"); 2184f6a835bbSGao feng goto err_deregister_misc; 2185d021c344SAndy King } 2186d021c344SAndy King 2187d021c344SAndy King err = sock_register(&vsock_family_ops); 2188d021c344SAndy King if (err) { 2189d021c344SAndy King pr_err("could not register af_vsock (%d) address family: %d\n", 2190d021c344SAndy King AF_VSOCK, err); 2191d021c344SAndy King goto err_unregister_proto; 2192d021c344SAndy King } 2193d021c344SAndy King 2194d021c344SAndy King return 0; 2195d021c344SAndy King 2196d021c344SAndy King err_unregister_proto: 2197d021c344SAndy King proto_unregister(&vsock_proto); 2198f6a835bbSGao feng err_deregister_misc: 2199d021c344SAndy King misc_deregister(&vsock_device); 2200f6a835bbSGao feng err_reset_transport: 2201d021c344SAndy King return err; 2202d021c344SAndy King } 2203d021c344SAndy King 2204c0cfa2d8SStefano Garzarella static void __exit vsock_exit(void) 2205d021c344SAndy King { 2206d021c344SAndy King misc_deregister(&vsock_device); 2207d021c344SAndy King sock_unregister(AF_VSOCK); 2208d021c344SAndy King proto_unregister(&vsock_proto); 2209d021c344SAndy King } 2210d021c344SAndy King 2211daabfbcaSStefano Garzarella const struct vsock_transport *vsock_core_get_transport(struct vsock_sock *vsk) 22120b01aeb3SStefan Hajnoczi { 2213daabfbcaSStefano Garzarella return vsk->transport; 22140b01aeb3SStefan Hajnoczi } 22150b01aeb3SStefan Hajnoczi EXPORT_SYMBOL_GPL(vsock_core_get_transport); 22160b01aeb3SStefan Hajnoczi 2217c0cfa2d8SStefano Garzarella int vsock_core_register(const struct vsock_transport *t, int features) 221805e489b1SStefan Hajnoczi { 22190e121905SStefano Garzarella const struct vsock_transport *t_h2g, *t_g2h, *t_dgram, *t_local; 2220c0cfa2d8SStefano Garzarella int err = mutex_lock_interruptible(&vsock_register_mutex); 2221c0cfa2d8SStefano Garzarella 2222c0cfa2d8SStefano Garzarella if (err) 2223c0cfa2d8SStefano Garzarella return err; 2224c0cfa2d8SStefano Garzarella 2225c0cfa2d8SStefano Garzarella t_h2g = transport_h2g; 2226c0cfa2d8SStefano Garzarella t_g2h = transport_g2h; 2227c0cfa2d8SStefano Garzarella t_dgram = transport_dgram; 22280e121905SStefano Garzarella t_local = transport_local; 2229c0cfa2d8SStefano Garzarella 2230c0cfa2d8SStefano Garzarella if (features & VSOCK_TRANSPORT_F_H2G) { 2231c0cfa2d8SStefano Garzarella if (t_h2g) { 2232c0cfa2d8SStefano Garzarella err = -EBUSY; 2233c0cfa2d8SStefano Garzarella goto err_busy; 2234c0cfa2d8SStefano Garzarella } 2235c0cfa2d8SStefano Garzarella t_h2g = t; 223605e489b1SStefan Hajnoczi } 223705e489b1SStefan Hajnoczi 2238c0cfa2d8SStefano Garzarella if (features & VSOCK_TRANSPORT_F_G2H) { 2239c0cfa2d8SStefano Garzarella if (t_g2h) { 2240c0cfa2d8SStefano Garzarella err = -EBUSY; 2241c0cfa2d8SStefano Garzarella goto err_busy; 2242c0cfa2d8SStefano Garzarella } 2243c0cfa2d8SStefano Garzarella t_g2h = t; 2244c0cfa2d8SStefano Garzarella } 2245c0cfa2d8SStefano Garzarella 2246c0cfa2d8SStefano Garzarella if (features & VSOCK_TRANSPORT_F_DGRAM) { 2247c0cfa2d8SStefano Garzarella if (t_dgram) { 2248c0cfa2d8SStefano Garzarella err = -EBUSY; 2249c0cfa2d8SStefano Garzarella goto err_busy; 2250c0cfa2d8SStefano Garzarella } 2251c0cfa2d8SStefano Garzarella t_dgram = t; 2252c0cfa2d8SStefano Garzarella } 2253c0cfa2d8SStefano Garzarella 22540e121905SStefano Garzarella if (features & VSOCK_TRANSPORT_F_LOCAL) { 22550e121905SStefano Garzarella if (t_local) { 22560e121905SStefano Garzarella err = -EBUSY; 22570e121905SStefano Garzarella goto err_busy; 22580e121905SStefano Garzarella } 22590e121905SStefano Garzarella t_local = t; 22600e121905SStefano Garzarella } 22610e121905SStefano Garzarella 2262c0cfa2d8SStefano Garzarella transport_h2g = t_h2g; 2263c0cfa2d8SStefano Garzarella transport_g2h = t_g2h; 2264c0cfa2d8SStefano Garzarella transport_dgram = t_dgram; 22650e121905SStefano Garzarella transport_local = t_local; 2266c0cfa2d8SStefano Garzarella 2267c0cfa2d8SStefano Garzarella err_busy: 2268c0cfa2d8SStefano Garzarella mutex_unlock(&vsock_register_mutex); 2269c0cfa2d8SStefano Garzarella return err; 2270c0cfa2d8SStefano Garzarella } 2271c0cfa2d8SStefano Garzarella EXPORT_SYMBOL_GPL(vsock_core_register); 2272c0cfa2d8SStefano Garzarella 2273c0cfa2d8SStefano Garzarella void vsock_core_unregister(const struct vsock_transport *t) 2274c0cfa2d8SStefano Garzarella { 2275c0cfa2d8SStefano Garzarella mutex_lock(&vsock_register_mutex); 2276c0cfa2d8SStefano Garzarella 2277c0cfa2d8SStefano Garzarella if (transport_h2g == t) 2278c0cfa2d8SStefano Garzarella transport_h2g = NULL; 2279c0cfa2d8SStefano Garzarella 2280c0cfa2d8SStefano Garzarella if (transport_g2h == t) 2281c0cfa2d8SStefano Garzarella transport_g2h = NULL; 2282c0cfa2d8SStefano Garzarella 2283c0cfa2d8SStefano Garzarella if (transport_dgram == t) 2284c0cfa2d8SStefano Garzarella transport_dgram = NULL; 2285c0cfa2d8SStefano Garzarella 22860e121905SStefano Garzarella if (transport_local == t) 22870e121905SStefano Garzarella transport_local = NULL; 22880e121905SStefano Garzarella 2289c0cfa2d8SStefano Garzarella mutex_unlock(&vsock_register_mutex); 2290c0cfa2d8SStefano Garzarella } 2291c0cfa2d8SStefano Garzarella EXPORT_SYMBOL_GPL(vsock_core_unregister); 2292c0cfa2d8SStefano Garzarella 2293c0cfa2d8SStefano Garzarella module_init(vsock_init); 229405e489b1SStefan Hajnoczi module_exit(vsock_exit); 2295c1eef220SCong Wang 2296d021c344SAndy King MODULE_AUTHOR("VMware, Inc."); 2297d021c344SAndy King MODULE_DESCRIPTION("VMware Virtual Socket Family"); 22981190cfdbSJorgen Hansen MODULE_VERSION("1.0.2.0-k"); 2299d021c344SAndy King MODULE_LICENSE("GPL v2"); 2300