1d021c344SAndy King /* 2d021c344SAndy King * VMware vSockets Driver 3d021c344SAndy King * 4d021c344SAndy King * Copyright (C) 2007-2013 VMware, Inc. All rights reserved. 5d021c344SAndy King * 6d021c344SAndy King * This program is free software; you can redistribute it and/or modify it 7d021c344SAndy King * under the terms of the GNU General Public License as published by the Free 8d021c344SAndy King * Software Foundation version 2 and no later version. 9d021c344SAndy King * 10d021c344SAndy King * This program is distributed in the hope that it will be useful, but WITHOUT 11d021c344SAndy King * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or 12d021c344SAndy King * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for 13d021c344SAndy King * more details. 14d021c344SAndy King */ 15d021c344SAndy King 16d021c344SAndy King /* Implementation notes: 17d021c344SAndy King * 18d021c344SAndy King * - There are two kinds of sockets: those created by user action (such as 19d021c344SAndy King * calling socket(2)) and those created by incoming connection request packets. 20d021c344SAndy King * 21d021c344SAndy King * - There are two "global" tables, one for bound sockets (sockets that have 22d021c344SAndy King * specified an address that they are responsible for) and one for connected 23d021c344SAndy King * sockets (sockets that have established a connection with another socket). 24d021c344SAndy King * These tables are "global" in that all sockets on the system are placed 25d021c344SAndy King * within them. - Note, though, that the bound table contains an extra entry 26d021c344SAndy King * for a list of unbound sockets and SOCK_DGRAM sockets will always remain in 27d021c344SAndy King * that list. The bound table is used solely for lookup of sockets when packets 28d021c344SAndy King * are received and that's not necessary for SOCK_DGRAM sockets since we create 29d021c344SAndy King * a datagram handle for each and need not perform a lookup. Keeping SOCK_DGRAM 30d021c344SAndy King * sockets out of the bound hash buckets will reduce the chance of collisions 31d021c344SAndy King * when looking for SOCK_STREAM sockets and prevents us from having to check the 32d021c344SAndy King * socket type in the hash table lookups. 33d021c344SAndy King * 34d021c344SAndy King * - Sockets created by user action will either be "client" sockets that 35d021c344SAndy King * initiate a connection or "server" sockets that listen for connections; we do 36d021c344SAndy King * not support simultaneous connects (two "client" sockets connecting). 37d021c344SAndy King * 38d021c344SAndy King * - "Server" sockets are referred to as listener sockets throughout this 393b4477d2SStefan Hajnoczi * implementation because they are in the TCP_LISTEN state. When a 40ea3803c1SStefan Hajnoczi * connection request is received (the second kind of socket mentioned above), 41ea3803c1SStefan Hajnoczi * we create a new socket and refer to it as a pending socket. These pending 42ea3803c1SStefan Hajnoczi * sockets are placed on the pending connection list of the listener socket. 43ea3803c1SStefan Hajnoczi * When future packets are received for the address the listener socket is 44ea3803c1SStefan Hajnoczi * bound to, we check if the source of the packet is from one that has an 45ea3803c1SStefan Hajnoczi * existing pending connection. If it does, we process the packet for the 46ea3803c1SStefan Hajnoczi * pending socket. When that socket reaches the connected state, it is removed 47ea3803c1SStefan Hajnoczi * from the listener socket's pending list and enqueued in the listener 48ea3803c1SStefan Hajnoczi * socket's accept queue. Callers of accept(2) will accept connected sockets 49ea3803c1SStefan Hajnoczi * from the listener socket's accept queue. If the socket cannot be accepted 50ea3803c1SStefan Hajnoczi * for some reason then it is marked rejected. Once the connection is 51ea3803c1SStefan Hajnoczi * accepted, it is owned by the user process and the responsibility for cleanup 52ea3803c1SStefan Hajnoczi * falls with that user process. 53d021c344SAndy King * 54d021c344SAndy King * - It is possible that these pending sockets will never reach the connected 55d021c344SAndy King * state; in fact, we may never receive another packet after the connection 56d021c344SAndy King * request. Because of this, we must schedule a cleanup function to run in the 57d021c344SAndy King * future, after some amount of time passes where a connection should have been 58d021c344SAndy King * established. This function ensures that the socket is off all lists so it 59d021c344SAndy King * cannot be retrieved, then drops all references to the socket so it is cleaned 60d021c344SAndy King * up (sock_put() -> sk_free() -> our sk_destruct implementation). Note this 61d021c344SAndy King * function will also cleanup rejected sockets, those that reach the connected 62d021c344SAndy King * state but leave it before they have been accepted. 63d021c344SAndy King * 644192f672SStefan Hajnoczi * - Lock ordering for pending or accept queue sockets is: 654192f672SStefan Hajnoczi * 664192f672SStefan Hajnoczi * lock_sock(listener); 674192f672SStefan Hajnoczi * lock_sock_nested(pending, SINGLE_DEPTH_NESTING); 684192f672SStefan Hajnoczi * 694192f672SStefan Hajnoczi * Using explicit nested locking keeps lockdep happy since normally only one 704192f672SStefan Hajnoczi * lock of a given class may be taken at a time. 714192f672SStefan Hajnoczi * 72d021c344SAndy King * - Sockets created by user action will be cleaned up when the user process 73d021c344SAndy King * calls close(2), causing our release implementation to be called. Our release 74d021c344SAndy King * implementation will perform some cleanup then drop the last reference so our 75d021c344SAndy King * sk_destruct implementation is invoked. Our sk_destruct implementation will 76d021c344SAndy King * perform additional cleanup that's common for both types of sockets. 77d021c344SAndy King * 78d021c344SAndy King * - A socket's reference count is what ensures that the structure won't be 79d021c344SAndy King * freed. Each entry in a list (such as the "global" bound and connected tables 80d021c344SAndy King * and the listener socket's pending list and connected queue) ensures a 81d021c344SAndy King * reference. When we defer work until process context and pass a socket as our 82d021c344SAndy King * argument, we must ensure the reference count is increased to ensure the 83d021c344SAndy King * socket isn't freed before the function is run; the deferred function will 84d021c344SAndy King * then drop the reference. 853b4477d2SStefan Hajnoczi * 863b4477d2SStefan Hajnoczi * - sk->sk_state uses the TCP state constants because they are widely used by 873b4477d2SStefan Hajnoczi * other address families and exposed to userspace tools like ss(8): 883b4477d2SStefan Hajnoczi * 893b4477d2SStefan Hajnoczi * TCP_CLOSE - unconnected 903b4477d2SStefan Hajnoczi * TCP_SYN_SENT - connecting 913b4477d2SStefan Hajnoczi * TCP_ESTABLISHED - connected 923b4477d2SStefan Hajnoczi * TCP_CLOSING - disconnecting 933b4477d2SStefan Hajnoczi * TCP_LISTEN - listening 94d021c344SAndy King */ 95d021c344SAndy King 96d021c344SAndy King #include <linux/types.h> 97d021c344SAndy King #include <linux/bitops.h> 98d021c344SAndy King #include <linux/cred.h> 99d021c344SAndy King #include <linux/init.h> 100d021c344SAndy King #include <linux/io.h> 101d021c344SAndy King #include <linux/kernel.h> 102174cd4b1SIngo Molnar #include <linux/sched/signal.h> 103d021c344SAndy King #include <linux/kmod.h> 104d021c344SAndy King #include <linux/list.h> 105d021c344SAndy King #include <linux/miscdevice.h> 106d021c344SAndy King #include <linux/module.h> 107d021c344SAndy King #include <linux/mutex.h> 108d021c344SAndy King #include <linux/net.h> 109d021c344SAndy King #include <linux/poll.h> 110d021c344SAndy King #include <linux/skbuff.h> 111d021c344SAndy King #include <linux/smp.h> 112d021c344SAndy King #include <linux/socket.h> 113d021c344SAndy King #include <linux/stddef.h> 114d021c344SAndy King #include <linux/unistd.h> 115d021c344SAndy King #include <linux/wait.h> 116d021c344SAndy King #include <linux/workqueue.h> 117d021c344SAndy King #include <net/sock.h> 11882a54d0eSAsias He #include <net/af_vsock.h> 119d021c344SAndy King 120d021c344SAndy King static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr); 121d021c344SAndy King static void vsock_sk_destruct(struct sock *sk); 122d021c344SAndy King static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb); 123d021c344SAndy King 124d021c344SAndy King /* Protocol family. */ 125d021c344SAndy King static struct proto vsock_proto = { 126d021c344SAndy King .name = "AF_VSOCK", 127d021c344SAndy King .owner = THIS_MODULE, 128d021c344SAndy King .obj_size = sizeof(struct vsock_sock), 129d021c344SAndy King }; 130d021c344SAndy King 131d021c344SAndy King /* The default peer timeout indicates how long we will wait for a peer response 132d021c344SAndy King * to a control message. 133d021c344SAndy King */ 134d021c344SAndy King #define VSOCK_DEFAULT_CONNECT_TIMEOUT (2 * HZ) 135d021c344SAndy King 136d021c344SAndy King static const struct vsock_transport *transport; 137d021c344SAndy King static DEFINE_MUTEX(vsock_register_mutex); 138d021c344SAndy King 139d021c344SAndy King /**** EXPORTS ****/ 140d021c344SAndy King 141d021c344SAndy King /* Get the ID of the local context. This is transport dependent. */ 142d021c344SAndy King 143d021c344SAndy King int vm_sockets_get_local_cid(void) 144d021c344SAndy King { 145d021c344SAndy King return transport->get_local_cid(); 146d021c344SAndy King } 147d021c344SAndy King EXPORT_SYMBOL_GPL(vm_sockets_get_local_cid); 148d021c344SAndy King 149d021c344SAndy King /**** UTILS ****/ 150d021c344SAndy King 151d021c344SAndy King /* Each bound VSocket is stored in the bind hash table and each connected 152d021c344SAndy King * VSocket is stored in the connected hash table. 153d021c344SAndy King * 154d021c344SAndy King * Unbound sockets are all put on the same list attached to the end of the hash 155d021c344SAndy King * table (vsock_unbound_sockets). Bound sockets are added to the hash table in 156d021c344SAndy King * the bucket that their local address hashes to (vsock_bound_sockets(addr) 157d021c344SAndy King * represents the list that addr hashes to). 158d021c344SAndy King * 159d021c344SAndy King * Specifically, we initialize the vsock_bind_table array to a size of 160d021c344SAndy King * VSOCK_HASH_SIZE + 1 so that vsock_bind_table[0] through 161d021c344SAndy King * vsock_bind_table[VSOCK_HASH_SIZE - 1] are for bound sockets and 162d021c344SAndy King * vsock_bind_table[VSOCK_HASH_SIZE] is for unbound sockets. The hash function 163a49dd9dcSAsias He * mods with VSOCK_HASH_SIZE to ensure this. 164d021c344SAndy King */ 165d021c344SAndy King #define MAX_PORT_RETRIES 24 166d021c344SAndy King 167a49dd9dcSAsias He #define VSOCK_HASH(addr) ((addr)->svm_port % VSOCK_HASH_SIZE) 168d021c344SAndy King #define vsock_bound_sockets(addr) (&vsock_bind_table[VSOCK_HASH(addr)]) 169d021c344SAndy King #define vsock_unbound_sockets (&vsock_bind_table[VSOCK_HASH_SIZE]) 170d021c344SAndy King 171d021c344SAndy King /* XXX This can probably be implemented in a better way. */ 172d021c344SAndy King #define VSOCK_CONN_HASH(src, dst) \ 173a49dd9dcSAsias He (((src)->svm_cid ^ (dst)->svm_port) % VSOCK_HASH_SIZE) 174d021c344SAndy King #define vsock_connected_sockets(src, dst) \ 175d021c344SAndy King (&vsock_connected_table[VSOCK_CONN_HASH(src, dst)]) 176d021c344SAndy King #define vsock_connected_sockets_vsk(vsk) \ 177d021c344SAndy King vsock_connected_sockets(&(vsk)->remote_addr, &(vsk)->local_addr) 178d021c344SAndy King 17944f20980SStefan Hajnoczi struct list_head vsock_bind_table[VSOCK_HASH_SIZE + 1]; 18044f20980SStefan Hajnoczi EXPORT_SYMBOL_GPL(vsock_bind_table); 18144f20980SStefan Hajnoczi struct list_head vsock_connected_table[VSOCK_HASH_SIZE]; 18244f20980SStefan Hajnoczi EXPORT_SYMBOL_GPL(vsock_connected_table); 18344f20980SStefan Hajnoczi DEFINE_SPINLOCK(vsock_table_lock); 18444f20980SStefan Hajnoczi EXPORT_SYMBOL_GPL(vsock_table_lock); 185d021c344SAndy King 186b3a6dfe8SAsias He /* Autobind this socket to the local address if necessary. */ 187b3a6dfe8SAsias He static int vsock_auto_bind(struct vsock_sock *vsk) 188b3a6dfe8SAsias He { 189b3a6dfe8SAsias He struct sock *sk = sk_vsock(vsk); 190b3a6dfe8SAsias He struct sockaddr_vm local_addr; 191b3a6dfe8SAsias He 192b3a6dfe8SAsias He if (vsock_addr_bound(&vsk->local_addr)) 193b3a6dfe8SAsias He return 0; 194b3a6dfe8SAsias He vsock_addr_init(&local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 195b3a6dfe8SAsias He return __vsock_bind(sk, &local_addr); 196b3a6dfe8SAsias He } 197b3a6dfe8SAsias He 198c1eef220SCong Wang static int __init vsock_init_tables(void) 199d021c344SAndy King { 200d021c344SAndy King int i; 201d021c344SAndy King 202d021c344SAndy King for (i = 0; i < ARRAY_SIZE(vsock_bind_table); i++) 203d021c344SAndy King INIT_LIST_HEAD(&vsock_bind_table[i]); 204d021c344SAndy King 205d021c344SAndy King for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) 206d021c344SAndy King INIT_LIST_HEAD(&vsock_connected_table[i]); 207c1eef220SCong Wang return 0; 208d021c344SAndy King } 209d021c344SAndy King 210d021c344SAndy King static void __vsock_insert_bound(struct list_head *list, 211d021c344SAndy King struct vsock_sock *vsk) 212d021c344SAndy King { 213d021c344SAndy King sock_hold(&vsk->sk); 214d021c344SAndy King list_add(&vsk->bound_table, list); 215d021c344SAndy King } 216d021c344SAndy King 217d021c344SAndy King static void __vsock_insert_connected(struct list_head *list, 218d021c344SAndy King struct vsock_sock *vsk) 219d021c344SAndy King { 220d021c344SAndy King sock_hold(&vsk->sk); 221d021c344SAndy King list_add(&vsk->connected_table, list); 222d021c344SAndy King } 223d021c344SAndy King 224d021c344SAndy King static void __vsock_remove_bound(struct vsock_sock *vsk) 225d021c344SAndy King { 226d021c344SAndy King list_del_init(&vsk->bound_table); 227d021c344SAndy King sock_put(&vsk->sk); 228d021c344SAndy King } 229d021c344SAndy King 230d021c344SAndy King static void __vsock_remove_connected(struct vsock_sock *vsk) 231d021c344SAndy King { 232d021c344SAndy King list_del_init(&vsk->connected_table); 233d021c344SAndy King sock_put(&vsk->sk); 234d021c344SAndy King } 235d021c344SAndy King 236d021c344SAndy King static struct sock *__vsock_find_bound_socket(struct sockaddr_vm *addr) 237d021c344SAndy King { 238d021c344SAndy King struct vsock_sock *vsk; 239d021c344SAndy King 240d021c344SAndy King list_for_each_entry(vsk, vsock_bound_sockets(addr), bound_table) 241990454b5SReilly Grant if (addr->svm_port == vsk->local_addr.svm_port) 242d021c344SAndy King return sk_vsock(vsk); 243d021c344SAndy King 244d021c344SAndy King return NULL; 245d021c344SAndy King } 246d021c344SAndy King 247d021c344SAndy King static struct sock *__vsock_find_connected_socket(struct sockaddr_vm *src, 248d021c344SAndy King struct sockaddr_vm *dst) 249d021c344SAndy King { 250d021c344SAndy King struct vsock_sock *vsk; 251d021c344SAndy King 252d021c344SAndy King list_for_each_entry(vsk, vsock_connected_sockets(src, dst), 253d021c344SAndy King connected_table) { 254990454b5SReilly Grant if (vsock_addr_equals_addr(src, &vsk->remote_addr) && 255990454b5SReilly Grant dst->svm_port == vsk->local_addr.svm_port) { 256d021c344SAndy King return sk_vsock(vsk); 257d021c344SAndy King } 258d021c344SAndy King } 259d021c344SAndy King 260d021c344SAndy King return NULL; 261d021c344SAndy King } 262d021c344SAndy King 263d021c344SAndy King static void vsock_insert_unbound(struct vsock_sock *vsk) 264d021c344SAndy King { 265d021c344SAndy King spin_lock_bh(&vsock_table_lock); 266d021c344SAndy King __vsock_insert_bound(vsock_unbound_sockets, vsk); 267d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 268d021c344SAndy King } 269d021c344SAndy King 270d021c344SAndy King void vsock_insert_connected(struct vsock_sock *vsk) 271d021c344SAndy King { 272d021c344SAndy King struct list_head *list = vsock_connected_sockets( 273d021c344SAndy King &vsk->remote_addr, &vsk->local_addr); 274d021c344SAndy King 275d021c344SAndy King spin_lock_bh(&vsock_table_lock); 276d021c344SAndy King __vsock_insert_connected(list, vsk); 277d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 278d021c344SAndy King } 279d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_insert_connected); 280d021c344SAndy King 281d021c344SAndy King void vsock_remove_bound(struct vsock_sock *vsk) 282d021c344SAndy King { 283d021c344SAndy King spin_lock_bh(&vsock_table_lock); 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); 292d021c344SAndy King __vsock_remove_connected(vsk); 293d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 294d021c344SAndy King } 295d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_remove_connected); 296d021c344SAndy King 297d021c344SAndy King struct sock *vsock_find_bound_socket(struct sockaddr_vm *addr) 298d021c344SAndy King { 299d021c344SAndy King struct sock *sk; 300d021c344SAndy King 301d021c344SAndy King spin_lock_bh(&vsock_table_lock); 302d021c344SAndy King sk = __vsock_find_bound_socket(addr); 303d021c344SAndy King if (sk) 304d021c344SAndy King sock_hold(sk); 305d021c344SAndy King 306d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 307d021c344SAndy King 308d021c344SAndy King return sk; 309d021c344SAndy King } 310d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_find_bound_socket); 311d021c344SAndy King 312d021c344SAndy King struct sock *vsock_find_connected_socket(struct sockaddr_vm *src, 313d021c344SAndy King struct sockaddr_vm *dst) 314d021c344SAndy King { 315d021c344SAndy King struct sock *sk; 316d021c344SAndy King 317d021c344SAndy King spin_lock_bh(&vsock_table_lock); 318d021c344SAndy King sk = __vsock_find_connected_socket(src, dst); 319d021c344SAndy King if (sk) 320d021c344SAndy King sock_hold(sk); 321d021c344SAndy King 322d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 323d021c344SAndy King 324d021c344SAndy King return sk; 325d021c344SAndy King } 326d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_find_connected_socket); 327d021c344SAndy King 328d021c344SAndy King static bool vsock_in_bound_table(struct vsock_sock *vsk) 329d021c344SAndy King { 330d021c344SAndy King bool ret; 331d021c344SAndy King 332d021c344SAndy King spin_lock_bh(&vsock_table_lock); 333d021c344SAndy King ret = __vsock_in_bound_table(vsk); 334d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 335d021c344SAndy King 336d021c344SAndy King return ret; 337d021c344SAndy King } 338d021c344SAndy King 339d021c344SAndy King static bool vsock_in_connected_table(struct vsock_sock *vsk) 340d021c344SAndy King { 341d021c344SAndy King bool ret; 342d021c344SAndy King 343d021c344SAndy King spin_lock_bh(&vsock_table_lock); 344d021c344SAndy King ret = __vsock_in_connected_table(vsk); 345d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 346d021c344SAndy King 347d021c344SAndy King return ret; 348d021c344SAndy King } 349d021c344SAndy King 3506773b7dcSStefan Hajnoczi void vsock_remove_sock(struct vsock_sock *vsk) 3516773b7dcSStefan Hajnoczi { 3526773b7dcSStefan Hajnoczi if (vsock_in_bound_table(vsk)) 3536773b7dcSStefan Hajnoczi vsock_remove_bound(vsk); 3546773b7dcSStefan Hajnoczi 3556773b7dcSStefan Hajnoczi if (vsock_in_connected_table(vsk)) 3566773b7dcSStefan Hajnoczi vsock_remove_connected(vsk); 3576773b7dcSStefan Hajnoczi } 3586773b7dcSStefan Hajnoczi EXPORT_SYMBOL_GPL(vsock_remove_sock); 3596773b7dcSStefan Hajnoczi 360d021c344SAndy King void vsock_for_each_connected_socket(void (*fn)(struct sock *sk)) 361d021c344SAndy King { 362d021c344SAndy King int i; 363d021c344SAndy King 364d021c344SAndy King spin_lock_bh(&vsock_table_lock); 365d021c344SAndy King 366d021c344SAndy King for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) { 367d021c344SAndy King struct vsock_sock *vsk; 368d021c344SAndy King list_for_each_entry(vsk, &vsock_connected_table[i], 369d9af2d67SJulia Lawall connected_table) 370d021c344SAndy King fn(sk_vsock(vsk)); 371d021c344SAndy King } 372d021c344SAndy King 373d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 374d021c344SAndy King } 375d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_for_each_connected_socket); 376d021c344SAndy King 377d021c344SAndy King void vsock_add_pending(struct sock *listener, struct sock *pending) 378d021c344SAndy King { 379d021c344SAndy King struct vsock_sock *vlistener; 380d021c344SAndy King struct vsock_sock *vpending; 381d021c344SAndy King 382d021c344SAndy King vlistener = vsock_sk(listener); 383d021c344SAndy King vpending = vsock_sk(pending); 384d021c344SAndy King 385d021c344SAndy King sock_hold(pending); 386d021c344SAndy King sock_hold(listener); 387d021c344SAndy King list_add_tail(&vpending->pending_links, &vlistener->pending_links); 388d021c344SAndy King } 389d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_add_pending); 390d021c344SAndy King 391d021c344SAndy King void vsock_remove_pending(struct sock *listener, struct sock *pending) 392d021c344SAndy King { 393d021c344SAndy King struct vsock_sock *vpending = vsock_sk(pending); 394d021c344SAndy King 395d021c344SAndy King list_del_init(&vpending->pending_links); 396d021c344SAndy King sock_put(listener); 397d021c344SAndy King sock_put(pending); 398d021c344SAndy King } 399d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_remove_pending); 400d021c344SAndy King 401d021c344SAndy King void vsock_enqueue_accept(struct sock *listener, struct sock *connected) 402d021c344SAndy King { 403d021c344SAndy King struct vsock_sock *vlistener; 404d021c344SAndy King struct vsock_sock *vconnected; 405d021c344SAndy King 406d021c344SAndy King vlistener = vsock_sk(listener); 407d021c344SAndy King vconnected = vsock_sk(connected); 408d021c344SAndy King 409d021c344SAndy King sock_hold(connected); 410d021c344SAndy King sock_hold(listener); 411d021c344SAndy King list_add_tail(&vconnected->accept_queue, &vlistener->accept_queue); 412d021c344SAndy King } 413d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_enqueue_accept); 414d021c344SAndy King 415d021c344SAndy King static struct sock *vsock_dequeue_accept(struct sock *listener) 416d021c344SAndy King { 417d021c344SAndy King struct vsock_sock *vlistener; 418d021c344SAndy King struct vsock_sock *vconnected; 419d021c344SAndy King 420d021c344SAndy King vlistener = vsock_sk(listener); 421d021c344SAndy King 422d021c344SAndy King if (list_empty(&vlistener->accept_queue)) 423d021c344SAndy King return NULL; 424d021c344SAndy King 425d021c344SAndy King vconnected = list_entry(vlistener->accept_queue.next, 426d021c344SAndy King struct vsock_sock, accept_queue); 427d021c344SAndy King 428d021c344SAndy King list_del_init(&vconnected->accept_queue); 429d021c344SAndy King sock_put(listener); 430d021c344SAndy King /* The caller will need a reference on the connected socket so we let 431d021c344SAndy King * it call sock_put(). 432d021c344SAndy King */ 433d021c344SAndy King 434d021c344SAndy King return sk_vsock(vconnected); 435d021c344SAndy King } 436d021c344SAndy King 437d021c344SAndy King static bool vsock_is_accept_queue_empty(struct sock *sk) 438d021c344SAndy King { 439d021c344SAndy King struct vsock_sock *vsk = vsock_sk(sk); 440d021c344SAndy King return list_empty(&vsk->accept_queue); 441d021c344SAndy King } 442d021c344SAndy King 443d021c344SAndy King static bool vsock_is_pending(struct sock *sk) 444d021c344SAndy King { 445d021c344SAndy King struct vsock_sock *vsk = vsock_sk(sk); 446d021c344SAndy King return !list_empty(&vsk->pending_links); 447d021c344SAndy King } 448d021c344SAndy King 449d021c344SAndy King static int vsock_send_shutdown(struct sock *sk, int mode) 450d021c344SAndy King { 451d021c344SAndy King return transport->shutdown(vsock_sk(sk), mode); 452d021c344SAndy King } 453d021c344SAndy King 454d021c344SAndy King void vsock_pending_work(struct work_struct *work) 455d021c344SAndy King { 456d021c344SAndy King struct sock *sk; 457d021c344SAndy King struct sock *listener; 458d021c344SAndy King struct vsock_sock *vsk; 459d021c344SAndy King bool cleanup; 460d021c344SAndy King 461d021c344SAndy King vsk = container_of(work, struct vsock_sock, dwork.work); 462d021c344SAndy King sk = sk_vsock(vsk); 463d021c344SAndy King listener = vsk->listener; 464d021c344SAndy King cleanup = true; 465d021c344SAndy King 466d021c344SAndy King lock_sock(listener); 4674192f672SStefan Hajnoczi lock_sock_nested(sk, SINGLE_DEPTH_NESTING); 468d021c344SAndy King 469d021c344SAndy King if (vsock_is_pending(sk)) { 470d021c344SAndy King vsock_remove_pending(listener, sk); 4711190cfdbSJorgen Hansen 4721190cfdbSJorgen Hansen listener->sk_ack_backlog--; 473d021c344SAndy King } else if (!vsk->rejected) { 474d021c344SAndy King /* We are not on the pending list and accept() did not reject 475d021c344SAndy King * us, so we must have been accepted by our user process. We 476d021c344SAndy King * just need to drop our references to the sockets and be on 477d021c344SAndy King * our way. 478d021c344SAndy King */ 479d021c344SAndy King cleanup = false; 480d021c344SAndy King goto out; 481d021c344SAndy King } 482d021c344SAndy King 483d021c344SAndy King /* We need to remove ourself from the global connected sockets list so 484d021c344SAndy King * incoming packets can't find this socket, and to reduce the reference 485d021c344SAndy King * count. 486d021c344SAndy King */ 487d021c344SAndy King if (vsock_in_connected_table(vsk)) 488d021c344SAndy King vsock_remove_connected(vsk); 489d021c344SAndy King 4903b4477d2SStefan Hajnoczi sk->sk_state = TCP_CLOSE; 491d021c344SAndy King 492d021c344SAndy King out: 493d021c344SAndy King release_sock(sk); 494d021c344SAndy King release_sock(listener); 495d021c344SAndy King if (cleanup) 496d021c344SAndy King sock_put(sk); 497d021c344SAndy King 498d021c344SAndy King sock_put(sk); 499d021c344SAndy King sock_put(listener); 500d021c344SAndy King } 501d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_pending_work); 502d021c344SAndy King 503d021c344SAndy King /**** SOCKET OPERATIONS ****/ 504d021c344SAndy King 505d021c344SAndy King static int __vsock_bind_stream(struct vsock_sock *vsk, 506d021c344SAndy King struct sockaddr_vm *addr) 507d021c344SAndy King { 508d021c344SAndy King static u32 port = LAST_RESERVED_PORT + 1; 509d021c344SAndy King struct sockaddr_vm new_addr; 510d021c344SAndy King 511d021c344SAndy King vsock_addr_init(&new_addr, addr->svm_cid, addr->svm_port); 512d021c344SAndy King 513d021c344SAndy King if (addr->svm_port == VMADDR_PORT_ANY) { 514d021c344SAndy King bool found = false; 515d021c344SAndy King unsigned int i; 516d021c344SAndy King 517d021c344SAndy King for (i = 0; i < MAX_PORT_RETRIES; i++) { 518d021c344SAndy King if (port <= LAST_RESERVED_PORT) 519d021c344SAndy King port = LAST_RESERVED_PORT + 1; 520d021c344SAndy King 521d021c344SAndy King new_addr.svm_port = port++; 522d021c344SAndy King 523d021c344SAndy King if (!__vsock_find_bound_socket(&new_addr)) { 524d021c344SAndy King found = true; 525d021c344SAndy King break; 526d021c344SAndy King } 527d021c344SAndy King } 528d021c344SAndy King 529d021c344SAndy King if (!found) 530d021c344SAndy King return -EADDRNOTAVAIL; 531d021c344SAndy King } else { 532d021c344SAndy King /* If port is in reserved range, ensure caller 533d021c344SAndy King * has necessary privileges. 534d021c344SAndy King */ 535d021c344SAndy King if (addr->svm_port <= LAST_RESERVED_PORT && 536d021c344SAndy King !capable(CAP_NET_BIND_SERVICE)) { 537d021c344SAndy King return -EACCES; 538d021c344SAndy King } 539d021c344SAndy King 540d021c344SAndy King if (__vsock_find_bound_socket(&new_addr)) 541d021c344SAndy King return -EADDRINUSE; 542d021c344SAndy King } 543d021c344SAndy King 544d021c344SAndy King vsock_addr_init(&vsk->local_addr, new_addr.svm_cid, new_addr.svm_port); 545d021c344SAndy King 546d021c344SAndy King /* Remove stream sockets from the unbound list and add them to the hash 547d021c344SAndy King * table for easy lookup by its address. The unbound list is simply an 548d021c344SAndy King * extra entry at the end of the hash table, a trick used by AF_UNIX. 549d021c344SAndy King */ 550d021c344SAndy King __vsock_remove_bound(vsk); 551d021c344SAndy King __vsock_insert_bound(vsock_bound_sockets(&vsk->local_addr), vsk); 552d021c344SAndy King 553d021c344SAndy King return 0; 554d021c344SAndy King } 555d021c344SAndy King 556d021c344SAndy King static int __vsock_bind_dgram(struct vsock_sock *vsk, 557d021c344SAndy King struct sockaddr_vm *addr) 558d021c344SAndy King { 559d021c344SAndy King return transport->dgram_bind(vsk, addr); 560d021c344SAndy King } 561d021c344SAndy King 562d021c344SAndy King static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr) 563d021c344SAndy King { 564d021c344SAndy King struct vsock_sock *vsk = vsock_sk(sk); 565d021c344SAndy King u32 cid; 566d021c344SAndy King int retval; 567d021c344SAndy King 568d021c344SAndy King /* First ensure this socket isn't already bound. */ 569d021c344SAndy King if (vsock_addr_bound(&vsk->local_addr)) 570d021c344SAndy King return -EINVAL; 571d021c344SAndy King 572d021c344SAndy King /* Now bind to the provided address or select appropriate values if 573d021c344SAndy King * none are provided (VMADDR_CID_ANY and VMADDR_PORT_ANY). Note that 574d021c344SAndy King * like AF_INET prevents binding to a non-local IP address (in most 575d021c344SAndy King * cases), we only allow binding to the local CID. 576d021c344SAndy King */ 577d021c344SAndy King cid = transport->get_local_cid(); 578d021c344SAndy King if (addr->svm_cid != cid && addr->svm_cid != VMADDR_CID_ANY) 579d021c344SAndy King return -EADDRNOTAVAIL; 580d021c344SAndy King 581d021c344SAndy King switch (sk->sk_socket->type) { 582d021c344SAndy King case SOCK_STREAM: 583d021c344SAndy King spin_lock_bh(&vsock_table_lock); 584d021c344SAndy King retval = __vsock_bind_stream(vsk, addr); 585d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 586d021c344SAndy King break; 587d021c344SAndy King 588d021c344SAndy King case SOCK_DGRAM: 589d021c344SAndy King retval = __vsock_bind_dgram(vsk, addr); 590d021c344SAndy King break; 591d021c344SAndy King 592d021c344SAndy King default: 593d021c344SAndy King retval = -EINVAL; 594d021c344SAndy King break; 595d021c344SAndy King } 596d021c344SAndy King 597d021c344SAndy King return retval; 598d021c344SAndy King } 599d021c344SAndy King 600d021c344SAndy King struct sock *__vsock_create(struct net *net, 601d021c344SAndy King struct socket *sock, 602d021c344SAndy King struct sock *parent, 603d021c344SAndy King gfp_t priority, 60411aa9c28SEric W. Biederman unsigned short type, 60511aa9c28SEric W. Biederman int kern) 606d021c344SAndy King { 607d021c344SAndy King struct sock *sk; 608d021c344SAndy King struct vsock_sock *psk; 609d021c344SAndy King struct vsock_sock *vsk; 610d021c344SAndy King 61111aa9c28SEric W. Biederman sk = sk_alloc(net, AF_VSOCK, priority, &vsock_proto, kern); 612d021c344SAndy King if (!sk) 613d021c344SAndy King return NULL; 614d021c344SAndy King 615d021c344SAndy King sock_init_data(sock, sk); 616d021c344SAndy King 617d021c344SAndy King /* sk->sk_type is normally set in sock_init_data, but only if sock is 618d021c344SAndy King * non-NULL. We make sure that our sockets always have a type by 619d021c344SAndy King * setting it here if needed. 620d021c344SAndy King */ 621d021c344SAndy King if (!sock) 622d021c344SAndy King sk->sk_type = type; 623d021c344SAndy King 624d021c344SAndy King vsk = vsock_sk(sk); 625d021c344SAndy King vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 626d021c344SAndy King vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 627d021c344SAndy King 628d021c344SAndy King sk->sk_destruct = vsock_sk_destruct; 629d021c344SAndy King sk->sk_backlog_rcv = vsock_queue_rcv_skb; 630d021c344SAndy King sock_reset_flag(sk, SOCK_DONE); 631d021c344SAndy King 632d021c344SAndy King INIT_LIST_HEAD(&vsk->bound_table); 633d021c344SAndy King INIT_LIST_HEAD(&vsk->connected_table); 634d021c344SAndy King vsk->listener = NULL; 635d021c344SAndy King INIT_LIST_HEAD(&vsk->pending_links); 636d021c344SAndy King INIT_LIST_HEAD(&vsk->accept_queue); 637d021c344SAndy King vsk->rejected = false; 638d021c344SAndy King vsk->sent_request = false; 639d021c344SAndy King vsk->ignore_connecting_rst = false; 640d021c344SAndy King vsk->peer_shutdown = 0; 641d021c344SAndy King 642d021c344SAndy King psk = parent ? vsock_sk(parent) : NULL; 643d021c344SAndy King if (parent) { 644d021c344SAndy King vsk->trusted = psk->trusted; 645d021c344SAndy King vsk->owner = get_cred(psk->owner); 646d021c344SAndy King vsk->connect_timeout = psk->connect_timeout; 647d021c344SAndy King } else { 648d021c344SAndy King vsk->trusted = capable(CAP_NET_ADMIN); 649d021c344SAndy King vsk->owner = get_current_cred(); 650d021c344SAndy King vsk->connect_timeout = VSOCK_DEFAULT_CONNECT_TIMEOUT; 651d021c344SAndy King } 652d021c344SAndy King 653d021c344SAndy King if (transport->init(vsk, psk) < 0) { 654d021c344SAndy King sk_free(sk); 655d021c344SAndy King return NULL; 656d021c344SAndy King } 657d021c344SAndy King 658d021c344SAndy King if (sock) 659d021c344SAndy King vsock_insert_unbound(vsk); 660d021c344SAndy King 661d021c344SAndy King return sk; 662d021c344SAndy King } 663d021c344SAndy King EXPORT_SYMBOL_GPL(__vsock_create); 664d021c344SAndy King 665d021c344SAndy King static void __vsock_release(struct sock *sk) 666d021c344SAndy King { 667d021c344SAndy King if (sk) { 668d021c344SAndy King struct sk_buff *skb; 669d021c344SAndy King struct sock *pending; 670d021c344SAndy King struct vsock_sock *vsk; 671d021c344SAndy King 672d021c344SAndy King vsk = vsock_sk(sk); 673d021c344SAndy King pending = NULL; /* Compiler warning. */ 674d021c344SAndy King 675d021c344SAndy King transport->release(vsk); 676d021c344SAndy King 677d021c344SAndy King lock_sock(sk); 678d021c344SAndy King sock_orphan(sk); 679d021c344SAndy King sk->sk_shutdown = SHUTDOWN_MASK; 680d021c344SAndy King 681d021c344SAndy King while ((skb = skb_dequeue(&sk->sk_receive_queue))) 682d021c344SAndy King kfree_skb(skb); 683d021c344SAndy King 684d021c344SAndy King /* Clean up any sockets that never were accepted. */ 685d021c344SAndy King while ((pending = vsock_dequeue_accept(sk)) != NULL) { 686d021c344SAndy King __vsock_release(pending); 687d021c344SAndy King sock_put(pending); 688d021c344SAndy King } 689d021c344SAndy King 690d021c344SAndy King release_sock(sk); 691d021c344SAndy King sock_put(sk); 692d021c344SAndy King } 693d021c344SAndy King } 694d021c344SAndy King 695d021c344SAndy King static void vsock_sk_destruct(struct sock *sk) 696d021c344SAndy King { 697d021c344SAndy King struct vsock_sock *vsk = vsock_sk(sk); 698d021c344SAndy King 699d021c344SAndy King transport->destruct(vsk); 700d021c344SAndy King 701d021c344SAndy King /* When clearing these addresses, there's no need to set the family and 702d021c344SAndy King * possibly register the address family with the kernel. 703d021c344SAndy King */ 704d021c344SAndy King vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 705d021c344SAndy King vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 706d021c344SAndy King 707d021c344SAndy King put_cred(vsk->owner); 708d021c344SAndy King } 709d021c344SAndy King 710d021c344SAndy King static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) 711d021c344SAndy King { 712d021c344SAndy King int err; 713d021c344SAndy King 714d021c344SAndy King err = sock_queue_rcv_skb(sk, skb); 715d021c344SAndy King if (err) 716d021c344SAndy King kfree_skb(skb); 717d021c344SAndy King 718d021c344SAndy King return err; 719d021c344SAndy King } 720d021c344SAndy King 721d021c344SAndy King s64 vsock_stream_has_data(struct vsock_sock *vsk) 722d021c344SAndy King { 723d021c344SAndy King return transport->stream_has_data(vsk); 724d021c344SAndy King } 725d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_stream_has_data); 726d021c344SAndy King 727d021c344SAndy King s64 vsock_stream_has_space(struct vsock_sock *vsk) 728d021c344SAndy King { 729d021c344SAndy King return transport->stream_has_space(vsk); 730d021c344SAndy King } 731d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_stream_has_space); 732d021c344SAndy King 733d021c344SAndy King static int vsock_release(struct socket *sock) 734d021c344SAndy King { 735d021c344SAndy King __vsock_release(sock->sk); 736d021c344SAndy King sock->sk = NULL; 737d021c344SAndy King sock->state = SS_FREE; 738d021c344SAndy King 739d021c344SAndy King return 0; 740d021c344SAndy King } 741d021c344SAndy King 742d021c344SAndy King static int 743d021c344SAndy King vsock_bind(struct socket *sock, struct sockaddr *addr, int addr_len) 744d021c344SAndy King { 745d021c344SAndy King int err; 746d021c344SAndy King struct sock *sk; 747d021c344SAndy King struct sockaddr_vm *vm_addr; 748d021c344SAndy King 749d021c344SAndy King sk = sock->sk; 750d021c344SAndy King 751d021c344SAndy King if (vsock_addr_cast(addr, addr_len, &vm_addr) != 0) 752d021c344SAndy King return -EINVAL; 753d021c344SAndy King 754d021c344SAndy King lock_sock(sk); 755d021c344SAndy King err = __vsock_bind(sk, vm_addr); 756d021c344SAndy King release_sock(sk); 757d021c344SAndy King 758d021c344SAndy King return err; 759d021c344SAndy King } 760d021c344SAndy King 761d021c344SAndy King static int vsock_getname(struct socket *sock, 762d021c344SAndy King struct sockaddr *addr, int *addr_len, int peer) 763d021c344SAndy King { 764d021c344SAndy King int err; 765d021c344SAndy King struct sock *sk; 766d021c344SAndy King struct vsock_sock *vsk; 767d021c344SAndy King struct sockaddr_vm *vm_addr; 768d021c344SAndy King 769d021c344SAndy King sk = sock->sk; 770d021c344SAndy King vsk = vsock_sk(sk); 771d021c344SAndy King err = 0; 772d021c344SAndy King 773d021c344SAndy King lock_sock(sk); 774d021c344SAndy King 775d021c344SAndy King if (peer) { 776d021c344SAndy King if (sock->state != SS_CONNECTED) { 777d021c344SAndy King err = -ENOTCONN; 778d021c344SAndy King goto out; 779d021c344SAndy King } 780d021c344SAndy King vm_addr = &vsk->remote_addr; 781d021c344SAndy King } else { 782d021c344SAndy King vm_addr = &vsk->local_addr; 783d021c344SAndy King } 784d021c344SAndy King 785d021c344SAndy King if (!vm_addr) { 786d021c344SAndy King err = -EINVAL; 787d021c344SAndy King goto out; 788d021c344SAndy King } 789d021c344SAndy King 790d021c344SAndy King /* sys_getsockname() and sys_getpeername() pass us a 791d021c344SAndy King * MAX_SOCK_ADDR-sized buffer and don't set addr_len. Unfortunately 792d021c344SAndy King * that macro is defined in socket.c instead of .h, so we hardcode its 793d021c344SAndy King * value here. 794d021c344SAndy King */ 795d021c344SAndy King BUILD_BUG_ON(sizeof(*vm_addr) > 128); 796d021c344SAndy King memcpy(addr, vm_addr, sizeof(*vm_addr)); 797d021c344SAndy King *addr_len = sizeof(*vm_addr); 798d021c344SAndy King 799d021c344SAndy King out: 800d021c344SAndy King release_sock(sk); 801d021c344SAndy King return err; 802d021c344SAndy King } 803d021c344SAndy King 804d021c344SAndy King static int vsock_shutdown(struct socket *sock, int mode) 805d021c344SAndy King { 806d021c344SAndy King int err; 807d021c344SAndy King struct sock *sk; 808d021c344SAndy King 809d021c344SAndy King /* User level uses SHUT_RD (0) and SHUT_WR (1), but the kernel uses 810d021c344SAndy King * RCV_SHUTDOWN (1) and SEND_SHUTDOWN (2), so we must increment mode 811d021c344SAndy King * here like the other address families do. Note also that the 812d021c344SAndy King * increment makes SHUT_RDWR (2) into RCV_SHUTDOWN | SEND_SHUTDOWN (3), 813d021c344SAndy King * which is what we want. 814d021c344SAndy King */ 815d021c344SAndy King mode++; 816d021c344SAndy King 817d021c344SAndy King if ((mode & ~SHUTDOWN_MASK) || !mode) 818d021c344SAndy King return -EINVAL; 819d021c344SAndy King 820d021c344SAndy King /* If this is a STREAM socket and it is not connected then bail out 821d021c344SAndy King * immediately. If it is a DGRAM socket then we must first kick the 822d021c344SAndy King * socket so that it wakes up from any sleeping calls, for example 823d021c344SAndy King * recv(), and then afterwards return the error. 824d021c344SAndy King */ 825d021c344SAndy King 826d021c344SAndy King sk = sock->sk; 827d021c344SAndy King if (sock->state == SS_UNCONNECTED) { 828d021c344SAndy King err = -ENOTCONN; 829d021c344SAndy King if (sk->sk_type == SOCK_STREAM) 830d021c344SAndy King return err; 831d021c344SAndy King } else { 832d021c344SAndy King sock->state = SS_DISCONNECTING; 833d021c344SAndy King err = 0; 834d021c344SAndy King } 835d021c344SAndy King 836d021c344SAndy King /* Receive and send shutdowns are treated alike. */ 837d021c344SAndy King mode = mode & (RCV_SHUTDOWN | SEND_SHUTDOWN); 838d021c344SAndy King if (mode) { 839d021c344SAndy King lock_sock(sk); 840d021c344SAndy King sk->sk_shutdown |= mode; 841d021c344SAndy King sk->sk_state_change(sk); 842d021c344SAndy King release_sock(sk); 843d021c344SAndy King 844d021c344SAndy King if (sk->sk_type == SOCK_STREAM) { 845d021c344SAndy King sock_reset_flag(sk, SOCK_DONE); 846d021c344SAndy King vsock_send_shutdown(sk, mode); 847d021c344SAndy King } 848d021c344SAndy King } 849d021c344SAndy King 850d021c344SAndy King return err; 851d021c344SAndy King } 852d021c344SAndy King 853d021c344SAndy King static unsigned int vsock_poll(struct file *file, struct socket *sock, 854d021c344SAndy King poll_table *wait) 855d021c344SAndy King { 856d021c344SAndy King struct sock *sk; 857d021c344SAndy King unsigned int mask; 858d021c344SAndy King struct vsock_sock *vsk; 859d021c344SAndy King 860d021c344SAndy King sk = sock->sk; 861d021c344SAndy King vsk = vsock_sk(sk); 862d021c344SAndy King 863d021c344SAndy King poll_wait(file, sk_sleep(sk), wait); 864d021c344SAndy King mask = 0; 865d021c344SAndy King 866d021c344SAndy King if (sk->sk_err) 867d021c344SAndy King /* Signify that there has been an error on this socket. */ 868d021c344SAndy King mask |= POLLERR; 869d021c344SAndy King 870d021c344SAndy King /* INET sockets treat local write shutdown and peer write shutdown as a 871d021c344SAndy King * case of POLLHUP set. 872d021c344SAndy King */ 873d021c344SAndy King if ((sk->sk_shutdown == SHUTDOWN_MASK) || 874d021c344SAndy King ((sk->sk_shutdown & SEND_SHUTDOWN) && 875d021c344SAndy King (vsk->peer_shutdown & SEND_SHUTDOWN))) { 876d021c344SAndy King mask |= POLLHUP; 877d021c344SAndy King } 878d021c344SAndy King 879d021c344SAndy King if (sk->sk_shutdown & RCV_SHUTDOWN || 880d021c344SAndy King vsk->peer_shutdown & SEND_SHUTDOWN) { 881d021c344SAndy King mask |= POLLRDHUP; 882d021c344SAndy King } 883d021c344SAndy King 884d021c344SAndy King if (sock->type == SOCK_DGRAM) { 885d021c344SAndy King /* For datagram sockets we can read if there is something in 886d021c344SAndy King * the queue and write as long as the socket isn't shutdown for 887d021c344SAndy King * sending. 888d021c344SAndy King */ 889d021c344SAndy King if (!skb_queue_empty(&sk->sk_receive_queue) || 890d021c344SAndy King (sk->sk_shutdown & RCV_SHUTDOWN)) { 891d021c344SAndy King mask |= POLLIN | POLLRDNORM; 892d021c344SAndy King } 893d021c344SAndy King 894d021c344SAndy King if (!(sk->sk_shutdown & SEND_SHUTDOWN)) 895d021c344SAndy King mask |= POLLOUT | POLLWRNORM | POLLWRBAND; 896d021c344SAndy King 897d021c344SAndy King } else if (sock->type == SOCK_STREAM) { 898d021c344SAndy King lock_sock(sk); 899d021c344SAndy King 900d021c344SAndy King /* Listening sockets that have connections in their accept 901d021c344SAndy King * queue can be read. 902d021c344SAndy King */ 9033b4477d2SStefan Hajnoczi if (sk->sk_state == TCP_LISTEN 904d021c344SAndy King && !vsock_is_accept_queue_empty(sk)) 905d021c344SAndy King mask |= POLLIN | POLLRDNORM; 906d021c344SAndy King 907d021c344SAndy King /* If there is something in the queue then we can read. */ 908d021c344SAndy King if (transport->stream_is_active(vsk) && 909d021c344SAndy King !(sk->sk_shutdown & RCV_SHUTDOWN)) { 910d021c344SAndy King bool data_ready_now = false; 911d021c344SAndy King int ret = transport->notify_poll_in( 912d021c344SAndy King vsk, 1, &data_ready_now); 913d021c344SAndy King if (ret < 0) { 914d021c344SAndy King mask |= POLLERR; 915d021c344SAndy King } else { 916d021c344SAndy King if (data_ready_now) 917d021c344SAndy King mask |= POLLIN | POLLRDNORM; 918d021c344SAndy King 919d021c344SAndy King } 920d021c344SAndy King } 921d021c344SAndy King 922d021c344SAndy King /* Sockets whose connections have been closed, reset, or 923d021c344SAndy King * terminated should also be considered read, and we check the 924d021c344SAndy King * shutdown flag for that. 925d021c344SAndy King */ 926d021c344SAndy King if (sk->sk_shutdown & RCV_SHUTDOWN || 927d021c344SAndy King vsk->peer_shutdown & SEND_SHUTDOWN) { 928d021c344SAndy King mask |= POLLIN | POLLRDNORM; 929d021c344SAndy King } 930d021c344SAndy King 931d021c344SAndy King /* Connected sockets that can produce data can be written. */ 9323b4477d2SStefan Hajnoczi if (sk->sk_state == TCP_ESTABLISHED) { 933d021c344SAndy King if (!(sk->sk_shutdown & SEND_SHUTDOWN)) { 934d021c344SAndy King bool space_avail_now = false; 935d021c344SAndy King int ret = transport->notify_poll_out( 936d021c344SAndy King vsk, 1, &space_avail_now); 937d021c344SAndy King if (ret < 0) { 938d021c344SAndy King mask |= POLLERR; 939d021c344SAndy King } else { 940d021c344SAndy King if (space_avail_now) 941d021c344SAndy King /* Remove POLLWRBAND since INET 942d021c344SAndy King * sockets are not setting it. 943d021c344SAndy King */ 944d021c344SAndy King mask |= POLLOUT | POLLWRNORM; 945d021c344SAndy King 946d021c344SAndy King } 947d021c344SAndy King } 948d021c344SAndy King } 949d021c344SAndy King 950d021c344SAndy King /* Simulate INET socket poll behaviors, which sets 951d021c344SAndy King * POLLOUT|POLLWRNORM when peer is closed and nothing to read, 952d021c344SAndy King * but local send is not shutdown. 953d021c344SAndy King */ 954ba3169fcSStefan Hajnoczi if (sk->sk_state == TCP_CLOSE || sk->sk_state == TCP_CLOSING) { 955d021c344SAndy King if (!(sk->sk_shutdown & SEND_SHUTDOWN)) 956d021c344SAndy King mask |= POLLOUT | POLLWRNORM; 957d021c344SAndy King 958d021c344SAndy King } 959d021c344SAndy King 960d021c344SAndy King release_sock(sk); 961d021c344SAndy King } 962d021c344SAndy King 963d021c344SAndy King return mask; 964d021c344SAndy King } 965d021c344SAndy King 9661b784140SYing Xue static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg, 9671b784140SYing Xue size_t len) 968d021c344SAndy King { 969d021c344SAndy King int err; 970d021c344SAndy King struct sock *sk; 971d021c344SAndy King struct vsock_sock *vsk; 972d021c344SAndy King struct sockaddr_vm *remote_addr; 973d021c344SAndy King 974d021c344SAndy King if (msg->msg_flags & MSG_OOB) 975d021c344SAndy King return -EOPNOTSUPP; 976d021c344SAndy King 977d021c344SAndy King /* For now, MSG_DONTWAIT is always assumed... */ 978d021c344SAndy King err = 0; 979d021c344SAndy King sk = sock->sk; 980d021c344SAndy King vsk = vsock_sk(sk); 981d021c344SAndy King 982d021c344SAndy King lock_sock(sk); 983d021c344SAndy King 984b3a6dfe8SAsias He err = vsock_auto_bind(vsk); 985b3a6dfe8SAsias He if (err) 986d021c344SAndy King goto out; 987d021c344SAndy King 988d021c344SAndy King 989d021c344SAndy King /* If the provided message contains an address, use that. Otherwise 990d021c344SAndy King * fall back on the socket's remote handle (if it has been connected). 991d021c344SAndy King */ 992d021c344SAndy King if (msg->msg_name && 993d021c344SAndy King vsock_addr_cast(msg->msg_name, msg->msg_namelen, 994d021c344SAndy King &remote_addr) == 0) { 995d021c344SAndy King /* Ensure this address is of the right type and is a valid 996d021c344SAndy King * destination. 997d021c344SAndy King */ 998d021c344SAndy King 999d021c344SAndy King if (remote_addr->svm_cid == VMADDR_CID_ANY) 1000d021c344SAndy King remote_addr->svm_cid = transport->get_local_cid(); 1001d021c344SAndy King 1002d021c344SAndy King if (!vsock_addr_bound(remote_addr)) { 1003d021c344SAndy King err = -EINVAL; 1004d021c344SAndy King goto out; 1005d021c344SAndy King } 1006d021c344SAndy King } else if (sock->state == SS_CONNECTED) { 1007d021c344SAndy King remote_addr = &vsk->remote_addr; 1008d021c344SAndy King 1009d021c344SAndy King if (remote_addr->svm_cid == VMADDR_CID_ANY) 1010d021c344SAndy King remote_addr->svm_cid = transport->get_local_cid(); 1011d021c344SAndy King 1012d021c344SAndy King /* XXX Should connect() or this function ensure remote_addr is 1013d021c344SAndy King * bound? 1014d021c344SAndy King */ 1015d021c344SAndy King if (!vsock_addr_bound(&vsk->remote_addr)) { 1016d021c344SAndy King err = -EINVAL; 1017d021c344SAndy King goto out; 1018d021c344SAndy King } 1019d021c344SAndy King } else { 1020d021c344SAndy King err = -EINVAL; 1021d021c344SAndy King goto out; 1022d021c344SAndy King } 1023d021c344SAndy King 1024d021c344SAndy King if (!transport->dgram_allow(remote_addr->svm_cid, 1025d021c344SAndy King remote_addr->svm_port)) { 1026d021c344SAndy King err = -EINVAL; 1027d021c344SAndy King goto out; 1028d021c344SAndy King } 1029d021c344SAndy King 10300f7db23aSAl Viro err = transport->dgram_enqueue(vsk, remote_addr, msg, len); 1031d021c344SAndy King 1032d021c344SAndy King out: 1033d021c344SAndy King release_sock(sk); 1034d021c344SAndy King return err; 1035d021c344SAndy King } 1036d021c344SAndy King 1037d021c344SAndy King static int vsock_dgram_connect(struct socket *sock, 1038d021c344SAndy King struct sockaddr *addr, int addr_len, int flags) 1039d021c344SAndy King { 1040d021c344SAndy King int err; 1041d021c344SAndy King struct sock *sk; 1042d021c344SAndy King struct vsock_sock *vsk; 1043d021c344SAndy King struct sockaddr_vm *remote_addr; 1044d021c344SAndy King 1045d021c344SAndy King sk = sock->sk; 1046d021c344SAndy King vsk = vsock_sk(sk); 1047d021c344SAndy King 1048d021c344SAndy King err = vsock_addr_cast(addr, addr_len, &remote_addr); 1049d021c344SAndy King if (err == -EAFNOSUPPORT && remote_addr->svm_family == AF_UNSPEC) { 1050d021c344SAndy King lock_sock(sk); 1051d021c344SAndy King vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, 1052d021c344SAndy King VMADDR_PORT_ANY); 1053d021c344SAndy King sock->state = SS_UNCONNECTED; 1054d021c344SAndy King release_sock(sk); 1055d021c344SAndy King return 0; 1056d021c344SAndy King } else if (err != 0) 1057d021c344SAndy King return -EINVAL; 1058d021c344SAndy King 1059d021c344SAndy King lock_sock(sk); 1060d021c344SAndy King 1061b3a6dfe8SAsias He err = vsock_auto_bind(vsk); 1062b3a6dfe8SAsias He if (err) 1063d021c344SAndy King goto out; 1064d021c344SAndy King 1065d021c344SAndy King if (!transport->dgram_allow(remote_addr->svm_cid, 1066d021c344SAndy King remote_addr->svm_port)) { 1067d021c344SAndy King err = -EINVAL; 1068d021c344SAndy King goto out; 1069d021c344SAndy King } 1070d021c344SAndy King 1071d021c344SAndy King memcpy(&vsk->remote_addr, remote_addr, sizeof(vsk->remote_addr)); 1072d021c344SAndy King sock->state = SS_CONNECTED; 1073d021c344SAndy King 1074d021c344SAndy King out: 1075d021c344SAndy King release_sock(sk); 1076d021c344SAndy King return err; 1077d021c344SAndy King } 1078d021c344SAndy King 10791b784140SYing Xue static int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg, 10801b784140SYing Xue size_t len, int flags) 1081d021c344SAndy King { 10821b784140SYing Xue return transport->dgram_dequeue(vsock_sk(sock->sk), msg, len, flags); 1083d021c344SAndy King } 1084d021c344SAndy King 1085d021c344SAndy King static const struct proto_ops vsock_dgram_ops = { 1086d021c344SAndy King .family = PF_VSOCK, 1087d021c344SAndy King .owner = THIS_MODULE, 1088d021c344SAndy King .release = vsock_release, 1089d021c344SAndy King .bind = vsock_bind, 1090d021c344SAndy King .connect = vsock_dgram_connect, 1091d021c344SAndy King .socketpair = sock_no_socketpair, 1092d021c344SAndy King .accept = sock_no_accept, 1093d021c344SAndy King .getname = vsock_getname, 1094d021c344SAndy King .poll = vsock_poll, 1095d021c344SAndy King .ioctl = sock_no_ioctl, 1096d021c344SAndy King .listen = sock_no_listen, 1097d021c344SAndy King .shutdown = vsock_shutdown, 1098d021c344SAndy King .setsockopt = sock_no_setsockopt, 1099d021c344SAndy King .getsockopt = sock_no_getsockopt, 1100d021c344SAndy King .sendmsg = vsock_dgram_sendmsg, 1101d021c344SAndy King .recvmsg = vsock_dgram_recvmsg, 1102d021c344SAndy King .mmap = sock_no_mmap, 1103d021c344SAndy King .sendpage = sock_no_sendpage, 1104d021c344SAndy King }; 1105d021c344SAndy King 1106380feae0SPeng Tao static int vsock_transport_cancel_pkt(struct vsock_sock *vsk) 1107380feae0SPeng Tao { 1108380feae0SPeng Tao if (!transport->cancel_pkt) 1109380feae0SPeng Tao return -EOPNOTSUPP; 1110380feae0SPeng Tao 1111380feae0SPeng Tao return transport->cancel_pkt(vsk); 1112380feae0SPeng Tao } 1113380feae0SPeng Tao 1114d021c344SAndy King static void vsock_connect_timeout(struct work_struct *work) 1115d021c344SAndy King { 1116d021c344SAndy King struct sock *sk; 1117d021c344SAndy King struct vsock_sock *vsk; 1118380feae0SPeng Tao int cancel = 0; 1119d021c344SAndy King 1120d021c344SAndy King vsk = container_of(work, struct vsock_sock, dwork.work); 1121d021c344SAndy King sk = sk_vsock(vsk); 1122d021c344SAndy King 1123d021c344SAndy King lock_sock(sk); 11243b4477d2SStefan Hajnoczi if (sk->sk_state == TCP_SYN_SENT && 1125d021c344SAndy King (sk->sk_shutdown != SHUTDOWN_MASK)) { 11263b4477d2SStefan Hajnoczi sk->sk_state = TCP_CLOSE; 1127d021c344SAndy King sk->sk_err = ETIMEDOUT; 1128d021c344SAndy King sk->sk_error_report(sk); 1129380feae0SPeng Tao cancel = 1; 1130d021c344SAndy King } 1131d021c344SAndy King release_sock(sk); 1132380feae0SPeng Tao if (cancel) 1133380feae0SPeng Tao vsock_transport_cancel_pkt(vsk); 1134d021c344SAndy King 1135d021c344SAndy King sock_put(sk); 1136d021c344SAndy King } 1137d021c344SAndy King 1138d021c344SAndy King static int vsock_stream_connect(struct socket *sock, struct sockaddr *addr, 1139d021c344SAndy King int addr_len, int flags) 1140d021c344SAndy King { 1141d021c344SAndy King int err; 1142d021c344SAndy King struct sock *sk; 1143d021c344SAndy King struct vsock_sock *vsk; 1144d021c344SAndy King struct sockaddr_vm *remote_addr; 1145d021c344SAndy King long timeout; 1146d021c344SAndy King DEFINE_WAIT(wait); 1147d021c344SAndy King 1148d021c344SAndy King err = 0; 1149d021c344SAndy King sk = sock->sk; 1150d021c344SAndy King vsk = vsock_sk(sk); 1151d021c344SAndy King 1152d021c344SAndy King lock_sock(sk); 1153d021c344SAndy King 1154d021c344SAndy King /* XXX AF_UNSPEC should make us disconnect like AF_INET. */ 1155d021c344SAndy King switch (sock->state) { 1156d021c344SAndy King case SS_CONNECTED: 1157d021c344SAndy King err = -EISCONN; 1158d021c344SAndy King goto out; 1159d021c344SAndy King case SS_DISCONNECTING: 1160d021c344SAndy King err = -EINVAL; 1161d021c344SAndy King goto out; 1162d021c344SAndy King case SS_CONNECTING: 1163d021c344SAndy King /* This continues on so we can move sock into the SS_CONNECTED 1164d021c344SAndy King * state once the connection has completed (at which point err 1165d021c344SAndy King * will be set to zero also). Otherwise, we will either wait 1166d021c344SAndy King * for the connection or return -EALREADY should this be a 1167d021c344SAndy King * non-blocking call. 1168d021c344SAndy King */ 1169d021c344SAndy King err = -EALREADY; 1170d021c344SAndy King break; 1171d021c344SAndy King default: 11723b4477d2SStefan Hajnoczi if ((sk->sk_state == TCP_LISTEN) || 1173d021c344SAndy King vsock_addr_cast(addr, addr_len, &remote_addr) != 0) { 1174d021c344SAndy King err = -EINVAL; 1175d021c344SAndy King goto out; 1176d021c344SAndy King } 1177d021c344SAndy King 1178d021c344SAndy King /* The hypervisor and well-known contexts do not have socket 1179d021c344SAndy King * endpoints. 1180d021c344SAndy King */ 1181d021c344SAndy King if (!transport->stream_allow(remote_addr->svm_cid, 1182d021c344SAndy King remote_addr->svm_port)) { 1183d021c344SAndy King err = -ENETUNREACH; 1184d021c344SAndy King goto out; 1185d021c344SAndy King } 1186d021c344SAndy King 1187d021c344SAndy King /* Set the remote address that we are connecting to. */ 1188d021c344SAndy King memcpy(&vsk->remote_addr, remote_addr, 1189d021c344SAndy King sizeof(vsk->remote_addr)); 1190d021c344SAndy King 1191b3a6dfe8SAsias He err = vsock_auto_bind(vsk); 1192b3a6dfe8SAsias He if (err) 1193d021c344SAndy King goto out; 1194d021c344SAndy King 11953b4477d2SStefan Hajnoczi sk->sk_state = TCP_SYN_SENT; 1196d021c344SAndy King 1197d021c344SAndy King err = transport->connect(vsk); 1198d021c344SAndy King if (err < 0) 1199d021c344SAndy King goto out; 1200d021c344SAndy King 1201d021c344SAndy King /* Mark sock as connecting and set the error code to in 1202d021c344SAndy King * progress in case this is a non-blocking connect. 1203d021c344SAndy King */ 1204d021c344SAndy King sock->state = SS_CONNECTING; 1205d021c344SAndy King err = -EINPROGRESS; 1206d021c344SAndy King } 1207d021c344SAndy King 1208d021c344SAndy King /* The receive path will handle all communication until we are able to 1209d021c344SAndy King * enter the connected state. Here we wait for the connection to be 1210d021c344SAndy King * completed or a notification of an error. 1211d021c344SAndy King */ 1212d021c344SAndy King timeout = vsk->connect_timeout; 1213d021c344SAndy King prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); 1214d021c344SAndy King 12153b4477d2SStefan Hajnoczi while (sk->sk_state != TCP_ESTABLISHED && sk->sk_err == 0) { 1216d021c344SAndy King if (flags & O_NONBLOCK) { 1217d021c344SAndy King /* If we're not going to block, we schedule a timeout 1218d021c344SAndy King * function to generate a timeout on the connection 1219d021c344SAndy King * attempt, in case the peer doesn't respond in a 1220d021c344SAndy King * timely manner. We hold on to the socket until the 1221d021c344SAndy King * timeout fires. 1222d021c344SAndy King */ 1223d021c344SAndy King sock_hold(sk); 1224d021c344SAndy King INIT_DELAYED_WORK(&vsk->dwork, 1225d021c344SAndy King vsock_connect_timeout); 1226d021c344SAndy King schedule_delayed_work(&vsk->dwork, timeout); 1227d021c344SAndy King 1228d021c344SAndy King /* Skip ahead to preserve error code set above. */ 1229d021c344SAndy King goto out_wait; 1230d021c344SAndy King } 1231d021c344SAndy King 1232d021c344SAndy King release_sock(sk); 1233d021c344SAndy King timeout = schedule_timeout(timeout); 1234d021c344SAndy King lock_sock(sk); 1235d021c344SAndy King 1236d021c344SAndy King if (signal_pending(current)) { 1237d021c344SAndy King err = sock_intr_errno(timeout); 12383b4477d2SStefan Hajnoczi sk->sk_state = TCP_CLOSE; 1239f7f9b5e7SClaudio Imbrenda sock->state = SS_UNCONNECTED; 1240380feae0SPeng Tao vsock_transport_cancel_pkt(vsk); 1241f7f9b5e7SClaudio Imbrenda goto out_wait; 1242d021c344SAndy King } else if (timeout == 0) { 1243d021c344SAndy King err = -ETIMEDOUT; 12443b4477d2SStefan Hajnoczi sk->sk_state = TCP_CLOSE; 1245f7f9b5e7SClaudio Imbrenda sock->state = SS_UNCONNECTED; 1246380feae0SPeng Tao vsock_transport_cancel_pkt(vsk); 1247f7f9b5e7SClaudio Imbrenda goto out_wait; 1248d021c344SAndy King } 1249d021c344SAndy King 1250d021c344SAndy King prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); 1251d021c344SAndy King } 1252d021c344SAndy King 1253d021c344SAndy King if (sk->sk_err) { 1254d021c344SAndy King err = -sk->sk_err; 12553b4477d2SStefan Hajnoczi sk->sk_state = TCP_CLOSE; 1256f7f9b5e7SClaudio Imbrenda sock->state = SS_UNCONNECTED; 1257f7f9b5e7SClaudio Imbrenda } else { 1258d021c344SAndy King err = 0; 1259f7f9b5e7SClaudio Imbrenda } 1260d021c344SAndy King 1261d021c344SAndy King out_wait: 1262d021c344SAndy King finish_wait(sk_sleep(sk), &wait); 1263d021c344SAndy King out: 1264d021c344SAndy King release_sock(sk); 1265d021c344SAndy King return err; 1266d021c344SAndy King } 1267d021c344SAndy King 1268cdfbabfbSDavid Howells static int vsock_accept(struct socket *sock, struct socket *newsock, int flags, 1269cdfbabfbSDavid Howells bool kern) 1270d021c344SAndy King { 1271d021c344SAndy King struct sock *listener; 1272d021c344SAndy King int err; 1273d021c344SAndy King struct sock *connected; 1274d021c344SAndy King struct vsock_sock *vconnected; 1275d021c344SAndy King long timeout; 1276d021c344SAndy King DEFINE_WAIT(wait); 1277d021c344SAndy King 1278d021c344SAndy King err = 0; 1279d021c344SAndy King listener = sock->sk; 1280d021c344SAndy King 1281d021c344SAndy King lock_sock(listener); 1282d021c344SAndy King 1283d021c344SAndy King if (sock->type != SOCK_STREAM) { 1284d021c344SAndy King err = -EOPNOTSUPP; 1285d021c344SAndy King goto out; 1286d021c344SAndy King } 1287d021c344SAndy King 12883b4477d2SStefan Hajnoczi if (listener->sk_state != TCP_LISTEN) { 1289d021c344SAndy King err = -EINVAL; 1290d021c344SAndy King goto out; 1291d021c344SAndy King } 1292d021c344SAndy King 1293d021c344SAndy King /* Wait for children sockets to appear; these are the new sockets 1294d021c344SAndy King * created upon connection establishment. 1295d021c344SAndy King */ 1296d021c344SAndy King timeout = sock_sndtimeo(listener, flags & O_NONBLOCK); 1297d021c344SAndy King prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE); 1298d021c344SAndy King 1299d021c344SAndy King while ((connected = vsock_dequeue_accept(listener)) == NULL && 1300d021c344SAndy King listener->sk_err == 0) { 1301d021c344SAndy King release_sock(listener); 1302d021c344SAndy King timeout = schedule_timeout(timeout); 1303f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(listener), &wait); 1304d021c344SAndy King lock_sock(listener); 1305d021c344SAndy King 1306d021c344SAndy King if (signal_pending(current)) { 1307d021c344SAndy King err = sock_intr_errno(timeout); 1308f7f9b5e7SClaudio Imbrenda goto out; 1309d021c344SAndy King } else if (timeout == 0) { 1310d021c344SAndy King err = -EAGAIN; 1311f7f9b5e7SClaudio Imbrenda goto out; 1312d021c344SAndy King } 1313d021c344SAndy King 1314d021c344SAndy King prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE); 1315d021c344SAndy King } 1316f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(listener), &wait); 1317d021c344SAndy King 1318d021c344SAndy King if (listener->sk_err) 1319d021c344SAndy King err = -listener->sk_err; 1320d021c344SAndy King 1321d021c344SAndy King if (connected) { 1322d021c344SAndy King listener->sk_ack_backlog--; 1323d021c344SAndy King 13244192f672SStefan Hajnoczi lock_sock_nested(connected, SINGLE_DEPTH_NESTING); 1325d021c344SAndy King vconnected = vsock_sk(connected); 1326d021c344SAndy King 1327d021c344SAndy King /* If the listener socket has received an error, then we should 1328d021c344SAndy King * reject this socket and return. Note that we simply mark the 1329d021c344SAndy King * socket rejected, drop our reference, and let the cleanup 1330d021c344SAndy King * function handle the cleanup; the fact that we found it in 1331d021c344SAndy King * the listener's accept queue guarantees that the cleanup 1332d021c344SAndy King * function hasn't run yet. 1333d021c344SAndy King */ 1334d021c344SAndy King if (err) { 1335d021c344SAndy King vconnected->rejected = true; 1336f7f9b5e7SClaudio Imbrenda } else { 1337d021c344SAndy King newsock->state = SS_CONNECTED; 1338d021c344SAndy King sock_graft(connected, newsock); 1339f7f9b5e7SClaudio Imbrenda } 1340f7f9b5e7SClaudio Imbrenda 1341d021c344SAndy King release_sock(connected); 1342d021c344SAndy King sock_put(connected); 1343d021c344SAndy King } 1344d021c344SAndy King 1345d021c344SAndy King out: 1346d021c344SAndy King release_sock(listener); 1347d021c344SAndy King return err; 1348d021c344SAndy King } 1349d021c344SAndy King 1350d021c344SAndy King static int vsock_listen(struct socket *sock, int backlog) 1351d021c344SAndy King { 1352d021c344SAndy King int err; 1353d021c344SAndy King struct sock *sk; 1354d021c344SAndy King struct vsock_sock *vsk; 1355d021c344SAndy King 1356d021c344SAndy King sk = sock->sk; 1357d021c344SAndy King 1358d021c344SAndy King lock_sock(sk); 1359d021c344SAndy King 1360d021c344SAndy King if (sock->type != SOCK_STREAM) { 1361d021c344SAndy King err = -EOPNOTSUPP; 1362d021c344SAndy King goto out; 1363d021c344SAndy King } 1364d021c344SAndy King 1365d021c344SAndy King if (sock->state != SS_UNCONNECTED) { 1366d021c344SAndy King err = -EINVAL; 1367d021c344SAndy King goto out; 1368d021c344SAndy King } 1369d021c344SAndy King 1370d021c344SAndy King vsk = vsock_sk(sk); 1371d021c344SAndy King 1372d021c344SAndy King if (!vsock_addr_bound(&vsk->local_addr)) { 1373d021c344SAndy King err = -EINVAL; 1374d021c344SAndy King goto out; 1375d021c344SAndy King } 1376d021c344SAndy King 1377d021c344SAndy King sk->sk_max_ack_backlog = backlog; 13783b4477d2SStefan Hajnoczi sk->sk_state = TCP_LISTEN; 1379d021c344SAndy King 1380d021c344SAndy King err = 0; 1381d021c344SAndy King 1382d021c344SAndy King out: 1383d021c344SAndy King release_sock(sk); 1384d021c344SAndy King return err; 1385d021c344SAndy King } 1386d021c344SAndy King 1387d021c344SAndy King static int vsock_stream_setsockopt(struct socket *sock, 1388d021c344SAndy King int level, 1389d021c344SAndy King int optname, 1390d021c344SAndy King char __user *optval, 1391d021c344SAndy King unsigned int optlen) 1392d021c344SAndy King { 1393d021c344SAndy King int err; 1394d021c344SAndy King struct sock *sk; 1395d021c344SAndy King struct vsock_sock *vsk; 1396d021c344SAndy King u64 val; 1397d021c344SAndy King 1398d021c344SAndy King if (level != AF_VSOCK) 1399d021c344SAndy King return -ENOPROTOOPT; 1400d021c344SAndy King 1401d021c344SAndy King #define COPY_IN(_v) \ 1402d021c344SAndy King do { \ 1403d021c344SAndy King if (optlen < sizeof(_v)) { \ 1404d021c344SAndy King err = -EINVAL; \ 1405d021c344SAndy King goto exit; \ 1406d021c344SAndy King } \ 1407d021c344SAndy King if (copy_from_user(&_v, optval, sizeof(_v)) != 0) { \ 1408d021c344SAndy King err = -EFAULT; \ 1409d021c344SAndy King goto exit; \ 1410d021c344SAndy King } \ 1411d021c344SAndy King } while (0) 1412d021c344SAndy King 1413d021c344SAndy King err = 0; 1414d021c344SAndy King sk = sock->sk; 1415d021c344SAndy King vsk = vsock_sk(sk); 1416d021c344SAndy King 1417d021c344SAndy King lock_sock(sk); 1418d021c344SAndy King 1419d021c344SAndy King switch (optname) { 1420d021c344SAndy King case SO_VM_SOCKETS_BUFFER_SIZE: 1421d021c344SAndy King COPY_IN(val); 1422d021c344SAndy King transport->set_buffer_size(vsk, val); 1423d021c344SAndy King break; 1424d021c344SAndy King 1425d021c344SAndy King case SO_VM_SOCKETS_BUFFER_MAX_SIZE: 1426d021c344SAndy King COPY_IN(val); 1427d021c344SAndy King transport->set_max_buffer_size(vsk, val); 1428d021c344SAndy King break; 1429d021c344SAndy King 1430d021c344SAndy King case SO_VM_SOCKETS_BUFFER_MIN_SIZE: 1431d021c344SAndy King COPY_IN(val); 1432d021c344SAndy King transport->set_min_buffer_size(vsk, val); 1433d021c344SAndy King break; 1434d021c344SAndy King 1435d021c344SAndy King case SO_VM_SOCKETS_CONNECT_TIMEOUT: { 1436d021c344SAndy King struct timeval tv; 1437d021c344SAndy King COPY_IN(tv); 1438d021c344SAndy King if (tv.tv_sec >= 0 && tv.tv_usec < USEC_PER_SEC && 1439d021c344SAndy King tv.tv_sec < (MAX_SCHEDULE_TIMEOUT / HZ - 1)) { 1440d021c344SAndy King vsk->connect_timeout = tv.tv_sec * HZ + 1441d021c344SAndy King DIV_ROUND_UP(tv.tv_usec, (1000000 / HZ)); 1442d021c344SAndy King if (vsk->connect_timeout == 0) 1443d021c344SAndy King vsk->connect_timeout = 1444d021c344SAndy King VSOCK_DEFAULT_CONNECT_TIMEOUT; 1445d021c344SAndy King 1446d021c344SAndy King } else { 1447d021c344SAndy King err = -ERANGE; 1448d021c344SAndy King } 1449d021c344SAndy King break; 1450d021c344SAndy King } 1451d021c344SAndy King 1452d021c344SAndy King default: 1453d021c344SAndy King err = -ENOPROTOOPT; 1454d021c344SAndy King break; 1455d021c344SAndy King } 1456d021c344SAndy King 1457d021c344SAndy King #undef COPY_IN 1458d021c344SAndy King 1459d021c344SAndy King exit: 1460d021c344SAndy King release_sock(sk); 1461d021c344SAndy King return err; 1462d021c344SAndy King } 1463d021c344SAndy King 1464d021c344SAndy King static int vsock_stream_getsockopt(struct socket *sock, 1465d021c344SAndy King int level, int optname, 1466d021c344SAndy King char __user *optval, 1467d021c344SAndy King int __user *optlen) 1468d021c344SAndy King { 1469d021c344SAndy King int err; 1470d021c344SAndy King int len; 1471d021c344SAndy King struct sock *sk; 1472d021c344SAndy King struct vsock_sock *vsk; 1473d021c344SAndy King u64 val; 1474d021c344SAndy King 1475d021c344SAndy King if (level != AF_VSOCK) 1476d021c344SAndy King return -ENOPROTOOPT; 1477d021c344SAndy King 1478d021c344SAndy King err = get_user(len, optlen); 1479d021c344SAndy King if (err != 0) 1480d021c344SAndy King return err; 1481d021c344SAndy King 1482d021c344SAndy King #define COPY_OUT(_v) \ 1483d021c344SAndy King do { \ 1484d021c344SAndy King if (len < sizeof(_v)) \ 1485d021c344SAndy King return -EINVAL; \ 1486d021c344SAndy King \ 1487d021c344SAndy King len = sizeof(_v); \ 1488d021c344SAndy King if (copy_to_user(optval, &_v, len) != 0) \ 1489d021c344SAndy King return -EFAULT; \ 1490d021c344SAndy King \ 1491d021c344SAndy King } while (0) 1492d021c344SAndy King 1493d021c344SAndy King err = 0; 1494d021c344SAndy King sk = sock->sk; 1495d021c344SAndy King vsk = vsock_sk(sk); 1496d021c344SAndy King 1497d021c344SAndy King switch (optname) { 1498d021c344SAndy King case SO_VM_SOCKETS_BUFFER_SIZE: 1499d021c344SAndy King val = transport->get_buffer_size(vsk); 1500d021c344SAndy King COPY_OUT(val); 1501d021c344SAndy King break; 1502d021c344SAndy King 1503d021c344SAndy King case SO_VM_SOCKETS_BUFFER_MAX_SIZE: 1504d021c344SAndy King val = transport->get_max_buffer_size(vsk); 1505d021c344SAndy King COPY_OUT(val); 1506d021c344SAndy King break; 1507d021c344SAndy King 1508d021c344SAndy King case SO_VM_SOCKETS_BUFFER_MIN_SIZE: 1509d021c344SAndy King val = transport->get_min_buffer_size(vsk); 1510d021c344SAndy King COPY_OUT(val); 1511d021c344SAndy King break; 1512d021c344SAndy King 1513d021c344SAndy King case SO_VM_SOCKETS_CONNECT_TIMEOUT: { 1514d021c344SAndy King struct timeval tv; 1515d021c344SAndy King tv.tv_sec = vsk->connect_timeout / HZ; 1516d021c344SAndy King tv.tv_usec = 1517d021c344SAndy King (vsk->connect_timeout - 1518d021c344SAndy King tv.tv_sec * HZ) * (1000000 / HZ); 1519d021c344SAndy King COPY_OUT(tv); 1520d021c344SAndy King break; 1521d021c344SAndy King } 1522d021c344SAndy King default: 1523d021c344SAndy King return -ENOPROTOOPT; 1524d021c344SAndy King } 1525d021c344SAndy King 1526d021c344SAndy King err = put_user(len, optlen); 1527d021c344SAndy King if (err != 0) 1528d021c344SAndy King return -EFAULT; 1529d021c344SAndy King 1530d021c344SAndy King #undef COPY_OUT 1531d021c344SAndy King 1532d021c344SAndy King return 0; 1533d021c344SAndy King } 1534d021c344SAndy King 15351b784140SYing Xue static int vsock_stream_sendmsg(struct socket *sock, struct msghdr *msg, 15361b784140SYing Xue size_t len) 1537d021c344SAndy King { 1538d021c344SAndy King struct sock *sk; 1539d021c344SAndy King struct vsock_sock *vsk; 1540d021c344SAndy King ssize_t total_written; 1541d021c344SAndy King long timeout; 1542d021c344SAndy King int err; 1543d021c344SAndy King struct vsock_transport_send_notify_data send_data; 1544499fde66SWANG Cong DEFINE_WAIT_FUNC(wait, woken_wake_function); 1545d021c344SAndy King 1546d021c344SAndy King sk = sock->sk; 1547d021c344SAndy King vsk = vsock_sk(sk); 1548d021c344SAndy King total_written = 0; 1549d021c344SAndy King err = 0; 1550d021c344SAndy King 1551d021c344SAndy King if (msg->msg_flags & MSG_OOB) 1552d021c344SAndy King return -EOPNOTSUPP; 1553d021c344SAndy King 1554d021c344SAndy King lock_sock(sk); 1555d021c344SAndy King 1556d021c344SAndy King /* Callers should not provide a destination with stream sockets. */ 1557d021c344SAndy King if (msg->msg_namelen) { 15583b4477d2SStefan Hajnoczi err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP; 1559d021c344SAndy King goto out; 1560d021c344SAndy King } 1561d021c344SAndy King 1562d021c344SAndy King /* Send data only if both sides are not shutdown in the direction. */ 1563d021c344SAndy King if (sk->sk_shutdown & SEND_SHUTDOWN || 1564d021c344SAndy King vsk->peer_shutdown & RCV_SHUTDOWN) { 1565d021c344SAndy King err = -EPIPE; 1566d021c344SAndy King goto out; 1567d021c344SAndy King } 1568d021c344SAndy King 15693b4477d2SStefan Hajnoczi if (sk->sk_state != TCP_ESTABLISHED || 1570d021c344SAndy King !vsock_addr_bound(&vsk->local_addr)) { 1571d021c344SAndy King err = -ENOTCONN; 1572d021c344SAndy King goto out; 1573d021c344SAndy King } 1574d021c344SAndy King 1575d021c344SAndy King if (!vsock_addr_bound(&vsk->remote_addr)) { 1576d021c344SAndy King err = -EDESTADDRREQ; 1577d021c344SAndy King goto out; 1578d021c344SAndy King } 1579d021c344SAndy King 1580d021c344SAndy King /* Wait for room in the produce queue to enqueue our user's data. */ 1581d021c344SAndy King timeout = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT); 1582d021c344SAndy King 1583d021c344SAndy King err = transport->notify_send_init(vsk, &send_data); 1584d021c344SAndy King if (err < 0) 1585d021c344SAndy King goto out; 1586d021c344SAndy King 1587d021c344SAndy King while (total_written < len) { 1588d021c344SAndy King ssize_t written; 1589d021c344SAndy King 1590499fde66SWANG Cong add_wait_queue(sk_sleep(sk), &wait); 1591d021c344SAndy King while (vsock_stream_has_space(vsk) == 0 && 1592d021c344SAndy King sk->sk_err == 0 && 1593d021c344SAndy King !(sk->sk_shutdown & SEND_SHUTDOWN) && 1594d021c344SAndy King !(vsk->peer_shutdown & RCV_SHUTDOWN)) { 1595d021c344SAndy King 1596d021c344SAndy King /* Don't wait for non-blocking sockets. */ 1597d021c344SAndy King if (timeout == 0) { 1598d021c344SAndy King err = -EAGAIN; 1599499fde66SWANG Cong remove_wait_queue(sk_sleep(sk), &wait); 1600f7f9b5e7SClaudio Imbrenda goto out_err; 1601d021c344SAndy King } 1602d021c344SAndy King 1603d021c344SAndy King err = transport->notify_send_pre_block(vsk, &send_data); 1604f7f9b5e7SClaudio Imbrenda if (err < 0) { 1605499fde66SWANG Cong remove_wait_queue(sk_sleep(sk), &wait); 1606f7f9b5e7SClaudio Imbrenda goto out_err; 1607f7f9b5e7SClaudio Imbrenda } 1608d021c344SAndy King 1609d021c344SAndy King release_sock(sk); 1610499fde66SWANG Cong timeout = wait_woken(&wait, TASK_INTERRUPTIBLE, timeout); 1611d021c344SAndy King lock_sock(sk); 1612d021c344SAndy King if (signal_pending(current)) { 1613d021c344SAndy King err = sock_intr_errno(timeout); 1614499fde66SWANG Cong remove_wait_queue(sk_sleep(sk), &wait); 1615f7f9b5e7SClaudio Imbrenda goto out_err; 1616d021c344SAndy King } else if (timeout == 0) { 1617d021c344SAndy King err = -EAGAIN; 1618499fde66SWANG Cong remove_wait_queue(sk_sleep(sk), &wait); 1619f7f9b5e7SClaudio Imbrenda goto out_err; 1620d021c344SAndy King } 1621d021c344SAndy King } 1622499fde66SWANG Cong remove_wait_queue(sk_sleep(sk), &wait); 1623d021c344SAndy King 1624d021c344SAndy King /* These checks occur both as part of and after the loop 1625d021c344SAndy King * conditional since we need to check before and after 1626d021c344SAndy King * sleeping. 1627d021c344SAndy King */ 1628d021c344SAndy King if (sk->sk_err) { 1629d021c344SAndy King err = -sk->sk_err; 1630f7f9b5e7SClaudio Imbrenda goto out_err; 1631d021c344SAndy King } else if ((sk->sk_shutdown & SEND_SHUTDOWN) || 1632d021c344SAndy King (vsk->peer_shutdown & RCV_SHUTDOWN)) { 1633d021c344SAndy King err = -EPIPE; 1634f7f9b5e7SClaudio Imbrenda goto out_err; 1635d021c344SAndy King } 1636d021c344SAndy King 1637d021c344SAndy King err = transport->notify_send_pre_enqueue(vsk, &send_data); 1638d021c344SAndy King if (err < 0) 1639f7f9b5e7SClaudio Imbrenda goto out_err; 1640d021c344SAndy King 1641d021c344SAndy King /* Note that enqueue will only write as many bytes as are free 1642d021c344SAndy King * in the produce queue, so we don't need to ensure len is 1643d021c344SAndy King * smaller than the queue size. It is the caller's 1644d021c344SAndy King * responsibility to check how many bytes we were able to send. 1645d021c344SAndy King */ 1646d021c344SAndy King 1647d021c344SAndy King written = transport->stream_enqueue( 16480f7db23aSAl Viro vsk, msg, 1649d021c344SAndy King len - total_written); 1650d021c344SAndy King if (written < 0) { 1651d021c344SAndy King err = -ENOMEM; 1652f7f9b5e7SClaudio Imbrenda goto out_err; 1653d021c344SAndy King } 1654d021c344SAndy King 1655d021c344SAndy King total_written += written; 1656d021c344SAndy King 1657d021c344SAndy King err = transport->notify_send_post_enqueue( 1658d021c344SAndy King vsk, written, &send_data); 1659d021c344SAndy King if (err < 0) 1660f7f9b5e7SClaudio Imbrenda goto out_err; 1661d021c344SAndy King 1662d021c344SAndy King } 1663d021c344SAndy King 1664f7f9b5e7SClaudio Imbrenda out_err: 1665d021c344SAndy King if (total_written > 0) 1666d021c344SAndy King err = total_written; 1667d021c344SAndy King out: 1668d021c344SAndy King release_sock(sk); 1669d021c344SAndy King return err; 1670d021c344SAndy King } 1671d021c344SAndy King 1672d021c344SAndy King 1673d021c344SAndy King static int 16741b784140SYing Xue vsock_stream_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, 16751b784140SYing Xue int flags) 1676d021c344SAndy King { 1677d021c344SAndy King struct sock *sk; 1678d021c344SAndy King struct vsock_sock *vsk; 1679d021c344SAndy King int err; 1680d021c344SAndy King size_t target; 1681d021c344SAndy King ssize_t copied; 1682d021c344SAndy King long timeout; 1683d021c344SAndy King struct vsock_transport_recv_notify_data recv_data; 1684d021c344SAndy King 1685d021c344SAndy King DEFINE_WAIT(wait); 1686d021c344SAndy King 1687d021c344SAndy King sk = sock->sk; 1688d021c344SAndy King vsk = vsock_sk(sk); 1689d021c344SAndy King err = 0; 1690d021c344SAndy King 1691d021c344SAndy King lock_sock(sk); 1692d021c344SAndy King 16933b4477d2SStefan Hajnoczi if (sk->sk_state != TCP_ESTABLISHED) { 1694d021c344SAndy King /* Recvmsg is supposed to return 0 if a peer performs an 1695d021c344SAndy King * orderly shutdown. Differentiate between that case and when a 1696d021c344SAndy King * peer has not connected or a local shutdown occured with the 1697d021c344SAndy King * SOCK_DONE flag. 1698d021c344SAndy King */ 1699d021c344SAndy King if (sock_flag(sk, SOCK_DONE)) 1700d021c344SAndy King err = 0; 1701d021c344SAndy King else 1702d021c344SAndy King err = -ENOTCONN; 1703d021c344SAndy King 1704d021c344SAndy King goto out; 1705d021c344SAndy King } 1706d021c344SAndy King 1707d021c344SAndy King if (flags & MSG_OOB) { 1708d021c344SAndy King err = -EOPNOTSUPP; 1709d021c344SAndy King goto out; 1710d021c344SAndy King } 1711d021c344SAndy King 1712d021c344SAndy King /* We don't check peer_shutdown flag here since peer may actually shut 1713d021c344SAndy King * down, but there can be data in the queue that a local socket can 1714d021c344SAndy King * receive. 1715d021c344SAndy King */ 1716d021c344SAndy King if (sk->sk_shutdown & RCV_SHUTDOWN) { 1717d021c344SAndy King err = 0; 1718d021c344SAndy King goto out; 1719d021c344SAndy King } 1720d021c344SAndy King 1721d021c344SAndy King /* It is valid on Linux to pass in a zero-length receive buffer. This 1722d021c344SAndy King * is not an error. We may as well bail out now. 1723d021c344SAndy King */ 1724d021c344SAndy King if (!len) { 1725d021c344SAndy King err = 0; 1726d021c344SAndy King goto out; 1727d021c344SAndy King } 1728d021c344SAndy King 1729d021c344SAndy King /* We must not copy less than target bytes into the user's buffer 1730d021c344SAndy King * before returning successfully, so we wait for the consume queue to 1731d021c344SAndy King * have that much data to consume before dequeueing. Note that this 1732d021c344SAndy King * makes it impossible to handle cases where target is greater than the 1733d021c344SAndy King * queue size. 1734d021c344SAndy King */ 1735d021c344SAndy King target = sock_rcvlowat(sk, flags & MSG_WAITALL, len); 1736d021c344SAndy King if (target >= transport->stream_rcvhiwat(vsk)) { 1737d021c344SAndy King err = -ENOMEM; 1738d021c344SAndy King goto out; 1739d021c344SAndy King } 1740d021c344SAndy King timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT); 1741d021c344SAndy King copied = 0; 1742d021c344SAndy King 1743d021c344SAndy King err = transport->notify_recv_init(vsk, target, &recv_data); 1744d021c344SAndy King if (err < 0) 1745d021c344SAndy King goto out; 1746d021c344SAndy King 1747d021c344SAndy King 1748d021c344SAndy King while (1) { 1749f7f9b5e7SClaudio Imbrenda s64 ready; 1750f7f9b5e7SClaudio Imbrenda 1751f7f9b5e7SClaudio Imbrenda prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); 1752f7f9b5e7SClaudio Imbrenda ready = vsock_stream_has_data(vsk); 1753f7f9b5e7SClaudio Imbrenda 1754f7f9b5e7SClaudio Imbrenda if (ready == 0) { 1755f7f9b5e7SClaudio Imbrenda if (sk->sk_err != 0 || 1756f7f9b5e7SClaudio Imbrenda (sk->sk_shutdown & RCV_SHUTDOWN) || 1757f7f9b5e7SClaudio Imbrenda (vsk->peer_shutdown & SEND_SHUTDOWN)) { 1758f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(sk), &wait); 1759f7f9b5e7SClaudio Imbrenda break; 1760f7f9b5e7SClaudio Imbrenda } 1761f7f9b5e7SClaudio Imbrenda /* Don't wait for non-blocking sockets. */ 1762f7f9b5e7SClaudio Imbrenda if (timeout == 0) { 1763f7f9b5e7SClaudio Imbrenda err = -EAGAIN; 1764f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(sk), &wait); 1765f7f9b5e7SClaudio Imbrenda break; 1766f7f9b5e7SClaudio Imbrenda } 1767f7f9b5e7SClaudio Imbrenda 1768f7f9b5e7SClaudio Imbrenda err = transport->notify_recv_pre_block( 1769f7f9b5e7SClaudio Imbrenda vsk, target, &recv_data); 1770f7f9b5e7SClaudio Imbrenda if (err < 0) { 1771f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(sk), &wait); 1772f7f9b5e7SClaudio Imbrenda break; 1773f7f9b5e7SClaudio Imbrenda } 1774f7f9b5e7SClaudio Imbrenda release_sock(sk); 1775f7f9b5e7SClaudio Imbrenda timeout = schedule_timeout(timeout); 1776f7f9b5e7SClaudio Imbrenda lock_sock(sk); 1777f7f9b5e7SClaudio Imbrenda 1778f7f9b5e7SClaudio Imbrenda if (signal_pending(current)) { 1779f7f9b5e7SClaudio Imbrenda err = sock_intr_errno(timeout); 1780f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(sk), &wait); 1781f7f9b5e7SClaudio Imbrenda break; 1782f7f9b5e7SClaudio Imbrenda } else if (timeout == 0) { 1783f7f9b5e7SClaudio Imbrenda err = -EAGAIN; 1784f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(sk), &wait); 1785f7f9b5e7SClaudio Imbrenda break; 1786f7f9b5e7SClaudio Imbrenda } 1787f7f9b5e7SClaudio Imbrenda } else { 1788f7f9b5e7SClaudio Imbrenda ssize_t read; 1789f7f9b5e7SClaudio Imbrenda 1790f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(sk), &wait); 1791d021c344SAndy King 1792d021c344SAndy King if (ready < 0) { 1793f7f9b5e7SClaudio Imbrenda /* Invalid queue pair content. XXX This should 1794f7f9b5e7SClaudio Imbrenda * be changed to a connection reset in a later 1795f7f9b5e7SClaudio Imbrenda * change. 1796d021c344SAndy King */ 1797d021c344SAndy King 1798d021c344SAndy King err = -ENOMEM; 1799f7f9b5e7SClaudio Imbrenda goto out; 1800f7f9b5e7SClaudio Imbrenda } 1801d021c344SAndy King 1802d021c344SAndy King err = transport->notify_recv_pre_dequeue( 1803d021c344SAndy King vsk, target, &recv_data); 1804d021c344SAndy King if (err < 0) 1805d021c344SAndy King break; 1806d021c344SAndy King 1807d021c344SAndy King read = transport->stream_dequeue( 18080f7db23aSAl Viro vsk, msg, 1809d021c344SAndy King len - copied, flags); 1810d021c344SAndy King if (read < 0) { 1811d021c344SAndy King err = -ENOMEM; 1812d021c344SAndy King break; 1813d021c344SAndy King } 1814d021c344SAndy King 1815d021c344SAndy King copied += read; 1816d021c344SAndy King 1817d021c344SAndy King err = transport->notify_recv_post_dequeue( 1818d021c344SAndy King vsk, target, read, 1819d021c344SAndy King !(flags & MSG_PEEK), &recv_data); 1820d021c344SAndy King if (err < 0) 1821f7f9b5e7SClaudio Imbrenda goto out; 1822d021c344SAndy King 1823d021c344SAndy King if (read >= target || flags & MSG_PEEK) 1824d021c344SAndy King break; 1825d021c344SAndy King 1826d021c344SAndy King target -= read; 1827d021c344SAndy King } 1828d021c344SAndy King } 1829d021c344SAndy King 1830d021c344SAndy King if (sk->sk_err) 1831d021c344SAndy King err = -sk->sk_err; 1832d021c344SAndy King else if (sk->sk_shutdown & RCV_SHUTDOWN) 1833d021c344SAndy King err = 0; 1834d021c344SAndy King 1835dedc58e0SIan Campbell if (copied > 0) 1836d021c344SAndy King err = copied; 1837d021c344SAndy King 1838d021c344SAndy King out: 1839d021c344SAndy King release_sock(sk); 1840d021c344SAndy King return err; 1841d021c344SAndy King } 1842d021c344SAndy King 1843d021c344SAndy King static const struct proto_ops vsock_stream_ops = { 1844d021c344SAndy King .family = PF_VSOCK, 1845d021c344SAndy King .owner = THIS_MODULE, 1846d021c344SAndy King .release = vsock_release, 1847d021c344SAndy King .bind = vsock_bind, 1848d021c344SAndy King .connect = vsock_stream_connect, 1849d021c344SAndy King .socketpair = sock_no_socketpair, 1850d021c344SAndy King .accept = vsock_accept, 1851d021c344SAndy King .getname = vsock_getname, 1852d021c344SAndy King .poll = vsock_poll, 1853d021c344SAndy King .ioctl = sock_no_ioctl, 1854d021c344SAndy King .listen = vsock_listen, 1855d021c344SAndy King .shutdown = vsock_shutdown, 1856d021c344SAndy King .setsockopt = vsock_stream_setsockopt, 1857d021c344SAndy King .getsockopt = vsock_stream_getsockopt, 1858d021c344SAndy King .sendmsg = vsock_stream_sendmsg, 1859d021c344SAndy King .recvmsg = vsock_stream_recvmsg, 1860d021c344SAndy King .mmap = sock_no_mmap, 1861d021c344SAndy King .sendpage = sock_no_sendpage, 1862d021c344SAndy King }; 1863d021c344SAndy King 1864d021c344SAndy King static int vsock_create(struct net *net, struct socket *sock, 1865d021c344SAndy King int protocol, int kern) 1866d021c344SAndy King { 1867d021c344SAndy King if (!sock) 1868d021c344SAndy King return -EINVAL; 1869d021c344SAndy King 18706cf1c5fcSAndy King if (protocol && protocol != PF_VSOCK) 1871d021c344SAndy King return -EPROTONOSUPPORT; 1872d021c344SAndy King 1873d021c344SAndy King switch (sock->type) { 1874d021c344SAndy King case SOCK_DGRAM: 1875d021c344SAndy King sock->ops = &vsock_dgram_ops; 1876d021c344SAndy King break; 1877d021c344SAndy King case SOCK_STREAM: 1878d021c344SAndy King sock->ops = &vsock_stream_ops; 1879d021c344SAndy King break; 1880d021c344SAndy King default: 1881d021c344SAndy King return -ESOCKTNOSUPPORT; 1882d021c344SAndy King } 1883d021c344SAndy King 1884d021c344SAndy King sock->state = SS_UNCONNECTED; 1885d021c344SAndy King 188611aa9c28SEric W. Biederman return __vsock_create(net, sock, NULL, GFP_KERNEL, 0, kern) ? 0 : -ENOMEM; 1887d021c344SAndy King } 1888d021c344SAndy King 1889d021c344SAndy King static const struct net_proto_family vsock_family_ops = { 1890d021c344SAndy King .family = AF_VSOCK, 1891d021c344SAndy King .create = vsock_create, 1892d021c344SAndy King .owner = THIS_MODULE, 1893d021c344SAndy King }; 1894d021c344SAndy King 1895d021c344SAndy King static long vsock_dev_do_ioctl(struct file *filp, 1896d021c344SAndy King unsigned int cmd, void __user *ptr) 1897d021c344SAndy King { 1898d021c344SAndy King u32 __user *p = ptr; 1899d021c344SAndy King int retval = 0; 1900d021c344SAndy King 1901d021c344SAndy King switch (cmd) { 1902d021c344SAndy King case IOCTL_VM_SOCKETS_GET_LOCAL_CID: 1903d021c344SAndy King if (put_user(transport->get_local_cid(), p) != 0) 1904d021c344SAndy King retval = -EFAULT; 1905d021c344SAndy King break; 1906d021c344SAndy King 1907d021c344SAndy King default: 1908d021c344SAndy King pr_err("Unknown ioctl %d\n", cmd); 1909d021c344SAndy King retval = -EINVAL; 1910d021c344SAndy King } 1911d021c344SAndy King 1912d021c344SAndy King return retval; 1913d021c344SAndy King } 1914d021c344SAndy King 1915d021c344SAndy King static long vsock_dev_ioctl(struct file *filp, 1916d021c344SAndy King unsigned int cmd, unsigned long arg) 1917d021c344SAndy King { 1918d021c344SAndy King return vsock_dev_do_ioctl(filp, cmd, (void __user *)arg); 1919d021c344SAndy King } 1920d021c344SAndy King 1921d021c344SAndy King #ifdef CONFIG_COMPAT 1922d021c344SAndy King static long vsock_dev_compat_ioctl(struct file *filp, 1923d021c344SAndy King unsigned int cmd, unsigned long arg) 1924d021c344SAndy King { 1925d021c344SAndy King return vsock_dev_do_ioctl(filp, cmd, compat_ptr(arg)); 1926d021c344SAndy King } 1927d021c344SAndy King #endif 1928d021c344SAndy King 1929d021c344SAndy King static const struct file_operations vsock_device_ops = { 1930d021c344SAndy King .owner = THIS_MODULE, 1931d021c344SAndy King .unlocked_ioctl = vsock_dev_ioctl, 1932d021c344SAndy King #ifdef CONFIG_COMPAT 1933d021c344SAndy King .compat_ioctl = vsock_dev_compat_ioctl, 1934d021c344SAndy King #endif 1935d021c344SAndy King .open = nonseekable_open, 1936d021c344SAndy King }; 1937d021c344SAndy King 1938d021c344SAndy King static struct miscdevice vsock_device = { 1939d021c344SAndy King .name = "vsock", 1940d021c344SAndy King .fops = &vsock_device_ops, 1941d021c344SAndy King }; 1942d021c344SAndy King 19432c4a336eSAndy King int __vsock_core_init(const struct vsock_transport *t, struct module *owner) 1944d021c344SAndy King { 19452c4a336eSAndy King int err = mutex_lock_interruptible(&vsock_register_mutex); 19462c4a336eSAndy King 19472c4a336eSAndy King if (err) 19482c4a336eSAndy King return err; 19492c4a336eSAndy King 19502c4a336eSAndy King if (transport) { 19512c4a336eSAndy King err = -EBUSY; 19522c4a336eSAndy King goto err_busy; 19532c4a336eSAndy King } 19542c4a336eSAndy King 19552c4a336eSAndy King /* Transport must be the owner of the protocol so that it can't 19562c4a336eSAndy King * unload while there are open sockets. 19572c4a336eSAndy King */ 19582c4a336eSAndy King vsock_proto.owner = owner; 19592c4a336eSAndy King transport = t; 1960d021c344SAndy King 19616ad0b2f7SAsias He vsock_device.minor = MISC_DYNAMIC_MINOR; 1962d021c344SAndy King err = misc_register(&vsock_device); 1963d021c344SAndy King if (err) { 1964d021c344SAndy King pr_err("Failed to register misc device\n"); 1965f6a835bbSGao feng goto err_reset_transport; 1966d021c344SAndy King } 1967d021c344SAndy King 1968d021c344SAndy King err = proto_register(&vsock_proto, 1); /* we want our slab */ 1969d021c344SAndy King if (err) { 1970d021c344SAndy King pr_err("Cannot register vsock protocol\n"); 1971f6a835bbSGao feng goto err_deregister_misc; 1972d021c344SAndy King } 1973d021c344SAndy King 1974d021c344SAndy King err = sock_register(&vsock_family_ops); 1975d021c344SAndy King if (err) { 1976d021c344SAndy King pr_err("could not register af_vsock (%d) address family: %d\n", 1977d021c344SAndy King AF_VSOCK, err); 1978d021c344SAndy King goto err_unregister_proto; 1979d021c344SAndy King } 1980d021c344SAndy King 19812c4a336eSAndy King mutex_unlock(&vsock_register_mutex); 1982d021c344SAndy King return 0; 1983d021c344SAndy King 1984d021c344SAndy King err_unregister_proto: 1985d021c344SAndy King proto_unregister(&vsock_proto); 1986f6a835bbSGao feng err_deregister_misc: 1987d021c344SAndy King misc_deregister(&vsock_device); 1988f6a835bbSGao feng err_reset_transport: 19892c4a336eSAndy King transport = NULL; 19902c4a336eSAndy King err_busy: 19912c4a336eSAndy King mutex_unlock(&vsock_register_mutex); 1992d021c344SAndy King return err; 1993d021c344SAndy King } 19942c4a336eSAndy King EXPORT_SYMBOL_GPL(__vsock_core_init); 1995d021c344SAndy King 1996d021c344SAndy King void vsock_core_exit(void) 1997d021c344SAndy King { 1998d021c344SAndy King mutex_lock(&vsock_register_mutex); 1999d021c344SAndy King 2000d021c344SAndy King misc_deregister(&vsock_device); 2001d021c344SAndy King sock_unregister(AF_VSOCK); 2002d021c344SAndy King proto_unregister(&vsock_proto); 2003d021c344SAndy King 2004d021c344SAndy King /* We do not want the assignment below re-ordered. */ 2005d021c344SAndy King mb(); 2006d021c344SAndy King transport = NULL; 2007d021c344SAndy King 2008d021c344SAndy King mutex_unlock(&vsock_register_mutex); 2009d021c344SAndy King } 2010d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_core_exit); 2011d021c344SAndy King 20120b01aeb3SStefan Hajnoczi const struct vsock_transport *vsock_core_get_transport(void) 20130b01aeb3SStefan Hajnoczi { 20140b01aeb3SStefan Hajnoczi /* vsock_register_mutex not taken since only the transport uses this 20150b01aeb3SStefan Hajnoczi * function and only while registered. 20160b01aeb3SStefan Hajnoczi */ 20170b01aeb3SStefan Hajnoczi return transport; 20180b01aeb3SStefan Hajnoczi } 20190b01aeb3SStefan Hajnoczi EXPORT_SYMBOL_GPL(vsock_core_get_transport); 20200b01aeb3SStefan Hajnoczi 2021c1eef220SCong Wang module_init(vsock_init_tables); 2022c1eef220SCong Wang 2023d021c344SAndy King MODULE_AUTHOR("VMware, Inc."); 2024d021c344SAndy King MODULE_DESCRIPTION("VMware Virtual Socket Family"); 20251190cfdbSJorgen Hansen MODULE_VERSION("1.0.2.0-k"); 2026d021c344SAndy King MODULE_LICENSE("GPL v2"); 2027