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> 1108236b08cSLepton Wu #include <linux/random.h> 111d021c344SAndy King #include <linux/skbuff.h> 112d021c344SAndy King #include <linux/smp.h> 113d021c344SAndy King #include <linux/socket.h> 114d021c344SAndy King #include <linux/stddef.h> 115d021c344SAndy King #include <linux/unistd.h> 116d021c344SAndy King #include <linux/wait.h> 117d021c344SAndy King #include <linux/workqueue.h> 118d021c344SAndy King #include <net/sock.h> 11982a54d0eSAsias He #include <net/af_vsock.h> 120d021c344SAndy King 121d021c344SAndy King static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr); 122d021c344SAndy King static void vsock_sk_destruct(struct sock *sk); 123d021c344SAndy King static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb); 124d021c344SAndy King 125d021c344SAndy King /* Protocol family. */ 126d021c344SAndy King static struct proto vsock_proto = { 127d021c344SAndy King .name = "AF_VSOCK", 128d021c344SAndy King .owner = THIS_MODULE, 129d021c344SAndy King .obj_size = sizeof(struct vsock_sock), 130d021c344SAndy King }; 131d021c344SAndy King 132d021c344SAndy King /* The default peer timeout indicates how long we will wait for a peer response 133d021c344SAndy King * to a control message. 134d021c344SAndy King */ 135d021c344SAndy King #define VSOCK_DEFAULT_CONNECT_TIMEOUT (2 * HZ) 136d021c344SAndy King 137d021c344SAndy King static const struct vsock_transport *transport; 138d021c344SAndy King static DEFINE_MUTEX(vsock_register_mutex); 139d021c344SAndy King 140d021c344SAndy King /**** EXPORTS ****/ 141d021c344SAndy King 142d021c344SAndy King /* Get the ID of the local context. This is transport dependent. */ 143d021c344SAndy King 144d021c344SAndy King int vm_sockets_get_local_cid(void) 145d021c344SAndy King { 146d021c344SAndy King return transport->get_local_cid(); 147d021c344SAndy King } 148d021c344SAndy King EXPORT_SYMBOL_GPL(vm_sockets_get_local_cid); 149d021c344SAndy King 150d021c344SAndy King /**** UTILS ****/ 151d021c344SAndy King 152d021c344SAndy King /* Each bound VSocket is stored in the bind hash table and each connected 153d021c344SAndy King * VSocket is stored in the connected hash table. 154d021c344SAndy King * 155d021c344SAndy King * Unbound sockets are all put on the same list attached to the end of the hash 156d021c344SAndy King * table (vsock_unbound_sockets). Bound sockets are added to the hash table in 157d021c344SAndy King * the bucket that their local address hashes to (vsock_bound_sockets(addr) 158d021c344SAndy King * represents the list that addr hashes to). 159d021c344SAndy King * 160d021c344SAndy King * Specifically, we initialize the vsock_bind_table array to a size of 161d021c344SAndy King * VSOCK_HASH_SIZE + 1 so that vsock_bind_table[0] through 162d021c344SAndy King * vsock_bind_table[VSOCK_HASH_SIZE - 1] are for bound sockets and 163d021c344SAndy King * vsock_bind_table[VSOCK_HASH_SIZE] is for unbound sockets. The hash function 164a49dd9dcSAsias He * mods with VSOCK_HASH_SIZE to ensure this. 165d021c344SAndy King */ 166d021c344SAndy King #define MAX_PORT_RETRIES 24 167d021c344SAndy King 168a49dd9dcSAsias He #define VSOCK_HASH(addr) ((addr)->svm_port % VSOCK_HASH_SIZE) 169d021c344SAndy King #define vsock_bound_sockets(addr) (&vsock_bind_table[VSOCK_HASH(addr)]) 170d021c344SAndy King #define vsock_unbound_sockets (&vsock_bind_table[VSOCK_HASH_SIZE]) 171d021c344SAndy King 172d021c344SAndy King /* XXX This can probably be implemented in a better way. */ 173d021c344SAndy King #define VSOCK_CONN_HASH(src, dst) \ 174a49dd9dcSAsias He (((src)->svm_cid ^ (dst)->svm_port) % VSOCK_HASH_SIZE) 175d021c344SAndy King #define vsock_connected_sockets(src, dst) \ 176d021c344SAndy King (&vsock_connected_table[VSOCK_CONN_HASH(src, dst)]) 177d021c344SAndy King #define vsock_connected_sockets_vsk(vsk) \ 178d021c344SAndy King vsock_connected_sockets(&(vsk)->remote_addr, &(vsk)->local_addr) 179d021c344SAndy King 18044f20980SStefan Hajnoczi struct list_head vsock_bind_table[VSOCK_HASH_SIZE + 1]; 18144f20980SStefan Hajnoczi EXPORT_SYMBOL_GPL(vsock_bind_table); 18244f20980SStefan Hajnoczi struct list_head vsock_connected_table[VSOCK_HASH_SIZE]; 18344f20980SStefan Hajnoczi EXPORT_SYMBOL_GPL(vsock_connected_table); 18444f20980SStefan Hajnoczi DEFINE_SPINLOCK(vsock_table_lock); 18544f20980SStefan Hajnoczi EXPORT_SYMBOL_GPL(vsock_table_lock); 186d021c344SAndy King 187b3a6dfe8SAsias He /* Autobind this socket to the local address if necessary. */ 188b3a6dfe8SAsias He static int vsock_auto_bind(struct vsock_sock *vsk) 189b3a6dfe8SAsias He { 190b3a6dfe8SAsias He struct sock *sk = sk_vsock(vsk); 191b3a6dfe8SAsias He struct sockaddr_vm local_addr; 192b3a6dfe8SAsias He 193b3a6dfe8SAsias He if (vsock_addr_bound(&vsk->local_addr)) 194b3a6dfe8SAsias He return 0; 195b3a6dfe8SAsias He vsock_addr_init(&local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 196b3a6dfe8SAsias He return __vsock_bind(sk, &local_addr); 197b3a6dfe8SAsias He } 198b3a6dfe8SAsias He 199c1eef220SCong Wang static int __init vsock_init_tables(void) 200d021c344SAndy King { 201d021c344SAndy King int i; 202d021c344SAndy King 203d021c344SAndy King for (i = 0; i < ARRAY_SIZE(vsock_bind_table); i++) 204d021c344SAndy King INIT_LIST_HEAD(&vsock_bind_table[i]); 205d021c344SAndy King 206d021c344SAndy King for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) 207d021c344SAndy King INIT_LIST_HEAD(&vsock_connected_table[i]); 208c1eef220SCong Wang return 0; 209d021c344SAndy King } 210d021c344SAndy King 211d021c344SAndy King static void __vsock_insert_bound(struct list_head *list, 212d021c344SAndy King struct vsock_sock *vsk) 213d021c344SAndy King { 214d021c344SAndy King sock_hold(&vsk->sk); 215d021c344SAndy King list_add(&vsk->bound_table, list); 216d021c344SAndy King } 217d021c344SAndy King 218d021c344SAndy King static void __vsock_insert_connected(struct list_head *list, 219d021c344SAndy King struct vsock_sock *vsk) 220d021c344SAndy King { 221d021c344SAndy King sock_hold(&vsk->sk); 222d021c344SAndy King list_add(&vsk->connected_table, list); 223d021c344SAndy King } 224d021c344SAndy King 225d021c344SAndy King static void __vsock_remove_bound(struct vsock_sock *vsk) 226d021c344SAndy King { 227d021c344SAndy King list_del_init(&vsk->bound_table); 228d021c344SAndy King sock_put(&vsk->sk); 229d021c344SAndy King } 230d021c344SAndy King 231d021c344SAndy King static void __vsock_remove_connected(struct vsock_sock *vsk) 232d021c344SAndy King { 233d021c344SAndy King list_del_init(&vsk->connected_table); 234d021c344SAndy King sock_put(&vsk->sk); 235d021c344SAndy King } 236d021c344SAndy King 237d021c344SAndy King static struct sock *__vsock_find_bound_socket(struct sockaddr_vm *addr) 238d021c344SAndy King { 239d021c344SAndy King struct vsock_sock *vsk; 240d021c344SAndy King 241d021c344SAndy King list_for_each_entry(vsk, vsock_bound_sockets(addr), bound_table) 242990454b5SReilly Grant if (addr->svm_port == vsk->local_addr.svm_port) 243d021c344SAndy King return sk_vsock(vsk); 244d021c344SAndy King 245d021c344SAndy King return NULL; 246d021c344SAndy King } 247d021c344SAndy King 248d021c344SAndy King static struct sock *__vsock_find_connected_socket(struct sockaddr_vm *src, 249d021c344SAndy King struct sockaddr_vm *dst) 250d021c344SAndy King { 251d021c344SAndy King struct vsock_sock *vsk; 252d021c344SAndy King 253d021c344SAndy King list_for_each_entry(vsk, vsock_connected_sockets(src, dst), 254d021c344SAndy King connected_table) { 255990454b5SReilly Grant if (vsock_addr_equals_addr(src, &vsk->remote_addr) && 256990454b5SReilly Grant dst->svm_port == vsk->local_addr.svm_port) { 257d021c344SAndy King return sk_vsock(vsk); 258d021c344SAndy King } 259d021c344SAndy King } 260d021c344SAndy King 261d021c344SAndy King return NULL; 262d021c344SAndy King } 263d021c344SAndy King 264d021c344SAndy King static void vsock_insert_unbound(struct vsock_sock *vsk) 265d021c344SAndy King { 266d021c344SAndy King spin_lock_bh(&vsock_table_lock); 267d021c344SAndy King __vsock_insert_bound(vsock_unbound_sockets, vsk); 268d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 269d021c344SAndy King } 270d021c344SAndy King 271d021c344SAndy King void vsock_insert_connected(struct vsock_sock *vsk) 272d021c344SAndy King { 273d021c344SAndy King struct list_head *list = vsock_connected_sockets( 274d021c344SAndy King &vsk->remote_addr, &vsk->local_addr); 275d021c344SAndy King 276d021c344SAndy King spin_lock_bh(&vsock_table_lock); 277d021c344SAndy King __vsock_insert_connected(list, vsk); 278d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 279d021c344SAndy King } 280d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_insert_connected); 281d021c344SAndy King 282d021c344SAndy King void vsock_remove_bound(struct vsock_sock *vsk) 283d021c344SAndy King { 284d021c344SAndy King spin_lock_bh(&vsock_table_lock); 285d021c344SAndy King __vsock_remove_bound(vsk); 286d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 287d021c344SAndy King } 288d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_remove_bound); 289d021c344SAndy King 290d021c344SAndy King void vsock_remove_connected(struct vsock_sock *vsk) 291d021c344SAndy King { 292d021c344SAndy King spin_lock_bh(&vsock_table_lock); 293d021c344SAndy King __vsock_remove_connected(vsk); 294d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 295d021c344SAndy King } 296d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_remove_connected); 297d021c344SAndy King 298d021c344SAndy King struct sock *vsock_find_bound_socket(struct sockaddr_vm *addr) 299d021c344SAndy King { 300d021c344SAndy King struct sock *sk; 301d021c344SAndy King 302d021c344SAndy King spin_lock_bh(&vsock_table_lock); 303d021c344SAndy King sk = __vsock_find_bound_socket(addr); 304d021c344SAndy King if (sk) 305d021c344SAndy King sock_hold(sk); 306d021c344SAndy King 307d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 308d021c344SAndy King 309d021c344SAndy King return sk; 310d021c344SAndy King } 311d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_find_bound_socket); 312d021c344SAndy King 313d021c344SAndy King struct sock *vsock_find_connected_socket(struct sockaddr_vm *src, 314d021c344SAndy King struct sockaddr_vm *dst) 315d021c344SAndy King { 316d021c344SAndy King struct sock *sk; 317d021c344SAndy King 318d021c344SAndy King spin_lock_bh(&vsock_table_lock); 319d021c344SAndy King sk = __vsock_find_connected_socket(src, dst); 320d021c344SAndy King if (sk) 321d021c344SAndy King sock_hold(sk); 322d021c344SAndy King 323d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 324d021c344SAndy King 325d021c344SAndy King return sk; 326d021c344SAndy King } 327d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_find_connected_socket); 328d021c344SAndy King 329d021c344SAndy King static bool vsock_in_bound_table(struct vsock_sock *vsk) 330d021c344SAndy King { 331d021c344SAndy King bool ret; 332d021c344SAndy King 333d021c344SAndy King spin_lock_bh(&vsock_table_lock); 334d021c344SAndy King ret = __vsock_in_bound_table(vsk); 335d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 336d021c344SAndy King 337d021c344SAndy King return ret; 338d021c344SAndy King } 339d021c344SAndy King 340d021c344SAndy King static bool vsock_in_connected_table(struct vsock_sock *vsk) 341d021c344SAndy King { 342d021c344SAndy King bool ret; 343d021c344SAndy King 344d021c344SAndy King spin_lock_bh(&vsock_table_lock); 345d021c344SAndy King ret = __vsock_in_connected_table(vsk); 346d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 347d021c344SAndy King 348d021c344SAndy King return ret; 349d021c344SAndy King } 350d021c344SAndy King 3516773b7dcSStefan Hajnoczi void vsock_remove_sock(struct vsock_sock *vsk) 3526773b7dcSStefan Hajnoczi { 3536773b7dcSStefan Hajnoczi if (vsock_in_bound_table(vsk)) 3546773b7dcSStefan Hajnoczi vsock_remove_bound(vsk); 3556773b7dcSStefan Hajnoczi 3566773b7dcSStefan Hajnoczi if (vsock_in_connected_table(vsk)) 3576773b7dcSStefan Hajnoczi vsock_remove_connected(vsk); 3586773b7dcSStefan Hajnoczi } 3596773b7dcSStefan Hajnoczi EXPORT_SYMBOL_GPL(vsock_remove_sock); 3606773b7dcSStefan Hajnoczi 361d021c344SAndy King void vsock_for_each_connected_socket(void (*fn)(struct sock *sk)) 362d021c344SAndy King { 363d021c344SAndy King int i; 364d021c344SAndy King 365d021c344SAndy King spin_lock_bh(&vsock_table_lock); 366d021c344SAndy King 367d021c344SAndy King for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) { 368d021c344SAndy King struct vsock_sock *vsk; 369d021c344SAndy King list_for_each_entry(vsk, &vsock_connected_table[i], 370d9af2d67SJulia Lawall connected_table) 371d021c344SAndy King fn(sk_vsock(vsk)); 372d021c344SAndy King } 373d021c344SAndy King 374d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 375d021c344SAndy King } 376d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_for_each_connected_socket); 377d021c344SAndy King 378d021c344SAndy King void vsock_add_pending(struct sock *listener, struct sock *pending) 379d021c344SAndy King { 380d021c344SAndy King struct vsock_sock *vlistener; 381d021c344SAndy King struct vsock_sock *vpending; 382d021c344SAndy King 383d021c344SAndy King vlistener = vsock_sk(listener); 384d021c344SAndy King vpending = vsock_sk(pending); 385d021c344SAndy King 386d021c344SAndy King sock_hold(pending); 387d021c344SAndy King sock_hold(listener); 388d021c344SAndy King list_add_tail(&vpending->pending_links, &vlistener->pending_links); 389d021c344SAndy King } 390d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_add_pending); 391d021c344SAndy King 392d021c344SAndy King void vsock_remove_pending(struct sock *listener, struct sock *pending) 393d021c344SAndy King { 394d021c344SAndy King struct vsock_sock *vpending = vsock_sk(pending); 395d021c344SAndy King 396d021c344SAndy King list_del_init(&vpending->pending_links); 397d021c344SAndy King sock_put(listener); 398d021c344SAndy King sock_put(pending); 399d021c344SAndy King } 400d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_remove_pending); 401d021c344SAndy King 402d021c344SAndy King void vsock_enqueue_accept(struct sock *listener, struct sock *connected) 403d021c344SAndy King { 404d021c344SAndy King struct vsock_sock *vlistener; 405d021c344SAndy King struct vsock_sock *vconnected; 406d021c344SAndy King 407d021c344SAndy King vlistener = vsock_sk(listener); 408d021c344SAndy King vconnected = vsock_sk(connected); 409d021c344SAndy King 410d021c344SAndy King sock_hold(connected); 411d021c344SAndy King sock_hold(listener); 412d021c344SAndy King list_add_tail(&vconnected->accept_queue, &vlistener->accept_queue); 413d021c344SAndy King } 414d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_enqueue_accept); 415d021c344SAndy King 416d021c344SAndy King static struct sock *vsock_dequeue_accept(struct sock *listener) 417d021c344SAndy King { 418d021c344SAndy King struct vsock_sock *vlistener; 419d021c344SAndy King struct vsock_sock *vconnected; 420d021c344SAndy King 421d021c344SAndy King vlistener = vsock_sk(listener); 422d021c344SAndy King 423d021c344SAndy King if (list_empty(&vlistener->accept_queue)) 424d021c344SAndy King return NULL; 425d021c344SAndy King 426d021c344SAndy King vconnected = list_entry(vlistener->accept_queue.next, 427d021c344SAndy King struct vsock_sock, accept_queue); 428d021c344SAndy King 429d021c344SAndy King list_del_init(&vconnected->accept_queue); 430d021c344SAndy King sock_put(listener); 431d021c344SAndy King /* The caller will need a reference on the connected socket so we let 432d021c344SAndy King * it call sock_put(). 433d021c344SAndy King */ 434d021c344SAndy King 435d021c344SAndy King return sk_vsock(vconnected); 436d021c344SAndy King } 437d021c344SAndy King 438d021c344SAndy King static bool vsock_is_accept_queue_empty(struct sock *sk) 439d021c344SAndy King { 440d021c344SAndy King struct vsock_sock *vsk = vsock_sk(sk); 441d021c344SAndy King return list_empty(&vsk->accept_queue); 442d021c344SAndy King } 443d021c344SAndy King 444d021c344SAndy King static bool vsock_is_pending(struct sock *sk) 445d021c344SAndy King { 446d021c344SAndy King struct vsock_sock *vsk = vsock_sk(sk); 447d021c344SAndy King return !list_empty(&vsk->pending_links); 448d021c344SAndy King } 449d021c344SAndy King 450d021c344SAndy King static int vsock_send_shutdown(struct sock *sk, int mode) 451d021c344SAndy King { 452d021c344SAndy King return transport->shutdown(vsock_sk(sk), mode); 453d021c344SAndy King } 454d021c344SAndy King 455455f05ecSCong Wang static void vsock_pending_work(struct work_struct *work) 456d021c344SAndy King { 457d021c344SAndy King struct sock *sk; 458d021c344SAndy King struct sock *listener; 459d021c344SAndy King struct vsock_sock *vsk; 460d021c344SAndy King bool cleanup; 461d021c344SAndy King 462455f05ecSCong Wang vsk = container_of(work, struct vsock_sock, pending_work.work); 463d021c344SAndy King sk = sk_vsock(vsk); 464d021c344SAndy King listener = vsk->listener; 465d021c344SAndy King cleanup = true; 466d021c344SAndy King 467d021c344SAndy King lock_sock(listener); 4684192f672SStefan Hajnoczi lock_sock_nested(sk, SINGLE_DEPTH_NESTING); 469d021c344SAndy King 470d021c344SAndy King if (vsock_is_pending(sk)) { 471d021c344SAndy King vsock_remove_pending(listener, sk); 4721190cfdbSJorgen Hansen 4731190cfdbSJorgen Hansen listener->sk_ack_backlog--; 474d021c344SAndy King } else if (!vsk->rejected) { 475d021c344SAndy King /* We are not on the pending list and accept() did not reject 476d021c344SAndy King * us, so we must have been accepted by our user process. We 477d021c344SAndy King * just need to drop our references to the sockets and be on 478d021c344SAndy King * our way. 479d021c344SAndy King */ 480d021c344SAndy King cleanup = false; 481d021c344SAndy King goto out; 482d021c344SAndy King } 483d021c344SAndy King 484d021c344SAndy King /* We need to remove ourself from the global connected sockets list so 485d021c344SAndy King * incoming packets can't find this socket, and to reduce the reference 486d021c344SAndy King * count. 487d021c344SAndy King */ 488d021c344SAndy King if (vsock_in_connected_table(vsk)) 489d021c344SAndy King vsock_remove_connected(vsk); 490d021c344SAndy King 4913b4477d2SStefan Hajnoczi sk->sk_state = TCP_CLOSE; 492d021c344SAndy King 493d021c344SAndy King out: 494d021c344SAndy King release_sock(sk); 495d021c344SAndy King release_sock(listener); 496d021c344SAndy King if (cleanup) 497d021c344SAndy King sock_put(sk); 498d021c344SAndy King 499d021c344SAndy King sock_put(sk); 500d021c344SAndy King sock_put(listener); 501d021c344SAndy King } 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 { 508a22d3251SLepton Wu static u32 port; 509d021c344SAndy King struct sockaddr_vm new_addr; 510d021c344SAndy King 5118236b08cSLepton Wu if (!port) 5128236b08cSLepton Wu port = LAST_RESERVED_PORT + 1 + 5138236b08cSLepton Wu prandom_u32_max(U32_MAX - LAST_RESERVED_PORT); 5148236b08cSLepton Wu 515d021c344SAndy King vsock_addr_init(&new_addr, addr->svm_cid, addr->svm_port); 516d021c344SAndy King 517d021c344SAndy King if (addr->svm_port == VMADDR_PORT_ANY) { 518d021c344SAndy King bool found = false; 519d021c344SAndy King unsigned int i; 520d021c344SAndy King 521d021c344SAndy King for (i = 0; i < MAX_PORT_RETRIES; i++) { 522d021c344SAndy King if (port <= LAST_RESERVED_PORT) 523d021c344SAndy King port = LAST_RESERVED_PORT + 1; 524d021c344SAndy King 525d021c344SAndy King new_addr.svm_port = port++; 526d021c344SAndy King 527d021c344SAndy King if (!__vsock_find_bound_socket(&new_addr)) { 528d021c344SAndy King found = true; 529d021c344SAndy King break; 530d021c344SAndy King } 531d021c344SAndy King } 532d021c344SAndy King 533d021c344SAndy King if (!found) 534d021c344SAndy King return -EADDRNOTAVAIL; 535d021c344SAndy King } else { 536d021c344SAndy King /* If port is in reserved range, ensure caller 537d021c344SAndy King * has necessary privileges. 538d021c344SAndy King */ 539d021c344SAndy King if (addr->svm_port <= LAST_RESERVED_PORT && 540d021c344SAndy King !capable(CAP_NET_BIND_SERVICE)) { 541d021c344SAndy King return -EACCES; 542d021c344SAndy King } 543d021c344SAndy King 544d021c344SAndy King if (__vsock_find_bound_socket(&new_addr)) 545d021c344SAndy King return -EADDRINUSE; 546d021c344SAndy King } 547d021c344SAndy King 548d021c344SAndy King vsock_addr_init(&vsk->local_addr, new_addr.svm_cid, new_addr.svm_port); 549d021c344SAndy King 550d021c344SAndy King /* Remove stream sockets from the unbound list and add them to the hash 551d021c344SAndy King * table for easy lookup by its address. The unbound list is simply an 552d021c344SAndy King * extra entry at the end of the hash table, a trick used by AF_UNIX. 553d021c344SAndy King */ 554d021c344SAndy King __vsock_remove_bound(vsk); 555d021c344SAndy King __vsock_insert_bound(vsock_bound_sockets(&vsk->local_addr), vsk); 556d021c344SAndy King 557d021c344SAndy King return 0; 558d021c344SAndy King } 559d021c344SAndy King 560d021c344SAndy King static int __vsock_bind_dgram(struct vsock_sock *vsk, 561d021c344SAndy King struct sockaddr_vm *addr) 562d021c344SAndy King { 563d021c344SAndy King return transport->dgram_bind(vsk, addr); 564d021c344SAndy King } 565d021c344SAndy King 566d021c344SAndy King static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr) 567d021c344SAndy King { 568d021c344SAndy King struct vsock_sock *vsk = vsock_sk(sk); 569d021c344SAndy King u32 cid; 570d021c344SAndy King int retval; 571d021c344SAndy King 572d021c344SAndy King /* First ensure this socket isn't already bound. */ 573d021c344SAndy King if (vsock_addr_bound(&vsk->local_addr)) 574d021c344SAndy King return -EINVAL; 575d021c344SAndy King 576d021c344SAndy King /* Now bind to the provided address or select appropriate values if 577d021c344SAndy King * none are provided (VMADDR_CID_ANY and VMADDR_PORT_ANY). Note that 578d021c344SAndy King * like AF_INET prevents binding to a non-local IP address (in most 579d021c344SAndy King * cases), we only allow binding to the local CID. 580d021c344SAndy King */ 581d021c344SAndy King cid = transport->get_local_cid(); 582d021c344SAndy King if (addr->svm_cid != cid && addr->svm_cid != VMADDR_CID_ANY) 583d021c344SAndy King return -EADDRNOTAVAIL; 584d021c344SAndy King 585d021c344SAndy King switch (sk->sk_socket->type) { 586d021c344SAndy King case SOCK_STREAM: 587d021c344SAndy King spin_lock_bh(&vsock_table_lock); 588d021c344SAndy King retval = __vsock_bind_stream(vsk, addr); 589d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 590d021c344SAndy King break; 591d021c344SAndy King 592d021c344SAndy King case SOCK_DGRAM: 593d021c344SAndy King retval = __vsock_bind_dgram(vsk, addr); 594d021c344SAndy King break; 595d021c344SAndy King 596d021c344SAndy King default: 597d021c344SAndy King retval = -EINVAL; 598d021c344SAndy King break; 599d021c344SAndy King } 600d021c344SAndy King 601d021c344SAndy King return retval; 602d021c344SAndy King } 603d021c344SAndy King 604455f05ecSCong Wang static void vsock_connect_timeout(struct work_struct *work); 605455f05ecSCong Wang 606d021c344SAndy King struct sock *__vsock_create(struct net *net, 607d021c344SAndy King struct socket *sock, 608d021c344SAndy King struct sock *parent, 609d021c344SAndy King gfp_t priority, 61011aa9c28SEric W. Biederman unsigned short type, 61111aa9c28SEric W. Biederman int kern) 612d021c344SAndy King { 613d021c344SAndy King struct sock *sk; 614d021c344SAndy King struct vsock_sock *psk; 615d021c344SAndy King struct vsock_sock *vsk; 616d021c344SAndy King 61711aa9c28SEric W. Biederman sk = sk_alloc(net, AF_VSOCK, priority, &vsock_proto, kern); 618d021c344SAndy King if (!sk) 619d021c344SAndy King return NULL; 620d021c344SAndy King 621d021c344SAndy King sock_init_data(sock, sk); 622d021c344SAndy King 623d021c344SAndy King /* sk->sk_type is normally set in sock_init_data, but only if sock is 624d021c344SAndy King * non-NULL. We make sure that our sockets always have a type by 625d021c344SAndy King * setting it here if needed. 626d021c344SAndy King */ 627d021c344SAndy King if (!sock) 628d021c344SAndy King sk->sk_type = type; 629d021c344SAndy King 630d021c344SAndy King vsk = vsock_sk(sk); 631d021c344SAndy King vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 632d021c344SAndy King vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 633d021c344SAndy King 634d021c344SAndy King sk->sk_destruct = vsock_sk_destruct; 635d021c344SAndy King sk->sk_backlog_rcv = vsock_queue_rcv_skb; 636d021c344SAndy King sock_reset_flag(sk, SOCK_DONE); 637d021c344SAndy King 638d021c344SAndy King INIT_LIST_HEAD(&vsk->bound_table); 639d021c344SAndy King INIT_LIST_HEAD(&vsk->connected_table); 640d021c344SAndy King vsk->listener = NULL; 641d021c344SAndy King INIT_LIST_HEAD(&vsk->pending_links); 642d021c344SAndy King INIT_LIST_HEAD(&vsk->accept_queue); 643d021c344SAndy King vsk->rejected = false; 644d021c344SAndy King vsk->sent_request = false; 645d021c344SAndy King vsk->ignore_connecting_rst = false; 646d021c344SAndy King vsk->peer_shutdown = 0; 647455f05ecSCong Wang INIT_DELAYED_WORK(&vsk->connect_work, vsock_connect_timeout); 648455f05ecSCong Wang INIT_DELAYED_WORK(&vsk->pending_work, vsock_pending_work); 649d021c344SAndy King 650d021c344SAndy King psk = parent ? vsock_sk(parent) : NULL; 651d021c344SAndy King if (parent) { 652d021c344SAndy King vsk->trusted = psk->trusted; 653d021c344SAndy King vsk->owner = get_cred(psk->owner); 654d021c344SAndy King vsk->connect_timeout = psk->connect_timeout; 655d021c344SAndy King } else { 656d021c344SAndy King vsk->trusted = capable(CAP_NET_ADMIN); 657d021c344SAndy King vsk->owner = get_current_cred(); 658d021c344SAndy King vsk->connect_timeout = VSOCK_DEFAULT_CONNECT_TIMEOUT; 659d021c344SAndy King } 660d021c344SAndy King 661d021c344SAndy King if (transport->init(vsk, psk) < 0) { 662d021c344SAndy King sk_free(sk); 663d021c344SAndy King return NULL; 664d021c344SAndy King } 665d021c344SAndy King 666d021c344SAndy King if (sock) 667d021c344SAndy King vsock_insert_unbound(vsk); 668d021c344SAndy King 669d021c344SAndy King return sk; 670d021c344SAndy King } 671d021c344SAndy King EXPORT_SYMBOL_GPL(__vsock_create); 672d021c344SAndy King 673d021c344SAndy King static void __vsock_release(struct sock *sk) 674d021c344SAndy King { 675d021c344SAndy King if (sk) { 676d021c344SAndy King struct sk_buff *skb; 677d021c344SAndy King struct sock *pending; 678d021c344SAndy King struct vsock_sock *vsk; 679d021c344SAndy King 680d021c344SAndy King vsk = vsock_sk(sk); 681d021c344SAndy King pending = NULL; /* Compiler warning. */ 682d021c344SAndy King 683d021c344SAndy King transport->release(vsk); 684d021c344SAndy King 685d021c344SAndy King lock_sock(sk); 686d021c344SAndy King sock_orphan(sk); 687d021c344SAndy King sk->sk_shutdown = SHUTDOWN_MASK; 688d021c344SAndy King 689d021c344SAndy King while ((skb = skb_dequeue(&sk->sk_receive_queue))) 690d021c344SAndy King kfree_skb(skb); 691d021c344SAndy King 692d021c344SAndy King /* Clean up any sockets that never were accepted. */ 693d021c344SAndy King while ((pending = vsock_dequeue_accept(sk)) != NULL) { 694d021c344SAndy King __vsock_release(pending); 695d021c344SAndy King sock_put(pending); 696d021c344SAndy King } 697d021c344SAndy King 698d021c344SAndy King release_sock(sk); 699d021c344SAndy King sock_put(sk); 700d021c344SAndy King } 701d021c344SAndy King } 702d021c344SAndy King 703d021c344SAndy King static void vsock_sk_destruct(struct sock *sk) 704d021c344SAndy King { 705d021c344SAndy King struct vsock_sock *vsk = vsock_sk(sk); 706d021c344SAndy King 707d021c344SAndy King transport->destruct(vsk); 708d021c344SAndy King 709d021c344SAndy King /* When clearing these addresses, there's no need to set the family and 710d021c344SAndy King * possibly register the address family with the kernel. 711d021c344SAndy King */ 712d021c344SAndy King vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 713d021c344SAndy King vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 714d021c344SAndy King 715d021c344SAndy King put_cred(vsk->owner); 716d021c344SAndy King } 717d021c344SAndy King 718d021c344SAndy King static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) 719d021c344SAndy King { 720d021c344SAndy King int err; 721d021c344SAndy King 722d021c344SAndy King err = sock_queue_rcv_skb(sk, skb); 723d021c344SAndy King if (err) 724d021c344SAndy King kfree_skb(skb); 725d021c344SAndy King 726d021c344SAndy King return err; 727d021c344SAndy King } 728d021c344SAndy King 729d021c344SAndy King s64 vsock_stream_has_data(struct vsock_sock *vsk) 730d021c344SAndy King { 731d021c344SAndy King return transport->stream_has_data(vsk); 732d021c344SAndy King } 733d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_stream_has_data); 734d021c344SAndy King 735d021c344SAndy King s64 vsock_stream_has_space(struct vsock_sock *vsk) 736d021c344SAndy King { 737d021c344SAndy King return transport->stream_has_space(vsk); 738d021c344SAndy King } 739d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_stream_has_space); 740d021c344SAndy King 741d021c344SAndy King static int vsock_release(struct socket *sock) 742d021c344SAndy King { 743d021c344SAndy King __vsock_release(sock->sk); 744d021c344SAndy King sock->sk = NULL; 745d021c344SAndy King sock->state = SS_FREE; 746d021c344SAndy King 747d021c344SAndy King return 0; 748d021c344SAndy King } 749d021c344SAndy King 750d021c344SAndy King static int 751d021c344SAndy King vsock_bind(struct socket *sock, struct sockaddr *addr, int addr_len) 752d021c344SAndy King { 753d021c344SAndy King int err; 754d021c344SAndy King struct sock *sk; 755d021c344SAndy King struct sockaddr_vm *vm_addr; 756d021c344SAndy King 757d021c344SAndy King sk = sock->sk; 758d021c344SAndy King 759d021c344SAndy King if (vsock_addr_cast(addr, addr_len, &vm_addr) != 0) 760d021c344SAndy King return -EINVAL; 761d021c344SAndy King 762d021c344SAndy King lock_sock(sk); 763d021c344SAndy King err = __vsock_bind(sk, vm_addr); 764d021c344SAndy King release_sock(sk); 765d021c344SAndy King 766d021c344SAndy King return err; 767d021c344SAndy King } 768d021c344SAndy King 769d021c344SAndy King static int vsock_getname(struct socket *sock, 7709b2c45d4SDenys Vlasenko struct sockaddr *addr, int peer) 771d021c344SAndy King { 772d021c344SAndy King int err; 773d021c344SAndy King struct sock *sk; 774d021c344SAndy King struct vsock_sock *vsk; 775d021c344SAndy King struct sockaddr_vm *vm_addr; 776d021c344SAndy King 777d021c344SAndy King sk = sock->sk; 778d021c344SAndy King vsk = vsock_sk(sk); 779d021c344SAndy King err = 0; 780d021c344SAndy King 781d021c344SAndy King lock_sock(sk); 782d021c344SAndy King 783d021c344SAndy King if (peer) { 784d021c344SAndy King if (sock->state != SS_CONNECTED) { 785d021c344SAndy King err = -ENOTCONN; 786d021c344SAndy King goto out; 787d021c344SAndy King } 788d021c344SAndy King vm_addr = &vsk->remote_addr; 789d021c344SAndy King } else { 790d021c344SAndy King vm_addr = &vsk->local_addr; 791d021c344SAndy King } 792d021c344SAndy King 793d021c344SAndy King if (!vm_addr) { 794d021c344SAndy King err = -EINVAL; 795d021c344SAndy King goto out; 796d021c344SAndy King } 797d021c344SAndy King 798d021c344SAndy King /* sys_getsockname() and sys_getpeername() pass us a 799d021c344SAndy King * MAX_SOCK_ADDR-sized buffer and don't set addr_len. Unfortunately 800d021c344SAndy King * that macro is defined in socket.c instead of .h, so we hardcode its 801d021c344SAndy King * value here. 802d021c344SAndy King */ 803d021c344SAndy King BUILD_BUG_ON(sizeof(*vm_addr) > 128); 804d021c344SAndy King memcpy(addr, vm_addr, sizeof(*vm_addr)); 8059b2c45d4SDenys Vlasenko err = sizeof(*vm_addr); 806d021c344SAndy King 807d021c344SAndy King out: 808d021c344SAndy King release_sock(sk); 809d021c344SAndy King return err; 810d021c344SAndy King } 811d021c344SAndy King 812d021c344SAndy King static int vsock_shutdown(struct socket *sock, int mode) 813d021c344SAndy King { 814d021c344SAndy King int err; 815d021c344SAndy King struct sock *sk; 816d021c344SAndy King 817d021c344SAndy King /* User level uses SHUT_RD (0) and SHUT_WR (1), but the kernel uses 818d021c344SAndy King * RCV_SHUTDOWN (1) and SEND_SHUTDOWN (2), so we must increment mode 819d021c344SAndy King * here like the other address families do. Note also that the 820d021c344SAndy King * increment makes SHUT_RDWR (2) into RCV_SHUTDOWN | SEND_SHUTDOWN (3), 821d021c344SAndy King * which is what we want. 822d021c344SAndy King */ 823d021c344SAndy King mode++; 824d021c344SAndy King 825d021c344SAndy King if ((mode & ~SHUTDOWN_MASK) || !mode) 826d021c344SAndy King return -EINVAL; 827d021c344SAndy King 828d021c344SAndy King /* If this is a STREAM socket and it is not connected then bail out 829d021c344SAndy King * immediately. If it is a DGRAM socket then we must first kick the 830d021c344SAndy King * socket so that it wakes up from any sleeping calls, for example 831d021c344SAndy King * recv(), and then afterwards return the error. 832d021c344SAndy King */ 833d021c344SAndy King 834d021c344SAndy King sk = sock->sk; 835d021c344SAndy King if (sock->state == SS_UNCONNECTED) { 836d021c344SAndy King err = -ENOTCONN; 837d021c344SAndy King if (sk->sk_type == SOCK_STREAM) 838d021c344SAndy King return err; 839d021c344SAndy King } else { 840d021c344SAndy King sock->state = SS_DISCONNECTING; 841d021c344SAndy King err = 0; 842d021c344SAndy King } 843d021c344SAndy King 844d021c344SAndy King /* Receive and send shutdowns are treated alike. */ 845d021c344SAndy King mode = mode & (RCV_SHUTDOWN | SEND_SHUTDOWN); 846d021c344SAndy King if (mode) { 847d021c344SAndy King lock_sock(sk); 848d021c344SAndy King sk->sk_shutdown |= mode; 849d021c344SAndy King sk->sk_state_change(sk); 850d021c344SAndy King release_sock(sk); 851d021c344SAndy King 852d021c344SAndy King if (sk->sk_type == SOCK_STREAM) { 853d021c344SAndy King sock_reset_flag(sk, SOCK_DONE); 854d021c344SAndy King vsock_send_shutdown(sk, mode); 855d021c344SAndy King } 856d021c344SAndy King } 857d021c344SAndy King 858d021c344SAndy King return err; 859d021c344SAndy King } 860d021c344SAndy King 861a11e1d43SLinus Torvalds static __poll_t vsock_poll(struct file *file, struct socket *sock, 862a11e1d43SLinus Torvalds poll_table *wait) 863d021c344SAndy King { 864a11e1d43SLinus Torvalds struct sock *sk; 865a11e1d43SLinus Torvalds __poll_t mask; 866a11e1d43SLinus Torvalds struct vsock_sock *vsk; 867a11e1d43SLinus Torvalds 868a11e1d43SLinus Torvalds sk = sock->sk; 869a11e1d43SLinus Torvalds vsk = vsock_sk(sk); 870a11e1d43SLinus Torvalds 871a11e1d43SLinus Torvalds poll_wait(file, sk_sleep(sk), wait); 872a11e1d43SLinus Torvalds mask = 0; 873d021c344SAndy King 874d021c344SAndy King if (sk->sk_err) 875d021c344SAndy King /* Signify that there has been an error on this socket. */ 876a9a08845SLinus Torvalds mask |= EPOLLERR; 877d021c344SAndy King 878d021c344SAndy King /* INET sockets treat local write shutdown and peer write shutdown as a 879a9a08845SLinus Torvalds * case of EPOLLHUP set. 880d021c344SAndy King */ 881d021c344SAndy King if ((sk->sk_shutdown == SHUTDOWN_MASK) || 882d021c344SAndy King ((sk->sk_shutdown & SEND_SHUTDOWN) && 883d021c344SAndy King (vsk->peer_shutdown & SEND_SHUTDOWN))) { 884a9a08845SLinus Torvalds mask |= EPOLLHUP; 885d021c344SAndy King } 886d021c344SAndy King 887d021c344SAndy King if (sk->sk_shutdown & RCV_SHUTDOWN || 888d021c344SAndy King vsk->peer_shutdown & SEND_SHUTDOWN) { 889a9a08845SLinus Torvalds mask |= EPOLLRDHUP; 890d021c344SAndy King } 891d021c344SAndy King 892d021c344SAndy King if (sock->type == SOCK_DGRAM) { 893d021c344SAndy King /* For datagram sockets we can read if there is something in 894d021c344SAndy King * the queue and write as long as the socket isn't shutdown for 895d021c344SAndy King * sending. 896d021c344SAndy King */ 897d021c344SAndy King if (!skb_queue_empty(&sk->sk_receive_queue) || 898d021c344SAndy King (sk->sk_shutdown & RCV_SHUTDOWN)) { 899a9a08845SLinus Torvalds mask |= EPOLLIN | EPOLLRDNORM; 900d021c344SAndy King } 901d021c344SAndy King 902d021c344SAndy King if (!(sk->sk_shutdown & SEND_SHUTDOWN)) 903a9a08845SLinus Torvalds mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND; 904d021c344SAndy King 905d021c344SAndy King } else if (sock->type == SOCK_STREAM) { 906d021c344SAndy King lock_sock(sk); 907d021c344SAndy King 908d021c344SAndy King /* Listening sockets that have connections in their accept 909d021c344SAndy King * queue can be read. 910d021c344SAndy King */ 9113b4477d2SStefan Hajnoczi if (sk->sk_state == TCP_LISTEN 912d021c344SAndy King && !vsock_is_accept_queue_empty(sk)) 913a9a08845SLinus Torvalds mask |= EPOLLIN | EPOLLRDNORM; 914d021c344SAndy King 915d021c344SAndy King /* If there is something in the queue then we can read. */ 916d021c344SAndy King if (transport->stream_is_active(vsk) && 917d021c344SAndy King !(sk->sk_shutdown & RCV_SHUTDOWN)) { 918d021c344SAndy King bool data_ready_now = false; 919d021c344SAndy King int ret = transport->notify_poll_in( 920d021c344SAndy King vsk, 1, &data_ready_now); 921d021c344SAndy King if (ret < 0) { 922a9a08845SLinus Torvalds mask |= EPOLLERR; 923d021c344SAndy King } else { 924d021c344SAndy King if (data_ready_now) 925a9a08845SLinus Torvalds mask |= EPOLLIN | EPOLLRDNORM; 926d021c344SAndy King 927d021c344SAndy King } 928d021c344SAndy King } 929d021c344SAndy King 930d021c344SAndy King /* Sockets whose connections have been closed, reset, or 931d021c344SAndy King * terminated should also be considered read, and we check the 932d021c344SAndy King * shutdown flag for that. 933d021c344SAndy King */ 934d021c344SAndy King if (sk->sk_shutdown & RCV_SHUTDOWN || 935d021c344SAndy King vsk->peer_shutdown & SEND_SHUTDOWN) { 936a9a08845SLinus Torvalds mask |= EPOLLIN | EPOLLRDNORM; 937d021c344SAndy King } 938d021c344SAndy King 939d021c344SAndy King /* Connected sockets that can produce data can be written. */ 9403b4477d2SStefan Hajnoczi if (sk->sk_state == TCP_ESTABLISHED) { 941d021c344SAndy King if (!(sk->sk_shutdown & SEND_SHUTDOWN)) { 942d021c344SAndy King bool space_avail_now = false; 943d021c344SAndy King int ret = transport->notify_poll_out( 944d021c344SAndy King vsk, 1, &space_avail_now); 945d021c344SAndy King if (ret < 0) { 946a9a08845SLinus Torvalds mask |= EPOLLERR; 947d021c344SAndy King } else { 948d021c344SAndy King if (space_avail_now) 949a9a08845SLinus Torvalds /* Remove EPOLLWRBAND since INET 950d021c344SAndy King * sockets are not setting it. 951d021c344SAndy King */ 952a9a08845SLinus Torvalds mask |= EPOLLOUT | EPOLLWRNORM; 953d021c344SAndy King 954d021c344SAndy King } 955d021c344SAndy King } 956d021c344SAndy King } 957d021c344SAndy King 958d021c344SAndy King /* Simulate INET socket poll behaviors, which sets 959a9a08845SLinus Torvalds * EPOLLOUT|EPOLLWRNORM when peer is closed and nothing to read, 960d021c344SAndy King * but local send is not shutdown. 961d021c344SAndy King */ 962ba3169fcSStefan Hajnoczi if (sk->sk_state == TCP_CLOSE || sk->sk_state == TCP_CLOSING) { 963d021c344SAndy King if (!(sk->sk_shutdown & SEND_SHUTDOWN)) 964a9a08845SLinus Torvalds mask |= EPOLLOUT | EPOLLWRNORM; 965d021c344SAndy King 966d021c344SAndy King } 967d021c344SAndy King 968d021c344SAndy King release_sock(sk); 969d021c344SAndy King } 970d021c344SAndy King 971d021c344SAndy King return mask; 972d021c344SAndy King } 973d021c344SAndy King 9741b784140SYing Xue static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg, 9751b784140SYing Xue size_t len) 976d021c344SAndy King { 977d021c344SAndy King int err; 978d021c344SAndy King struct sock *sk; 979d021c344SAndy King struct vsock_sock *vsk; 980d021c344SAndy King struct sockaddr_vm *remote_addr; 981d021c344SAndy King 982d021c344SAndy King if (msg->msg_flags & MSG_OOB) 983d021c344SAndy King return -EOPNOTSUPP; 984d021c344SAndy King 985d021c344SAndy King /* For now, MSG_DONTWAIT is always assumed... */ 986d021c344SAndy King err = 0; 987d021c344SAndy King sk = sock->sk; 988d021c344SAndy King vsk = vsock_sk(sk); 989d021c344SAndy King 990d021c344SAndy King lock_sock(sk); 991d021c344SAndy King 992b3a6dfe8SAsias He err = vsock_auto_bind(vsk); 993b3a6dfe8SAsias He if (err) 994d021c344SAndy King goto out; 995d021c344SAndy King 996d021c344SAndy King 997d021c344SAndy King /* If the provided message contains an address, use that. Otherwise 998d021c344SAndy King * fall back on the socket's remote handle (if it has been connected). 999d021c344SAndy King */ 1000d021c344SAndy King if (msg->msg_name && 1001d021c344SAndy King vsock_addr_cast(msg->msg_name, msg->msg_namelen, 1002d021c344SAndy King &remote_addr) == 0) { 1003d021c344SAndy King /* Ensure this address is of the right type and is a valid 1004d021c344SAndy King * destination. 1005d021c344SAndy King */ 1006d021c344SAndy King 1007d021c344SAndy King if (remote_addr->svm_cid == VMADDR_CID_ANY) 1008d021c344SAndy King remote_addr->svm_cid = transport->get_local_cid(); 1009d021c344SAndy King 1010d021c344SAndy King if (!vsock_addr_bound(remote_addr)) { 1011d021c344SAndy King err = -EINVAL; 1012d021c344SAndy King goto out; 1013d021c344SAndy King } 1014d021c344SAndy King } else if (sock->state == SS_CONNECTED) { 1015d021c344SAndy King remote_addr = &vsk->remote_addr; 1016d021c344SAndy King 1017d021c344SAndy King if (remote_addr->svm_cid == VMADDR_CID_ANY) 1018d021c344SAndy King remote_addr->svm_cid = transport->get_local_cid(); 1019d021c344SAndy King 1020d021c344SAndy King /* XXX Should connect() or this function ensure remote_addr is 1021d021c344SAndy King * bound? 1022d021c344SAndy King */ 1023d021c344SAndy King if (!vsock_addr_bound(&vsk->remote_addr)) { 1024d021c344SAndy King err = -EINVAL; 1025d021c344SAndy King goto out; 1026d021c344SAndy King } 1027d021c344SAndy King } else { 1028d021c344SAndy King err = -EINVAL; 1029d021c344SAndy King goto out; 1030d021c344SAndy King } 1031d021c344SAndy King 1032d021c344SAndy King if (!transport->dgram_allow(remote_addr->svm_cid, 1033d021c344SAndy King remote_addr->svm_port)) { 1034d021c344SAndy King err = -EINVAL; 1035d021c344SAndy King goto out; 1036d021c344SAndy King } 1037d021c344SAndy King 10380f7db23aSAl Viro err = transport->dgram_enqueue(vsk, remote_addr, msg, len); 1039d021c344SAndy King 1040d021c344SAndy King out: 1041d021c344SAndy King release_sock(sk); 1042d021c344SAndy King return err; 1043d021c344SAndy King } 1044d021c344SAndy King 1045d021c344SAndy King static int vsock_dgram_connect(struct socket *sock, 1046d021c344SAndy King struct sockaddr *addr, int addr_len, int flags) 1047d021c344SAndy King { 1048d021c344SAndy King int err; 1049d021c344SAndy King struct sock *sk; 1050d021c344SAndy King struct vsock_sock *vsk; 1051d021c344SAndy King struct sockaddr_vm *remote_addr; 1052d021c344SAndy King 1053d021c344SAndy King sk = sock->sk; 1054d021c344SAndy King vsk = vsock_sk(sk); 1055d021c344SAndy King 1056d021c344SAndy King err = vsock_addr_cast(addr, addr_len, &remote_addr); 1057d021c344SAndy King if (err == -EAFNOSUPPORT && remote_addr->svm_family == AF_UNSPEC) { 1058d021c344SAndy King lock_sock(sk); 1059d021c344SAndy King vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, 1060d021c344SAndy King VMADDR_PORT_ANY); 1061d021c344SAndy King sock->state = SS_UNCONNECTED; 1062d021c344SAndy King release_sock(sk); 1063d021c344SAndy King return 0; 1064d021c344SAndy King } else if (err != 0) 1065d021c344SAndy King return -EINVAL; 1066d021c344SAndy King 1067d021c344SAndy King lock_sock(sk); 1068d021c344SAndy King 1069b3a6dfe8SAsias He err = vsock_auto_bind(vsk); 1070b3a6dfe8SAsias He if (err) 1071d021c344SAndy King goto out; 1072d021c344SAndy King 1073d021c344SAndy King if (!transport->dgram_allow(remote_addr->svm_cid, 1074d021c344SAndy King remote_addr->svm_port)) { 1075d021c344SAndy King err = -EINVAL; 1076d021c344SAndy King goto out; 1077d021c344SAndy King } 1078d021c344SAndy King 1079d021c344SAndy King memcpy(&vsk->remote_addr, remote_addr, sizeof(vsk->remote_addr)); 1080d021c344SAndy King sock->state = SS_CONNECTED; 1081d021c344SAndy King 1082d021c344SAndy King out: 1083d021c344SAndy King release_sock(sk); 1084d021c344SAndy King return err; 1085d021c344SAndy King } 1086d021c344SAndy King 10871b784140SYing Xue static int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg, 10881b784140SYing Xue size_t len, int flags) 1089d021c344SAndy King { 10901b784140SYing Xue return transport->dgram_dequeue(vsock_sk(sock->sk), msg, len, flags); 1091d021c344SAndy King } 1092d021c344SAndy King 1093d021c344SAndy King static const struct proto_ops vsock_dgram_ops = { 1094d021c344SAndy King .family = PF_VSOCK, 1095d021c344SAndy King .owner = THIS_MODULE, 1096d021c344SAndy King .release = vsock_release, 1097d021c344SAndy King .bind = vsock_bind, 1098d021c344SAndy King .connect = vsock_dgram_connect, 1099d021c344SAndy King .socketpair = sock_no_socketpair, 1100d021c344SAndy King .accept = sock_no_accept, 1101d021c344SAndy King .getname = vsock_getname, 1102a11e1d43SLinus Torvalds .poll = vsock_poll, 1103d021c344SAndy King .ioctl = sock_no_ioctl, 1104d021c344SAndy King .listen = sock_no_listen, 1105d021c344SAndy King .shutdown = vsock_shutdown, 1106d021c344SAndy King .setsockopt = sock_no_setsockopt, 1107d021c344SAndy King .getsockopt = sock_no_getsockopt, 1108d021c344SAndy King .sendmsg = vsock_dgram_sendmsg, 1109d021c344SAndy King .recvmsg = vsock_dgram_recvmsg, 1110d021c344SAndy King .mmap = sock_no_mmap, 1111d021c344SAndy King .sendpage = sock_no_sendpage, 1112d021c344SAndy King }; 1113d021c344SAndy King 1114380feae0SPeng Tao static int vsock_transport_cancel_pkt(struct vsock_sock *vsk) 1115380feae0SPeng Tao { 1116380feae0SPeng Tao if (!transport->cancel_pkt) 1117380feae0SPeng Tao return -EOPNOTSUPP; 1118380feae0SPeng Tao 1119380feae0SPeng Tao return transport->cancel_pkt(vsk); 1120380feae0SPeng Tao } 1121380feae0SPeng Tao 1122d021c344SAndy King static void vsock_connect_timeout(struct work_struct *work) 1123d021c344SAndy King { 1124d021c344SAndy King struct sock *sk; 1125d021c344SAndy King struct vsock_sock *vsk; 1126380feae0SPeng Tao int cancel = 0; 1127d021c344SAndy King 1128455f05ecSCong Wang vsk = container_of(work, struct vsock_sock, connect_work.work); 1129d021c344SAndy King sk = sk_vsock(vsk); 1130d021c344SAndy King 1131d021c344SAndy King lock_sock(sk); 11323b4477d2SStefan Hajnoczi if (sk->sk_state == TCP_SYN_SENT && 1133d021c344SAndy King (sk->sk_shutdown != SHUTDOWN_MASK)) { 11343b4477d2SStefan Hajnoczi sk->sk_state = TCP_CLOSE; 1135d021c344SAndy King sk->sk_err = ETIMEDOUT; 1136d021c344SAndy King sk->sk_error_report(sk); 1137380feae0SPeng Tao cancel = 1; 1138d021c344SAndy King } 1139d021c344SAndy King release_sock(sk); 1140380feae0SPeng Tao if (cancel) 1141380feae0SPeng Tao vsock_transport_cancel_pkt(vsk); 1142d021c344SAndy King 1143d021c344SAndy King sock_put(sk); 1144d021c344SAndy King } 1145d021c344SAndy King 1146d021c344SAndy King static int vsock_stream_connect(struct socket *sock, struct sockaddr *addr, 1147d021c344SAndy King int addr_len, int flags) 1148d021c344SAndy King { 1149d021c344SAndy King int err; 1150d021c344SAndy King struct sock *sk; 1151d021c344SAndy King struct vsock_sock *vsk; 1152d021c344SAndy King struct sockaddr_vm *remote_addr; 1153d021c344SAndy King long timeout; 1154d021c344SAndy King DEFINE_WAIT(wait); 1155d021c344SAndy King 1156d021c344SAndy King err = 0; 1157d021c344SAndy King sk = sock->sk; 1158d021c344SAndy King vsk = vsock_sk(sk); 1159d021c344SAndy King 1160d021c344SAndy King lock_sock(sk); 1161d021c344SAndy King 1162d021c344SAndy King /* XXX AF_UNSPEC should make us disconnect like AF_INET. */ 1163d021c344SAndy King switch (sock->state) { 1164d021c344SAndy King case SS_CONNECTED: 1165d021c344SAndy King err = -EISCONN; 1166d021c344SAndy King goto out; 1167d021c344SAndy King case SS_DISCONNECTING: 1168d021c344SAndy King err = -EINVAL; 1169d021c344SAndy King goto out; 1170d021c344SAndy King case SS_CONNECTING: 1171d021c344SAndy King /* This continues on so we can move sock into the SS_CONNECTED 1172d021c344SAndy King * state once the connection has completed (at which point err 1173d021c344SAndy King * will be set to zero also). Otherwise, we will either wait 1174d021c344SAndy King * for the connection or return -EALREADY should this be a 1175d021c344SAndy King * non-blocking call. 1176d021c344SAndy King */ 1177d021c344SAndy King err = -EALREADY; 1178d021c344SAndy King break; 1179d021c344SAndy King default: 11803b4477d2SStefan Hajnoczi if ((sk->sk_state == TCP_LISTEN) || 1181d021c344SAndy King vsock_addr_cast(addr, addr_len, &remote_addr) != 0) { 1182d021c344SAndy King err = -EINVAL; 1183d021c344SAndy King goto out; 1184d021c344SAndy King } 1185d021c344SAndy King 1186d021c344SAndy King /* The hypervisor and well-known contexts do not have socket 1187d021c344SAndy King * endpoints. 1188d021c344SAndy King */ 1189d021c344SAndy King if (!transport->stream_allow(remote_addr->svm_cid, 1190d021c344SAndy King remote_addr->svm_port)) { 1191d021c344SAndy King err = -ENETUNREACH; 1192d021c344SAndy King goto out; 1193d021c344SAndy King } 1194d021c344SAndy King 1195d021c344SAndy King /* Set the remote address that we are connecting to. */ 1196d021c344SAndy King memcpy(&vsk->remote_addr, remote_addr, 1197d021c344SAndy King sizeof(vsk->remote_addr)); 1198d021c344SAndy King 1199b3a6dfe8SAsias He err = vsock_auto_bind(vsk); 1200b3a6dfe8SAsias He if (err) 1201d021c344SAndy King goto out; 1202d021c344SAndy King 12033b4477d2SStefan Hajnoczi sk->sk_state = TCP_SYN_SENT; 1204d021c344SAndy King 1205d021c344SAndy King err = transport->connect(vsk); 1206d021c344SAndy King if (err < 0) 1207d021c344SAndy King goto out; 1208d021c344SAndy King 1209d021c344SAndy King /* Mark sock as connecting and set the error code to in 1210d021c344SAndy King * progress in case this is a non-blocking connect. 1211d021c344SAndy King */ 1212d021c344SAndy King sock->state = SS_CONNECTING; 1213d021c344SAndy King err = -EINPROGRESS; 1214d021c344SAndy King } 1215d021c344SAndy King 1216d021c344SAndy King /* The receive path will handle all communication until we are able to 1217d021c344SAndy King * enter the connected state. Here we wait for the connection to be 1218d021c344SAndy King * completed or a notification of an error. 1219d021c344SAndy King */ 1220d021c344SAndy King timeout = vsk->connect_timeout; 1221d021c344SAndy King prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); 1222d021c344SAndy King 12233b4477d2SStefan Hajnoczi while (sk->sk_state != TCP_ESTABLISHED && sk->sk_err == 0) { 1224d021c344SAndy King if (flags & O_NONBLOCK) { 1225d021c344SAndy King /* If we're not going to block, we schedule a timeout 1226d021c344SAndy King * function to generate a timeout on the connection 1227d021c344SAndy King * attempt, in case the peer doesn't respond in a 1228d021c344SAndy King * timely manner. We hold on to the socket until the 1229d021c344SAndy King * timeout fires. 1230d021c344SAndy King */ 1231d021c344SAndy King sock_hold(sk); 1232455f05ecSCong Wang schedule_delayed_work(&vsk->connect_work, timeout); 1233d021c344SAndy King 1234d021c344SAndy King /* Skip ahead to preserve error code set above. */ 1235d021c344SAndy King goto out_wait; 1236d021c344SAndy King } 1237d021c344SAndy King 1238d021c344SAndy King release_sock(sk); 1239d021c344SAndy King timeout = schedule_timeout(timeout); 1240d021c344SAndy King lock_sock(sk); 1241d021c344SAndy King 1242d021c344SAndy King if (signal_pending(current)) { 1243d021c344SAndy King err = sock_intr_errno(timeout); 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 } else if (timeout == 0) { 1249d021c344SAndy King err = -ETIMEDOUT; 12503b4477d2SStefan Hajnoczi sk->sk_state = TCP_CLOSE; 1251f7f9b5e7SClaudio Imbrenda sock->state = SS_UNCONNECTED; 1252380feae0SPeng Tao vsock_transport_cancel_pkt(vsk); 1253f7f9b5e7SClaudio Imbrenda goto out_wait; 1254d021c344SAndy King } 1255d021c344SAndy King 1256d021c344SAndy King prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); 1257d021c344SAndy King } 1258d021c344SAndy King 1259d021c344SAndy King if (sk->sk_err) { 1260d021c344SAndy King err = -sk->sk_err; 12613b4477d2SStefan Hajnoczi sk->sk_state = TCP_CLOSE; 1262f7f9b5e7SClaudio Imbrenda sock->state = SS_UNCONNECTED; 1263f7f9b5e7SClaudio Imbrenda } else { 1264d021c344SAndy King err = 0; 1265f7f9b5e7SClaudio Imbrenda } 1266d021c344SAndy King 1267d021c344SAndy King out_wait: 1268d021c344SAndy King finish_wait(sk_sleep(sk), &wait); 1269d021c344SAndy King out: 1270d021c344SAndy King release_sock(sk); 1271d021c344SAndy King return err; 1272d021c344SAndy King } 1273d021c344SAndy King 1274cdfbabfbSDavid Howells static int vsock_accept(struct socket *sock, struct socket *newsock, int flags, 1275cdfbabfbSDavid Howells bool kern) 1276d021c344SAndy King { 1277d021c344SAndy King struct sock *listener; 1278d021c344SAndy King int err; 1279d021c344SAndy King struct sock *connected; 1280d021c344SAndy King struct vsock_sock *vconnected; 1281d021c344SAndy King long timeout; 1282d021c344SAndy King DEFINE_WAIT(wait); 1283d021c344SAndy King 1284d021c344SAndy King err = 0; 1285d021c344SAndy King listener = sock->sk; 1286d021c344SAndy King 1287d021c344SAndy King lock_sock(listener); 1288d021c344SAndy King 1289d021c344SAndy King if (sock->type != SOCK_STREAM) { 1290d021c344SAndy King err = -EOPNOTSUPP; 1291d021c344SAndy King goto out; 1292d021c344SAndy King } 1293d021c344SAndy King 12943b4477d2SStefan Hajnoczi if (listener->sk_state != TCP_LISTEN) { 1295d021c344SAndy King err = -EINVAL; 1296d021c344SAndy King goto out; 1297d021c344SAndy King } 1298d021c344SAndy King 1299d021c344SAndy King /* Wait for children sockets to appear; these are the new sockets 1300d021c344SAndy King * created upon connection establishment. 1301d021c344SAndy King */ 1302d021c344SAndy King timeout = sock_sndtimeo(listener, flags & O_NONBLOCK); 1303d021c344SAndy King prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE); 1304d021c344SAndy King 1305d021c344SAndy King while ((connected = vsock_dequeue_accept(listener)) == NULL && 1306d021c344SAndy King listener->sk_err == 0) { 1307d021c344SAndy King release_sock(listener); 1308d021c344SAndy King timeout = schedule_timeout(timeout); 1309f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(listener), &wait); 1310d021c344SAndy King lock_sock(listener); 1311d021c344SAndy King 1312d021c344SAndy King if (signal_pending(current)) { 1313d021c344SAndy King err = sock_intr_errno(timeout); 1314f7f9b5e7SClaudio Imbrenda goto out; 1315d021c344SAndy King } else if (timeout == 0) { 1316d021c344SAndy King err = -EAGAIN; 1317f7f9b5e7SClaudio Imbrenda goto out; 1318d021c344SAndy King } 1319d021c344SAndy King 1320d021c344SAndy King prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE); 1321d021c344SAndy King } 1322f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(listener), &wait); 1323d021c344SAndy King 1324d021c344SAndy King if (listener->sk_err) 1325d021c344SAndy King err = -listener->sk_err; 1326d021c344SAndy King 1327d021c344SAndy King if (connected) { 1328d021c344SAndy King listener->sk_ack_backlog--; 1329d021c344SAndy King 13304192f672SStefan Hajnoczi lock_sock_nested(connected, SINGLE_DEPTH_NESTING); 1331d021c344SAndy King vconnected = vsock_sk(connected); 1332d021c344SAndy King 1333d021c344SAndy King /* If the listener socket has received an error, then we should 1334d021c344SAndy King * reject this socket and return. Note that we simply mark the 1335d021c344SAndy King * socket rejected, drop our reference, and let the cleanup 1336d021c344SAndy King * function handle the cleanup; the fact that we found it in 1337d021c344SAndy King * the listener's accept queue guarantees that the cleanup 1338d021c344SAndy King * function hasn't run yet. 1339d021c344SAndy King */ 1340d021c344SAndy King if (err) { 1341d021c344SAndy King vconnected->rejected = true; 1342f7f9b5e7SClaudio Imbrenda } else { 1343d021c344SAndy King newsock->state = SS_CONNECTED; 1344d021c344SAndy King sock_graft(connected, newsock); 1345f7f9b5e7SClaudio Imbrenda } 1346f7f9b5e7SClaudio Imbrenda 1347d021c344SAndy King release_sock(connected); 1348d021c344SAndy King sock_put(connected); 1349d021c344SAndy King } 1350d021c344SAndy King 1351d021c344SAndy King out: 1352d021c344SAndy King release_sock(listener); 1353d021c344SAndy King return err; 1354d021c344SAndy King } 1355d021c344SAndy King 1356d021c344SAndy King static int vsock_listen(struct socket *sock, int backlog) 1357d021c344SAndy King { 1358d021c344SAndy King int err; 1359d021c344SAndy King struct sock *sk; 1360d021c344SAndy King struct vsock_sock *vsk; 1361d021c344SAndy King 1362d021c344SAndy King sk = sock->sk; 1363d021c344SAndy King 1364d021c344SAndy King lock_sock(sk); 1365d021c344SAndy King 1366d021c344SAndy King if (sock->type != SOCK_STREAM) { 1367d021c344SAndy King err = -EOPNOTSUPP; 1368d021c344SAndy King goto out; 1369d021c344SAndy King } 1370d021c344SAndy King 1371d021c344SAndy King if (sock->state != SS_UNCONNECTED) { 1372d021c344SAndy King err = -EINVAL; 1373d021c344SAndy King goto out; 1374d021c344SAndy King } 1375d021c344SAndy King 1376d021c344SAndy King vsk = vsock_sk(sk); 1377d021c344SAndy King 1378d021c344SAndy King if (!vsock_addr_bound(&vsk->local_addr)) { 1379d021c344SAndy King err = -EINVAL; 1380d021c344SAndy King goto out; 1381d021c344SAndy King } 1382d021c344SAndy King 1383d021c344SAndy King sk->sk_max_ack_backlog = backlog; 13843b4477d2SStefan Hajnoczi sk->sk_state = TCP_LISTEN; 1385d021c344SAndy King 1386d021c344SAndy King err = 0; 1387d021c344SAndy King 1388d021c344SAndy King out: 1389d021c344SAndy King release_sock(sk); 1390d021c344SAndy King return err; 1391d021c344SAndy King } 1392d021c344SAndy King 1393d021c344SAndy King static int vsock_stream_setsockopt(struct socket *sock, 1394d021c344SAndy King int level, 1395d021c344SAndy King int optname, 1396d021c344SAndy King char __user *optval, 1397d021c344SAndy King unsigned int optlen) 1398d021c344SAndy King { 1399d021c344SAndy King int err; 1400d021c344SAndy King struct sock *sk; 1401d021c344SAndy King struct vsock_sock *vsk; 1402d021c344SAndy King u64 val; 1403d021c344SAndy King 1404d021c344SAndy King if (level != AF_VSOCK) 1405d021c344SAndy King return -ENOPROTOOPT; 1406d021c344SAndy King 1407d021c344SAndy King #define COPY_IN(_v) \ 1408d021c344SAndy King do { \ 1409d021c344SAndy King if (optlen < sizeof(_v)) { \ 1410d021c344SAndy King err = -EINVAL; \ 1411d021c344SAndy King goto exit; \ 1412d021c344SAndy King } \ 1413d021c344SAndy King if (copy_from_user(&_v, optval, sizeof(_v)) != 0) { \ 1414d021c344SAndy King err = -EFAULT; \ 1415d021c344SAndy King goto exit; \ 1416d021c344SAndy King } \ 1417d021c344SAndy King } while (0) 1418d021c344SAndy King 1419d021c344SAndy King err = 0; 1420d021c344SAndy King sk = sock->sk; 1421d021c344SAndy King vsk = vsock_sk(sk); 1422d021c344SAndy King 1423d021c344SAndy King lock_sock(sk); 1424d021c344SAndy King 1425d021c344SAndy King switch (optname) { 1426d021c344SAndy King case SO_VM_SOCKETS_BUFFER_SIZE: 1427d021c344SAndy King COPY_IN(val); 1428d021c344SAndy King transport->set_buffer_size(vsk, val); 1429d021c344SAndy King break; 1430d021c344SAndy King 1431d021c344SAndy King case SO_VM_SOCKETS_BUFFER_MAX_SIZE: 1432d021c344SAndy King COPY_IN(val); 1433d021c344SAndy King transport->set_max_buffer_size(vsk, val); 1434d021c344SAndy King break; 1435d021c344SAndy King 1436d021c344SAndy King case SO_VM_SOCKETS_BUFFER_MIN_SIZE: 1437d021c344SAndy King COPY_IN(val); 1438d021c344SAndy King transport->set_min_buffer_size(vsk, val); 1439d021c344SAndy King break; 1440d021c344SAndy King 1441d021c344SAndy King case SO_VM_SOCKETS_CONNECT_TIMEOUT: { 1442d021c344SAndy King struct timeval tv; 1443d021c344SAndy King COPY_IN(tv); 1444d021c344SAndy King if (tv.tv_sec >= 0 && tv.tv_usec < USEC_PER_SEC && 1445d021c344SAndy King tv.tv_sec < (MAX_SCHEDULE_TIMEOUT / HZ - 1)) { 1446d021c344SAndy King vsk->connect_timeout = tv.tv_sec * HZ + 1447d021c344SAndy King DIV_ROUND_UP(tv.tv_usec, (1000000 / HZ)); 1448d021c344SAndy King if (vsk->connect_timeout == 0) 1449d021c344SAndy King vsk->connect_timeout = 1450d021c344SAndy King VSOCK_DEFAULT_CONNECT_TIMEOUT; 1451d021c344SAndy King 1452d021c344SAndy King } else { 1453d021c344SAndy King err = -ERANGE; 1454d021c344SAndy King } 1455d021c344SAndy King break; 1456d021c344SAndy King } 1457d021c344SAndy King 1458d021c344SAndy King default: 1459d021c344SAndy King err = -ENOPROTOOPT; 1460d021c344SAndy King break; 1461d021c344SAndy King } 1462d021c344SAndy King 1463d021c344SAndy King #undef COPY_IN 1464d021c344SAndy King 1465d021c344SAndy King exit: 1466d021c344SAndy King release_sock(sk); 1467d021c344SAndy King return err; 1468d021c344SAndy King } 1469d021c344SAndy King 1470d021c344SAndy King static int vsock_stream_getsockopt(struct socket *sock, 1471d021c344SAndy King int level, int optname, 1472d021c344SAndy King char __user *optval, 1473d021c344SAndy King int __user *optlen) 1474d021c344SAndy King { 1475d021c344SAndy King int err; 1476d021c344SAndy King int len; 1477d021c344SAndy King struct sock *sk; 1478d021c344SAndy King struct vsock_sock *vsk; 1479d021c344SAndy King u64 val; 1480d021c344SAndy King 1481d021c344SAndy King if (level != AF_VSOCK) 1482d021c344SAndy King return -ENOPROTOOPT; 1483d021c344SAndy King 1484d021c344SAndy King err = get_user(len, optlen); 1485d021c344SAndy King if (err != 0) 1486d021c344SAndy King return err; 1487d021c344SAndy King 1488d021c344SAndy King #define COPY_OUT(_v) \ 1489d021c344SAndy King do { \ 1490d021c344SAndy King if (len < sizeof(_v)) \ 1491d021c344SAndy King return -EINVAL; \ 1492d021c344SAndy King \ 1493d021c344SAndy King len = sizeof(_v); \ 1494d021c344SAndy King if (copy_to_user(optval, &_v, len) != 0) \ 1495d021c344SAndy King return -EFAULT; \ 1496d021c344SAndy King \ 1497d021c344SAndy King } while (0) 1498d021c344SAndy King 1499d021c344SAndy King err = 0; 1500d021c344SAndy King sk = sock->sk; 1501d021c344SAndy King vsk = vsock_sk(sk); 1502d021c344SAndy King 1503d021c344SAndy King switch (optname) { 1504d021c344SAndy King case SO_VM_SOCKETS_BUFFER_SIZE: 1505d021c344SAndy King val = transport->get_buffer_size(vsk); 1506d021c344SAndy King COPY_OUT(val); 1507d021c344SAndy King break; 1508d021c344SAndy King 1509d021c344SAndy King case SO_VM_SOCKETS_BUFFER_MAX_SIZE: 1510d021c344SAndy King val = transport->get_max_buffer_size(vsk); 1511d021c344SAndy King COPY_OUT(val); 1512d021c344SAndy King break; 1513d021c344SAndy King 1514d021c344SAndy King case SO_VM_SOCKETS_BUFFER_MIN_SIZE: 1515d021c344SAndy King val = transport->get_min_buffer_size(vsk); 1516d021c344SAndy King COPY_OUT(val); 1517d021c344SAndy King break; 1518d021c344SAndy King 1519d021c344SAndy King case SO_VM_SOCKETS_CONNECT_TIMEOUT: { 1520d021c344SAndy King struct timeval tv; 1521d021c344SAndy King tv.tv_sec = vsk->connect_timeout / HZ; 1522d021c344SAndy King tv.tv_usec = 1523d021c344SAndy King (vsk->connect_timeout - 1524d021c344SAndy King tv.tv_sec * HZ) * (1000000 / HZ); 1525d021c344SAndy King COPY_OUT(tv); 1526d021c344SAndy King break; 1527d021c344SAndy King } 1528d021c344SAndy King default: 1529d021c344SAndy King return -ENOPROTOOPT; 1530d021c344SAndy King } 1531d021c344SAndy King 1532d021c344SAndy King err = put_user(len, optlen); 1533d021c344SAndy King if (err != 0) 1534d021c344SAndy King return -EFAULT; 1535d021c344SAndy King 1536d021c344SAndy King #undef COPY_OUT 1537d021c344SAndy King 1538d021c344SAndy King return 0; 1539d021c344SAndy King } 1540d021c344SAndy King 15411b784140SYing Xue static int vsock_stream_sendmsg(struct socket *sock, struct msghdr *msg, 15421b784140SYing Xue size_t len) 1543d021c344SAndy King { 1544d021c344SAndy King struct sock *sk; 1545d021c344SAndy King struct vsock_sock *vsk; 1546d021c344SAndy King ssize_t total_written; 1547d021c344SAndy King long timeout; 1548d021c344SAndy King int err; 1549d021c344SAndy King struct vsock_transport_send_notify_data send_data; 1550499fde66SWANG Cong DEFINE_WAIT_FUNC(wait, woken_wake_function); 1551d021c344SAndy King 1552d021c344SAndy King sk = sock->sk; 1553d021c344SAndy King vsk = vsock_sk(sk); 1554d021c344SAndy King total_written = 0; 1555d021c344SAndy King err = 0; 1556d021c344SAndy King 1557d021c344SAndy King if (msg->msg_flags & MSG_OOB) 1558d021c344SAndy King return -EOPNOTSUPP; 1559d021c344SAndy King 1560d021c344SAndy King lock_sock(sk); 1561d021c344SAndy King 1562d021c344SAndy King /* Callers should not provide a destination with stream sockets. */ 1563d021c344SAndy King if (msg->msg_namelen) { 15643b4477d2SStefan Hajnoczi err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP; 1565d021c344SAndy King goto out; 1566d021c344SAndy King } 1567d021c344SAndy King 1568d021c344SAndy King /* Send data only if both sides are not shutdown in the direction. */ 1569d021c344SAndy King if (sk->sk_shutdown & SEND_SHUTDOWN || 1570d021c344SAndy King vsk->peer_shutdown & RCV_SHUTDOWN) { 1571d021c344SAndy King err = -EPIPE; 1572d021c344SAndy King goto out; 1573d021c344SAndy King } 1574d021c344SAndy King 15753b4477d2SStefan Hajnoczi if (sk->sk_state != TCP_ESTABLISHED || 1576d021c344SAndy King !vsock_addr_bound(&vsk->local_addr)) { 1577d021c344SAndy King err = -ENOTCONN; 1578d021c344SAndy King goto out; 1579d021c344SAndy King } 1580d021c344SAndy King 1581d021c344SAndy King if (!vsock_addr_bound(&vsk->remote_addr)) { 1582d021c344SAndy King err = -EDESTADDRREQ; 1583d021c344SAndy King goto out; 1584d021c344SAndy King } 1585d021c344SAndy King 1586d021c344SAndy King /* Wait for room in the produce queue to enqueue our user's data. */ 1587d021c344SAndy King timeout = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT); 1588d021c344SAndy King 1589d021c344SAndy King err = transport->notify_send_init(vsk, &send_data); 1590d021c344SAndy King if (err < 0) 1591d021c344SAndy King goto out; 1592d021c344SAndy King 1593d021c344SAndy King while (total_written < len) { 1594d021c344SAndy King ssize_t written; 1595d021c344SAndy King 1596499fde66SWANG Cong add_wait_queue(sk_sleep(sk), &wait); 1597d021c344SAndy King while (vsock_stream_has_space(vsk) == 0 && 1598d021c344SAndy King sk->sk_err == 0 && 1599d021c344SAndy King !(sk->sk_shutdown & SEND_SHUTDOWN) && 1600d021c344SAndy King !(vsk->peer_shutdown & RCV_SHUTDOWN)) { 1601d021c344SAndy King 1602d021c344SAndy King /* Don't wait for non-blocking sockets. */ 1603d021c344SAndy King if (timeout == 0) { 1604d021c344SAndy King err = -EAGAIN; 1605499fde66SWANG Cong remove_wait_queue(sk_sleep(sk), &wait); 1606f7f9b5e7SClaudio Imbrenda goto out_err; 1607d021c344SAndy King } 1608d021c344SAndy King 1609d021c344SAndy King err = transport->notify_send_pre_block(vsk, &send_data); 1610f7f9b5e7SClaudio Imbrenda if (err < 0) { 1611499fde66SWANG Cong remove_wait_queue(sk_sleep(sk), &wait); 1612f7f9b5e7SClaudio Imbrenda goto out_err; 1613f7f9b5e7SClaudio Imbrenda } 1614d021c344SAndy King 1615d021c344SAndy King release_sock(sk); 1616499fde66SWANG Cong timeout = wait_woken(&wait, TASK_INTERRUPTIBLE, timeout); 1617d021c344SAndy King lock_sock(sk); 1618d021c344SAndy King if (signal_pending(current)) { 1619d021c344SAndy King err = sock_intr_errno(timeout); 1620499fde66SWANG Cong remove_wait_queue(sk_sleep(sk), &wait); 1621f7f9b5e7SClaudio Imbrenda goto out_err; 1622d021c344SAndy King } else if (timeout == 0) { 1623d021c344SAndy King err = -EAGAIN; 1624499fde66SWANG Cong remove_wait_queue(sk_sleep(sk), &wait); 1625f7f9b5e7SClaudio Imbrenda goto out_err; 1626d021c344SAndy King } 1627d021c344SAndy King } 1628499fde66SWANG Cong remove_wait_queue(sk_sleep(sk), &wait); 1629d021c344SAndy King 1630d021c344SAndy King /* These checks occur both as part of and after the loop 1631d021c344SAndy King * conditional since we need to check before and after 1632d021c344SAndy King * sleeping. 1633d021c344SAndy King */ 1634d021c344SAndy King if (sk->sk_err) { 1635d021c344SAndy King err = -sk->sk_err; 1636f7f9b5e7SClaudio Imbrenda goto out_err; 1637d021c344SAndy King } else if ((sk->sk_shutdown & SEND_SHUTDOWN) || 1638d021c344SAndy King (vsk->peer_shutdown & RCV_SHUTDOWN)) { 1639d021c344SAndy King err = -EPIPE; 1640f7f9b5e7SClaudio Imbrenda goto out_err; 1641d021c344SAndy King } 1642d021c344SAndy King 1643d021c344SAndy King err = transport->notify_send_pre_enqueue(vsk, &send_data); 1644d021c344SAndy King if (err < 0) 1645f7f9b5e7SClaudio Imbrenda goto out_err; 1646d021c344SAndy King 1647d021c344SAndy King /* Note that enqueue will only write as many bytes as are free 1648d021c344SAndy King * in the produce queue, so we don't need to ensure len is 1649d021c344SAndy King * smaller than the queue size. It is the caller's 1650d021c344SAndy King * responsibility to check how many bytes we were able to send. 1651d021c344SAndy King */ 1652d021c344SAndy King 1653d021c344SAndy King written = transport->stream_enqueue( 16540f7db23aSAl Viro vsk, msg, 1655d021c344SAndy King len - total_written); 1656d021c344SAndy King if (written < 0) { 1657d021c344SAndy King err = -ENOMEM; 1658f7f9b5e7SClaudio Imbrenda goto out_err; 1659d021c344SAndy King } 1660d021c344SAndy King 1661d021c344SAndy King total_written += written; 1662d021c344SAndy King 1663d021c344SAndy King err = transport->notify_send_post_enqueue( 1664d021c344SAndy King vsk, written, &send_data); 1665d021c344SAndy King if (err < 0) 1666f7f9b5e7SClaudio Imbrenda goto out_err; 1667d021c344SAndy King 1668d021c344SAndy King } 1669d021c344SAndy King 1670f7f9b5e7SClaudio Imbrenda out_err: 1671d021c344SAndy King if (total_written > 0) 1672d021c344SAndy King err = total_written; 1673d021c344SAndy King out: 1674d021c344SAndy King release_sock(sk); 1675d021c344SAndy King return err; 1676d021c344SAndy King } 1677d021c344SAndy King 1678d021c344SAndy King 1679d021c344SAndy King static int 16801b784140SYing Xue vsock_stream_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, 16811b784140SYing Xue int flags) 1682d021c344SAndy King { 1683d021c344SAndy King struct sock *sk; 1684d021c344SAndy King struct vsock_sock *vsk; 1685d021c344SAndy King int err; 1686d021c344SAndy King size_t target; 1687d021c344SAndy King ssize_t copied; 1688d021c344SAndy King long timeout; 1689d021c344SAndy King struct vsock_transport_recv_notify_data recv_data; 1690d021c344SAndy King 1691d021c344SAndy King DEFINE_WAIT(wait); 1692d021c344SAndy King 1693d021c344SAndy King sk = sock->sk; 1694d021c344SAndy King vsk = vsock_sk(sk); 1695d021c344SAndy King err = 0; 1696d021c344SAndy King 1697d021c344SAndy King lock_sock(sk); 1698d021c344SAndy King 16993b4477d2SStefan Hajnoczi if (sk->sk_state != TCP_ESTABLISHED) { 1700d021c344SAndy King /* Recvmsg is supposed to return 0 if a peer performs an 1701d021c344SAndy King * orderly shutdown. Differentiate between that case and when a 1702d021c344SAndy King * peer has not connected or a local shutdown occured with the 1703d021c344SAndy King * SOCK_DONE flag. 1704d021c344SAndy King */ 1705d021c344SAndy King if (sock_flag(sk, SOCK_DONE)) 1706d021c344SAndy King err = 0; 1707d021c344SAndy King else 1708d021c344SAndy King err = -ENOTCONN; 1709d021c344SAndy King 1710d021c344SAndy King goto out; 1711d021c344SAndy King } 1712d021c344SAndy King 1713d021c344SAndy King if (flags & MSG_OOB) { 1714d021c344SAndy King err = -EOPNOTSUPP; 1715d021c344SAndy King goto out; 1716d021c344SAndy King } 1717d021c344SAndy King 1718d021c344SAndy King /* We don't check peer_shutdown flag here since peer may actually shut 1719d021c344SAndy King * down, but there can be data in the queue that a local socket can 1720d021c344SAndy King * receive. 1721d021c344SAndy King */ 1722d021c344SAndy King if (sk->sk_shutdown & RCV_SHUTDOWN) { 1723d021c344SAndy King err = 0; 1724d021c344SAndy King goto out; 1725d021c344SAndy King } 1726d021c344SAndy King 1727d021c344SAndy King /* It is valid on Linux to pass in a zero-length receive buffer. This 1728d021c344SAndy King * is not an error. We may as well bail out now. 1729d021c344SAndy King */ 1730d021c344SAndy King if (!len) { 1731d021c344SAndy King err = 0; 1732d021c344SAndy King goto out; 1733d021c344SAndy King } 1734d021c344SAndy King 1735d021c344SAndy King /* We must not copy less than target bytes into the user's buffer 1736d021c344SAndy King * before returning successfully, so we wait for the consume queue to 1737d021c344SAndy King * have that much data to consume before dequeueing. Note that this 1738d021c344SAndy King * makes it impossible to handle cases where target is greater than the 1739d021c344SAndy King * queue size. 1740d021c344SAndy King */ 1741d021c344SAndy King target = sock_rcvlowat(sk, flags & MSG_WAITALL, len); 1742d021c344SAndy King if (target >= transport->stream_rcvhiwat(vsk)) { 1743d021c344SAndy King err = -ENOMEM; 1744d021c344SAndy King goto out; 1745d021c344SAndy King } 1746d021c344SAndy King timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT); 1747d021c344SAndy King copied = 0; 1748d021c344SAndy King 1749d021c344SAndy King err = transport->notify_recv_init(vsk, target, &recv_data); 1750d021c344SAndy King if (err < 0) 1751d021c344SAndy King goto out; 1752d021c344SAndy King 1753d021c344SAndy King 1754d021c344SAndy King while (1) { 1755f7f9b5e7SClaudio Imbrenda s64 ready; 1756f7f9b5e7SClaudio Imbrenda 1757f7f9b5e7SClaudio Imbrenda prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); 1758f7f9b5e7SClaudio Imbrenda ready = vsock_stream_has_data(vsk); 1759f7f9b5e7SClaudio Imbrenda 1760f7f9b5e7SClaudio Imbrenda if (ready == 0) { 1761f7f9b5e7SClaudio Imbrenda if (sk->sk_err != 0 || 1762f7f9b5e7SClaudio Imbrenda (sk->sk_shutdown & RCV_SHUTDOWN) || 1763f7f9b5e7SClaudio Imbrenda (vsk->peer_shutdown & SEND_SHUTDOWN)) { 1764f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(sk), &wait); 1765f7f9b5e7SClaudio Imbrenda break; 1766f7f9b5e7SClaudio Imbrenda } 1767f7f9b5e7SClaudio Imbrenda /* Don't wait for non-blocking sockets. */ 1768f7f9b5e7SClaudio Imbrenda if (timeout == 0) { 1769f7f9b5e7SClaudio Imbrenda err = -EAGAIN; 1770f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(sk), &wait); 1771f7f9b5e7SClaudio Imbrenda break; 1772f7f9b5e7SClaudio Imbrenda } 1773f7f9b5e7SClaudio Imbrenda 1774f7f9b5e7SClaudio Imbrenda err = transport->notify_recv_pre_block( 1775f7f9b5e7SClaudio Imbrenda vsk, target, &recv_data); 1776f7f9b5e7SClaudio Imbrenda if (err < 0) { 1777f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(sk), &wait); 1778f7f9b5e7SClaudio Imbrenda break; 1779f7f9b5e7SClaudio Imbrenda } 1780f7f9b5e7SClaudio Imbrenda release_sock(sk); 1781f7f9b5e7SClaudio Imbrenda timeout = schedule_timeout(timeout); 1782f7f9b5e7SClaudio Imbrenda lock_sock(sk); 1783f7f9b5e7SClaudio Imbrenda 1784f7f9b5e7SClaudio Imbrenda if (signal_pending(current)) { 1785f7f9b5e7SClaudio Imbrenda err = sock_intr_errno(timeout); 1786f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(sk), &wait); 1787f7f9b5e7SClaudio Imbrenda break; 1788f7f9b5e7SClaudio Imbrenda } else if (timeout == 0) { 1789f7f9b5e7SClaudio Imbrenda err = -EAGAIN; 1790f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(sk), &wait); 1791f7f9b5e7SClaudio Imbrenda break; 1792f7f9b5e7SClaudio Imbrenda } 1793f7f9b5e7SClaudio Imbrenda } else { 1794f7f9b5e7SClaudio Imbrenda ssize_t read; 1795f7f9b5e7SClaudio Imbrenda 1796f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(sk), &wait); 1797d021c344SAndy King 1798d021c344SAndy King if (ready < 0) { 1799f7f9b5e7SClaudio Imbrenda /* Invalid queue pair content. XXX This should 1800f7f9b5e7SClaudio Imbrenda * be changed to a connection reset in a later 1801f7f9b5e7SClaudio Imbrenda * change. 1802d021c344SAndy King */ 1803d021c344SAndy King 1804d021c344SAndy King err = -ENOMEM; 1805f7f9b5e7SClaudio Imbrenda goto out; 1806f7f9b5e7SClaudio Imbrenda } 1807d021c344SAndy King 1808d021c344SAndy King err = transport->notify_recv_pre_dequeue( 1809d021c344SAndy King vsk, target, &recv_data); 1810d021c344SAndy King if (err < 0) 1811d021c344SAndy King break; 1812d021c344SAndy King 1813d021c344SAndy King read = transport->stream_dequeue( 18140f7db23aSAl Viro vsk, msg, 1815d021c344SAndy King len - copied, flags); 1816d021c344SAndy King if (read < 0) { 1817d021c344SAndy King err = -ENOMEM; 1818d021c344SAndy King break; 1819d021c344SAndy King } 1820d021c344SAndy King 1821d021c344SAndy King copied += read; 1822d021c344SAndy King 1823d021c344SAndy King err = transport->notify_recv_post_dequeue( 1824d021c344SAndy King vsk, target, read, 1825d021c344SAndy King !(flags & MSG_PEEK), &recv_data); 1826d021c344SAndy King if (err < 0) 1827f7f9b5e7SClaudio Imbrenda goto out; 1828d021c344SAndy King 1829d021c344SAndy King if (read >= target || flags & MSG_PEEK) 1830d021c344SAndy King break; 1831d021c344SAndy King 1832d021c344SAndy King target -= read; 1833d021c344SAndy King } 1834d021c344SAndy King } 1835d021c344SAndy King 1836d021c344SAndy King if (sk->sk_err) 1837d021c344SAndy King err = -sk->sk_err; 1838d021c344SAndy King else if (sk->sk_shutdown & RCV_SHUTDOWN) 1839d021c344SAndy King err = 0; 1840d021c344SAndy King 1841dedc58e0SIan Campbell if (copied > 0) 1842d021c344SAndy King err = copied; 1843d021c344SAndy King 1844d021c344SAndy King out: 1845d021c344SAndy King release_sock(sk); 1846d021c344SAndy King return err; 1847d021c344SAndy King } 1848d021c344SAndy King 1849d021c344SAndy King static const struct proto_ops vsock_stream_ops = { 1850d021c344SAndy King .family = PF_VSOCK, 1851d021c344SAndy King .owner = THIS_MODULE, 1852d021c344SAndy King .release = vsock_release, 1853d021c344SAndy King .bind = vsock_bind, 1854d021c344SAndy King .connect = vsock_stream_connect, 1855d021c344SAndy King .socketpair = sock_no_socketpair, 1856d021c344SAndy King .accept = vsock_accept, 1857d021c344SAndy King .getname = vsock_getname, 1858a11e1d43SLinus Torvalds .poll = vsock_poll, 1859d021c344SAndy King .ioctl = sock_no_ioctl, 1860d021c344SAndy King .listen = vsock_listen, 1861d021c344SAndy King .shutdown = vsock_shutdown, 1862d021c344SAndy King .setsockopt = vsock_stream_setsockopt, 1863d021c344SAndy King .getsockopt = vsock_stream_getsockopt, 1864d021c344SAndy King .sendmsg = vsock_stream_sendmsg, 1865d021c344SAndy King .recvmsg = vsock_stream_recvmsg, 1866d021c344SAndy King .mmap = sock_no_mmap, 1867d021c344SAndy King .sendpage = sock_no_sendpage, 1868d021c344SAndy King }; 1869d021c344SAndy King 1870d021c344SAndy King static int vsock_create(struct net *net, struct socket *sock, 1871d021c344SAndy King int protocol, int kern) 1872d021c344SAndy King { 1873d021c344SAndy King if (!sock) 1874d021c344SAndy King return -EINVAL; 1875d021c344SAndy King 18766cf1c5fcSAndy King if (protocol && protocol != PF_VSOCK) 1877d021c344SAndy King return -EPROTONOSUPPORT; 1878d021c344SAndy King 1879d021c344SAndy King switch (sock->type) { 1880d021c344SAndy King case SOCK_DGRAM: 1881d021c344SAndy King sock->ops = &vsock_dgram_ops; 1882d021c344SAndy King break; 1883d021c344SAndy King case SOCK_STREAM: 1884d021c344SAndy King sock->ops = &vsock_stream_ops; 1885d021c344SAndy King break; 1886d021c344SAndy King default: 1887d021c344SAndy King return -ESOCKTNOSUPPORT; 1888d021c344SAndy King } 1889d021c344SAndy King 1890d021c344SAndy King sock->state = SS_UNCONNECTED; 1891d021c344SAndy King 189211aa9c28SEric W. Biederman return __vsock_create(net, sock, NULL, GFP_KERNEL, 0, kern) ? 0 : -ENOMEM; 1893d021c344SAndy King } 1894d021c344SAndy King 1895d021c344SAndy King static const struct net_proto_family vsock_family_ops = { 1896d021c344SAndy King .family = AF_VSOCK, 1897d021c344SAndy King .create = vsock_create, 1898d021c344SAndy King .owner = THIS_MODULE, 1899d021c344SAndy King }; 1900d021c344SAndy King 1901d021c344SAndy King static long vsock_dev_do_ioctl(struct file *filp, 1902d021c344SAndy King unsigned int cmd, void __user *ptr) 1903d021c344SAndy King { 1904d021c344SAndy King u32 __user *p = ptr; 1905d021c344SAndy King int retval = 0; 1906d021c344SAndy King 1907d021c344SAndy King switch (cmd) { 1908d021c344SAndy King case IOCTL_VM_SOCKETS_GET_LOCAL_CID: 1909d021c344SAndy King if (put_user(transport->get_local_cid(), p) != 0) 1910d021c344SAndy King retval = -EFAULT; 1911d021c344SAndy King break; 1912d021c344SAndy King 1913d021c344SAndy King default: 1914d021c344SAndy King pr_err("Unknown ioctl %d\n", cmd); 1915d021c344SAndy King retval = -EINVAL; 1916d021c344SAndy King } 1917d021c344SAndy King 1918d021c344SAndy King return retval; 1919d021c344SAndy King } 1920d021c344SAndy King 1921d021c344SAndy King static long vsock_dev_ioctl(struct file *filp, 1922d021c344SAndy King unsigned int cmd, unsigned long arg) 1923d021c344SAndy King { 1924d021c344SAndy King return vsock_dev_do_ioctl(filp, cmd, (void __user *)arg); 1925d021c344SAndy King } 1926d021c344SAndy King 1927d021c344SAndy King #ifdef CONFIG_COMPAT 1928d021c344SAndy King static long vsock_dev_compat_ioctl(struct file *filp, 1929d021c344SAndy King unsigned int cmd, unsigned long arg) 1930d021c344SAndy King { 1931d021c344SAndy King return vsock_dev_do_ioctl(filp, cmd, compat_ptr(arg)); 1932d021c344SAndy King } 1933d021c344SAndy King #endif 1934d021c344SAndy King 1935d021c344SAndy King static const struct file_operations vsock_device_ops = { 1936d021c344SAndy King .owner = THIS_MODULE, 1937d021c344SAndy King .unlocked_ioctl = vsock_dev_ioctl, 1938d021c344SAndy King #ifdef CONFIG_COMPAT 1939d021c344SAndy King .compat_ioctl = vsock_dev_compat_ioctl, 1940d021c344SAndy King #endif 1941d021c344SAndy King .open = nonseekable_open, 1942d021c344SAndy King }; 1943d021c344SAndy King 1944d021c344SAndy King static struct miscdevice vsock_device = { 1945d021c344SAndy King .name = "vsock", 1946d021c344SAndy King .fops = &vsock_device_ops, 1947d021c344SAndy King }; 1948d021c344SAndy King 19492c4a336eSAndy King int __vsock_core_init(const struct vsock_transport *t, struct module *owner) 1950d021c344SAndy King { 19512c4a336eSAndy King int err = mutex_lock_interruptible(&vsock_register_mutex); 19522c4a336eSAndy King 19532c4a336eSAndy King if (err) 19542c4a336eSAndy King return err; 19552c4a336eSAndy King 19562c4a336eSAndy King if (transport) { 19572c4a336eSAndy King err = -EBUSY; 19582c4a336eSAndy King goto err_busy; 19592c4a336eSAndy King } 19602c4a336eSAndy King 19612c4a336eSAndy King /* Transport must be the owner of the protocol so that it can't 19622c4a336eSAndy King * unload while there are open sockets. 19632c4a336eSAndy King */ 19642c4a336eSAndy King vsock_proto.owner = owner; 19652c4a336eSAndy King transport = t; 1966d021c344SAndy King 19676ad0b2f7SAsias He vsock_device.minor = MISC_DYNAMIC_MINOR; 1968d021c344SAndy King err = misc_register(&vsock_device); 1969d021c344SAndy King if (err) { 1970d021c344SAndy King pr_err("Failed to register misc device\n"); 1971f6a835bbSGao feng goto err_reset_transport; 1972d021c344SAndy King } 1973d021c344SAndy King 1974d021c344SAndy King err = proto_register(&vsock_proto, 1); /* we want our slab */ 1975d021c344SAndy King if (err) { 1976d021c344SAndy King pr_err("Cannot register vsock protocol\n"); 1977f6a835bbSGao feng goto err_deregister_misc; 1978d021c344SAndy King } 1979d021c344SAndy King 1980d021c344SAndy King err = sock_register(&vsock_family_ops); 1981d021c344SAndy King if (err) { 1982d021c344SAndy King pr_err("could not register af_vsock (%d) address family: %d\n", 1983d021c344SAndy King AF_VSOCK, err); 1984d021c344SAndy King goto err_unregister_proto; 1985d021c344SAndy King } 1986d021c344SAndy King 19872c4a336eSAndy King mutex_unlock(&vsock_register_mutex); 1988d021c344SAndy King return 0; 1989d021c344SAndy King 1990d021c344SAndy King err_unregister_proto: 1991d021c344SAndy King proto_unregister(&vsock_proto); 1992f6a835bbSGao feng err_deregister_misc: 1993d021c344SAndy King misc_deregister(&vsock_device); 1994f6a835bbSGao feng err_reset_transport: 19952c4a336eSAndy King transport = NULL; 19962c4a336eSAndy King err_busy: 19972c4a336eSAndy King mutex_unlock(&vsock_register_mutex); 1998d021c344SAndy King return err; 1999d021c344SAndy King } 20002c4a336eSAndy King EXPORT_SYMBOL_GPL(__vsock_core_init); 2001d021c344SAndy King 2002d021c344SAndy King void vsock_core_exit(void) 2003d021c344SAndy King { 2004d021c344SAndy King mutex_lock(&vsock_register_mutex); 2005d021c344SAndy King 2006d021c344SAndy King misc_deregister(&vsock_device); 2007d021c344SAndy King sock_unregister(AF_VSOCK); 2008d021c344SAndy King proto_unregister(&vsock_proto); 2009d021c344SAndy King 2010d021c344SAndy King /* We do not want the assignment below re-ordered. */ 2011d021c344SAndy King mb(); 2012d021c344SAndy King transport = NULL; 2013d021c344SAndy King 2014d021c344SAndy King mutex_unlock(&vsock_register_mutex); 2015d021c344SAndy King } 2016d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_core_exit); 2017d021c344SAndy King 20180b01aeb3SStefan Hajnoczi const struct vsock_transport *vsock_core_get_transport(void) 20190b01aeb3SStefan Hajnoczi { 20200b01aeb3SStefan Hajnoczi /* vsock_register_mutex not taken since only the transport uses this 20210b01aeb3SStefan Hajnoczi * function and only while registered. 20220b01aeb3SStefan Hajnoczi */ 20230b01aeb3SStefan Hajnoczi return transport; 20240b01aeb3SStefan Hajnoczi } 20250b01aeb3SStefan Hajnoczi EXPORT_SYMBOL_GPL(vsock_core_get_transport); 20260b01aeb3SStefan Hajnoczi 202705e489b1SStefan Hajnoczi static void __exit vsock_exit(void) 202805e489b1SStefan Hajnoczi { 202905e489b1SStefan Hajnoczi /* Do nothing. This function makes this module removable. */ 203005e489b1SStefan Hajnoczi } 203105e489b1SStefan Hajnoczi 2032c1eef220SCong Wang module_init(vsock_init_tables); 203305e489b1SStefan Hajnoczi module_exit(vsock_exit); 2034c1eef220SCong Wang 2035d021c344SAndy King MODULE_AUTHOR("VMware, Inc."); 2036d021c344SAndy King MODULE_DESCRIPTION("VMware Virtual Socket Family"); 20371190cfdbSJorgen Hansen MODULE_VERSION("1.0.2.0-k"); 2038d021c344SAndy King MODULE_LICENSE("GPL v2"); 2039