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 39ea3803c1SStefan Hajnoczi * implementation because they are in the VSOCK_SS_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. 85d021c344SAndy King */ 86d021c344SAndy King 87d021c344SAndy King #include <linux/types.h> 88d021c344SAndy King #include <linux/bitops.h> 89d021c344SAndy King #include <linux/cred.h> 90d021c344SAndy King #include <linux/init.h> 91d021c344SAndy King #include <linux/io.h> 92d021c344SAndy King #include <linux/kernel.h> 93174cd4b1SIngo Molnar #include <linux/sched/signal.h> 94d021c344SAndy King #include <linux/kmod.h> 95d021c344SAndy King #include <linux/list.h> 96d021c344SAndy King #include <linux/miscdevice.h> 97d021c344SAndy King #include <linux/module.h> 98d021c344SAndy King #include <linux/mutex.h> 99d021c344SAndy King #include <linux/net.h> 100d021c344SAndy King #include <linux/poll.h> 101d021c344SAndy King #include <linux/skbuff.h> 102d021c344SAndy King #include <linux/smp.h> 103d021c344SAndy King #include <linux/socket.h> 104d021c344SAndy King #include <linux/stddef.h> 105d021c344SAndy King #include <linux/unistd.h> 106d021c344SAndy King #include <linux/wait.h> 107d021c344SAndy King #include <linux/workqueue.h> 108d021c344SAndy King #include <net/sock.h> 10982a54d0eSAsias He #include <net/af_vsock.h> 110d021c344SAndy King 111d021c344SAndy King static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr); 112d021c344SAndy King static void vsock_sk_destruct(struct sock *sk); 113d021c344SAndy King static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb); 114d021c344SAndy King 115d021c344SAndy King /* Protocol family. */ 116d021c344SAndy King static struct proto vsock_proto = { 117d021c344SAndy King .name = "AF_VSOCK", 118d021c344SAndy King .owner = THIS_MODULE, 119d021c344SAndy King .obj_size = sizeof(struct vsock_sock), 120d021c344SAndy King }; 121d021c344SAndy King 122d021c344SAndy King /* The default peer timeout indicates how long we will wait for a peer response 123d021c344SAndy King * to a control message. 124d021c344SAndy King */ 125d021c344SAndy King #define VSOCK_DEFAULT_CONNECT_TIMEOUT (2 * HZ) 126d021c344SAndy King 127d021c344SAndy King static const struct vsock_transport *transport; 128d021c344SAndy King static DEFINE_MUTEX(vsock_register_mutex); 129d021c344SAndy King 130d021c344SAndy King /**** EXPORTS ****/ 131d021c344SAndy King 132d021c344SAndy King /* Get the ID of the local context. This is transport dependent. */ 133d021c344SAndy King 134d021c344SAndy King int vm_sockets_get_local_cid(void) 135d021c344SAndy King { 136d021c344SAndy King return transport->get_local_cid(); 137d021c344SAndy King } 138d021c344SAndy King EXPORT_SYMBOL_GPL(vm_sockets_get_local_cid); 139d021c344SAndy King 140d021c344SAndy King /**** UTILS ****/ 141d021c344SAndy King 142d021c344SAndy King /* Each bound VSocket is stored in the bind hash table and each connected 143d021c344SAndy King * VSocket is stored in the connected hash table. 144d021c344SAndy King * 145d021c344SAndy King * Unbound sockets are all put on the same list attached to the end of the hash 146d021c344SAndy King * table (vsock_unbound_sockets). Bound sockets are added to the hash table in 147d021c344SAndy King * the bucket that their local address hashes to (vsock_bound_sockets(addr) 148d021c344SAndy King * represents the list that addr hashes to). 149d021c344SAndy King * 150d021c344SAndy King * Specifically, we initialize the vsock_bind_table array to a size of 151d021c344SAndy King * VSOCK_HASH_SIZE + 1 so that vsock_bind_table[0] through 152d021c344SAndy King * vsock_bind_table[VSOCK_HASH_SIZE - 1] are for bound sockets and 153d021c344SAndy King * vsock_bind_table[VSOCK_HASH_SIZE] is for unbound sockets. The hash function 154a49dd9dcSAsias He * mods with VSOCK_HASH_SIZE to ensure this. 155d021c344SAndy King */ 156d021c344SAndy King #define VSOCK_HASH_SIZE 251 157d021c344SAndy King #define MAX_PORT_RETRIES 24 158d021c344SAndy King 159a49dd9dcSAsias He #define VSOCK_HASH(addr) ((addr)->svm_port % VSOCK_HASH_SIZE) 160d021c344SAndy King #define vsock_bound_sockets(addr) (&vsock_bind_table[VSOCK_HASH(addr)]) 161d021c344SAndy King #define vsock_unbound_sockets (&vsock_bind_table[VSOCK_HASH_SIZE]) 162d021c344SAndy King 163d021c344SAndy King /* XXX This can probably be implemented in a better way. */ 164d021c344SAndy King #define VSOCK_CONN_HASH(src, dst) \ 165a49dd9dcSAsias He (((src)->svm_cid ^ (dst)->svm_port) % VSOCK_HASH_SIZE) 166d021c344SAndy King #define vsock_connected_sockets(src, dst) \ 167d021c344SAndy King (&vsock_connected_table[VSOCK_CONN_HASH(src, dst)]) 168d021c344SAndy King #define vsock_connected_sockets_vsk(vsk) \ 169d021c344SAndy King vsock_connected_sockets(&(vsk)->remote_addr, &(vsk)->local_addr) 170d021c344SAndy King 171d021c344SAndy King static struct list_head vsock_bind_table[VSOCK_HASH_SIZE + 1]; 172d021c344SAndy King static struct list_head vsock_connected_table[VSOCK_HASH_SIZE]; 173d021c344SAndy King static DEFINE_SPINLOCK(vsock_table_lock); 174d021c344SAndy King 175b3a6dfe8SAsias He /* Autobind this socket to the local address if necessary. */ 176b3a6dfe8SAsias He static int vsock_auto_bind(struct vsock_sock *vsk) 177b3a6dfe8SAsias He { 178b3a6dfe8SAsias He struct sock *sk = sk_vsock(vsk); 179b3a6dfe8SAsias He struct sockaddr_vm local_addr; 180b3a6dfe8SAsias He 181b3a6dfe8SAsias He if (vsock_addr_bound(&vsk->local_addr)) 182b3a6dfe8SAsias He return 0; 183b3a6dfe8SAsias He vsock_addr_init(&local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 184b3a6dfe8SAsias He return __vsock_bind(sk, &local_addr); 185b3a6dfe8SAsias He } 186b3a6dfe8SAsias He 18722ee3b57SGeert Uytterhoeven static void vsock_init_tables(void) 188d021c344SAndy King { 189d021c344SAndy King int i; 190d021c344SAndy King 191d021c344SAndy King for (i = 0; i < ARRAY_SIZE(vsock_bind_table); i++) 192d021c344SAndy King INIT_LIST_HEAD(&vsock_bind_table[i]); 193d021c344SAndy King 194d021c344SAndy King for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) 195d021c344SAndy King INIT_LIST_HEAD(&vsock_connected_table[i]); 196d021c344SAndy King } 197d021c344SAndy King 198d021c344SAndy King static void __vsock_insert_bound(struct list_head *list, 199d021c344SAndy King struct vsock_sock *vsk) 200d021c344SAndy King { 201d021c344SAndy King sock_hold(&vsk->sk); 202d021c344SAndy King list_add(&vsk->bound_table, list); 203d021c344SAndy King } 204d021c344SAndy King 205d021c344SAndy King static void __vsock_insert_connected(struct list_head *list, 206d021c344SAndy King struct vsock_sock *vsk) 207d021c344SAndy King { 208d021c344SAndy King sock_hold(&vsk->sk); 209d021c344SAndy King list_add(&vsk->connected_table, list); 210d021c344SAndy King } 211d021c344SAndy King 212d021c344SAndy King static void __vsock_remove_bound(struct vsock_sock *vsk) 213d021c344SAndy King { 214d021c344SAndy King list_del_init(&vsk->bound_table); 215d021c344SAndy King sock_put(&vsk->sk); 216d021c344SAndy King } 217d021c344SAndy King 218d021c344SAndy King static void __vsock_remove_connected(struct vsock_sock *vsk) 219d021c344SAndy King { 220d021c344SAndy King list_del_init(&vsk->connected_table); 221d021c344SAndy King sock_put(&vsk->sk); 222d021c344SAndy King } 223d021c344SAndy King 224d021c344SAndy King static struct sock *__vsock_find_bound_socket(struct sockaddr_vm *addr) 225d021c344SAndy King { 226d021c344SAndy King struct vsock_sock *vsk; 227d021c344SAndy King 228d021c344SAndy King list_for_each_entry(vsk, vsock_bound_sockets(addr), bound_table) 229990454b5SReilly Grant if (addr->svm_port == vsk->local_addr.svm_port) 230d021c344SAndy King return sk_vsock(vsk); 231d021c344SAndy King 232d021c344SAndy King return NULL; 233d021c344SAndy King } 234d021c344SAndy King 235d021c344SAndy King static struct sock *__vsock_find_connected_socket(struct sockaddr_vm *src, 236d021c344SAndy King struct sockaddr_vm *dst) 237d021c344SAndy King { 238d021c344SAndy King struct vsock_sock *vsk; 239d021c344SAndy King 240d021c344SAndy King list_for_each_entry(vsk, vsock_connected_sockets(src, dst), 241d021c344SAndy King connected_table) { 242990454b5SReilly Grant if (vsock_addr_equals_addr(src, &vsk->remote_addr) && 243990454b5SReilly Grant dst->svm_port == vsk->local_addr.svm_port) { 244d021c344SAndy King return sk_vsock(vsk); 245d021c344SAndy King } 246d021c344SAndy King } 247d021c344SAndy King 248d021c344SAndy King return NULL; 249d021c344SAndy King } 250d021c344SAndy King 251d021c344SAndy King static bool __vsock_in_bound_table(struct vsock_sock *vsk) 252d021c344SAndy King { 253d021c344SAndy King return !list_empty(&vsk->bound_table); 254d021c344SAndy King } 255d021c344SAndy King 256d021c344SAndy King static bool __vsock_in_connected_table(struct vsock_sock *vsk) 257d021c344SAndy King { 258d021c344SAndy King return !list_empty(&vsk->connected_table); 259d021c344SAndy King } 260d021c344SAndy King 261d021c344SAndy King static void vsock_insert_unbound(struct vsock_sock *vsk) 262d021c344SAndy King { 263d021c344SAndy King spin_lock_bh(&vsock_table_lock); 264d021c344SAndy King __vsock_insert_bound(vsock_unbound_sockets, vsk); 265d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 266d021c344SAndy King } 267d021c344SAndy King 268d021c344SAndy King void vsock_insert_connected(struct vsock_sock *vsk) 269d021c344SAndy King { 270d021c344SAndy King struct list_head *list = vsock_connected_sockets( 271d021c344SAndy King &vsk->remote_addr, &vsk->local_addr); 272d021c344SAndy King 273d021c344SAndy King spin_lock_bh(&vsock_table_lock); 274d021c344SAndy King __vsock_insert_connected(list, vsk); 275d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 276d021c344SAndy King } 277d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_insert_connected); 278d021c344SAndy King 279d021c344SAndy King void vsock_remove_bound(struct vsock_sock *vsk) 280d021c344SAndy King { 281d021c344SAndy King spin_lock_bh(&vsock_table_lock); 282d021c344SAndy King __vsock_remove_bound(vsk); 283d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 284d021c344SAndy King } 285d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_remove_bound); 286d021c344SAndy King 287d021c344SAndy King void vsock_remove_connected(struct vsock_sock *vsk) 288d021c344SAndy King { 289d021c344SAndy King spin_lock_bh(&vsock_table_lock); 290d021c344SAndy King __vsock_remove_connected(vsk); 291d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 292d021c344SAndy King } 293d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_remove_connected); 294d021c344SAndy King 295d021c344SAndy King struct sock *vsock_find_bound_socket(struct sockaddr_vm *addr) 296d021c344SAndy King { 297d021c344SAndy King struct sock *sk; 298d021c344SAndy King 299d021c344SAndy King spin_lock_bh(&vsock_table_lock); 300d021c344SAndy King sk = __vsock_find_bound_socket(addr); 301d021c344SAndy King if (sk) 302d021c344SAndy King sock_hold(sk); 303d021c344SAndy King 304d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 305d021c344SAndy King 306d021c344SAndy King return sk; 307d021c344SAndy King } 308d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_find_bound_socket); 309d021c344SAndy King 310d021c344SAndy King struct sock *vsock_find_connected_socket(struct sockaddr_vm *src, 311d021c344SAndy King struct sockaddr_vm *dst) 312d021c344SAndy King { 313d021c344SAndy King struct sock *sk; 314d021c344SAndy King 315d021c344SAndy King spin_lock_bh(&vsock_table_lock); 316d021c344SAndy King sk = __vsock_find_connected_socket(src, dst); 317d021c344SAndy King if (sk) 318d021c344SAndy King sock_hold(sk); 319d021c344SAndy King 320d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 321d021c344SAndy King 322d021c344SAndy King return sk; 323d021c344SAndy King } 324d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_find_connected_socket); 325d021c344SAndy King 326d021c344SAndy King static bool vsock_in_bound_table(struct vsock_sock *vsk) 327d021c344SAndy King { 328d021c344SAndy King bool ret; 329d021c344SAndy King 330d021c344SAndy King spin_lock_bh(&vsock_table_lock); 331d021c344SAndy King ret = __vsock_in_bound_table(vsk); 332d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 333d021c344SAndy King 334d021c344SAndy King return ret; 335d021c344SAndy King } 336d021c344SAndy King 337d021c344SAndy King static bool vsock_in_connected_table(struct vsock_sock *vsk) 338d021c344SAndy King { 339d021c344SAndy King bool ret; 340d021c344SAndy King 341d021c344SAndy King spin_lock_bh(&vsock_table_lock); 342d021c344SAndy King ret = __vsock_in_connected_table(vsk); 343d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 344d021c344SAndy King 345d021c344SAndy King return ret; 346d021c344SAndy King } 347d021c344SAndy King 3486773b7dcSStefan Hajnoczi void vsock_remove_sock(struct vsock_sock *vsk) 3496773b7dcSStefan Hajnoczi { 3506773b7dcSStefan Hajnoczi if (vsock_in_bound_table(vsk)) 3516773b7dcSStefan Hajnoczi vsock_remove_bound(vsk); 3526773b7dcSStefan Hajnoczi 3536773b7dcSStefan Hajnoczi if (vsock_in_connected_table(vsk)) 3546773b7dcSStefan Hajnoczi vsock_remove_connected(vsk); 3556773b7dcSStefan Hajnoczi } 3566773b7dcSStefan Hajnoczi EXPORT_SYMBOL_GPL(vsock_remove_sock); 3576773b7dcSStefan Hajnoczi 358d021c344SAndy King void vsock_for_each_connected_socket(void (*fn)(struct sock *sk)) 359d021c344SAndy King { 360d021c344SAndy King int i; 361d021c344SAndy King 362d021c344SAndy King spin_lock_bh(&vsock_table_lock); 363d021c344SAndy King 364d021c344SAndy King for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) { 365d021c344SAndy King struct vsock_sock *vsk; 366d021c344SAndy King list_for_each_entry(vsk, &vsock_connected_table[i], 367d9af2d67SJulia Lawall connected_table) 368d021c344SAndy King fn(sk_vsock(vsk)); 369d021c344SAndy King } 370d021c344SAndy King 371d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 372d021c344SAndy King } 373d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_for_each_connected_socket); 374d021c344SAndy King 375d021c344SAndy King void vsock_add_pending(struct sock *listener, struct sock *pending) 376d021c344SAndy King { 377d021c344SAndy King struct vsock_sock *vlistener; 378d021c344SAndy King struct vsock_sock *vpending; 379d021c344SAndy King 380d021c344SAndy King vlistener = vsock_sk(listener); 381d021c344SAndy King vpending = vsock_sk(pending); 382d021c344SAndy King 383d021c344SAndy King sock_hold(pending); 384d021c344SAndy King sock_hold(listener); 385d021c344SAndy King list_add_tail(&vpending->pending_links, &vlistener->pending_links); 386d021c344SAndy King } 387d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_add_pending); 388d021c344SAndy King 389d021c344SAndy King void vsock_remove_pending(struct sock *listener, struct sock *pending) 390d021c344SAndy King { 391d021c344SAndy King struct vsock_sock *vpending = vsock_sk(pending); 392d021c344SAndy King 393d021c344SAndy King list_del_init(&vpending->pending_links); 394d021c344SAndy King sock_put(listener); 395d021c344SAndy King sock_put(pending); 396d021c344SAndy King } 397d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_remove_pending); 398d021c344SAndy King 399d021c344SAndy King void vsock_enqueue_accept(struct sock *listener, struct sock *connected) 400d021c344SAndy King { 401d021c344SAndy King struct vsock_sock *vlistener; 402d021c344SAndy King struct vsock_sock *vconnected; 403d021c344SAndy King 404d021c344SAndy King vlistener = vsock_sk(listener); 405d021c344SAndy King vconnected = vsock_sk(connected); 406d021c344SAndy King 407d021c344SAndy King sock_hold(connected); 408d021c344SAndy King sock_hold(listener); 409d021c344SAndy King list_add_tail(&vconnected->accept_queue, &vlistener->accept_queue); 410d021c344SAndy King } 411d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_enqueue_accept); 412d021c344SAndy King 413d021c344SAndy King static struct sock *vsock_dequeue_accept(struct sock *listener) 414d021c344SAndy King { 415d021c344SAndy King struct vsock_sock *vlistener; 416d021c344SAndy King struct vsock_sock *vconnected; 417d021c344SAndy King 418d021c344SAndy King vlistener = vsock_sk(listener); 419d021c344SAndy King 420d021c344SAndy King if (list_empty(&vlistener->accept_queue)) 421d021c344SAndy King return NULL; 422d021c344SAndy King 423d021c344SAndy King vconnected = list_entry(vlistener->accept_queue.next, 424d021c344SAndy King struct vsock_sock, accept_queue); 425d021c344SAndy King 426d021c344SAndy King list_del_init(&vconnected->accept_queue); 427d021c344SAndy King sock_put(listener); 428d021c344SAndy King /* The caller will need a reference on the connected socket so we let 429d021c344SAndy King * it call sock_put(). 430d021c344SAndy King */ 431d021c344SAndy King 432d021c344SAndy King return sk_vsock(vconnected); 433d021c344SAndy King } 434d021c344SAndy King 435d021c344SAndy King static bool vsock_is_accept_queue_empty(struct sock *sk) 436d021c344SAndy King { 437d021c344SAndy King struct vsock_sock *vsk = vsock_sk(sk); 438d021c344SAndy King return list_empty(&vsk->accept_queue); 439d021c344SAndy King } 440d021c344SAndy King 441d021c344SAndy King static bool vsock_is_pending(struct sock *sk) 442d021c344SAndy King { 443d021c344SAndy King struct vsock_sock *vsk = vsock_sk(sk); 444d021c344SAndy King return !list_empty(&vsk->pending_links); 445d021c344SAndy King } 446d021c344SAndy King 447d021c344SAndy King static int vsock_send_shutdown(struct sock *sk, int mode) 448d021c344SAndy King { 449d021c344SAndy King return transport->shutdown(vsock_sk(sk), mode); 450d021c344SAndy King } 451d021c344SAndy King 452d021c344SAndy King void vsock_pending_work(struct work_struct *work) 453d021c344SAndy King { 454d021c344SAndy King struct sock *sk; 455d021c344SAndy King struct sock *listener; 456d021c344SAndy King struct vsock_sock *vsk; 457d021c344SAndy King bool cleanup; 458d021c344SAndy King 459d021c344SAndy King vsk = container_of(work, struct vsock_sock, dwork.work); 460d021c344SAndy King sk = sk_vsock(vsk); 461d021c344SAndy King listener = vsk->listener; 462d021c344SAndy King cleanup = true; 463d021c344SAndy King 464d021c344SAndy King lock_sock(listener); 4654192f672SStefan Hajnoczi lock_sock_nested(sk, SINGLE_DEPTH_NESTING); 466d021c344SAndy King 467d021c344SAndy King if (vsock_is_pending(sk)) { 468d021c344SAndy King vsock_remove_pending(listener, sk); 4691190cfdbSJorgen Hansen 4701190cfdbSJorgen Hansen listener->sk_ack_backlog--; 471d021c344SAndy King } else if (!vsk->rejected) { 472d021c344SAndy King /* We are not on the pending list and accept() did not reject 473d021c344SAndy King * us, so we must have been accepted by our user process. We 474d021c344SAndy King * just need to drop our references to the sockets and be on 475d021c344SAndy King * our way. 476d021c344SAndy King */ 477d021c344SAndy King cleanup = false; 478d021c344SAndy King goto out; 479d021c344SAndy King } 480d021c344SAndy King 481d021c344SAndy King /* We need to remove ourself from the global connected sockets list so 482d021c344SAndy King * incoming packets can't find this socket, and to reduce the reference 483d021c344SAndy King * count. 484d021c344SAndy King */ 485d021c344SAndy King if (vsock_in_connected_table(vsk)) 486d021c344SAndy King vsock_remove_connected(vsk); 487d021c344SAndy King 488d021c344SAndy King sk->sk_state = SS_FREE; 489d021c344SAndy King 490d021c344SAndy King out: 491d021c344SAndy King release_sock(sk); 492d021c344SAndy King release_sock(listener); 493d021c344SAndy King if (cleanup) 494d021c344SAndy King sock_put(sk); 495d021c344SAndy King 496d021c344SAndy King sock_put(sk); 497d021c344SAndy King sock_put(listener); 498d021c344SAndy King } 499d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_pending_work); 500d021c344SAndy King 501d021c344SAndy King /**** SOCKET OPERATIONS ****/ 502d021c344SAndy King 503d021c344SAndy King static int __vsock_bind_stream(struct vsock_sock *vsk, 504d021c344SAndy King struct sockaddr_vm *addr) 505d021c344SAndy King { 506d021c344SAndy King static u32 port = LAST_RESERVED_PORT + 1; 507d021c344SAndy King struct sockaddr_vm new_addr; 508d021c344SAndy King 509d021c344SAndy King vsock_addr_init(&new_addr, addr->svm_cid, addr->svm_port); 510d021c344SAndy King 511d021c344SAndy King if (addr->svm_port == VMADDR_PORT_ANY) { 512d021c344SAndy King bool found = false; 513d021c344SAndy King unsigned int i; 514d021c344SAndy King 515d021c344SAndy King for (i = 0; i < MAX_PORT_RETRIES; i++) { 516d021c344SAndy King if (port <= LAST_RESERVED_PORT) 517d021c344SAndy King port = LAST_RESERVED_PORT + 1; 518d021c344SAndy King 519d021c344SAndy King new_addr.svm_port = port++; 520d021c344SAndy King 521d021c344SAndy King if (!__vsock_find_bound_socket(&new_addr)) { 522d021c344SAndy King found = true; 523d021c344SAndy King break; 524d021c344SAndy King } 525d021c344SAndy King } 526d021c344SAndy King 527d021c344SAndy King if (!found) 528d021c344SAndy King return -EADDRNOTAVAIL; 529d021c344SAndy King } else { 530d021c344SAndy King /* If port is in reserved range, ensure caller 531d021c344SAndy King * has necessary privileges. 532d021c344SAndy King */ 533d021c344SAndy King if (addr->svm_port <= LAST_RESERVED_PORT && 534d021c344SAndy King !capable(CAP_NET_BIND_SERVICE)) { 535d021c344SAndy King return -EACCES; 536d021c344SAndy King } 537d021c344SAndy King 538d021c344SAndy King if (__vsock_find_bound_socket(&new_addr)) 539d021c344SAndy King return -EADDRINUSE; 540d021c344SAndy King } 541d021c344SAndy King 542d021c344SAndy King vsock_addr_init(&vsk->local_addr, new_addr.svm_cid, new_addr.svm_port); 543d021c344SAndy King 544d021c344SAndy King /* Remove stream sockets from the unbound list and add them to the hash 545d021c344SAndy King * table for easy lookup by its address. The unbound list is simply an 546d021c344SAndy King * extra entry at the end of the hash table, a trick used by AF_UNIX. 547d021c344SAndy King */ 548d021c344SAndy King __vsock_remove_bound(vsk); 549d021c344SAndy King __vsock_insert_bound(vsock_bound_sockets(&vsk->local_addr), vsk); 550d021c344SAndy King 551d021c344SAndy King return 0; 552d021c344SAndy King } 553d021c344SAndy King 554d021c344SAndy King static int __vsock_bind_dgram(struct vsock_sock *vsk, 555d021c344SAndy King struct sockaddr_vm *addr) 556d021c344SAndy King { 557d021c344SAndy King return transport->dgram_bind(vsk, addr); 558d021c344SAndy King } 559d021c344SAndy King 560d021c344SAndy King static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr) 561d021c344SAndy King { 562d021c344SAndy King struct vsock_sock *vsk = vsock_sk(sk); 563d021c344SAndy King u32 cid; 564d021c344SAndy King int retval; 565d021c344SAndy King 566d021c344SAndy King /* First ensure this socket isn't already bound. */ 567d021c344SAndy King if (vsock_addr_bound(&vsk->local_addr)) 568d021c344SAndy King return -EINVAL; 569d021c344SAndy King 570d021c344SAndy King /* Now bind to the provided address or select appropriate values if 571d021c344SAndy King * none are provided (VMADDR_CID_ANY and VMADDR_PORT_ANY). Note that 572d021c344SAndy King * like AF_INET prevents binding to a non-local IP address (in most 573d021c344SAndy King * cases), we only allow binding to the local CID. 574d021c344SAndy King */ 575d021c344SAndy King cid = transport->get_local_cid(); 576d021c344SAndy King if (addr->svm_cid != cid && addr->svm_cid != VMADDR_CID_ANY) 577d021c344SAndy King return -EADDRNOTAVAIL; 578d021c344SAndy King 579d021c344SAndy King switch (sk->sk_socket->type) { 580d021c344SAndy King case SOCK_STREAM: 581d021c344SAndy King spin_lock_bh(&vsock_table_lock); 582d021c344SAndy King retval = __vsock_bind_stream(vsk, addr); 583d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 584d021c344SAndy King break; 585d021c344SAndy King 586d021c344SAndy King case SOCK_DGRAM: 587d021c344SAndy King retval = __vsock_bind_dgram(vsk, addr); 588d021c344SAndy King break; 589d021c344SAndy King 590d021c344SAndy King default: 591d021c344SAndy King retval = -EINVAL; 592d021c344SAndy King break; 593d021c344SAndy King } 594d021c344SAndy King 595d021c344SAndy King return retval; 596d021c344SAndy King } 597d021c344SAndy King 598d021c344SAndy King struct sock *__vsock_create(struct net *net, 599d021c344SAndy King struct socket *sock, 600d021c344SAndy King struct sock *parent, 601d021c344SAndy King gfp_t priority, 60211aa9c28SEric W. Biederman unsigned short type, 60311aa9c28SEric W. Biederman int kern) 604d021c344SAndy King { 605d021c344SAndy King struct sock *sk; 606d021c344SAndy King struct vsock_sock *psk; 607d021c344SAndy King struct vsock_sock *vsk; 608d021c344SAndy King 60911aa9c28SEric W. Biederman sk = sk_alloc(net, AF_VSOCK, priority, &vsock_proto, kern); 610d021c344SAndy King if (!sk) 611d021c344SAndy King return NULL; 612d021c344SAndy King 613d021c344SAndy King sock_init_data(sock, sk); 614d021c344SAndy King 615d021c344SAndy King /* sk->sk_type is normally set in sock_init_data, but only if sock is 616d021c344SAndy King * non-NULL. We make sure that our sockets always have a type by 617d021c344SAndy King * setting it here if needed. 618d021c344SAndy King */ 619d021c344SAndy King if (!sock) 620d021c344SAndy King sk->sk_type = type; 621d021c344SAndy King 622d021c344SAndy King vsk = vsock_sk(sk); 623d021c344SAndy King vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 624d021c344SAndy King vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 625d021c344SAndy King 626d021c344SAndy King sk->sk_destruct = vsock_sk_destruct; 627d021c344SAndy King sk->sk_backlog_rcv = vsock_queue_rcv_skb; 628d021c344SAndy King sk->sk_state = 0; 629d021c344SAndy King sock_reset_flag(sk, SOCK_DONE); 630d021c344SAndy King 631d021c344SAndy King INIT_LIST_HEAD(&vsk->bound_table); 632d021c344SAndy King INIT_LIST_HEAD(&vsk->connected_table); 633d021c344SAndy King vsk->listener = NULL; 634d021c344SAndy King INIT_LIST_HEAD(&vsk->pending_links); 635d021c344SAndy King INIT_LIST_HEAD(&vsk->accept_queue); 636d021c344SAndy King vsk->rejected = false; 637d021c344SAndy King vsk->sent_request = false; 638d021c344SAndy King vsk->ignore_connecting_rst = false; 639d021c344SAndy King vsk->peer_shutdown = 0; 640d021c344SAndy King 641d021c344SAndy King psk = parent ? vsock_sk(parent) : NULL; 642d021c344SAndy King if (parent) { 643d021c344SAndy King vsk->trusted = psk->trusted; 644d021c344SAndy King vsk->owner = get_cred(psk->owner); 645d021c344SAndy King vsk->connect_timeout = psk->connect_timeout; 646d021c344SAndy King } else { 647d021c344SAndy King vsk->trusted = capable(CAP_NET_ADMIN); 648d021c344SAndy King vsk->owner = get_current_cred(); 649d021c344SAndy King vsk->connect_timeout = VSOCK_DEFAULT_CONNECT_TIMEOUT; 650d021c344SAndy King } 651d021c344SAndy King 652d021c344SAndy King if (transport->init(vsk, psk) < 0) { 653d021c344SAndy King sk_free(sk); 654d021c344SAndy King return NULL; 655d021c344SAndy King } 656d021c344SAndy King 657d021c344SAndy King if (sock) 658d021c344SAndy King vsock_insert_unbound(vsk); 659d021c344SAndy King 660d021c344SAndy King return sk; 661d021c344SAndy King } 662d021c344SAndy King EXPORT_SYMBOL_GPL(__vsock_create); 663d021c344SAndy King 664d021c344SAndy King static void __vsock_release(struct sock *sk) 665d021c344SAndy King { 666d021c344SAndy King if (sk) { 667d021c344SAndy King struct sk_buff *skb; 668d021c344SAndy King struct sock *pending; 669d021c344SAndy King struct vsock_sock *vsk; 670d021c344SAndy King 671d021c344SAndy King vsk = vsock_sk(sk); 672d021c344SAndy King pending = NULL; /* Compiler warning. */ 673d021c344SAndy King 674d021c344SAndy King transport->release(vsk); 675d021c344SAndy King 676d021c344SAndy King lock_sock(sk); 677d021c344SAndy King sock_orphan(sk); 678d021c344SAndy King sk->sk_shutdown = SHUTDOWN_MASK; 679d021c344SAndy King 680d021c344SAndy King while ((skb = skb_dequeue(&sk->sk_receive_queue))) 681d021c344SAndy King kfree_skb(skb); 682d021c344SAndy King 683d021c344SAndy King /* Clean up any sockets that never were accepted. */ 684d021c344SAndy King while ((pending = vsock_dequeue_accept(sk)) != NULL) { 685d021c344SAndy King __vsock_release(pending); 686d021c344SAndy King sock_put(pending); 687d021c344SAndy King } 688d021c344SAndy King 689d021c344SAndy King release_sock(sk); 690d021c344SAndy King sock_put(sk); 691d021c344SAndy King } 692d021c344SAndy King } 693d021c344SAndy King 694d021c344SAndy King static void vsock_sk_destruct(struct sock *sk) 695d021c344SAndy King { 696d021c344SAndy King struct vsock_sock *vsk = vsock_sk(sk); 697d021c344SAndy King 698d021c344SAndy King transport->destruct(vsk); 699d021c344SAndy King 700d021c344SAndy King /* When clearing these addresses, there's no need to set the family and 701d021c344SAndy King * possibly register the address family with the kernel. 702d021c344SAndy King */ 703d021c344SAndy King vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 704d021c344SAndy King vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 705d021c344SAndy King 706d021c344SAndy King put_cred(vsk->owner); 707d021c344SAndy King } 708d021c344SAndy King 709d021c344SAndy King static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) 710d021c344SAndy King { 711d021c344SAndy King int err; 712d021c344SAndy King 713d021c344SAndy King err = sock_queue_rcv_skb(sk, skb); 714d021c344SAndy King if (err) 715d021c344SAndy King kfree_skb(skb); 716d021c344SAndy King 717d021c344SAndy King return err; 718d021c344SAndy King } 719d021c344SAndy King 720d021c344SAndy King s64 vsock_stream_has_data(struct vsock_sock *vsk) 721d021c344SAndy King { 722d021c344SAndy King return transport->stream_has_data(vsk); 723d021c344SAndy King } 724d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_stream_has_data); 725d021c344SAndy King 726d021c344SAndy King s64 vsock_stream_has_space(struct vsock_sock *vsk) 727d021c344SAndy King { 728d021c344SAndy King return transport->stream_has_space(vsk); 729d021c344SAndy King } 730d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_stream_has_space); 731d021c344SAndy King 732d021c344SAndy King static int vsock_release(struct socket *sock) 733d021c344SAndy King { 734d021c344SAndy King __vsock_release(sock->sk); 735d021c344SAndy King sock->sk = NULL; 736d021c344SAndy King sock->state = SS_FREE; 737d021c344SAndy King 738d021c344SAndy King return 0; 739d021c344SAndy King } 740d021c344SAndy King 741d021c344SAndy King static int 742d021c344SAndy King vsock_bind(struct socket *sock, struct sockaddr *addr, int addr_len) 743d021c344SAndy King { 744d021c344SAndy King int err; 745d021c344SAndy King struct sock *sk; 746d021c344SAndy King struct sockaddr_vm *vm_addr; 747d021c344SAndy King 748d021c344SAndy King sk = sock->sk; 749d021c344SAndy King 750d021c344SAndy King if (vsock_addr_cast(addr, addr_len, &vm_addr) != 0) 751d021c344SAndy King return -EINVAL; 752d021c344SAndy King 753d021c344SAndy King lock_sock(sk); 754d021c344SAndy King err = __vsock_bind(sk, vm_addr); 755d021c344SAndy King release_sock(sk); 756d021c344SAndy King 757d021c344SAndy King return err; 758d021c344SAndy King } 759d021c344SAndy King 760d021c344SAndy King static int vsock_getname(struct socket *sock, 761d021c344SAndy King struct sockaddr *addr, int *addr_len, int peer) 762d021c344SAndy King { 763d021c344SAndy King int err; 764d021c344SAndy King struct sock *sk; 765d021c344SAndy King struct vsock_sock *vsk; 766d021c344SAndy King struct sockaddr_vm *vm_addr; 767d021c344SAndy King 768d021c344SAndy King sk = sock->sk; 769d021c344SAndy King vsk = vsock_sk(sk); 770d021c344SAndy King err = 0; 771d021c344SAndy King 772d021c344SAndy King lock_sock(sk); 773d021c344SAndy King 774d021c344SAndy King if (peer) { 775d021c344SAndy King if (sock->state != SS_CONNECTED) { 776d021c344SAndy King err = -ENOTCONN; 777d021c344SAndy King goto out; 778d021c344SAndy King } 779d021c344SAndy King vm_addr = &vsk->remote_addr; 780d021c344SAndy King } else { 781d021c344SAndy King vm_addr = &vsk->local_addr; 782d021c344SAndy King } 783d021c344SAndy King 784d021c344SAndy King if (!vm_addr) { 785d021c344SAndy King err = -EINVAL; 786d021c344SAndy King goto out; 787d021c344SAndy King } 788d021c344SAndy King 789d021c344SAndy King /* sys_getsockname() and sys_getpeername() pass us a 790d021c344SAndy King * MAX_SOCK_ADDR-sized buffer and don't set addr_len. Unfortunately 791d021c344SAndy King * that macro is defined in socket.c instead of .h, so we hardcode its 792d021c344SAndy King * value here. 793d021c344SAndy King */ 794d021c344SAndy King BUILD_BUG_ON(sizeof(*vm_addr) > 128); 795d021c344SAndy King memcpy(addr, vm_addr, sizeof(*vm_addr)); 796d021c344SAndy King *addr_len = sizeof(*vm_addr); 797d021c344SAndy King 798d021c344SAndy King out: 799d021c344SAndy King release_sock(sk); 800d021c344SAndy King return err; 801d021c344SAndy King } 802d021c344SAndy King 803d021c344SAndy King static int vsock_shutdown(struct socket *sock, int mode) 804d021c344SAndy King { 805d021c344SAndy King int err; 806d021c344SAndy King struct sock *sk; 807d021c344SAndy King 808d021c344SAndy King /* User level uses SHUT_RD (0) and SHUT_WR (1), but the kernel uses 809d021c344SAndy King * RCV_SHUTDOWN (1) and SEND_SHUTDOWN (2), so we must increment mode 810d021c344SAndy King * here like the other address families do. Note also that the 811d021c344SAndy King * increment makes SHUT_RDWR (2) into RCV_SHUTDOWN | SEND_SHUTDOWN (3), 812d021c344SAndy King * which is what we want. 813d021c344SAndy King */ 814d021c344SAndy King mode++; 815d021c344SAndy King 816d021c344SAndy King if ((mode & ~SHUTDOWN_MASK) || !mode) 817d021c344SAndy King return -EINVAL; 818d021c344SAndy King 819d021c344SAndy King /* If this is a STREAM socket and it is not connected then bail out 820d021c344SAndy King * immediately. If it is a DGRAM socket then we must first kick the 821d021c344SAndy King * socket so that it wakes up from any sleeping calls, for example 822d021c344SAndy King * recv(), and then afterwards return the error. 823d021c344SAndy King */ 824d021c344SAndy King 825d021c344SAndy King sk = sock->sk; 826d021c344SAndy King if (sock->state == SS_UNCONNECTED) { 827d021c344SAndy King err = -ENOTCONN; 828d021c344SAndy King if (sk->sk_type == SOCK_STREAM) 829d021c344SAndy King return err; 830d021c344SAndy King } else { 831d021c344SAndy King sock->state = SS_DISCONNECTING; 832d021c344SAndy King err = 0; 833d021c344SAndy King } 834d021c344SAndy King 835d021c344SAndy King /* Receive and send shutdowns are treated alike. */ 836d021c344SAndy King mode = mode & (RCV_SHUTDOWN | SEND_SHUTDOWN); 837d021c344SAndy King if (mode) { 838d021c344SAndy King lock_sock(sk); 839d021c344SAndy King sk->sk_shutdown |= mode; 840d021c344SAndy King sk->sk_state_change(sk); 841d021c344SAndy King release_sock(sk); 842d021c344SAndy King 843d021c344SAndy King if (sk->sk_type == SOCK_STREAM) { 844d021c344SAndy King sock_reset_flag(sk, SOCK_DONE); 845d021c344SAndy King vsock_send_shutdown(sk, mode); 846d021c344SAndy King } 847d021c344SAndy King } 848d021c344SAndy King 849d021c344SAndy King return err; 850d021c344SAndy King } 851d021c344SAndy King 852d021c344SAndy King static unsigned int vsock_poll(struct file *file, struct socket *sock, 853d021c344SAndy King poll_table *wait) 854d021c344SAndy King { 855d021c344SAndy King struct sock *sk; 856d021c344SAndy King unsigned int mask; 857d021c344SAndy King struct vsock_sock *vsk; 858d021c344SAndy King 859d021c344SAndy King sk = sock->sk; 860d021c344SAndy King vsk = vsock_sk(sk); 861d021c344SAndy King 862d021c344SAndy King poll_wait(file, sk_sleep(sk), wait); 863d021c344SAndy King mask = 0; 864d021c344SAndy King 865d021c344SAndy King if (sk->sk_err) 866d021c344SAndy King /* Signify that there has been an error on this socket. */ 867d021c344SAndy King mask |= POLLERR; 868d021c344SAndy King 869d021c344SAndy King /* INET sockets treat local write shutdown and peer write shutdown as a 870d021c344SAndy King * case of POLLHUP set. 871d021c344SAndy King */ 872d021c344SAndy King if ((sk->sk_shutdown == SHUTDOWN_MASK) || 873d021c344SAndy King ((sk->sk_shutdown & SEND_SHUTDOWN) && 874d021c344SAndy King (vsk->peer_shutdown & SEND_SHUTDOWN))) { 875d021c344SAndy King mask |= POLLHUP; 876d021c344SAndy King } 877d021c344SAndy King 878d021c344SAndy King if (sk->sk_shutdown & RCV_SHUTDOWN || 879d021c344SAndy King vsk->peer_shutdown & SEND_SHUTDOWN) { 880d021c344SAndy King mask |= POLLRDHUP; 881d021c344SAndy King } 882d021c344SAndy King 883d021c344SAndy King if (sock->type == SOCK_DGRAM) { 884d021c344SAndy King /* For datagram sockets we can read if there is something in 885d021c344SAndy King * the queue and write as long as the socket isn't shutdown for 886d021c344SAndy King * sending. 887d021c344SAndy King */ 888d021c344SAndy King if (!skb_queue_empty(&sk->sk_receive_queue) || 889d021c344SAndy King (sk->sk_shutdown & RCV_SHUTDOWN)) { 890d021c344SAndy King mask |= POLLIN | POLLRDNORM; 891d021c344SAndy King } 892d021c344SAndy King 893d021c344SAndy King if (!(sk->sk_shutdown & SEND_SHUTDOWN)) 894d021c344SAndy King mask |= POLLOUT | POLLWRNORM | POLLWRBAND; 895d021c344SAndy King 896d021c344SAndy King } else if (sock->type == SOCK_STREAM) { 897d021c344SAndy King lock_sock(sk); 898d021c344SAndy King 899d021c344SAndy King /* Listening sockets that have connections in their accept 900d021c344SAndy King * queue can be read. 901d021c344SAndy King */ 902ea3803c1SStefan Hajnoczi if (sk->sk_state == VSOCK_SS_LISTEN 903d021c344SAndy King && !vsock_is_accept_queue_empty(sk)) 904d021c344SAndy King mask |= POLLIN | POLLRDNORM; 905d021c344SAndy King 906d021c344SAndy King /* If there is something in the queue then we can read. */ 907d021c344SAndy King if (transport->stream_is_active(vsk) && 908d021c344SAndy King !(sk->sk_shutdown & RCV_SHUTDOWN)) { 909d021c344SAndy King bool data_ready_now = false; 910d021c344SAndy King int ret = transport->notify_poll_in( 911d021c344SAndy King vsk, 1, &data_ready_now); 912d021c344SAndy King if (ret < 0) { 913d021c344SAndy King mask |= POLLERR; 914d021c344SAndy King } else { 915d021c344SAndy King if (data_ready_now) 916d021c344SAndy King mask |= POLLIN | POLLRDNORM; 917d021c344SAndy King 918d021c344SAndy King } 919d021c344SAndy King } 920d021c344SAndy King 921d021c344SAndy King /* Sockets whose connections have been closed, reset, or 922d021c344SAndy King * terminated should also be considered read, and we check the 923d021c344SAndy King * shutdown flag for that. 924d021c344SAndy King */ 925d021c344SAndy King if (sk->sk_shutdown & RCV_SHUTDOWN || 926d021c344SAndy King vsk->peer_shutdown & SEND_SHUTDOWN) { 927d021c344SAndy King mask |= POLLIN | POLLRDNORM; 928d021c344SAndy King } 929d021c344SAndy King 930d021c344SAndy King /* Connected sockets that can produce data can be written. */ 931d021c344SAndy King if (sk->sk_state == SS_CONNECTED) { 932d021c344SAndy King if (!(sk->sk_shutdown & SEND_SHUTDOWN)) { 933d021c344SAndy King bool space_avail_now = false; 934d021c344SAndy King int ret = transport->notify_poll_out( 935d021c344SAndy King vsk, 1, &space_avail_now); 936d021c344SAndy King if (ret < 0) { 937d021c344SAndy King mask |= POLLERR; 938d021c344SAndy King } else { 939d021c344SAndy King if (space_avail_now) 940d021c344SAndy King /* Remove POLLWRBAND since INET 941d021c344SAndy King * sockets are not setting it. 942d021c344SAndy King */ 943d021c344SAndy King mask |= POLLOUT | POLLWRNORM; 944d021c344SAndy King 945d021c344SAndy King } 946d021c344SAndy King } 947d021c344SAndy King } 948d021c344SAndy King 949d021c344SAndy King /* Simulate INET socket poll behaviors, which sets 950d021c344SAndy King * POLLOUT|POLLWRNORM when peer is closed and nothing to read, 951d021c344SAndy King * but local send is not shutdown. 952d021c344SAndy King */ 953d021c344SAndy King if (sk->sk_state == SS_UNCONNECTED) { 954d021c344SAndy King if (!(sk->sk_shutdown & SEND_SHUTDOWN)) 955d021c344SAndy King mask |= POLLOUT | POLLWRNORM; 956d021c344SAndy King 957d021c344SAndy King } 958d021c344SAndy King 959d021c344SAndy King release_sock(sk); 960d021c344SAndy King } 961d021c344SAndy King 962d021c344SAndy King return mask; 963d021c344SAndy King } 964d021c344SAndy King 9651b784140SYing Xue static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg, 9661b784140SYing Xue size_t len) 967d021c344SAndy King { 968d021c344SAndy King int err; 969d021c344SAndy King struct sock *sk; 970d021c344SAndy King struct vsock_sock *vsk; 971d021c344SAndy King struct sockaddr_vm *remote_addr; 972d021c344SAndy King 973d021c344SAndy King if (msg->msg_flags & MSG_OOB) 974d021c344SAndy King return -EOPNOTSUPP; 975d021c344SAndy King 976d021c344SAndy King /* For now, MSG_DONTWAIT is always assumed... */ 977d021c344SAndy King err = 0; 978d021c344SAndy King sk = sock->sk; 979d021c344SAndy King vsk = vsock_sk(sk); 980d021c344SAndy King 981d021c344SAndy King lock_sock(sk); 982d021c344SAndy King 983b3a6dfe8SAsias He err = vsock_auto_bind(vsk); 984b3a6dfe8SAsias He if (err) 985d021c344SAndy King goto out; 986d021c344SAndy King 987d021c344SAndy King 988d021c344SAndy King /* If the provided message contains an address, use that. Otherwise 989d021c344SAndy King * fall back on the socket's remote handle (if it has been connected). 990d021c344SAndy King */ 991d021c344SAndy King if (msg->msg_name && 992d021c344SAndy King vsock_addr_cast(msg->msg_name, msg->msg_namelen, 993d021c344SAndy King &remote_addr) == 0) { 994d021c344SAndy King /* Ensure this address is of the right type and is a valid 995d021c344SAndy King * destination. 996d021c344SAndy King */ 997d021c344SAndy King 998d021c344SAndy King if (remote_addr->svm_cid == VMADDR_CID_ANY) 999d021c344SAndy King remote_addr->svm_cid = transport->get_local_cid(); 1000d021c344SAndy King 1001d021c344SAndy King if (!vsock_addr_bound(remote_addr)) { 1002d021c344SAndy King err = -EINVAL; 1003d021c344SAndy King goto out; 1004d021c344SAndy King } 1005d021c344SAndy King } else if (sock->state == SS_CONNECTED) { 1006d021c344SAndy King remote_addr = &vsk->remote_addr; 1007d021c344SAndy King 1008d021c344SAndy King if (remote_addr->svm_cid == VMADDR_CID_ANY) 1009d021c344SAndy King remote_addr->svm_cid = transport->get_local_cid(); 1010d021c344SAndy King 1011d021c344SAndy King /* XXX Should connect() or this function ensure remote_addr is 1012d021c344SAndy King * bound? 1013d021c344SAndy King */ 1014d021c344SAndy King if (!vsock_addr_bound(&vsk->remote_addr)) { 1015d021c344SAndy King err = -EINVAL; 1016d021c344SAndy King goto out; 1017d021c344SAndy King } 1018d021c344SAndy King } else { 1019d021c344SAndy King err = -EINVAL; 1020d021c344SAndy King goto out; 1021d021c344SAndy King } 1022d021c344SAndy King 1023d021c344SAndy King if (!transport->dgram_allow(remote_addr->svm_cid, 1024d021c344SAndy King remote_addr->svm_port)) { 1025d021c344SAndy King err = -EINVAL; 1026d021c344SAndy King goto out; 1027d021c344SAndy King } 1028d021c344SAndy King 10290f7db23aSAl Viro err = transport->dgram_enqueue(vsk, remote_addr, msg, len); 1030d021c344SAndy King 1031d021c344SAndy King out: 1032d021c344SAndy King release_sock(sk); 1033d021c344SAndy King return err; 1034d021c344SAndy King } 1035d021c344SAndy King 1036d021c344SAndy King static int vsock_dgram_connect(struct socket *sock, 1037d021c344SAndy King struct sockaddr *addr, int addr_len, int flags) 1038d021c344SAndy King { 1039d021c344SAndy King int err; 1040d021c344SAndy King struct sock *sk; 1041d021c344SAndy King struct vsock_sock *vsk; 1042d021c344SAndy King struct sockaddr_vm *remote_addr; 1043d021c344SAndy King 1044d021c344SAndy King sk = sock->sk; 1045d021c344SAndy King vsk = vsock_sk(sk); 1046d021c344SAndy King 1047d021c344SAndy King err = vsock_addr_cast(addr, addr_len, &remote_addr); 1048d021c344SAndy King if (err == -EAFNOSUPPORT && remote_addr->svm_family == AF_UNSPEC) { 1049d021c344SAndy King lock_sock(sk); 1050d021c344SAndy King vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, 1051d021c344SAndy King VMADDR_PORT_ANY); 1052d021c344SAndy King sock->state = SS_UNCONNECTED; 1053d021c344SAndy King release_sock(sk); 1054d021c344SAndy King return 0; 1055d021c344SAndy King } else if (err != 0) 1056d021c344SAndy King return -EINVAL; 1057d021c344SAndy King 1058d021c344SAndy King lock_sock(sk); 1059d021c344SAndy King 1060b3a6dfe8SAsias He err = vsock_auto_bind(vsk); 1061b3a6dfe8SAsias He if (err) 1062d021c344SAndy King goto out; 1063d021c344SAndy King 1064d021c344SAndy King if (!transport->dgram_allow(remote_addr->svm_cid, 1065d021c344SAndy King remote_addr->svm_port)) { 1066d021c344SAndy King err = -EINVAL; 1067d021c344SAndy King goto out; 1068d021c344SAndy King } 1069d021c344SAndy King 1070d021c344SAndy King memcpy(&vsk->remote_addr, remote_addr, sizeof(vsk->remote_addr)); 1071d021c344SAndy King sock->state = SS_CONNECTED; 1072d021c344SAndy King 1073d021c344SAndy King out: 1074d021c344SAndy King release_sock(sk); 1075d021c344SAndy King return err; 1076d021c344SAndy King } 1077d021c344SAndy King 10781b784140SYing Xue static int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg, 10791b784140SYing Xue size_t len, int flags) 1080d021c344SAndy King { 10811b784140SYing Xue return transport->dgram_dequeue(vsock_sk(sock->sk), msg, len, flags); 1082d021c344SAndy King } 1083d021c344SAndy King 1084d021c344SAndy King static const struct proto_ops vsock_dgram_ops = { 1085d021c344SAndy King .family = PF_VSOCK, 1086d021c344SAndy King .owner = THIS_MODULE, 1087d021c344SAndy King .release = vsock_release, 1088d021c344SAndy King .bind = vsock_bind, 1089d021c344SAndy King .connect = vsock_dgram_connect, 1090d021c344SAndy King .socketpair = sock_no_socketpair, 1091d021c344SAndy King .accept = sock_no_accept, 1092d021c344SAndy King .getname = vsock_getname, 1093d021c344SAndy King .poll = vsock_poll, 1094d021c344SAndy King .ioctl = sock_no_ioctl, 1095d021c344SAndy King .listen = sock_no_listen, 1096d021c344SAndy King .shutdown = vsock_shutdown, 1097d021c344SAndy King .setsockopt = sock_no_setsockopt, 1098d021c344SAndy King .getsockopt = sock_no_getsockopt, 1099d021c344SAndy King .sendmsg = vsock_dgram_sendmsg, 1100d021c344SAndy King .recvmsg = vsock_dgram_recvmsg, 1101d021c344SAndy King .mmap = sock_no_mmap, 1102d021c344SAndy King .sendpage = sock_no_sendpage, 1103d021c344SAndy King }; 1104d021c344SAndy King 1105d021c344SAndy King static void vsock_connect_timeout(struct work_struct *work) 1106d021c344SAndy King { 1107d021c344SAndy King struct sock *sk; 1108d021c344SAndy King struct vsock_sock *vsk; 1109d021c344SAndy King 1110d021c344SAndy King vsk = container_of(work, struct vsock_sock, dwork.work); 1111d021c344SAndy King sk = sk_vsock(vsk); 1112d021c344SAndy King 1113d021c344SAndy King lock_sock(sk); 1114d021c344SAndy King if (sk->sk_state == SS_CONNECTING && 1115d021c344SAndy King (sk->sk_shutdown != SHUTDOWN_MASK)) { 1116d021c344SAndy King sk->sk_state = SS_UNCONNECTED; 1117d021c344SAndy King sk->sk_err = ETIMEDOUT; 1118d021c344SAndy King sk->sk_error_report(sk); 1119d021c344SAndy King } 1120d021c344SAndy King release_sock(sk); 1121d021c344SAndy King 1122d021c344SAndy King sock_put(sk); 1123d021c344SAndy King } 1124d021c344SAndy King 1125d021c344SAndy King static int vsock_stream_connect(struct socket *sock, struct sockaddr *addr, 1126d021c344SAndy King int addr_len, int flags) 1127d021c344SAndy King { 1128d021c344SAndy King int err; 1129d021c344SAndy King struct sock *sk; 1130d021c344SAndy King struct vsock_sock *vsk; 1131d021c344SAndy King struct sockaddr_vm *remote_addr; 1132d021c344SAndy King long timeout; 1133d021c344SAndy King DEFINE_WAIT(wait); 1134d021c344SAndy King 1135d021c344SAndy King err = 0; 1136d021c344SAndy King sk = sock->sk; 1137d021c344SAndy King vsk = vsock_sk(sk); 1138d021c344SAndy King 1139d021c344SAndy King lock_sock(sk); 1140d021c344SAndy King 1141d021c344SAndy King /* XXX AF_UNSPEC should make us disconnect like AF_INET. */ 1142d021c344SAndy King switch (sock->state) { 1143d021c344SAndy King case SS_CONNECTED: 1144d021c344SAndy King err = -EISCONN; 1145d021c344SAndy King goto out; 1146d021c344SAndy King case SS_DISCONNECTING: 1147d021c344SAndy King err = -EINVAL; 1148d021c344SAndy King goto out; 1149d021c344SAndy King case SS_CONNECTING: 1150d021c344SAndy King /* This continues on so we can move sock into the SS_CONNECTED 1151d021c344SAndy King * state once the connection has completed (at which point err 1152d021c344SAndy King * will be set to zero also). Otherwise, we will either wait 1153d021c344SAndy King * for the connection or return -EALREADY should this be a 1154d021c344SAndy King * non-blocking call. 1155d021c344SAndy King */ 1156d021c344SAndy King err = -EALREADY; 1157d021c344SAndy King break; 1158d021c344SAndy King default: 1159ea3803c1SStefan Hajnoczi if ((sk->sk_state == VSOCK_SS_LISTEN) || 1160d021c344SAndy King vsock_addr_cast(addr, addr_len, &remote_addr) != 0) { 1161d021c344SAndy King err = -EINVAL; 1162d021c344SAndy King goto out; 1163d021c344SAndy King } 1164d021c344SAndy King 1165d021c344SAndy King /* The hypervisor and well-known contexts do not have socket 1166d021c344SAndy King * endpoints. 1167d021c344SAndy King */ 1168d021c344SAndy King if (!transport->stream_allow(remote_addr->svm_cid, 1169d021c344SAndy King remote_addr->svm_port)) { 1170d021c344SAndy King err = -ENETUNREACH; 1171d021c344SAndy King goto out; 1172d021c344SAndy King } 1173d021c344SAndy King 1174d021c344SAndy King /* Set the remote address that we are connecting to. */ 1175d021c344SAndy King memcpy(&vsk->remote_addr, remote_addr, 1176d021c344SAndy King sizeof(vsk->remote_addr)); 1177d021c344SAndy King 1178b3a6dfe8SAsias He err = vsock_auto_bind(vsk); 1179b3a6dfe8SAsias He if (err) 1180d021c344SAndy King goto out; 1181d021c344SAndy King 1182d021c344SAndy King sk->sk_state = SS_CONNECTING; 1183d021c344SAndy King 1184d021c344SAndy King err = transport->connect(vsk); 1185d021c344SAndy King if (err < 0) 1186d021c344SAndy King goto out; 1187d021c344SAndy King 1188d021c344SAndy King /* Mark sock as connecting and set the error code to in 1189d021c344SAndy King * progress in case this is a non-blocking connect. 1190d021c344SAndy King */ 1191d021c344SAndy King sock->state = SS_CONNECTING; 1192d021c344SAndy King err = -EINPROGRESS; 1193d021c344SAndy King } 1194d021c344SAndy King 1195d021c344SAndy King /* The receive path will handle all communication until we are able to 1196d021c344SAndy King * enter the connected state. Here we wait for the connection to be 1197d021c344SAndy King * completed or a notification of an error. 1198d021c344SAndy King */ 1199d021c344SAndy King timeout = vsk->connect_timeout; 1200d021c344SAndy King prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); 1201d021c344SAndy King 1202d021c344SAndy King while (sk->sk_state != SS_CONNECTED && sk->sk_err == 0) { 1203d021c344SAndy King if (flags & O_NONBLOCK) { 1204d021c344SAndy King /* If we're not going to block, we schedule a timeout 1205d021c344SAndy King * function to generate a timeout on the connection 1206d021c344SAndy King * attempt, in case the peer doesn't respond in a 1207d021c344SAndy King * timely manner. We hold on to the socket until the 1208d021c344SAndy King * timeout fires. 1209d021c344SAndy King */ 1210d021c344SAndy King sock_hold(sk); 1211d021c344SAndy King INIT_DELAYED_WORK(&vsk->dwork, 1212d021c344SAndy King vsock_connect_timeout); 1213d021c344SAndy King schedule_delayed_work(&vsk->dwork, timeout); 1214d021c344SAndy King 1215d021c344SAndy King /* Skip ahead to preserve error code set above. */ 1216d021c344SAndy King goto out_wait; 1217d021c344SAndy King } 1218d021c344SAndy King 1219d021c344SAndy King release_sock(sk); 1220d021c344SAndy King timeout = schedule_timeout(timeout); 1221d021c344SAndy King lock_sock(sk); 1222d021c344SAndy King 1223d021c344SAndy King if (signal_pending(current)) { 1224d021c344SAndy King err = sock_intr_errno(timeout); 1225f7f9b5e7SClaudio Imbrenda sk->sk_state = SS_UNCONNECTED; 1226f7f9b5e7SClaudio Imbrenda sock->state = SS_UNCONNECTED; 1227f7f9b5e7SClaudio Imbrenda goto out_wait; 1228d021c344SAndy King } else if (timeout == 0) { 1229d021c344SAndy King err = -ETIMEDOUT; 1230f7f9b5e7SClaudio Imbrenda sk->sk_state = SS_UNCONNECTED; 1231f7f9b5e7SClaudio Imbrenda sock->state = SS_UNCONNECTED; 1232f7f9b5e7SClaudio Imbrenda goto out_wait; 1233d021c344SAndy King } 1234d021c344SAndy King 1235d021c344SAndy King prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); 1236d021c344SAndy King } 1237d021c344SAndy King 1238d021c344SAndy King if (sk->sk_err) { 1239d021c344SAndy King err = -sk->sk_err; 1240f7f9b5e7SClaudio Imbrenda sk->sk_state = SS_UNCONNECTED; 1241f7f9b5e7SClaudio Imbrenda sock->state = SS_UNCONNECTED; 1242f7f9b5e7SClaudio Imbrenda } else { 1243d021c344SAndy King err = 0; 1244f7f9b5e7SClaudio Imbrenda } 1245d021c344SAndy King 1246d021c344SAndy King out_wait: 1247d021c344SAndy King finish_wait(sk_sleep(sk), &wait); 1248d021c344SAndy King out: 1249d021c344SAndy King release_sock(sk); 1250d021c344SAndy King return err; 1251d021c344SAndy King } 1252d021c344SAndy King 1253d021c344SAndy King static int vsock_accept(struct socket *sock, struct socket *newsock, int flags) 1254d021c344SAndy King { 1255d021c344SAndy King struct sock *listener; 1256d021c344SAndy King int err; 1257d021c344SAndy King struct sock *connected; 1258d021c344SAndy King struct vsock_sock *vconnected; 1259d021c344SAndy King long timeout; 1260d021c344SAndy King DEFINE_WAIT(wait); 1261d021c344SAndy King 1262d021c344SAndy King err = 0; 1263d021c344SAndy King listener = sock->sk; 1264d021c344SAndy King 1265d021c344SAndy King lock_sock(listener); 1266d021c344SAndy King 1267d021c344SAndy King if (sock->type != SOCK_STREAM) { 1268d021c344SAndy King err = -EOPNOTSUPP; 1269d021c344SAndy King goto out; 1270d021c344SAndy King } 1271d021c344SAndy King 1272ea3803c1SStefan Hajnoczi if (listener->sk_state != VSOCK_SS_LISTEN) { 1273d021c344SAndy King err = -EINVAL; 1274d021c344SAndy King goto out; 1275d021c344SAndy King } 1276d021c344SAndy King 1277d021c344SAndy King /* Wait for children sockets to appear; these are the new sockets 1278d021c344SAndy King * created upon connection establishment. 1279d021c344SAndy King */ 1280d021c344SAndy King timeout = sock_sndtimeo(listener, flags & O_NONBLOCK); 1281d021c344SAndy King prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE); 1282d021c344SAndy King 1283d021c344SAndy King while ((connected = vsock_dequeue_accept(listener)) == NULL && 1284d021c344SAndy King listener->sk_err == 0) { 1285d021c344SAndy King release_sock(listener); 1286d021c344SAndy King timeout = schedule_timeout(timeout); 1287f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(listener), &wait); 1288d021c344SAndy King lock_sock(listener); 1289d021c344SAndy King 1290d021c344SAndy King if (signal_pending(current)) { 1291d021c344SAndy King err = sock_intr_errno(timeout); 1292f7f9b5e7SClaudio Imbrenda goto out; 1293d021c344SAndy King } else if (timeout == 0) { 1294d021c344SAndy King err = -EAGAIN; 1295f7f9b5e7SClaudio Imbrenda goto out; 1296d021c344SAndy King } 1297d021c344SAndy King 1298d021c344SAndy King prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE); 1299d021c344SAndy King } 1300f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(listener), &wait); 1301d021c344SAndy King 1302d021c344SAndy King if (listener->sk_err) 1303d021c344SAndy King err = -listener->sk_err; 1304d021c344SAndy King 1305d021c344SAndy King if (connected) { 1306d021c344SAndy King listener->sk_ack_backlog--; 1307d021c344SAndy King 13084192f672SStefan Hajnoczi lock_sock_nested(connected, SINGLE_DEPTH_NESTING); 1309d021c344SAndy King vconnected = vsock_sk(connected); 1310d021c344SAndy King 1311d021c344SAndy King /* If the listener socket has received an error, then we should 1312d021c344SAndy King * reject this socket and return. Note that we simply mark the 1313d021c344SAndy King * socket rejected, drop our reference, and let the cleanup 1314d021c344SAndy King * function handle the cleanup; the fact that we found it in 1315d021c344SAndy King * the listener's accept queue guarantees that the cleanup 1316d021c344SAndy King * function hasn't run yet. 1317d021c344SAndy King */ 1318d021c344SAndy King if (err) { 1319d021c344SAndy King vconnected->rejected = true; 1320f7f9b5e7SClaudio Imbrenda } else { 1321d021c344SAndy King newsock->state = SS_CONNECTED; 1322d021c344SAndy King sock_graft(connected, newsock); 1323f7f9b5e7SClaudio Imbrenda } 1324f7f9b5e7SClaudio Imbrenda 1325d021c344SAndy King release_sock(connected); 1326d021c344SAndy King sock_put(connected); 1327d021c344SAndy King } 1328d021c344SAndy King 1329d021c344SAndy King out: 1330d021c344SAndy King release_sock(listener); 1331d021c344SAndy King return err; 1332d021c344SAndy King } 1333d021c344SAndy King 1334d021c344SAndy King static int vsock_listen(struct socket *sock, int backlog) 1335d021c344SAndy King { 1336d021c344SAndy King int err; 1337d021c344SAndy King struct sock *sk; 1338d021c344SAndy King struct vsock_sock *vsk; 1339d021c344SAndy King 1340d021c344SAndy King sk = sock->sk; 1341d021c344SAndy King 1342d021c344SAndy King lock_sock(sk); 1343d021c344SAndy King 1344d021c344SAndy King if (sock->type != SOCK_STREAM) { 1345d021c344SAndy King err = -EOPNOTSUPP; 1346d021c344SAndy King goto out; 1347d021c344SAndy King } 1348d021c344SAndy King 1349d021c344SAndy King if (sock->state != SS_UNCONNECTED) { 1350d021c344SAndy King err = -EINVAL; 1351d021c344SAndy King goto out; 1352d021c344SAndy King } 1353d021c344SAndy King 1354d021c344SAndy King vsk = vsock_sk(sk); 1355d021c344SAndy King 1356d021c344SAndy King if (!vsock_addr_bound(&vsk->local_addr)) { 1357d021c344SAndy King err = -EINVAL; 1358d021c344SAndy King goto out; 1359d021c344SAndy King } 1360d021c344SAndy King 1361d021c344SAndy King sk->sk_max_ack_backlog = backlog; 1362ea3803c1SStefan Hajnoczi sk->sk_state = VSOCK_SS_LISTEN; 1363d021c344SAndy King 1364d021c344SAndy King err = 0; 1365d021c344SAndy King 1366d021c344SAndy King out: 1367d021c344SAndy King release_sock(sk); 1368d021c344SAndy King return err; 1369d021c344SAndy King } 1370d021c344SAndy King 1371d021c344SAndy King static int vsock_stream_setsockopt(struct socket *sock, 1372d021c344SAndy King int level, 1373d021c344SAndy King int optname, 1374d021c344SAndy King char __user *optval, 1375d021c344SAndy King unsigned int optlen) 1376d021c344SAndy King { 1377d021c344SAndy King int err; 1378d021c344SAndy King struct sock *sk; 1379d021c344SAndy King struct vsock_sock *vsk; 1380d021c344SAndy King u64 val; 1381d021c344SAndy King 1382d021c344SAndy King if (level != AF_VSOCK) 1383d021c344SAndy King return -ENOPROTOOPT; 1384d021c344SAndy King 1385d021c344SAndy King #define COPY_IN(_v) \ 1386d021c344SAndy King do { \ 1387d021c344SAndy King if (optlen < sizeof(_v)) { \ 1388d021c344SAndy King err = -EINVAL; \ 1389d021c344SAndy King goto exit; \ 1390d021c344SAndy King } \ 1391d021c344SAndy King if (copy_from_user(&_v, optval, sizeof(_v)) != 0) { \ 1392d021c344SAndy King err = -EFAULT; \ 1393d021c344SAndy King goto exit; \ 1394d021c344SAndy King } \ 1395d021c344SAndy King } while (0) 1396d021c344SAndy King 1397d021c344SAndy King err = 0; 1398d021c344SAndy King sk = sock->sk; 1399d021c344SAndy King vsk = vsock_sk(sk); 1400d021c344SAndy King 1401d021c344SAndy King lock_sock(sk); 1402d021c344SAndy King 1403d021c344SAndy King switch (optname) { 1404d021c344SAndy King case SO_VM_SOCKETS_BUFFER_SIZE: 1405d021c344SAndy King COPY_IN(val); 1406d021c344SAndy King transport->set_buffer_size(vsk, val); 1407d021c344SAndy King break; 1408d021c344SAndy King 1409d021c344SAndy King case SO_VM_SOCKETS_BUFFER_MAX_SIZE: 1410d021c344SAndy King COPY_IN(val); 1411d021c344SAndy King transport->set_max_buffer_size(vsk, val); 1412d021c344SAndy King break; 1413d021c344SAndy King 1414d021c344SAndy King case SO_VM_SOCKETS_BUFFER_MIN_SIZE: 1415d021c344SAndy King COPY_IN(val); 1416d021c344SAndy King transport->set_min_buffer_size(vsk, val); 1417d021c344SAndy King break; 1418d021c344SAndy King 1419d021c344SAndy King case SO_VM_SOCKETS_CONNECT_TIMEOUT: { 1420d021c344SAndy King struct timeval tv; 1421d021c344SAndy King COPY_IN(tv); 1422d021c344SAndy King if (tv.tv_sec >= 0 && tv.tv_usec < USEC_PER_SEC && 1423d021c344SAndy King tv.tv_sec < (MAX_SCHEDULE_TIMEOUT / HZ - 1)) { 1424d021c344SAndy King vsk->connect_timeout = tv.tv_sec * HZ + 1425d021c344SAndy King DIV_ROUND_UP(tv.tv_usec, (1000000 / HZ)); 1426d021c344SAndy King if (vsk->connect_timeout == 0) 1427d021c344SAndy King vsk->connect_timeout = 1428d021c344SAndy King VSOCK_DEFAULT_CONNECT_TIMEOUT; 1429d021c344SAndy King 1430d021c344SAndy King } else { 1431d021c344SAndy King err = -ERANGE; 1432d021c344SAndy King } 1433d021c344SAndy King break; 1434d021c344SAndy King } 1435d021c344SAndy King 1436d021c344SAndy King default: 1437d021c344SAndy King err = -ENOPROTOOPT; 1438d021c344SAndy King break; 1439d021c344SAndy King } 1440d021c344SAndy King 1441d021c344SAndy King #undef COPY_IN 1442d021c344SAndy King 1443d021c344SAndy King exit: 1444d021c344SAndy King release_sock(sk); 1445d021c344SAndy King return err; 1446d021c344SAndy King } 1447d021c344SAndy King 1448d021c344SAndy King static int vsock_stream_getsockopt(struct socket *sock, 1449d021c344SAndy King int level, int optname, 1450d021c344SAndy King char __user *optval, 1451d021c344SAndy King int __user *optlen) 1452d021c344SAndy King { 1453d021c344SAndy King int err; 1454d021c344SAndy King int len; 1455d021c344SAndy King struct sock *sk; 1456d021c344SAndy King struct vsock_sock *vsk; 1457d021c344SAndy King u64 val; 1458d021c344SAndy King 1459d021c344SAndy King if (level != AF_VSOCK) 1460d021c344SAndy King return -ENOPROTOOPT; 1461d021c344SAndy King 1462d021c344SAndy King err = get_user(len, optlen); 1463d021c344SAndy King if (err != 0) 1464d021c344SAndy King return err; 1465d021c344SAndy King 1466d021c344SAndy King #define COPY_OUT(_v) \ 1467d021c344SAndy King do { \ 1468d021c344SAndy King if (len < sizeof(_v)) \ 1469d021c344SAndy King return -EINVAL; \ 1470d021c344SAndy King \ 1471d021c344SAndy King len = sizeof(_v); \ 1472d021c344SAndy King if (copy_to_user(optval, &_v, len) != 0) \ 1473d021c344SAndy King return -EFAULT; \ 1474d021c344SAndy King \ 1475d021c344SAndy King } while (0) 1476d021c344SAndy King 1477d021c344SAndy King err = 0; 1478d021c344SAndy King sk = sock->sk; 1479d021c344SAndy King vsk = vsock_sk(sk); 1480d021c344SAndy King 1481d021c344SAndy King switch (optname) { 1482d021c344SAndy King case SO_VM_SOCKETS_BUFFER_SIZE: 1483d021c344SAndy King val = transport->get_buffer_size(vsk); 1484d021c344SAndy King COPY_OUT(val); 1485d021c344SAndy King break; 1486d021c344SAndy King 1487d021c344SAndy King case SO_VM_SOCKETS_BUFFER_MAX_SIZE: 1488d021c344SAndy King val = transport->get_max_buffer_size(vsk); 1489d021c344SAndy King COPY_OUT(val); 1490d021c344SAndy King break; 1491d021c344SAndy King 1492d021c344SAndy King case SO_VM_SOCKETS_BUFFER_MIN_SIZE: 1493d021c344SAndy King val = transport->get_min_buffer_size(vsk); 1494d021c344SAndy King COPY_OUT(val); 1495d021c344SAndy King break; 1496d021c344SAndy King 1497d021c344SAndy King case SO_VM_SOCKETS_CONNECT_TIMEOUT: { 1498d021c344SAndy King struct timeval tv; 1499d021c344SAndy King tv.tv_sec = vsk->connect_timeout / HZ; 1500d021c344SAndy King tv.tv_usec = 1501d021c344SAndy King (vsk->connect_timeout - 1502d021c344SAndy King tv.tv_sec * HZ) * (1000000 / HZ); 1503d021c344SAndy King COPY_OUT(tv); 1504d021c344SAndy King break; 1505d021c344SAndy King } 1506d021c344SAndy King default: 1507d021c344SAndy King return -ENOPROTOOPT; 1508d021c344SAndy King } 1509d021c344SAndy King 1510d021c344SAndy King err = put_user(len, optlen); 1511d021c344SAndy King if (err != 0) 1512d021c344SAndy King return -EFAULT; 1513d021c344SAndy King 1514d021c344SAndy King #undef COPY_OUT 1515d021c344SAndy King 1516d021c344SAndy King return 0; 1517d021c344SAndy King } 1518d021c344SAndy King 15191b784140SYing Xue static int vsock_stream_sendmsg(struct socket *sock, struct msghdr *msg, 15201b784140SYing Xue size_t len) 1521d021c344SAndy King { 1522d021c344SAndy King struct sock *sk; 1523d021c344SAndy King struct vsock_sock *vsk; 1524d021c344SAndy King ssize_t total_written; 1525d021c344SAndy King long timeout; 1526d021c344SAndy King int err; 1527d021c344SAndy King struct vsock_transport_send_notify_data send_data; 1528d021c344SAndy King 1529d021c344SAndy King DEFINE_WAIT(wait); 1530d021c344SAndy King 1531d021c344SAndy King sk = sock->sk; 1532d021c344SAndy King vsk = vsock_sk(sk); 1533d021c344SAndy King total_written = 0; 1534d021c344SAndy King err = 0; 1535d021c344SAndy King 1536d021c344SAndy King if (msg->msg_flags & MSG_OOB) 1537d021c344SAndy King return -EOPNOTSUPP; 1538d021c344SAndy King 1539d021c344SAndy King lock_sock(sk); 1540d021c344SAndy King 1541d021c344SAndy King /* Callers should not provide a destination with stream sockets. */ 1542d021c344SAndy King if (msg->msg_namelen) { 1543d021c344SAndy King err = sk->sk_state == SS_CONNECTED ? -EISCONN : -EOPNOTSUPP; 1544d021c344SAndy King goto out; 1545d021c344SAndy King } 1546d021c344SAndy King 1547d021c344SAndy King /* Send data only if both sides are not shutdown in the direction. */ 1548d021c344SAndy King if (sk->sk_shutdown & SEND_SHUTDOWN || 1549d021c344SAndy King vsk->peer_shutdown & RCV_SHUTDOWN) { 1550d021c344SAndy King err = -EPIPE; 1551d021c344SAndy King goto out; 1552d021c344SAndy King } 1553d021c344SAndy King 1554d021c344SAndy King if (sk->sk_state != SS_CONNECTED || 1555d021c344SAndy King !vsock_addr_bound(&vsk->local_addr)) { 1556d021c344SAndy King err = -ENOTCONN; 1557d021c344SAndy King goto out; 1558d021c344SAndy King } 1559d021c344SAndy King 1560d021c344SAndy King if (!vsock_addr_bound(&vsk->remote_addr)) { 1561d021c344SAndy King err = -EDESTADDRREQ; 1562d021c344SAndy King goto out; 1563d021c344SAndy King } 1564d021c344SAndy King 1565d021c344SAndy King /* Wait for room in the produce queue to enqueue our user's data. */ 1566d021c344SAndy King timeout = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT); 1567d021c344SAndy King 1568d021c344SAndy King err = transport->notify_send_init(vsk, &send_data); 1569d021c344SAndy King if (err < 0) 1570d021c344SAndy King goto out; 1571d021c344SAndy King 15726f57e56aSClaudio Imbrenda 1573d021c344SAndy King while (total_written < len) { 1574d021c344SAndy King ssize_t written; 1575d021c344SAndy King 1576f7f9b5e7SClaudio Imbrenda prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); 1577d021c344SAndy King while (vsock_stream_has_space(vsk) == 0 && 1578d021c344SAndy King sk->sk_err == 0 && 1579d021c344SAndy King !(sk->sk_shutdown & SEND_SHUTDOWN) && 1580d021c344SAndy King !(vsk->peer_shutdown & RCV_SHUTDOWN)) { 1581d021c344SAndy King 1582d021c344SAndy King /* Don't wait for non-blocking sockets. */ 1583d021c344SAndy King if (timeout == 0) { 1584d021c344SAndy King err = -EAGAIN; 1585f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(sk), &wait); 1586f7f9b5e7SClaudio Imbrenda goto out_err; 1587d021c344SAndy King } 1588d021c344SAndy King 1589d021c344SAndy King err = transport->notify_send_pre_block(vsk, &send_data); 1590f7f9b5e7SClaudio Imbrenda if (err < 0) { 1591f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(sk), &wait); 1592f7f9b5e7SClaudio Imbrenda goto out_err; 1593f7f9b5e7SClaudio Imbrenda } 1594d021c344SAndy King 1595d021c344SAndy King release_sock(sk); 1596d021c344SAndy King timeout = schedule_timeout(timeout); 1597d021c344SAndy King lock_sock(sk); 1598d021c344SAndy King if (signal_pending(current)) { 1599d021c344SAndy King err = sock_intr_errno(timeout); 1600f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(sk), &wait); 1601f7f9b5e7SClaudio Imbrenda goto out_err; 1602d021c344SAndy King } else if (timeout == 0) { 1603d021c344SAndy King err = -EAGAIN; 1604f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(sk), &wait); 1605f7f9b5e7SClaudio Imbrenda goto out_err; 1606d021c344SAndy King } 1607d021c344SAndy King 16086f57e56aSClaudio Imbrenda prepare_to_wait(sk_sleep(sk), &wait, 16096f57e56aSClaudio Imbrenda TASK_INTERRUPTIBLE); 1610d021c344SAndy King } 1611f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(sk), &wait); 1612d021c344SAndy King 1613d021c344SAndy King /* These checks occur both as part of and after the loop 1614d021c344SAndy King * conditional since we need to check before and after 1615d021c344SAndy King * sleeping. 1616d021c344SAndy King */ 1617d021c344SAndy King if (sk->sk_err) { 1618d021c344SAndy King err = -sk->sk_err; 1619f7f9b5e7SClaudio Imbrenda goto out_err; 1620d021c344SAndy King } else if ((sk->sk_shutdown & SEND_SHUTDOWN) || 1621d021c344SAndy King (vsk->peer_shutdown & RCV_SHUTDOWN)) { 1622d021c344SAndy King err = -EPIPE; 1623f7f9b5e7SClaudio Imbrenda goto out_err; 1624d021c344SAndy King } 1625d021c344SAndy King 1626d021c344SAndy King err = transport->notify_send_pre_enqueue(vsk, &send_data); 1627d021c344SAndy King if (err < 0) 1628f7f9b5e7SClaudio Imbrenda goto out_err; 1629d021c344SAndy King 1630d021c344SAndy King /* Note that enqueue will only write as many bytes as are free 1631d021c344SAndy King * in the produce queue, so we don't need to ensure len is 1632d021c344SAndy King * smaller than the queue size. It is the caller's 1633d021c344SAndy King * responsibility to check how many bytes we were able to send. 1634d021c344SAndy King */ 1635d021c344SAndy King 1636d021c344SAndy King written = transport->stream_enqueue( 16370f7db23aSAl Viro vsk, msg, 1638d021c344SAndy King len - total_written); 1639d021c344SAndy King if (written < 0) { 1640d021c344SAndy King err = -ENOMEM; 1641f7f9b5e7SClaudio Imbrenda goto out_err; 1642d021c344SAndy King } 1643d021c344SAndy King 1644d021c344SAndy King total_written += written; 1645d021c344SAndy King 1646d021c344SAndy King err = transport->notify_send_post_enqueue( 1647d021c344SAndy King vsk, written, &send_data); 1648d021c344SAndy King if (err < 0) 1649f7f9b5e7SClaudio Imbrenda goto out_err; 1650d021c344SAndy King 1651d021c344SAndy King } 1652d021c344SAndy King 1653f7f9b5e7SClaudio Imbrenda out_err: 1654d021c344SAndy King if (total_written > 0) 1655d021c344SAndy King err = total_written; 1656d021c344SAndy King out: 1657d021c344SAndy King release_sock(sk); 1658d021c344SAndy King return err; 1659d021c344SAndy King } 1660d021c344SAndy King 1661d021c344SAndy King 1662d021c344SAndy King static int 16631b784140SYing Xue vsock_stream_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, 16641b784140SYing Xue int flags) 1665d021c344SAndy King { 1666d021c344SAndy King struct sock *sk; 1667d021c344SAndy King struct vsock_sock *vsk; 1668d021c344SAndy King int err; 1669d021c344SAndy King size_t target; 1670d021c344SAndy King ssize_t copied; 1671d021c344SAndy King long timeout; 1672d021c344SAndy King struct vsock_transport_recv_notify_data recv_data; 1673d021c344SAndy King 1674d021c344SAndy King DEFINE_WAIT(wait); 1675d021c344SAndy King 1676d021c344SAndy King sk = sock->sk; 1677d021c344SAndy King vsk = vsock_sk(sk); 1678d021c344SAndy King err = 0; 1679d021c344SAndy King 1680d021c344SAndy King lock_sock(sk); 1681d021c344SAndy King 1682d021c344SAndy King if (sk->sk_state != SS_CONNECTED) { 1683d021c344SAndy King /* Recvmsg is supposed to return 0 if a peer performs an 1684d021c344SAndy King * orderly shutdown. Differentiate between that case and when a 1685d021c344SAndy King * peer has not connected or a local shutdown occured with the 1686d021c344SAndy King * SOCK_DONE flag. 1687d021c344SAndy King */ 1688d021c344SAndy King if (sock_flag(sk, SOCK_DONE)) 1689d021c344SAndy King err = 0; 1690d021c344SAndy King else 1691d021c344SAndy King err = -ENOTCONN; 1692d021c344SAndy King 1693d021c344SAndy King goto out; 1694d021c344SAndy King } 1695d021c344SAndy King 1696d021c344SAndy King if (flags & MSG_OOB) { 1697d021c344SAndy King err = -EOPNOTSUPP; 1698d021c344SAndy King goto out; 1699d021c344SAndy King } 1700d021c344SAndy King 1701d021c344SAndy King /* We don't check peer_shutdown flag here since peer may actually shut 1702d021c344SAndy King * down, but there can be data in the queue that a local socket can 1703d021c344SAndy King * receive. 1704d021c344SAndy King */ 1705d021c344SAndy King if (sk->sk_shutdown & RCV_SHUTDOWN) { 1706d021c344SAndy King err = 0; 1707d021c344SAndy King goto out; 1708d021c344SAndy King } 1709d021c344SAndy King 1710d021c344SAndy King /* It is valid on Linux to pass in a zero-length receive buffer. This 1711d021c344SAndy King * is not an error. We may as well bail out now. 1712d021c344SAndy King */ 1713d021c344SAndy King if (!len) { 1714d021c344SAndy King err = 0; 1715d021c344SAndy King goto out; 1716d021c344SAndy King } 1717d021c344SAndy King 1718d021c344SAndy King /* We must not copy less than target bytes into the user's buffer 1719d021c344SAndy King * before returning successfully, so we wait for the consume queue to 1720d021c344SAndy King * have that much data to consume before dequeueing. Note that this 1721d021c344SAndy King * makes it impossible to handle cases where target is greater than the 1722d021c344SAndy King * queue size. 1723d021c344SAndy King */ 1724d021c344SAndy King target = sock_rcvlowat(sk, flags & MSG_WAITALL, len); 1725d021c344SAndy King if (target >= transport->stream_rcvhiwat(vsk)) { 1726d021c344SAndy King err = -ENOMEM; 1727d021c344SAndy King goto out; 1728d021c344SAndy King } 1729d021c344SAndy King timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT); 1730d021c344SAndy King copied = 0; 1731d021c344SAndy King 1732d021c344SAndy King err = transport->notify_recv_init(vsk, target, &recv_data); 1733d021c344SAndy King if (err < 0) 1734d021c344SAndy King goto out; 1735d021c344SAndy King 1736d021c344SAndy King 1737d021c344SAndy King while (1) { 1738f7f9b5e7SClaudio Imbrenda s64 ready; 1739f7f9b5e7SClaudio Imbrenda 1740f7f9b5e7SClaudio Imbrenda prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); 1741f7f9b5e7SClaudio Imbrenda ready = vsock_stream_has_data(vsk); 1742f7f9b5e7SClaudio Imbrenda 1743f7f9b5e7SClaudio Imbrenda if (ready == 0) { 1744f7f9b5e7SClaudio Imbrenda if (sk->sk_err != 0 || 1745f7f9b5e7SClaudio Imbrenda (sk->sk_shutdown & RCV_SHUTDOWN) || 1746f7f9b5e7SClaudio Imbrenda (vsk->peer_shutdown & SEND_SHUTDOWN)) { 1747f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(sk), &wait); 1748f7f9b5e7SClaudio Imbrenda break; 1749f7f9b5e7SClaudio Imbrenda } 1750f7f9b5e7SClaudio Imbrenda /* Don't wait for non-blocking sockets. */ 1751f7f9b5e7SClaudio Imbrenda if (timeout == 0) { 1752f7f9b5e7SClaudio Imbrenda err = -EAGAIN; 1753f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(sk), &wait); 1754f7f9b5e7SClaudio Imbrenda break; 1755f7f9b5e7SClaudio Imbrenda } 1756f7f9b5e7SClaudio Imbrenda 1757f7f9b5e7SClaudio Imbrenda err = transport->notify_recv_pre_block( 1758f7f9b5e7SClaudio Imbrenda vsk, target, &recv_data); 1759f7f9b5e7SClaudio Imbrenda if (err < 0) { 1760f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(sk), &wait); 1761f7f9b5e7SClaudio Imbrenda break; 1762f7f9b5e7SClaudio Imbrenda } 1763f7f9b5e7SClaudio Imbrenda release_sock(sk); 1764f7f9b5e7SClaudio Imbrenda timeout = schedule_timeout(timeout); 1765f7f9b5e7SClaudio Imbrenda lock_sock(sk); 1766f7f9b5e7SClaudio Imbrenda 1767f7f9b5e7SClaudio Imbrenda if (signal_pending(current)) { 1768f7f9b5e7SClaudio Imbrenda err = sock_intr_errno(timeout); 1769f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(sk), &wait); 1770f7f9b5e7SClaudio Imbrenda break; 1771f7f9b5e7SClaudio Imbrenda } else if (timeout == 0) { 1772f7f9b5e7SClaudio Imbrenda err = -EAGAIN; 1773f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(sk), &wait); 1774f7f9b5e7SClaudio Imbrenda break; 1775f7f9b5e7SClaudio Imbrenda } 1776f7f9b5e7SClaudio Imbrenda } else { 1777f7f9b5e7SClaudio Imbrenda ssize_t read; 1778f7f9b5e7SClaudio Imbrenda 1779f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(sk), &wait); 1780d021c344SAndy King 1781d021c344SAndy King if (ready < 0) { 1782f7f9b5e7SClaudio Imbrenda /* Invalid queue pair content. XXX This should 1783f7f9b5e7SClaudio Imbrenda * be changed to a connection reset in a later 1784f7f9b5e7SClaudio Imbrenda * change. 1785d021c344SAndy King */ 1786d021c344SAndy King 1787d021c344SAndy King err = -ENOMEM; 1788f7f9b5e7SClaudio Imbrenda goto out; 1789f7f9b5e7SClaudio Imbrenda } 1790d021c344SAndy King 1791d021c344SAndy King err = transport->notify_recv_pre_dequeue( 1792d021c344SAndy King vsk, target, &recv_data); 1793d021c344SAndy King if (err < 0) 1794d021c344SAndy King break; 1795d021c344SAndy King 1796d021c344SAndy King read = transport->stream_dequeue( 17970f7db23aSAl Viro vsk, msg, 1798d021c344SAndy King len - copied, flags); 1799d021c344SAndy King if (read < 0) { 1800d021c344SAndy King err = -ENOMEM; 1801d021c344SAndy King break; 1802d021c344SAndy King } 1803d021c344SAndy King 1804d021c344SAndy King copied += read; 1805d021c344SAndy King 1806d021c344SAndy King err = transport->notify_recv_post_dequeue( 1807d021c344SAndy King vsk, target, read, 1808d021c344SAndy King !(flags & MSG_PEEK), &recv_data); 1809d021c344SAndy King if (err < 0) 1810f7f9b5e7SClaudio Imbrenda goto out; 1811d021c344SAndy King 1812d021c344SAndy King if (read >= target || flags & MSG_PEEK) 1813d021c344SAndy King break; 1814d021c344SAndy King 1815d021c344SAndy King target -= read; 1816d021c344SAndy King } 1817d021c344SAndy King } 1818d021c344SAndy King 1819d021c344SAndy King if (sk->sk_err) 1820d021c344SAndy King err = -sk->sk_err; 1821d021c344SAndy King else if (sk->sk_shutdown & RCV_SHUTDOWN) 1822d021c344SAndy King err = 0; 1823d021c344SAndy King 1824dedc58e0SIan Campbell if (copied > 0) 1825d021c344SAndy King err = copied; 1826d021c344SAndy King 1827d021c344SAndy King out: 1828d021c344SAndy King release_sock(sk); 1829d021c344SAndy King return err; 1830d021c344SAndy King } 1831d021c344SAndy King 1832d021c344SAndy King static const struct proto_ops vsock_stream_ops = { 1833d021c344SAndy King .family = PF_VSOCK, 1834d021c344SAndy King .owner = THIS_MODULE, 1835d021c344SAndy King .release = vsock_release, 1836d021c344SAndy King .bind = vsock_bind, 1837d021c344SAndy King .connect = vsock_stream_connect, 1838d021c344SAndy King .socketpair = sock_no_socketpair, 1839d021c344SAndy King .accept = vsock_accept, 1840d021c344SAndy King .getname = vsock_getname, 1841d021c344SAndy King .poll = vsock_poll, 1842d021c344SAndy King .ioctl = sock_no_ioctl, 1843d021c344SAndy King .listen = vsock_listen, 1844d021c344SAndy King .shutdown = vsock_shutdown, 1845d021c344SAndy King .setsockopt = vsock_stream_setsockopt, 1846d021c344SAndy King .getsockopt = vsock_stream_getsockopt, 1847d021c344SAndy King .sendmsg = vsock_stream_sendmsg, 1848d021c344SAndy King .recvmsg = vsock_stream_recvmsg, 1849d021c344SAndy King .mmap = sock_no_mmap, 1850d021c344SAndy King .sendpage = sock_no_sendpage, 1851d021c344SAndy King }; 1852d021c344SAndy King 1853d021c344SAndy King static int vsock_create(struct net *net, struct socket *sock, 1854d021c344SAndy King int protocol, int kern) 1855d021c344SAndy King { 1856d021c344SAndy King if (!sock) 1857d021c344SAndy King return -EINVAL; 1858d021c344SAndy King 18596cf1c5fcSAndy King if (protocol && protocol != PF_VSOCK) 1860d021c344SAndy King return -EPROTONOSUPPORT; 1861d021c344SAndy King 1862d021c344SAndy King switch (sock->type) { 1863d021c344SAndy King case SOCK_DGRAM: 1864d021c344SAndy King sock->ops = &vsock_dgram_ops; 1865d021c344SAndy King break; 1866d021c344SAndy King case SOCK_STREAM: 1867d021c344SAndy King sock->ops = &vsock_stream_ops; 1868d021c344SAndy King break; 1869d021c344SAndy King default: 1870d021c344SAndy King return -ESOCKTNOSUPPORT; 1871d021c344SAndy King } 1872d021c344SAndy King 1873d021c344SAndy King sock->state = SS_UNCONNECTED; 1874d021c344SAndy King 187511aa9c28SEric W. Biederman return __vsock_create(net, sock, NULL, GFP_KERNEL, 0, kern) ? 0 : -ENOMEM; 1876d021c344SAndy King } 1877d021c344SAndy King 1878d021c344SAndy King static const struct net_proto_family vsock_family_ops = { 1879d021c344SAndy King .family = AF_VSOCK, 1880d021c344SAndy King .create = vsock_create, 1881d021c344SAndy King .owner = THIS_MODULE, 1882d021c344SAndy King }; 1883d021c344SAndy King 1884d021c344SAndy King static long vsock_dev_do_ioctl(struct file *filp, 1885d021c344SAndy King unsigned int cmd, void __user *ptr) 1886d021c344SAndy King { 1887d021c344SAndy King u32 __user *p = ptr; 1888d021c344SAndy King int retval = 0; 1889d021c344SAndy King 1890d021c344SAndy King switch (cmd) { 1891d021c344SAndy King case IOCTL_VM_SOCKETS_GET_LOCAL_CID: 1892d021c344SAndy King if (put_user(transport->get_local_cid(), p) != 0) 1893d021c344SAndy King retval = -EFAULT; 1894d021c344SAndy King break; 1895d021c344SAndy King 1896d021c344SAndy King default: 1897d021c344SAndy King pr_err("Unknown ioctl %d\n", cmd); 1898d021c344SAndy King retval = -EINVAL; 1899d021c344SAndy King } 1900d021c344SAndy King 1901d021c344SAndy King return retval; 1902d021c344SAndy King } 1903d021c344SAndy King 1904d021c344SAndy King static long vsock_dev_ioctl(struct file *filp, 1905d021c344SAndy King unsigned int cmd, unsigned long arg) 1906d021c344SAndy King { 1907d021c344SAndy King return vsock_dev_do_ioctl(filp, cmd, (void __user *)arg); 1908d021c344SAndy King } 1909d021c344SAndy King 1910d021c344SAndy King #ifdef CONFIG_COMPAT 1911d021c344SAndy King static long vsock_dev_compat_ioctl(struct file *filp, 1912d021c344SAndy King unsigned int cmd, unsigned long arg) 1913d021c344SAndy King { 1914d021c344SAndy King return vsock_dev_do_ioctl(filp, cmd, compat_ptr(arg)); 1915d021c344SAndy King } 1916d021c344SAndy King #endif 1917d021c344SAndy King 1918d021c344SAndy King static const struct file_operations vsock_device_ops = { 1919d021c344SAndy King .owner = THIS_MODULE, 1920d021c344SAndy King .unlocked_ioctl = vsock_dev_ioctl, 1921d021c344SAndy King #ifdef CONFIG_COMPAT 1922d021c344SAndy King .compat_ioctl = vsock_dev_compat_ioctl, 1923d021c344SAndy King #endif 1924d021c344SAndy King .open = nonseekable_open, 1925d021c344SAndy King }; 1926d021c344SAndy King 1927d021c344SAndy King static struct miscdevice vsock_device = { 1928d021c344SAndy King .name = "vsock", 1929d021c344SAndy King .fops = &vsock_device_ops, 1930d021c344SAndy King }; 1931d021c344SAndy King 19322c4a336eSAndy King int __vsock_core_init(const struct vsock_transport *t, struct module *owner) 1933d021c344SAndy King { 19342c4a336eSAndy King int err = mutex_lock_interruptible(&vsock_register_mutex); 19352c4a336eSAndy King 19362c4a336eSAndy King if (err) 19372c4a336eSAndy King return err; 19382c4a336eSAndy King 19392c4a336eSAndy King if (transport) { 19402c4a336eSAndy King err = -EBUSY; 19412c4a336eSAndy King goto err_busy; 19422c4a336eSAndy King } 19432c4a336eSAndy King 19442c4a336eSAndy King /* Transport must be the owner of the protocol so that it can't 19452c4a336eSAndy King * unload while there are open sockets. 19462c4a336eSAndy King */ 19472c4a336eSAndy King vsock_proto.owner = owner; 19482c4a336eSAndy King transport = t; 1949d021c344SAndy King 1950d021c344SAndy King vsock_init_tables(); 1951d021c344SAndy King 19526ad0b2f7SAsias He vsock_device.minor = MISC_DYNAMIC_MINOR; 1953d021c344SAndy King err = misc_register(&vsock_device); 1954d021c344SAndy King if (err) { 1955d021c344SAndy King pr_err("Failed to register misc device\n"); 1956f6a835bbSGao feng goto err_reset_transport; 1957d021c344SAndy King } 1958d021c344SAndy King 1959d021c344SAndy King err = proto_register(&vsock_proto, 1); /* we want our slab */ 1960d021c344SAndy King if (err) { 1961d021c344SAndy King pr_err("Cannot register vsock protocol\n"); 1962f6a835bbSGao feng goto err_deregister_misc; 1963d021c344SAndy King } 1964d021c344SAndy King 1965d021c344SAndy King err = sock_register(&vsock_family_ops); 1966d021c344SAndy King if (err) { 1967d021c344SAndy King pr_err("could not register af_vsock (%d) address family: %d\n", 1968d021c344SAndy King AF_VSOCK, err); 1969d021c344SAndy King goto err_unregister_proto; 1970d021c344SAndy King } 1971d021c344SAndy King 19722c4a336eSAndy King mutex_unlock(&vsock_register_mutex); 1973d021c344SAndy King return 0; 1974d021c344SAndy King 1975d021c344SAndy King err_unregister_proto: 1976d021c344SAndy King proto_unregister(&vsock_proto); 1977f6a835bbSGao feng err_deregister_misc: 1978d021c344SAndy King misc_deregister(&vsock_device); 1979f6a835bbSGao feng err_reset_transport: 19802c4a336eSAndy King transport = NULL; 19812c4a336eSAndy King err_busy: 19822c4a336eSAndy King mutex_unlock(&vsock_register_mutex); 1983d021c344SAndy King return err; 1984d021c344SAndy King } 19852c4a336eSAndy King EXPORT_SYMBOL_GPL(__vsock_core_init); 1986d021c344SAndy King 1987d021c344SAndy King void vsock_core_exit(void) 1988d021c344SAndy King { 1989d021c344SAndy King mutex_lock(&vsock_register_mutex); 1990d021c344SAndy King 1991d021c344SAndy King misc_deregister(&vsock_device); 1992d021c344SAndy King sock_unregister(AF_VSOCK); 1993d021c344SAndy King proto_unregister(&vsock_proto); 1994d021c344SAndy King 1995d021c344SAndy King /* We do not want the assignment below re-ordered. */ 1996d021c344SAndy King mb(); 1997d021c344SAndy King transport = NULL; 1998d021c344SAndy King 1999d021c344SAndy King mutex_unlock(&vsock_register_mutex); 2000d021c344SAndy King } 2001d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_core_exit); 2002d021c344SAndy King 20030b01aeb3SStefan Hajnoczi const struct vsock_transport *vsock_core_get_transport(void) 20040b01aeb3SStefan Hajnoczi { 20050b01aeb3SStefan Hajnoczi /* vsock_register_mutex not taken since only the transport uses this 20060b01aeb3SStefan Hajnoczi * function and only while registered. 20070b01aeb3SStefan Hajnoczi */ 20080b01aeb3SStefan Hajnoczi return transport; 20090b01aeb3SStefan Hajnoczi } 20100b01aeb3SStefan Hajnoczi EXPORT_SYMBOL_GPL(vsock_core_get_transport); 20110b01aeb3SStefan Hajnoczi 2012d021c344SAndy King MODULE_AUTHOR("VMware, Inc."); 2013d021c344SAndy King MODULE_DESCRIPTION("VMware Virtual Socket Family"); 20141190cfdbSJorgen Hansen MODULE_VERSION("1.0.2.0-k"); 2015d021c344SAndy King MODULE_LICENSE("GPL v2"); 2016