1685a6bf8SThomas Gleixner // SPDX-License-Identifier: GPL-2.0-only 2d021c344SAndy King /* 3d021c344SAndy King * VMware vSockets Driver 4d021c344SAndy King * 5d021c344SAndy King * Copyright (C) 2007-2013 VMware, Inc. All rights reserved. 6d021c344SAndy King */ 7d021c344SAndy King 8d021c344SAndy King /* Implementation notes: 9d021c344SAndy King * 10d021c344SAndy King * - There are two kinds of sockets: those created by user action (such as 11d021c344SAndy King * calling socket(2)) and those created by incoming connection request packets. 12d021c344SAndy King * 13d021c344SAndy King * - There are two "global" tables, one for bound sockets (sockets that have 14d021c344SAndy King * specified an address that they are responsible for) and one for connected 15d021c344SAndy King * sockets (sockets that have established a connection with another socket). 16d021c344SAndy King * These tables are "global" in that all sockets on the system are placed 17d021c344SAndy King * within them. - Note, though, that the bound table contains an extra entry 18d021c344SAndy King * for a list of unbound sockets and SOCK_DGRAM sockets will always remain in 19d021c344SAndy King * that list. The bound table is used solely for lookup of sockets when packets 20d021c344SAndy King * are received and that's not necessary for SOCK_DGRAM sockets since we create 21d021c344SAndy King * a datagram handle for each and need not perform a lookup. Keeping SOCK_DGRAM 22d021c344SAndy King * sockets out of the bound hash buckets will reduce the chance of collisions 23d021c344SAndy King * when looking for SOCK_STREAM sockets and prevents us from having to check the 24d021c344SAndy King * socket type in the hash table lookups. 25d021c344SAndy King * 26d021c344SAndy King * - Sockets created by user action will either be "client" sockets that 27d021c344SAndy King * initiate a connection or "server" sockets that listen for connections; we do 28d021c344SAndy King * not support simultaneous connects (two "client" sockets connecting). 29d021c344SAndy King * 30d021c344SAndy King * - "Server" sockets are referred to as listener sockets throughout this 313b4477d2SStefan Hajnoczi * implementation because they are in the TCP_LISTEN state. When a 32ea3803c1SStefan Hajnoczi * connection request is received (the second kind of socket mentioned above), 33ea3803c1SStefan Hajnoczi * we create a new socket and refer to it as a pending socket. These pending 34ea3803c1SStefan Hajnoczi * sockets are placed on the pending connection list of the listener socket. 35ea3803c1SStefan Hajnoczi * When future packets are received for the address the listener socket is 36ea3803c1SStefan Hajnoczi * bound to, we check if the source of the packet is from one that has an 37ea3803c1SStefan Hajnoczi * existing pending connection. If it does, we process the packet for the 38ea3803c1SStefan Hajnoczi * pending socket. When that socket reaches the connected state, it is removed 39ea3803c1SStefan Hajnoczi * from the listener socket's pending list and enqueued in the listener 40ea3803c1SStefan Hajnoczi * socket's accept queue. Callers of accept(2) will accept connected sockets 41ea3803c1SStefan Hajnoczi * from the listener socket's accept queue. If the socket cannot be accepted 42ea3803c1SStefan Hajnoczi * for some reason then it is marked rejected. Once the connection is 43ea3803c1SStefan Hajnoczi * accepted, it is owned by the user process and the responsibility for cleanup 44ea3803c1SStefan Hajnoczi * falls with that user process. 45d021c344SAndy King * 46d021c344SAndy King * - It is possible that these pending sockets will never reach the connected 47d021c344SAndy King * state; in fact, we may never receive another packet after the connection 48d021c344SAndy King * request. Because of this, we must schedule a cleanup function to run in the 49d021c344SAndy King * future, after some amount of time passes where a connection should have been 50d021c344SAndy King * established. This function ensures that the socket is off all lists so it 51d021c344SAndy King * cannot be retrieved, then drops all references to the socket so it is cleaned 52d021c344SAndy King * up (sock_put() -> sk_free() -> our sk_destruct implementation). Note this 53d021c344SAndy King * function will also cleanup rejected sockets, those that reach the connected 54d021c344SAndy King * state but leave it before they have been accepted. 55d021c344SAndy King * 564192f672SStefan Hajnoczi * - Lock ordering for pending or accept queue sockets is: 574192f672SStefan Hajnoczi * 584192f672SStefan Hajnoczi * lock_sock(listener); 594192f672SStefan Hajnoczi * lock_sock_nested(pending, SINGLE_DEPTH_NESTING); 604192f672SStefan Hajnoczi * 614192f672SStefan Hajnoczi * Using explicit nested locking keeps lockdep happy since normally only one 624192f672SStefan Hajnoczi * lock of a given class may be taken at a time. 634192f672SStefan Hajnoczi * 64d021c344SAndy King * - Sockets created by user action will be cleaned up when the user process 65d021c344SAndy King * calls close(2), causing our release implementation to be called. Our release 66d021c344SAndy King * implementation will perform some cleanup then drop the last reference so our 67d021c344SAndy King * sk_destruct implementation is invoked. Our sk_destruct implementation will 68d021c344SAndy King * perform additional cleanup that's common for both types of sockets. 69d021c344SAndy King * 70d021c344SAndy King * - A socket's reference count is what ensures that the structure won't be 71d021c344SAndy King * freed. Each entry in a list (such as the "global" bound and connected tables 72d021c344SAndy King * and the listener socket's pending list and connected queue) ensures a 73d021c344SAndy King * reference. When we defer work until process context and pass a socket as our 74d021c344SAndy King * argument, we must ensure the reference count is increased to ensure the 75d021c344SAndy King * socket isn't freed before the function is run; the deferred function will 76d021c344SAndy King * then drop the reference. 773b4477d2SStefan Hajnoczi * 783b4477d2SStefan Hajnoczi * - sk->sk_state uses the TCP state constants because they are widely used by 793b4477d2SStefan Hajnoczi * other address families and exposed to userspace tools like ss(8): 803b4477d2SStefan Hajnoczi * 813b4477d2SStefan Hajnoczi * TCP_CLOSE - unconnected 823b4477d2SStefan Hajnoczi * TCP_SYN_SENT - connecting 833b4477d2SStefan Hajnoczi * TCP_ESTABLISHED - connected 843b4477d2SStefan Hajnoczi * TCP_CLOSING - disconnecting 853b4477d2SStefan Hajnoczi * TCP_LISTEN - listening 86d021c344SAndy King */ 87d021c344SAndy King 88b6459415SJakub Kicinski #include <linux/compat.h> 89d021c344SAndy King #include <linux/types.h> 90d021c344SAndy King #include <linux/bitops.h> 91d021c344SAndy King #include <linux/cred.h> 92d021c344SAndy King #include <linux/init.h> 93d021c344SAndy King #include <linux/io.h> 94d021c344SAndy King #include <linux/kernel.h> 95174cd4b1SIngo Molnar #include <linux/sched/signal.h> 96d021c344SAndy King #include <linux/kmod.h> 97d021c344SAndy King #include <linux/list.h> 98d021c344SAndy King #include <linux/miscdevice.h> 99d021c344SAndy King #include <linux/module.h> 100d021c344SAndy King #include <linux/mutex.h> 101d021c344SAndy King #include <linux/net.h> 102d021c344SAndy King #include <linux/poll.h> 1038236b08cSLepton Wu #include <linux/random.h> 104d021c344SAndy King #include <linux/skbuff.h> 105d021c344SAndy King #include <linux/smp.h> 106d021c344SAndy King #include <linux/socket.h> 107d021c344SAndy King #include <linux/stddef.h> 108d021c344SAndy King #include <linux/unistd.h> 109d021c344SAndy King #include <linux/wait.h> 110d021c344SAndy King #include <linux/workqueue.h> 111d021c344SAndy King #include <net/sock.h> 11282a54d0eSAsias He #include <net/af_vsock.h> 113d021c344SAndy King 114d021c344SAndy King static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr); 115d021c344SAndy King static void vsock_sk_destruct(struct sock *sk); 116d021c344SAndy King static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb); 117d021c344SAndy King 118d021c344SAndy King /* Protocol family. */ 119d021c344SAndy King static struct proto vsock_proto = { 120d021c344SAndy King .name = "AF_VSOCK", 121d021c344SAndy King .owner = THIS_MODULE, 122d021c344SAndy King .obj_size = sizeof(struct vsock_sock), 123d021c344SAndy King }; 124d021c344SAndy King 125d021c344SAndy King /* The default peer timeout indicates how long we will wait for a peer response 126d021c344SAndy King * to a control message. 127d021c344SAndy King */ 128d021c344SAndy King #define VSOCK_DEFAULT_CONNECT_TIMEOUT (2 * HZ) 129d021c344SAndy King 130b9f2b0ffSStefano Garzarella #define VSOCK_DEFAULT_BUFFER_SIZE (1024 * 256) 131b9f2b0ffSStefano Garzarella #define VSOCK_DEFAULT_BUFFER_MAX_SIZE (1024 * 256) 132b9f2b0ffSStefano Garzarella #define VSOCK_DEFAULT_BUFFER_MIN_SIZE 128 133b9f2b0ffSStefano Garzarella 134c0cfa2d8SStefano Garzarella /* Transport used for host->guest communication */ 135c0cfa2d8SStefano Garzarella static const struct vsock_transport *transport_h2g; 136c0cfa2d8SStefano Garzarella /* Transport used for guest->host communication */ 137c0cfa2d8SStefano Garzarella static const struct vsock_transport *transport_g2h; 138c0cfa2d8SStefano Garzarella /* Transport used for DGRAM communication */ 139c0cfa2d8SStefano Garzarella static const struct vsock_transport *transport_dgram; 1400e121905SStefano Garzarella /* Transport used for local communication */ 1410e121905SStefano Garzarella static const struct vsock_transport *transport_local; 142d021c344SAndy King static DEFINE_MUTEX(vsock_register_mutex); 143d021c344SAndy King 144d021c344SAndy King /**** UTILS ****/ 145d021c344SAndy King 146d021c344SAndy King /* Each bound VSocket is stored in the bind hash table and each connected 147d021c344SAndy King * VSocket is stored in the connected hash table. 148d021c344SAndy King * 149d021c344SAndy King * Unbound sockets are all put on the same list attached to the end of the hash 150d021c344SAndy King * table (vsock_unbound_sockets). Bound sockets are added to the hash table in 151d021c344SAndy King * the bucket that their local address hashes to (vsock_bound_sockets(addr) 152d021c344SAndy King * represents the list that addr hashes to). 153d021c344SAndy King * 154d021c344SAndy King * Specifically, we initialize the vsock_bind_table array to a size of 155d021c344SAndy King * VSOCK_HASH_SIZE + 1 so that vsock_bind_table[0] through 156d021c344SAndy King * vsock_bind_table[VSOCK_HASH_SIZE - 1] are for bound sockets and 157d021c344SAndy King * vsock_bind_table[VSOCK_HASH_SIZE] is for unbound sockets. The hash function 158a49dd9dcSAsias He * mods with VSOCK_HASH_SIZE to ensure this. 159d021c344SAndy King */ 160d021c344SAndy King #define MAX_PORT_RETRIES 24 161d021c344SAndy King 162a49dd9dcSAsias He #define VSOCK_HASH(addr) ((addr)->svm_port % VSOCK_HASH_SIZE) 163d021c344SAndy King #define vsock_bound_sockets(addr) (&vsock_bind_table[VSOCK_HASH(addr)]) 164d021c344SAndy King #define vsock_unbound_sockets (&vsock_bind_table[VSOCK_HASH_SIZE]) 165d021c344SAndy King 166d021c344SAndy King /* XXX This can probably be implemented in a better way. */ 167d021c344SAndy King #define VSOCK_CONN_HASH(src, dst) \ 168a49dd9dcSAsias He (((src)->svm_cid ^ (dst)->svm_port) % VSOCK_HASH_SIZE) 169d021c344SAndy King #define vsock_connected_sockets(src, dst) \ 170d021c344SAndy King (&vsock_connected_table[VSOCK_CONN_HASH(src, dst)]) 171d021c344SAndy King #define vsock_connected_sockets_vsk(vsk) \ 172d021c344SAndy King vsock_connected_sockets(&(vsk)->remote_addr, &(vsk)->local_addr) 173d021c344SAndy King 17444f20980SStefan Hajnoczi struct list_head vsock_bind_table[VSOCK_HASH_SIZE + 1]; 17544f20980SStefan Hajnoczi EXPORT_SYMBOL_GPL(vsock_bind_table); 17644f20980SStefan Hajnoczi struct list_head vsock_connected_table[VSOCK_HASH_SIZE]; 17744f20980SStefan Hajnoczi EXPORT_SYMBOL_GPL(vsock_connected_table); 17844f20980SStefan Hajnoczi DEFINE_SPINLOCK(vsock_table_lock); 17944f20980SStefan Hajnoczi EXPORT_SYMBOL_GPL(vsock_table_lock); 180d021c344SAndy King 181b3a6dfe8SAsias He /* Autobind this socket to the local address if necessary. */ 182b3a6dfe8SAsias He static int vsock_auto_bind(struct vsock_sock *vsk) 183b3a6dfe8SAsias He { 184b3a6dfe8SAsias He struct sock *sk = sk_vsock(vsk); 185b3a6dfe8SAsias He struct sockaddr_vm local_addr; 186b3a6dfe8SAsias He 187b3a6dfe8SAsias He if (vsock_addr_bound(&vsk->local_addr)) 188b3a6dfe8SAsias He return 0; 189b3a6dfe8SAsias He vsock_addr_init(&local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 190b3a6dfe8SAsias He return __vsock_bind(sk, &local_addr); 191b3a6dfe8SAsias He } 192b3a6dfe8SAsias He 193c0cfa2d8SStefano Garzarella static void vsock_init_tables(void) 194d021c344SAndy King { 195d021c344SAndy King int i; 196d021c344SAndy King 197d021c344SAndy King for (i = 0; i < ARRAY_SIZE(vsock_bind_table); i++) 198d021c344SAndy King INIT_LIST_HEAD(&vsock_bind_table[i]); 199d021c344SAndy King 200d021c344SAndy King for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) 201d021c344SAndy King INIT_LIST_HEAD(&vsock_connected_table[i]); 202d021c344SAndy King } 203d021c344SAndy King 204d021c344SAndy King static void __vsock_insert_bound(struct list_head *list, 205d021c344SAndy King struct vsock_sock *vsk) 206d021c344SAndy King { 207d021c344SAndy King sock_hold(&vsk->sk); 208d021c344SAndy King list_add(&vsk->bound_table, list); 209d021c344SAndy King } 210d021c344SAndy King 211d021c344SAndy King static void __vsock_insert_connected(struct list_head *list, 212d021c344SAndy King struct vsock_sock *vsk) 213d021c344SAndy King { 214d021c344SAndy King sock_hold(&vsk->sk); 215d021c344SAndy King list_add(&vsk->connected_table, list); 216d021c344SAndy King } 217d021c344SAndy King 218d021c344SAndy King static void __vsock_remove_bound(struct vsock_sock *vsk) 219d021c344SAndy King { 220d021c344SAndy King list_del_init(&vsk->bound_table); 221d021c344SAndy King sock_put(&vsk->sk); 222d021c344SAndy King } 223d021c344SAndy King 224d021c344SAndy King static void __vsock_remove_connected(struct vsock_sock *vsk) 225d021c344SAndy King { 226d021c344SAndy King list_del_init(&vsk->connected_table); 227d021c344SAndy King sock_put(&vsk->sk); 228d021c344SAndy King } 229d021c344SAndy King 230d021c344SAndy King static struct sock *__vsock_find_bound_socket(struct sockaddr_vm *addr) 231d021c344SAndy King { 232d021c344SAndy King struct vsock_sock *vsk; 233d021c344SAndy King 23436c5b48bSStefano Garzarella list_for_each_entry(vsk, vsock_bound_sockets(addr), bound_table) { 23536c5b48bSStefano Garzarella if (vsock_addr_equals_addr(addr, &vsk->local_addr)) 236d021c344SAndy King return sk_vsock(vsk); 237d021c344SAndy King 23836c5b48bSStefano Garzarella if (addr->svm_port == vsk->local_addr.svm_port && 23936c5b48bSStefano Garzarella (vsk->local_addr.svm_cid == VMADDR_CID_ANY || 24036c5b48bSStefano Garzarella addr->svm_cid == VMADDR_CID_ANY)) 24136c5b48bSStefano Garzarella return sk_vsock(vsk); 24236c5b48bSStefano Garzarella } 24336c5b48bSStefano Garzarella 244d021c344SAndy King return NULL; 245d021c344SAndy King } 246d021c344SAndy King 247d021c344SAndy King static struct sock *__vsock_find_connected_socket(struct sockaddr_vm *src, 248d021c344SAndy King struct sockaddr_vm *dst) 249d021c344SAndy King { 250d021c344SAndy King struct vsock_sock *vsk; 251d021c344SAndy King 252d021c344SAndy King list_for_each_entry(vsk, vsock_connected_sockets(src, dst), 253d021c344SAndy King connected_table) { 254990454b5SReilly Grant if (vsock_addr_equals_addr(src, &vsk->remote_addr) && 255990454b5SReilly Grant dst->svm_port == vsk->local_addr.svm_port) { 256d021c344SAndy King return sk_vsock(vsk); 257d021c344SAndy King } 258d021c344SAndy King } 259d021c344SAndy King 260d021c344SAndy King return NULL; 261d021c344SAndy King } 262d021c344SAndy King 263d021c344SAndy King static void vsock_insert_unbound(struct vsock_sock *vsk) 264d021c344SAndy King { 265d021c344SAndy King spin_lock_bh(&vsock_table_lock); 266d021c344SAndy King __vsock_insert_bound(vsock_unbound_sockets, vsk); 267d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 268d021c344SAndy King } 269d021c344SAndy King 270d021c344SAndy King void vsock_insert_connected(struct vsock_sock *vsk) 271d021c344SAndy King { 272d021c344SAndy King struct list_head *list = vsock_connected_sockets( 273d021c344SAndy King &vsk->remote_addr, &vsk->local_addr); 274d021c344SAndy King 275d021c344SAndy King spin_lock_bh(&vsock_table_lock); 276d021c344SAndy King __vsock_insert_connected(list, vsk); 277d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 278d021c344SAndy King } 279d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_insert_connected); 280d021c344SAndy King 281d021c344SAndy King void vsock_remove_bound(struct vsock_sock *vsk) 282d021c344SAndy King { 283d021c344SAndy King spin_lock_bh(&vsock_table_lock); 284d5afa82cSSunil Muthuswamy if (__vsock_in_bound_table(vsk)) 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); 293d5afa82cSSunil Muthuswamy if (__vsock_in_connected_table(vsk)) 294d021c344SAndy King __vsock_remove_connected(vsk); 295d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 296d021c344SAndy King } 297d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_remove_connected); 298d021c344SAndy King 299d021c344SAndy King struct sock *vsock_find_bound_socket(struct sockaddr_vm *addr) 300d021c344SAndy King { 301d021c344SAndy King struct sock *sk; 302d021c344SAndy King 303d021c344SAndy King spin_lock_bh(&vsock_table_lock); 304d021c344SAndy King sk = __vsock_find_bound_socket(addr); 305d021c344SAndy King if (sk) 306d021c344SAndy King sock_hold(sk); 307d021c344SAndy King 308d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 309d021c344SAndy King 310d021c344SAndy King return sk; 311d021c344SAndy King } 312d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_find_bound_socket); 313d021c344SAndy King 314d021c344SAndy King struct sock *vsock_find_connected_socket(struct sockaddr_vm *src, 315d021c344SAndy King struct sockaddr_vm *dst) 316d021c344SAndy King { 317d021c344SAndy King struct sock *sk; 318d021c344SAndy King 319d021c344SAndy King spin_lock_bh(&vsock_table_lock); 320d021c344SAndy King sk = __vsock_find_connected_socket(src, dst); 321d021c344SAndy King if (sk) 322d021c344SAndy King sock_hold(sk); 323d021c344SAndy King 324d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 325d021c344SAndy King 326d021c344SAndy King return sk; 327d021c344SAndy King } 328d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_find_connected_socket); 329d021c344SAndy King 3306773b7dcSStefan Hajnoczi void vsock_remove_sock(struct vsock_sock *vsk) 3316773b7dcSStefan Hajnoczi { 3326773b7dcSStefan Hajnoczi vsock_remove_bound(vsk); 3336773b7dcSStefan Hajnoczi vsock_remove_connected(vsk); 3346773b7dcSStefan Hajnoczi } 3356773b7dcSStefan Hajnoczi EXPORT_SYMBOL_GPL(vsock_remove_sock); 3366773b7dcSStefan Hajnoczi 3378e6ed963SJiyong Park void vsock_for_each_connected_socket(struct vsock_transport *transport, 3388e6ed963SJiyong Park void (*fn)(struct sock *sk)) 339d021c344SAndy King { 340d021c344SAndy King int i; 341d021c344SAndy King 342d021c344SAndy King spin_lock_bh(&vsock_table_lock); 343d021c344SAndy King 344d021c344SAndy King for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) { 345d021c344SAndy King struct vsock_sock *vsk; 346d021c344SAndy King list_for_each_entry(vsk, &vsock_connected_table[i], 3478e6ed963SJiyong Park connected_table) { 3488e6ed963SJiyong Park if (vsk->transport != transport) 3498e6ed963SJiyong Park continue; 3508e6ed963SJiyong Park 351d021c344SAndy King fn(sk_vsock(vsk)); 352d021c344SAndy King } 3538e6ed963SJiyong Park } 354d021c344SAndy King 355d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 356d021c344SAndy King } 357d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_for_each_connected_socket); 358d021c344SAndy King 359d021c344SAndy King void vsock_add_pending(struct sock *listener, struct sock *pending) 360d021c344SAndy King { 361d021c344SAndy King struct vsock_sock *vlistener; 362d021c344SAndy King struct vsock_sock *vpending; 363d021c344SAndy King 364d021c344SAndy King vlistener = vsock_sk(listener); 365d021c344SAndy King vpending = vsock_sk(pending); 366d021c344SAndy King 367d021c344SAndy King sock_hold(pending); 368d021c344SAndy King sock_hold(listener); 369d021c344SAndy King list_add_tail(&vpending->pending_links, &vlistener->pending_links); 370d021c344SAndy King } 371d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_add_pending); 372d021c344SAndy King 373d021c344SAndy King void vsock_remove_pending(struct sock *listener, struct sock *pending) 374d021c344SAndy King { 375d021c344SAndy King struct vsock_sock *vpending = vsock_sk(pending); 376d021c344SAndy King 377d021c344SAndy King list_del_init(&vpending->pending_links); 378d021c344SAndy King sock_put(listener); 379d021c344SAndy King sock_put(pending); 380d021c344SAndy King } 381d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_remove_pending); 382d021c344SAndy King 383d021c344SAndy King void vsock_enqueue_accept(struct sock *listener, struct sock *connected) 384d021c344SAndy King { 385d021c344SAndy King struct vsock_sock *vlistener; 386d021c344SAndy King struct vsock_sock *vconnected; 387d021c344SAndy King 388d021c344SAndy King vlistener = vsock_sk(listener); 389d021c344SAndy King vconnected = vsock_sk(connected); 390d021c344SAndy King 391d021c344SAndy King sock_hold(connected); 392d021c344SAndy King sock_hold(listener); 393d021c344SAndy King list_add_tail(&vconnected->accept_queue, &vlistener->accept_queue); 394d021c344SAndy King } 395d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_enqueue_accept); 396d021c344SAndy King 397408624afSStefano Garzarella static bool vsock_use_local_transport(unsigned int remote_cid) 398408624afSStefano Garzarella { 399408624afSStefano Garzarella if (!transport_local) 400408624afSStefano Garzarella return false; 401408624afSStefano Garzarella 402408624afSStefano Garzarella if (remote_cid == VMADDR_CID_LOCAL) 403408624afSStefano Garzarella return true; 404408624afSStefano Garzarella 405408624afSStefano Garzarella if (transport_g2h) { 406408624afSStefano Garzarella return remote_cid == transport_g2h->get_local_cid(); 407408624afSStefano Garzarella } else { 408408624afSStefano Garzarella return remote_cid == VMADDR_CID_HOST; 409408624afSStefano Garzarella } 410408624afSStefano Garzarella } 411408624afSStefano Garzarella 4126a2c0962SStefano Garzarella static void vsock_deassign_transport(struct vsock_sock *vsk) 4136a2c0962SStefano Garzarella { 4146a2c0962SStefano Garzarella if (!vsk->transport) 4156a2c0962SStefano Garzarella return; 4166a2c0962SStefano Garzarella 4176a2c0962SStefano Garzarella vsk->transport->destruct(vsk); 4186a2c0962SStefano Garzarella module_put(vsk->transport->module); 4196a2c0962SStefano Garzarella vsk->transport = NULL; 4206a2c0962SStefano Garzarella } 4216a2c0962SStefano Garzarella 422c0cfa2d8SStefano Garzarella /* Assign a transport to a socket and call the .init transport callback. 423c0cfa2d8SStefano Garzarella * 4248cb48554SArseny Krasnov * Note: for connection oriented socket this must be called when vsk->remote_addr 4258cb48554SArseny Krasnov * is set (e.g. during the connect() or when a connection request on a listener 426c0cfa2d8SStefano Garzarella * socket is received). 427c0cfa2d8SStefano Garzarella * The vsk->remote_addr is used to decide which transport to use: 428408624afSStefano Garzarella * - remote CID == VMADDR_CID_LOCAL or g2h->local_cid or VMADDR_CID_HOST if 429408624afSStefano Garzarella * g2h is not loaded, will use local transport; 4307f816984SAndra Paraschiv * - remote CID <= VMADDR_CID_HOST or h2g is not loaded or remote flags field 4317f816984SAndra Paraschiv * includes VMADDR_FLAG_TO_HOST flag value, will use guest->host transport; 432c0cfa2d8SStefano Garzarella * - remote CID > VMADDR_CID_HOST will use host->guest transport; 433c0cfa2d8SStefano Garzarella */ 434c0cfa2d8SStefano Garzarella int vsock_assign_transport(struct vsock_sock *vsk, struct vsock_sock *psk) 435c0cfa2d8SStefano Garzarella { 436c0cfa2d8SStefano Garzarella const struct vsock_transport *new_transport; 437c0cfa2d8SStefano Garzarella struct sock *sk = sk_vsock(vsk); 438c0cfa2d8SStefano Garzarella unsigned int remote_cid = vsk->remote_addr.svm_cid; 4397f816984SAndra Paraschiv __u8 remote_flags; 440039fcccaSStefano Garzarella int ret; 441c0cfa2d8SStefano Garzarella 4421b5f2ab9SAndra Paraschiv /* If the packet is coming with the source and destination CIDs higher 4431b5f2ab9SAndra Paraschiv * than VMADDR_CID_HOST, then a vsock channel where all the packets are 4441b5f2ab9SAndra Paraschiv * forwarded to the host should be established. Then the host will 4451b5f2ab9SAndra Paraschiv * need to forward the packets to the guest. 4461b5f2ab9SAndra Paraschiv * 4471b5f2ab9SAndra Paraschiv * The flag is set on the (listen) receive path (psk is not NULL). On 4481b5f2ab9SAndra Paraschiv * the connect path the flag can be set by the user space application. 4491b5f2ab9SAndra Paraschiv */ 4501b5f2ab9SAndra Paraschiv if (psk && vsk->local_addr.svm_cid > VMADDR_CID_HOST && 4511b5f2ab9SAndra Paraschiv vsk->remote_addr.svm_cid > VMADDR_CID_HOST) 4521b5f2ab9SAndra Paraschiv vsk->remote_addr.svm_flags |= VMADDR_FLAG_TO_HOST; 4531b5f2ab9SAndra Paraschiv 4547f816984SAndra Paraschiv remote_flags = vsk->remote_addr.svm_flags; 4557f816984SAndra Paraschiv 456c0cfa2d8SStefano Garzarella switch (sk->sk_type) { 457c0cfa2d8SStefano Garzarella case SOCK_DGRAM: 458c0cfa2d8SStefano Garzarella new_transport = transport_dgram; 459c0cfa2d8SStefano Garzarella break; 460c0cfa2d8SStefano Garzarella case SOCK_STREAM: 4610798e78bSArseny Krasnov case SOCK_SEQPACKET: 462408624afSStefano Garzarella if (vsock_use_local_transport(remote_cid)) 463408624afSStefano Garzarella new_transport = transport_local; 4647f816984SAndra Paraschiv else if (remote_cid <= VMADDR_CID_HOST || !transport_h2g || 4657f816984SAndra Paraschiv (remote_flags & VMADDR_FLAG_TO_HOST)) 466c0cfa2d8SStefano Garzarella new_transport = transport_g2h; 467c0cfa2d8SStefano Garzarella else 468c0cfa2d8SStefano Garzarella new_transport = transport_h2g; 469c0cfa2d8SStefano Garzarella break; 470c0cfa2d8SStefano Garzarella default: 471c0cfa2d8SStefano Garzarella return -ESOCKTNOSUPPORT; 472c0cfa2d8SStefano Garzarella } 473c0cfa2d8SStefano Garzarella 474c0cfa2d8SStefano Garzarella if (vsk->transport) { 475c0cfa2d8SStefano Garzarella if (vsk->transport == new_transport) 476c0cfa2d8SStefano Garzarella return 0; 477c0cfa2d8SStefano Garzarella 4783f74957fSStefano Garzarella /* transport->release() must be called with sock lock acquired. 4798cb48554SArseny Krasnov * This path can only be taken during vsock_connect(), where we 4808cb48554SArseny Krasnov * have already held the sock lock. In the other cases, this 4818cb48554SArseny Krasnov * function is called on a new socket which is not assigned to 4828cb48554SArseny Krasnov * any transport. 4833f74957fSStefano Garzarella */ 484c0cfa2d8SStefano Garzarella vsk->transport->release(vsk); 4856a2c0962SStefano Garzarella vsock_deassign_transport(vsk); 486c0cfa2d8SStefano Garzarella } 487c0cfa2d8SStefano Garzarella 4886a2c0962SStefano Garzarella /* We increase the module refcnt to prevent the transport unloading 4896a2c0962SStefano Garzarella * while there are open sockets assigned to it. 4906a2c0962SStefano Garzarella */ 4916a2c0962SStefano Garzarella if (!new_transport || !try_module_get(new_transport->module)) 492c0cfa2d8SStefano Garzarella return -ENODEV; 493c0cfa2d8SStefano Garzarella 4940798e78bSArseny Krasnov if (sk->sk_type == SOCK_SEQPACKET) { 4950798e78bSArseny Krasnov if (!new_transport->seqpacket_allow || 4960798e78bSArseny Krasnov !new_transport->seqpacket_allow(remote_cid)) { 4970798e78bSArseny Krasnov module_put(new_transport->module); 4980798e78bSArseny Krasnov return -ESOCKTNOSUPPORT; 4990798e78bSArseny Krasnov } 5000798e78bSArseny Krasnov } 5010798e78bSArseny Krasnov 502039fcccaSStefano Garzarella ret = new_transport->init(vsk, psk); 503039fcccaSStefano Garzarella if (ret) { 504039fcccaSStefano Garzarella module_put(new_transport->module); 505039fcccaSStefano Garzarella return ret; 506039fcccaSStefano Garzarella } 507039fcccaSStefano Garzarella 508c0cfa2d8SStefano Garzarella vsk->transport = new_transport; 509c0cfa2d8SStefano Garzarella 510039fcccaSStefano Garzarella return 0; 511c0cfa2d8SStefano Garzarella } 512c0cfa2d8SStefano Garzarella EXPORT_SYMBOL_GPL(vsock_assign_transport); 513c0cfa2d8SStefano Garzarella 514c0cfa2d8SStefano Garzarella bool vsock_find_cid(unsigned int cid) 515c0cfa2d8SStefano Garzarella { 516c0cfa2d8SStefano Garzarella if (transport_g2h && cid == transport_g2h->get_local_cid()) 517c0cfa2d8SStefano Garzarella return true; 518c0cfa2d8SStefano Garzarella 519c0cfa2d8SStefano Garzarella if (transport_h2g && cid == VMADDR_CID_HOST) 520c0cfa2d8SStefano Garzarella return true; 521c0cfa2d8SStefano Garzarella 522408624afSStefano Garzarella if (transport_local && cid == VMADDR_CID_LOCAL) 523408624afSStefano Garzarella return true; 524408624afSStefano Garzarella 525c0cfa2d8SStefano Garzarella return false; 526c0cfa2d8SStefano Garzarella } 527c0cfa2d8SStefano Garzarella EXPORT_SYMBOL_GPL(vsock_find_cid); 528c0cfa2d8SStefano Garzarella 529d021c344SAndy King static struct sock *vsock_dequeue_accept(struct sock *listener) 530d021c344SAndy King { 531d021c344SAndy King struct vsock_sock *vlistener; 532d021c344SAndy King struct vsock_sock *vconnected; 533d021c344SAndy King 534d021c344SAndy King vlistener = vsock_sk(listener); 535d021c344SAndy King 536d021c344SAndy King if (list_empty(&vlistener->accept_queue)) 537d021c344SAndy King return NULL; 538d021c344SAndy King 539d021c344SAndy King vconnected = list_entry(vlistener->accept_queue.next, 540d021c344SAndy King struct vsock_sock, accept_queue); 541d021c344SAndy King 542d021c344SAndy King list_del_init(&vconnected->accept_queue); 543d021c344SAndy King sock_put(listener); 544d021c344SAndy King /* The caller will need a reference on the connected socket so we let 545d021c344SAndy King * it call sock_put(). 546d021c344SAndy King */ 547d021c344SAndy King 548d021c344SAndy King return sk_vsock(vconnected); 549d021c344SAndy King } 550d021c344SAndy King 551d021c344SAndy King static bool vsock_is_accept_queue_empty(struct sock *sk) 552d021c344SAndy King { 553d021c344SAndy King struct vsock_sock *vsk = vsock_sk(sk); 554d021c344SAndy King return list_empty(&vsk->accept_queue); 555d021c344SAndy King } 556d021c344SAndy King 557d021c344SAndy King static bool vsock_is_pending(struct sock *sk) 558d021c344SAndy King { 559d021c344SAndy King struct vsock_sock *vsk = vsock_sk(sk); 560d021c344SAndy King return !list_empty(&vsk->pending_links); 561d021c344SAndy King } 562d021c344SAndy King 563d021c344SAndy King static int vsock_send_shutdown(struct sock *sk, int mode) 564d021c344SAndy King { 565fe502c4aSStefano Garzarella struct vsock_sock *vsk = vsock_sk(sk); 566fe502c4aSStefano Garzarella 567c0cfa2d8SStefano Garzarella if (!vsk->transport) 568c0cfa2d8SStefano Garzarella return -ENODEV; 569c0cfa2d8SStefano Garzarella 570fe502c4aSStefano Garzarella return vsk->transport->shutdown(vsk, mode); 571d021c344SAndy King } 572d021c344SAndy King 573455f05ecSCong Wang static void vsock_pending_work(struct work_struct *work) 574d021c344SAndy King { 575d021c344SAndy King struct sock *sk; 576d021c344SAndy King struct sock *listener; 577d021c344SAndy King struct vsock_sock *vsk; 578d021c344SAndy King bool cleanup; 579d021c344SAndy King 580455f05ecSCong Wang vsk = container_of(work, struct vsock_sock, pending_work.work); 581d021c344SAndy King sk = sk_vsock(vsk); 582d021c344SAndy King listener = vsk->listener; 583d021c344SAndy King cleanup = true; 584d021c344SAndy King 585d021c344SAndy King lock_sock(listener); 5864192f672SStefan Hajnoczi lock_sock_nested(sk, SINGLE_DEPTH_NESTING); 587d021c344SAndy King 588d021c344SAndy King if (vsock_is_pending(sk)) { 589d021c344SAndy King vsock_remove_pending(listener, sk); 5901190cfdbSJorgen Hansen 5917976a11bSEric Dumazet sk_acceptq_removed(listener); 592d021c344SAndy King } else if (!vsk->rejected) { 593d021c344SAndy King /* We are not on the pending list and accept() did not reject 594d021c344SAndy King * us, so we must have been accepted by our user process. We 595d021c344SAndy King * just need to drop our references to the sockets and be on 596d021c344SAndy King * our way. 597d021c344SAndy King */ 598d021c344SAndy King cleanup = false; 599d021c344SAndy King goto out; 600d021c344SAndy King } 601d021c344SAndy King 602d021c344SAndy King /* We need to remove ourself from the global connected sockets list so 603d021c344SAndy King * incoming packets can't find this socket, and to reduce the reference 604d021c344SAndy King * count. 605d021c344SAndy King */ 606d021c344SAndy King vsock_remove_connected(vsk); 607d021c344SAndy King 6083b4477d2SStefan Hajnoczi sk->sk_state = TCP_CLOSE; 609d021c344SAndy King 610d021c344SAndy King out: 611d021c344SAndy King release_sock(sk); 612d021c344SAndy King release_sock(listener); 613d021c344SAndy King if (cleanup) 614d021c344SAndy King sock_put(sk); 615d021c344SAndy King 616d021c344SAndy King sock_put(sk); 617d021c344SAndy King sock_put(listener); 618d021c344SAndy King } 619d021c344SAndy King 620d021c344SAndy King /**** SOCKET OPERATIONS ****/ 621d021c344SAndy King 622a9e29e55SArseny Krasnov static int __vsock_bind_connectible(struct vsock_sock *vsk, 623d021c344SAndy King struct sockaddr_vm *addr) 624d021c344SAndy King { 625a22d3251SLepton Wu static u32 port; 626d021c344SAndy King struct sockaddr_vm new_addr; 627d021c344SAndy King 6288236b08cSLepton Wu if (!port) 629d247aabdSJason A. Donenfeld port = get_random_u32_above(LAST_RESERVED_PORT); 6308236b08cSLepton Wu 631d021c344SAndy King vsock_addr_init(&new_addr, addr->svm_cid, addr->svm_port); 632d021c344SAndy King 633d021c344SAndy King if (addr->svm_port == VMADDR_PORT_ANY) { 634d021c344SAndy King bool found = false; 635d021c344SAndy King unsigned int i; 636d021c344SAndy King 637d021c344SAndy King for (i = 0; i < MAX_PORT_RETRIES; i++) { 638d021c344SAndy King if (port <= LAST_RESERVED_PORT) 639d021c344SAndy King port = LAST_RESERVED_PORT + 1; 640d021c344SAndy King 641d021c344SAndy King new_addr.svm_port = port++; 642d021c344SAndy King 643d021c344SAndy King if (!__vsock_find_bound_socket(&new_addr)) { 644d021c344SAndy King found = true; 645d021c344SAndy King break; 646d021c344SAndy King } 647d021c344SAndy King } 648d021c344SAndy King 649d021c344SAndy King if (!found) 650d021c344SAndy King return -EADDRNOTAVAIL; 651d021c344SAndy King } else { 652d021c344SAndy King /* If port is in reserved range, ensure caller 653d021c344SAndy King * has necessary privileges. 654d021c344SAndy King */ 655d021c344SAndy King if (addr->svm_port <= LAST_RESERVED_PORT && 656d021c344SAndy King !capable(CAP_NET_BIND_SERVICE)) { 657d021c344SAndy King return -EACCES; 658d021c344SAndy King } 659d021c344SAndy King 660d021c344SAndy King if (__vsock_find_bound_socket(&new_addr)) 661d021c344SAndy King return -EADDRINUSE; 662d021c344SAndy King } 663d021c344SAndy King 664d021c344SAndy King vsock_addr_init(&vsk->local_addr, new_addr.svm_cid, new_addr.svm_port); 665d021c344SAndy King 6668cb48554SArseny Krasnov /* Remove connection oriented sockets from the unbound list and add them 6678cb48554SArseny Krasnov * to the hash table for easy lookup by its address. The unbound list 6688cb48554SArseny Krasnov * is simply an extra entry at the end of the hash table, a trick used 6698cb48554SArseny Krasnov * by AF_UNIX. 670d021c344SAndy King */ 671d021c344SAndy King __vsock_remove_bound(vsk); 672d021c344SAndy King __vsock_insert_bound(vsock_bound_sockets(&vsk->local_addr), vsk); 673d021c344SAndy King 674d021c344SAndy King return 0; 675d021c344SAndy King } 676d021c344SAndy King 677d021c344SAndy King static int __vsock_bind_dgram(struct vsock_sock *vsk, 678d021c344SAndy King struct sockaddr_vm *addr) 679d021c344SAndy King { 680fe502c4aSStefano Garzarella return vsk->transport->dgram_bind(vsk, addr); 681d021c344SAndy King } 682d021c344SAndy King 683d021c344SAndy King static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr) 684d021c344SAndy King { 685d021c344SAndy King struct vsock_sock *vsk = vsock_sk(sk); 686d021c344SAndy King int retval; 687d021c344SAndy King 688d021c344SAndy King /* First ensure this socket isn't already bound. */ 689d021c344SAndy King if (vsock_addr_bound(&vsk->local_addr)) 690d021c344SAndy King return -EINVAL; 691d021c344SAndy King 692d021c344SAndy King /* Now bind to the provided address or select appropriate values if 693d021c344SAndy King * none are provided (VMADDR_CID_ANY and VMADDR_PORT_ANY). Note that 694d021c344SAndy King * like AF_INET prevents binding to a non-local IP address (in most 695c0cfa2d8SStefano Garzarella * cases), we only allow binding to a local CID. 696d021c344SAndy King */ 697c0cfa2d8SStefano Garzarella if (addr->svm_cid != VMADDR_CID_ANY && !vsock_find_cid(addr->svm_cid)) 698d021c344SAndy King return -EADDRNOTAVAIL; 699d021c344SAndy King 700d021c344SAndy King switch (sk->sk_socket->type) { 701d021c344SAndy King case SOCK_STREAM: 7020798e78bSArseny Krasnov case SOCK_SEQPACKET: 703d021c344SAndy King spin_lock_bh(&vsock_table_lock); 704a9e29e55SArseny Krasnov retval = __vsock_bind_connectible(vsk, addr); 705d021c344SAndy King spin_unlock_bh(&vsock_table_lock); 706d021c344SAndy King break; 707d021c344SAndy King 708d021c344SAndy King case SOCK_DGRAM: 709d021c344SAndy King retval = __vsock_bind_dgram(vsk, addr); 710d021c344SAndy King break; 711d021c344SAndy King 712d021c344SAndy King default: 713d021c344SAndy King retval = -EINVAL; 714d021c344SAndy King break; 715d021c344SAndy King } 716d021c344SAndy King 717d021c344SAndy King return retval; 718d021c344SAndy King } 719d021c344SAndy King 720455f05ecSCong Wang static void vsock_connect_timeout(struct work_struct *work); 721455f05ecSCong Wang 722b9ca2f5fSStefano Garzarella static struct sock *__vsock_create(struct net *net, 723d021c344SAndy King struct socket *sock, 724d021c344SAndy King struct sock *parent, 725d021c344SAndy King gfp_t priority, 72611aa9c28SEric W. Biederman unsigned short type, 72711aa9c28SEric W. Biederman int kern) 728d021c344SAndy King { 729d021c344SAndy King struct sock *sk; 730d021c344SAndy King struct vsock_sock *psk; 731d021c344SAndy King struct vsock_sock *vsk; 732d021c344SAndy King 73311aa9c28SEric W. Biederman sk = sk_alloc(net, AF_VSOCK, priority, &vsock_proto, kern); 734d021c344SAndy King if (!sk) 735d021c344SAndy King return NULL; 736d021c344SAndy King 737d021c344SAndy King sock_init_data(sock, sk); 738d021c344SAndy King 739d021c344SAndy King /* sk->sk_type is normally set in sock_init_data, but only if sock is 740d021c344SAndy King * non-NULL. We make sure that our sockets always have a type by 741d021c344SAndy King * setting it here if needed. 742d021c344SAndy King */ 743d021c344SAndy King if (!sock) 744d021c344SAndy King sk->sk_type = type; 745d021c344SAndy King 746d021c344SAndy King vsk = vsock_sk(sk); 747d021c344SAndy King vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 748d021c344SAndy King vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 749d021c344SAndy King 750d021c344SAndy King sk->sk_destruct = vsock_sk_destruct; 751d021c344SAndy King sk->sk_backlog_rcv = vsock_queue_rcv_skb; 752d021c344SAndy King sock_reset_flag(sk, SOCK_DONE); 753d021c344SAndy King 754d021c344SAndy King INIT_LIST_HEAD(&vsk->bound_table); 755d021c344SAndy King INIT_LIST_HEAD(&vsk->connected_table); 756d021c344SAndy King vsk->listener = NULL; 757d021c344SAndy King INIT_LIST_HEAD(&vsk->pending_links); 758d021c344SAndy King INIT_LIST_HEAD(&vsk->accept_queue); 759d021c344SAndy King vsk->rejected = false; 760d021c344SAndy King vsk->sent_request = false; 761d021c344SAndy King vsk->ignore_connecting_rst = false; 762d021c344SAndy King vsk->peer_shutdown = 0; 763455f05ecSCong Wang INIT_DELAYED_WORK(&vsk->connect_work, vsock_connect_timeout); 764455f05ecSCong Wang INIT_DELAYED_WORK(&vsk->pending_work, vsock_pending_work); 765d021c344SAndy King 766d021c344SAndy King psk = parent ? vsock_sk(parent) : NULL; 767d021c344SAndy King if (parent) { 768d021c344SAndy King vsk->trusted = psk->trusted; 769d021c344SAndy King vsk->owner = get_cred(psk->owner); 770d021c344SAndy King vsk->connect_timeout = psk->connect_timeout; 771b9f2b0ffSStefano Garzarella vsk->buffer_size = psk->buffer_size; 772b9f2b0ffSStefano Garzarella vsk->buffer_min_size = psk->buffer_min_size; 773b9f2b0ffSStefano Garzarella vsk->buffer_max_size = psk->buffer_max_size; 7741f935e8eSDavid Brazdil security_sk_clone(parent, sk); 775d021c344SAndy King } else { 776af545bb5SJeff Vander Stoep vsk->trusted = ns_capable_noaudit(&init_user_ns, CAP_NET_ADMIN); 777d021c344SAndy King vsk->owner = get_current_cred(); 778d021c344SAndy King vsk->connect_timeout = VSOCK_DEFAULT_CONNECT_TIMEOUT; 779b9f2b0ffSStefano Garzarella vsk->buffer_size = VSOCK_DEFAULT_BUFFER_SIZE; 780b9f2b0ffSStefano Garzarella vsk->buffer_min_size = VSOCK_DEFAULT_BUFFER_MIN_SIZE; 781b9f2b0ffSStefano Garzarella vsk->buffer_max_size = VSOCK_DEFAULT_BUFFER_MAX_SIZE; 782d021c344SAndy King } 783d021c344SAndy King 784d021c344SAndy King return sk; 785d021c344SAndy King } 786d021c344SAndy King 787a9e29e55SArseny Krasnov static bool sock_type_connectible(u16 type) 788a9e29e55SArseny Krasnov { 7890798e78bSArseny Krasnov return (type == SOCK_STREAM) || (type == SOCK_SEQPACKET); 790a9e29e55SArseny Krasnov } 791a9e29e55SArseny Krasnov 7920d9138ffSDexuan Cui static void __vsock_release(struct sock *sk, int level) 793d021c344SAndy King { 794d021c344SAndy King if (sk) { 795d021c344SAndy King struct sock *pending; 796d021c344SAndy King struct vsock_sock *vsk; 797d021c344SAndy King 798d021c344SAndy King vsk = vsock_sk(sk); 799d021c344SAndy King pending = NULL; /* Compiler warning. */ 800d021c344SAndy King 8010d9138ffSDexuan Cui /* When "level" is SINGLE_DEPTH_NESTING, use the nested 8020d9138ffSDexuan Cui * version to avoid the warning "possible recursive locking 8030d9138ffSDexuan Cui * detected". When "level" is 0, lock_sock_nested(sk, level) 8040d9138ffSDexuan Cui * is the same as lock_sock(sk). 8050d9138ffSDexuan Cui */ 8060d9138ffSDexuan Cui lock_sock_nested(sk, level); 8073f74957fSStefano Garzarella 8083f74957fSStefano Garzarella if (vsk->transport) 8093f74957fSStefano Garzarella vsk->transport->release(vsk); 810a9e29e55SArseny Krasnov else if (sock_type_connectible(sk->sk_type)) 8113f74957fSStefano Garzarella vsock_remove_sock(vsk); 8123f74957fSStefano Garzarella 813d021c344SAndy King sock_orphan(sk); 814d021c344SAndy King sk->sk_shutdown = SHUTDOWN_MASK; 815d021c344SAndy King 8163b7ad08bSChristophe JAILLET skb_queue_purge(&sk->sk_receive_queue); 817d021c344SAndy King 818d021c344SAndy King /* Clean up any sockets that never were accepted. */ 819d021c344SAndy King while ((pending = vsock_dequeue_accept(sk)) != NULL) { 8200d9138ffSDexuan Cui __vsock_release(pending, SINGLE_DEPTH_NESTING); 821d021c344SAndy King sock_put(pending); 822d021c344SAndy King } 823d021c344SAndy King 824d021c344SAndy King release_sock(sk); 825d021c344SAndy King sock_put(sk); 826d021c344SAndy King } 827d021c344SAndy King } 828d021c344SAndy King 829d021c344SAndy King static void vsock_sk_destruct(struct sock *sk) 830d021c344SAndy King { 831d021c344SAndy King struct vsock_sock *vsk = vsock_sk(sk); 832d021c344SAndy King 8336a2c0962SStefano Garzarella vsock_deassign_transport(vsk); 834d021c344SAndy King 835d021c344SAndy King /* When clearing these addresses, there's no need to set the family and 836d021c344SAndy King * possibly register the address family with the kernel. 837d021c344SAndy King */ 838d021c344SAndy King vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 839d021c344SAndy King vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 840d021c344SAndy King 841d021c344SAndy King put_cred(vsk->owner); 842d021c344SAndy King } 843d021c344SAndy King 844d021c344SAndy King static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) 845d021c344SAndy King { 846d021c344SAndy King int err; 847d021c344SAndy King 848d021c344SAndy King err = sock_queue_rcv_skb(sk, skb); 849d021c344SAndy King if (err) 850d021c344SAndy King kfree_skb(skb); 851d021c344SAndy King 852d021c344SAndy King return err; 853d021c344SAndy King } 854d021c344SAndy King 855b9ca2f5fSStefano Garzarella struct sock *vsock_create_connected(struct sock *parent) 856b9ca2f5fSStefano Garzarella { 857b9ca2f5fSStefano Garzarella return __vsock_create(sock_net(parent), NULL, parent, GFP_KERNEL, 858b9ca2f5fSStefano Garzarella parent->sk_type, 0); 859b9ca2f5fSStefano Garzarella } 860b9ca2f5fSStefano Garzarella EXPORT_SYMBOL_GPL(vsock_create_connected); 861b9ca2f5fSStefano Garzarella 862d021c344SAndy King s64 vsock_stream_has_data(struct vsock_sock *vsk) 863d021c344SAndy King { 864fe502c4aSStefano Garzarella return vsk->transport->stream_has_data(vsk); 865d021c344SAndy King } 866d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_stream_has_data); 867d021c344SAndy King 868cc97141aSStefano Garzarella static s64 vsock_connectible_has_data(struct vsock_sock *vsk) 8690798e78bSArseny Krasnov { 8700798e78bSArseny Krasnov struct sock *sk = sk_vsock(vsk); 8710798e78bSArseny Krasnov 8720798e78bSArseny Krasnov if (sk->sk_type == SOCK_SEQPACKET) 8730798e78bSArseny Krasnov return vsk->transport->seqpacket_has_data(vsk); 8740798e78bSArseny Krasnov else 8750798e78bSArseny Krasnov return vsock_stream_has_data(vsk); 8760798e78bSArseny Krasnov } 8770798e78bSArseny Krasnov 878d021c344SAndy King s64 vsock_stream_has_space(struct vsock_sock *vsk) 879d021c344SAndy King { 880fe502c4aSStefano Garzarella return vsk->transport->stream_has_space(vsk); 881d021c344SAndy King } 882d021c344SAndy King EXPORT_SYMBOL_GPL(vsock_stream_has_space); 883d021c344SAndy King 884f2fdcf67SArseniy Krasnov void vsock_data_ready(struct sock *sk) 885f2fdcf67SArseniy Krasnov { 886f2fdcf67SArseniy Krasnov struct vsock_sock *vsk = vsock_sk(sk); 887f2fdcf67SArseniy Krasnov 888f2fdcf67SArseniy Krasnov if (vsock_stream_has_data(vsk) >= sk->sk_rcvlowat || 889f2fdcf67SArseniy Krasnov sock_flag(sk, SOCK_DONE)) 890f2fdcf67SArseniy Krasnov sk->sk_data_ready(sk); 891f2fdcf67SArseniy Krasnov } 892f2fdcf67SArseniy Krasnov EXPORT_SYMBOL_GPL(vsock_data_ready); 893f2fdcf67SArseniy Krasnov 894d021c344SAndy King static int vsock_release(struct socket *sock) 895d021c344SAndy King { 8960d9138ffSDexuan Cui __vsock_release(sock->sk, 0); 897d021c344SAndy King sock->sk = NULL; 898d021c344SAndy King sock->state = SS_FREE; 899d021c344SAndy King 900d021c344SAndy King return 0; 901d021c344SAndy King } 902d021c344SAndy King 903d021c344SAndy King static int 904d021c344SAndy King vsock_bind(struct socket *sock, struct sockaddr *addr, int addr_len) 905d021c344SAndy King { 906d021c344SAndy King int err; 907d021c344SAndy King struct sock *sk; 908d021c344SAndy King struct sockaddr_vm *vm_addr; 909d021c344SAndy King 910d021c344SAndy King sk = sock->sk; 911d021c344SAndy King 912d021c344SAndy King if (vsock_addr_cast(addr, addr_len, &vm_addr) != 0) 913d021c344SAndy King return -EINVAL; 914d021c344SAndy King 915d021c344SAndy King lock_sock(sk); 916d021c344SAndy King err = __vsock_bind(sk, vm_addr); 917d021c344SAndy King release_sock(sk); 918d021c344SAndy King 919d021c344SAndy King return err; 920d021c344SAndy King } 921d021c344SAndy King 922d021c344SAndy King static int vsock_getname(struct socket *sock, 9239b2c45d4SDenys Vlasenko struct sockaddr *addr, int peer) 924d021c344SAndy King { 925d021c344SAndy King int err; 926d021c344SAndy King struct sock *sk; 927d021c344SAndy King struct vsock_sock *vsk; 928d021c344SAndy King struct sockaddr_vm *vm_addr; 929d021c344SAndy King 930d021c344SAndy King sk = sock->sk; 931d021c344SAndy King vsk = vsock_sk(sk); 932d021c344SAndy King err = 0; 933d021c344SAndy King 934d021c344SAndy King lock_sock(sk); 935d021c344SAndy King 936d021c344SAndy King if (peer) { 937d021c344SAndy King if (sock->state != SS_CONNECTED) { 938d021c344SAndy King err = -ENOTCONN; 939d021c344SAndy King goto out; 940d021c344SAndy King } 941d021c344SAndy King vm_addr = &vsk->remote_addr; 942d021c344SAndy King } else { 943d021c344SAndy King vm_addr = &vsk->local_addr; 944d021c344SAndy King } 945d021c344SAndy King 946d021c344SAndy King if (!vm_addr) { 947d021c344SAndy King err = -EINVAL; 948d021c344SAndy King goto out; 949d021c344SAndy King } 950d021c344SAndy King 951d021c344SAndy King /* sys_getsockname() and sys_getpeername() pass us a 952d021c344SAndy King * MAX_SOCK_ADDR-sized buffer and don't set addr_len. Unfortunately 953d021c344SAndy King * that macro is defined in socket.c instead of .h, so we hardcode its 954d021c344SAndy King * value here. 955d021c344SAndy King */ 956d021c344SAndy King BUILD_BUG_ON(sizeof(*vm_addr) > 128); 957d021c344SAndy King memcpy(addr, vm_addr, sizeof(*vm_addr)); 9589b2c45d4SDenys Vlasenko err = sizeof(*vm_addr); 959d021c344SAndy King 960d021c344SAndy King out: 961d021c344SAndy King release_sock(sk); 962d021c344SAndy King return err; 963d021c344SAndy King } 964d021c344SAndy King 965d021c344SAndy King static int vsock_shutdown(struct socket *sock, int mode) 966d021c344SAndy King { 967d021c344SAndy King int err; 968d021c344SAndy King struct sock *sk; 969d021c344SAndy King 970d021c344SAndy King /* User level uses SHUT_RD (0) and SHUT_WR (1), but the kernel uses 971d021c344SAndy King * RCV_SHUTDOWN (1) and SEND_SHUTDOWN (2), so we must increment mode 972d021c344SAndy King * here like the other address families do. Note also that the 973d021c344SAndy King * increment makes SHUT_RDWR (2) into RCV_SHUTDOWN | SEND_SHUTDOWN (3), 974d021c344SAndy King * which is what we want. 975d021c344SAndy King */ 976d021c344SAndy King mode++; 977d021c344SAndy King 978d021c344SAndy King if ((mode & ~SHUTDOWN_MASK) || !mode) 979d021c344SAndy King return -EINVAL; 980d021c344SAndy King 9818cb48554SArseny Krasnov /* If this is a connection oriented socket and it is not connected then 9828cb48554SArseny Krasnov * bail out immediately. If it is a DGRAM socket then we must first 9838cb48554SArseny Krasnov * kick the socket so that it wakes up from any sleeping calls, for 9848cb48554SArseny Krasnov * example recv(), and then afterwards return the error. 985d021c344SAndy King */ 986d021c344SAndy King 987d021c344SAndy King sk = sock->sk; 9881c5fae9cSStefano Garzarella 9891c5fae9cSStefano Garzarella lock_sock(sk); 990d021c344SAndy King if (sock->state == SS_UNCONNECTED) { 991d021c344SAndy King err = -ENOTCONN; 992a9e29e55SArseny Krasnov if (sock_type_connectible(sk->sk_type)) 9931c5fae9cSStefano Garzarella goto out; 994d021c344SAndy King } else { 995d021c344SAndy King sock->state = SS_DISCONNECTING; 996d021c344SAndy King err = 0; 997d021c344SAndy King } 998d021c344SAndy King 999d021c344SAndy King /* Receive and send shutdowns are treated alike. */ 1000d021c344SAndy King mode = mode & (RCV_SHUTDOWN | SEND_SHUTDOWN); 1001d021c344SAndy King if (mode) { 1002d021c344SAndy King sk->sk_shutdown |= mode; 1003d021c344SAndy King sk->sk_state_change(sk); 1004d021c344SAndy King 1005a9e29e55SArseny Krasnov if (sock_type_connectible(sk->sk_type)) { 1006d021c344SAndy King sock_reset_flag(sk, SOCK_DONE); 1007d021c344SAndy King vsock_send_shutdown(sk, mode); 1008d021c344SAndy King } 1009d021c344SAndy King } 1010d021c344SAndy King 10111c5fae9cSStefano Garzarella out: 10121c5fae9cSStefano Garzarella release_sock(sk); 1013d021c344SAndy King return err; 1014d021c344SAndy King } 1015d021c344SAndy King 1016a11e1d43SLinus Torvalds static __poll_t vsock_poll(struct file *file, struct socket *sock, 1017a11e1d43SLinus Torvalds poll_table *wait) 1018d021c344SAndy King { 1019a11e1d43SLinus Torvalds struct sock *sk; 1020a11e1d43SLinus Torvalds __poll_t mask; 1021a11e1d43SLinus Torvalds struct vsock_sock *vsk; 1022a11e1d43SLinus Torvalds 1023a11e1d43SLinus Torvalds sk = sock->sk; 1024a11e1d43SLinus Torvalds vsk = vsock_sk(sk); 1025a11e1d43SLinus Torvalds 1026a11e1d43SLinus Torvalds poll_wait(file, sk_sleep(sk), wait); 1027a11e1d43SLinus Torvalds mask = 0; 1028d021c344SAndy King 1029d021c344SAndy King if (sk->sk_err) 1030d021c344SAndy King /* Signify that there has been an error on this socket. */ 1031a9a08845SLinus Torvalds mask |= EPOLLERR; 1032d021c344SAndy King 1033d021c344SAndy King /* INET sockets treat local write shutdown and peer write shutdown as a 1034a9a08845SLinus Torvalds * case of EPOLLHUP set. 1035d021c344SAndy King */ 1036d021c344SAndy King if ((sk->sk_shutdown == SHUTDOWN_MASK) || 1037d021c344SAndy King ((sk->sk_shutdown & SEND_SHUTDOWN) && 1038d021c344SAndy King (vsk->peer_shutdown & SEND_SHUTDOWN))) { 1039a9a08845SLinus Torvalds mask |= EPOLLHUP; 1040d021c344SAndy King } 1041d021c344SAndy King 1042d021c344SAndy King if (sk->sk_shutdown & RCV_SHUTDOWN || 1043d021c344SAndy King vsk->peer_shutdown & SEND_SHUTDOWN) { 1044a9a08845SLinus Torvalds mask |= EPOLLRDHUP; 1045d021c344SAndy King } 1046d021c344SAndy King 1047d021c344SAndy King if (sock->type == SOCK_DGRAM) { 1048d021c344SAndy King /* For datagram sockets we can read if there is something in 1049d021c344SAndy King * the queue and write as long as the socket isn't shutdown for 1050d021c344SAndy King * sending. 1051d021c344SAndy King */ 10523ef7cf57SEric Dumazet if (!skb_queue_empty_lockless(&sk->sk_receive_queue) || 1053d021c344SAndy King (sk->sk_shutdown & RCV_SHUTDOWN)) { 1054a9a08845SLinus Torvalds mask |= EPOLLIN | EPOLLRDNORM; 1055d021c344SAndy King } 1056d021c344SAndy King 1057d021c344SAndy King if (!(sk->sk_shutdown & SEND_SHUTDOWN)) 1058a9a08845SLinus Torvalds mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND; 1059d021c344SAndy King 1060a9e29e55SArseny Krasnov } else if (sock_type_connectible(sk->sk_type)) { 1061c518adafSAlexander Popov const struct vsock_transport *transport; 1062c518adafSAlexander Popov 1063d021c344SAndy King lock_sock(sk); 1064d021c344SAndy King 1065c518adafSAlexander Popov transport = vsk->transport; 1066c518adafSAlexander Popov 1067d021c344SAndy King /* Listening sockets that have connections in their accept 1068d021c344SAndy King * queue can be read. 1069d021c344SAndy King */ 10703b4477d2SStefan Hajnoczi if (sk->sk_state == TCP_LISTEN 1071d021c344SAndy King && !vsock_is_accept_queue_empty(sk)) 1072a9a08845SLinus Torvalds mask |= EPOLLIN | EPOLLRDNORM; 1073d021c344SAndy King 1074d021c344SAndy King /* If there is something in the queue then we can read. */ 1075c0cfa2d8SStefano Garzarella if (transport && transport->stream_is_active(vsk) && 1076d021c344SAndy King !(sk->sk_shutdown & RCV_SHUTDOWN)) { 1077d021c344SAndy King bool data_ready_now = false; 1078ee0b3843SArseniy Krasnov int target = sock_rcvlowat(sk, 0, INT_MAX); 1079d021c344SAndy King int ret = transport->notify_poll_in( 1080ee0b3843SArseniy Krasnov vsk, target, &data_ready_now); 1081d021c344SAndy King if (ret < 0) { 1082a9a08845SLinus Torvalds mask |= EPOLLERR; 1083d021c344SAndy King } else { 1084d021c344SAndy King if (data_ready_now) 1085a9a08845SLinus Torvalds mask |= EPOLLIN | EPOLLRDNORM; 1086d021c344SAndy King 1087d021c344SAndy King } 1088d021c344SAndy King } 1089d021c344SAndy King 1090d021c344SAndy King /* Sockets whose connections have been closed, reset, or 1091d021c344SAndy King * terminated should also be considered read, and we check the 1092d021c344SAndy King * shutdown flag for that. 1093d021c344SAndy King */ 1094d021c344SAndy King if (sk->sk_shutdown & RCV_SHUTDOWN || 1095d021c344SAndy King vsk->peer_shutdown & SEND_SHUTDOWN) { 1096a9a08845SLinus Torvalds mask |= EPOLLIN | EPOLLRDNORM; 1097d021c344SAndy King } 1098d021c344SAndy King 1099d021c344SAndy King /* Connected sockets that can produce data can be written. */ 11001980c058SStefano Garzarella if (transport && sk->sk_state == TCP_ESTABLISHED) { 1101d021c344SAndy King if (!(sk->sk_shutdown & SEND_SHUTDOWN)) { 1102d021c344SAndy King bool space_avail_now = false; 1103d021c344SAndy King int ret = transport->notify_poll_out( 1104d021c344SAndy King vsk, 1, &space_avail_now); 1105d021c344SAndy King if (ret < 0) { 1106a9a08845SLinus Torvalds mask |= EPOLLERR; 1107d021c344SAndy King } else { 1108d021c344SAndy King if (space_avail_now) 1109a9a08845SLinus Torvalds /* Remove EPOLLWRBAND since INET 1110d021c344SAndy King * sockets are not setting it. 1111d021c344SAndy King */ 1112a9a08845SLinus Torvalds mask |= EPOLLOUT | EPOLLWRNORM; 1113d021c344SAndy King 1114d021c344SAndy King } 1115d021c344SAndy King } 1116d021c344SAndy King } 1117d021c344SAndy King 1118d021c344SAndy King /* Simulate INET socket poll behaviors, which sets 1119a9a08845SLinus Torvalds * EPOLLOUT|EPOLLWRNORM when peer is closed and nothing to read, 1120d021c344SAndy King * but local send is not shutdown. 1121d021c344SAndy King */ 1122ba3169fcSStefan Hajnoczi if (sk->sk_state == TCP_CLOSE || sk->sk_state == TCP_CLOSING) { 1123d021c344SAndy King if (!(sk->sk_shutdown & SEND_SHUTDOWN)) 1124a9a08845SLinus Torvalds mask |= EPOLLOUT | EPOLLWRNORM; 1125d021c344SAndy King 1126d021c344SAndy King } 1127d021c344SAndy King 1128d021c344SAndy King release_sock(sk); 1129d021c344SAndy King } 1130d021c344SAndy King 1131d021c344SAndy King return mask; 1132d021c344SAndy King } 1133d021c344SAndy King 11341b784140SYing Xue static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg, 11351b784140SYing Xue size_t len) 1136d021c344SAndy King { 1137d021c344SAndy King int err; 1138d021c344SAndy King struct sock *sk; 1139d021c344SAndy King struct vsock_sock *vsk; 1140d021c344SAndy King struct sockaddr_vm *remote_addr; 1141fe502c4aSStefano Garzarella const struct vsock_transport *transport; 1142d021c344SAndy King 1143d021c344SAndy King if (msg->msg_flags & MSG_OOB) 1144d021c344SAndy King return -EOPNOTSUPP; 1145d021c344SAndy King 1146d021c344SAndy King /* For now, MSG_DONTWAIT is always assumed... */ 1147d021c344SAndy King err = 0; 1148d021c344SAndy King sk = sock->sk; 1149d021c344SAndy King vsk = vsock_sk(sk); 1150d021c344SAndy King 1151d021c344SAndy King lock_sock(sk); 1152d021c344SAndy King 1153c518adafSAlexander Popov transport = vsk->transport; 1154c518adafSAlexander Popov 1155b3a6dfe8SAsias He err = vsock_auto_bind(vsk); 1156b3a6dfe8SAsias He if (err) 1157d021c344SAndy King goto out; 1158d021c344SAndy King 1159d021c344SAndy King 1160d021c344SAndy King /* If the provided message contains an address, use that. Otherwise 1161d021c344SAndy King * fall back on the socket's remote handle (if it has been connected). 1162d021c344SAndy King */ 1163d021c344SAndy King if (msg->msg_name && 1164d021c344SAndy King vsock_addr_cast(msg->msg_name, msg->msg_namelen, 1165d021c344SAndy King &remote_addr) == 0) { 1166d021c344SAndy King /* Ensure this address is of the right type and is a valid 1167d021c344SAndy King * destination. 1168d021c344SAndy King */ 1169d021c344SAndy King 1170d021c344SAndy King if (remote_addr->svm_cid == VMADDR_CID_ANY) 1171d021c344SAndy King remote_addr->svm_cid = transport->get_local_cid(); 1172d021c344SAndy King 1173d021c344SAndy King if (!vsock_addr_bound(remote_addr)) { 1174d021c344SAndy King err = -EINVAL; 1175d021c344SAndy King goto out; 1176d021c344SAndy King } 1177d021c344SAndy King } else if (sock->state == SS_CONNECTED) { 1178d021c344SAndy King remote_addr = &vsk->remote_addr; 1179d021c344SAndy King 1180d021c344SAndy King if (remote_addr->svm_cid == VMADDR_CID_ANY) 1181d021c344SAndy King remote_addr->svm_cid = transport->get_local_cid(); 1182d021c344SAndy King 1183d021c344SAndy King /* XXX Should connect() or this function ensure remote_addr is 1184d021c344SAndy King * bound? 1185d021c344SAndy King */ 1186d021c344SAndy King if (!vsock_addr_bound(&vsk->remote_addr)) { 1187d021c344SAndy King err = -EINVAL; 1188d021c344SAndy King goto out; 1189d021c344SAndy King } 1190d021c344SAndy King } else { 1191d021c344SAndy King err = -EINVAL; 1192d021c344SAndy King goto out; 1193d021c344SAndy King } 1194d021c344SAndy King 1195d021c344SAndy King if (!transport->dgram_allow(remote_addr->svm_cid, 1196d021c344SAndy King remote_addr->svm_port)) { 1197d021c344SAndy King err = -EINVAL; 1198d021c344SAndy King goto out; 1199d021c344SAndy King } 1200d021c344SAndy King 12010f7db23aSAl Viro err = transport->dgram_enqueue(vsk, remote_addr, msg, len); 1202d021c344SAndy King 1203d021c344SAndy King out: 1204d021c344SAndy King release_sock(sk); 1205d021c344SAndy King return err; 1206d021c344SAndy King } 1207d021c344SAndy King 1208d021c344SAndy King static int vsock_dgram_connect(struct socket *sock, 1209d021c344SAndy King struct sockaddr *addr, int addr_len, int flags) 1210d021c344SAndy King { 1211d021c344SAndy King int err; 1212d021c344SAndy King struct sock *sk; 1213d021c344SAndy King struct vsock_sock *vsk; 1214d021c344SAndy King struct sockaddr_vm *remote_addr; 1215d021c344SAndy King 1216d021c344SAndy King sk = sock->sk; 1217d021c344SAndy King vsk = vsock_sk(sk); 1218d021c344SAndy King 1219d021c344SAndy King err = vsock_addr_cast(addr, addr_len, &remote_addr); 1220d021c344SAndy King if (err == -EAFNOSUPPORT && remote_addr->svm_family == AF_UNSPEC) { 1221d021c344SAndy King lock_sock(sk); 1222d021c344SAndy King vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, 1223d021c344SAndy King VMADDR_PORT_ANY); 1224d021c344SAndy King sock->state = SS_UNCONNECTED; 1225d021c344SAndy King release_sock(sk); 1226d021c344SAndy King return 0; 1227d021c344SAndy King } else if (err != 0) 1228d021c344SAndy King return -EINVAL; 1229d021c344SAndy King 1230d021c344SAndy King lock_sock(sk); 1231d021c344SAndy King 1232b3a6dfe8SAsias He err = vsock_auto_bind(vsk); 1233b3a6dfe8SAsias He if (err) 1234d021c344SAndy King goto out; 1235d021c344SAndy King 1236fe502c4aSStefano Garzarella if (!vsk->transport->dgram_allow(remote_addr->svm_cid, 1237d021c344SAndy King remote_addr->svm_port)) { 1238d021c344SAndy King err = -EINVAL; 1239d021c344SAndy King goto out; 1240d021c344SAndy King } 1241d021c344SAndy King 1242d021c344SAndy King memcpy(&vsk->remote_addr, remote_addr, sizeof(vsk->remote_addr)); 1243d021c344SAndy King sock->state = SS_CONNECTED; 1244d021c344SAndy King 1245d021c344SAndy King out: 1246d021c344SAndy King release_sock(sk); 1247d021c344SAndy King return err; 1248d021c344SAndy King } 1249d021c344SAndy King 12501b784140SYing Xue static int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg, 12511b784140SYing Xue size_t len, int flags) 1252d021c344SAndy King { 1253fe502c4aSStefano Garzarella struct vsock_sock *vsk = vsock_sk(sock->sk); 1254fe502c4aSStefano Garzarella 1255fe502c4aSStefano Garzarella return vsk->transport->dgram_dequeue(vsk, msg, len, flags); 1256d021c344SAndy King } 1257d021c344SAndy King 1258d021c344SAndy King static const struct proto_ops vsock_dgram_ops = { 1259d021c344SAndy King .family = PF_VSOCK, 1260d021c344SAndy King .owner = THIS_MODULE, 1261d021c344SAndy King .release = vsock_release, 1262d021c344SAndy King .bind = vsock_bind, 1263d021c344SAndy King .connect = vsock_dgram_connect, 1264d021c344SAndy King .socketpair = sock_no_socketpair, 1265d021c344SAndy King .accept = sock_no_accept, 1266d021c344SAndy King .getname = vsock_getname, 1267a11e1d43SLinus Torvalds .poll = vsock_poll, 1268d021c344SAndy King .ioctl = sock_no_ioctl, 1269d021c344SAndy King .listen = sock_no_listen, 1270d021c344SAndy King .shutdown = vsock_shutdown, 1271d021c344SAndy King .sendmsg = vsock_dgram_sendmsg, 1272d021c344SAndy King .recvmsg = vsock_dgram_recvmsg, 1273d021c344SAndy King .mmap = sock_no_mmap, 1274d021c344SAndy King .sendpage = sock_no_sendpage, 1275d021c344SAndy King }; 1276d021c344SAndy King 1277380feae0SPeng Tao static int vsock_transport_cancel_pkt(struct vsock_sock *vsk) 1278380feae0SPeng Tao { 1279fe502c4aSStefano Garzarella const struct vsock_transport *transport = vsk->transport; 1280fe502c4aSStefano Garzarella 12815d1cbcc9SNorbert Slusarek if (!transport || !transport->cancel_pkt) 1282380feae0SPeng Tao return -EOPNOTSUPP; 1283380feae0SPeng Tao 1284380feae0SPeng Tao return transport->cancel_pkt(vsk); 1285380feae0SPeng Tao } 1286380feae0SPeng Tao 1287d021c344SAndy King static void vsock_connect_timeout(struct work_struct *work) 1288d021c344SAndy King { 1289d021c344SAndy King struct sock *sk; 1290d021c344SAndy King struct vsock_sock *vsk; 1291d021c344SAndy King 1292455f05ecSCong Wang vsk = container_of(work, struct vsock_sock, connect_work.work); 1293d021c344SAndy King sk = sk_vsock(vsk); 1294d021c344SAndy King 1295d021c344SAndy King lock_sock(sk); 12963b4477d2SStefan Hajnoczi if (sk->sk_state == TCP_SYN_SENT && 1297d021c344SAndy King (sk->sk_shutdown != SHUTDOWN_MASK)) { 12983b4477d2SStefan Hajnoczi sk->sk_state = TCP_CLOSE; 1299a3e7b29eSPeilin Ye sk->sk_socket->state = SS_UNCONNECTED; 1300d021c344SAndy King sk->sk_err = ETIMEDOUT; 1301e3ae2365SAlexander Aring sk_error_report(sk); 13023d0bc44dSNorbert Slusarek vsock_transport_cancel_pkt(vsk); 1303d021c344SAndy King } 1304d021c344SAndy King release_sock(sk); 1305d021c344SAndy King 1306d021c344SAndy King sock_put(sk); 1307d021c344SAndy King } 1308d021c344SAndy King 1309a9e29e55SArseny Krasnov static int vsock_connect(struct socket *sock, struct sockaddr *addr, 1310d021c344SAndy King int addr_len, int flags) 1311d021c344SAndy King { 1312d021c344SAndy King int err; 1313d021c344SAndy King struct sock *sk; 1314d021c344SAndy King struct vsock_sock *vsk; 1315fe502c4aSStefano Garzarella const struct vsock_transport *transport; 1316d021c344SAndy King struct sockaddr_vm *remote_addr; 1317d021c344SAndy King long timeout; 1318d021c344SAndy King DEFINE_WAIT(wait); 1319d021c344SAndy King 1320d021c344SAndy King err = 0; 1321d021c344SAndy King sk = sock->sk; 1322d021c344SAndy King vsk = vsock_sk(sk); 1323d021c344SAndy King 1324d021c344SAndy King lock_sock(sk); 1325d021c344SAndy King 1326d021c344SAndy King /* XXX AF_UNSPEC should make us disconnect like AF_INET. */ 1327d021c344SAndy King switch (sock->state) { 1328d021c344SAndy King case SS_CONNECTED: 1329d021c344SAndy King err = -EISCONN; 1330d021c344SAndy King goto out; 1331d021c344SAndy King case SS_DISCONNECTING: 1332d021c344SAndy King err = -EINVAL; 1333d021c344SAndy King goto out; 1334d021c344SAndy King case SS_CONNECTING: 1335d021c344SAndy King /* This continues on so we can move sock into the SS_CONNECTED 1336d021c344SAndy King * state once the connection has completed (at which point err 1337d021c344SAndy King * will be set to zero also). Otherwise, we will either wait 1338d021c344SAndy King * for the connection or return -EALREADY should this be a 1339d021c344SAndy King * non-blocking call. 1340d021c344SAndy King */ 1341d021c344SAndy King err = -EALREADY; 1342c7cd82b9SEiichi Tsukata if (flags & O_NONBLOCK) 1343c7cd82b9SEiichi Tsukata goto out; 1344d021c344SAndy King break; 1345d021c344SAndy King default: 13463b4477d2SStefan Hajnoczi if ((sk->sk_state == TCP_LISTEN) || 1347d021c344SAndy King vsock_addr_cast(addr, addr_len, &remote_addr) != 0) { 1348d021c344SAndy King err = -EINVAL; 1349d021c344SAndy King goto out; 1350d021c344SAndy King } 1351d021c344SAndy King 1352c0cfa2d8SStefano Garzarella /* Set the remote address that we are connecting to. */ 1353c0cfa2d8SStefano Garzarella memcpy(&vsk->remote_addr, remote_addr, 1354c0cfa2d8SStefano Garzarella sizeof(vsk->remote_addr)); 1355c0cfa2d8SStefano Garzarella 1356c0cfa2d8SStefano Garzarella err = vsock_assign_transport(vsk, NULL); 1357c0cfa2d8SStefano Garzarella if (err) 1358c0cfa2d8SStefano Garzarella goto out; 1359c0cfa2d8SStefano Garzarella 1360c0cfa2d8SStefano Garzarella transport = vsk->transport; 1361c0cfa2d8SStefano Garzarella 1362d021c344SAndy King /* The hypervisor and well-known contexts do not have socket 1363d021c344SAndy King * endpoints. 1364d021c344SAndy King */ 1365c0cfa2d8SStefano Garzarella if (!transport || 1366c0cfa2d8SStefano Garzarella !transport->stream_allow(remote_addr->svm_cid, 1367d021c344SAndy King remote_addr->svm_port)) { 1368d021c344SAndy King err = -ENETUNREACH; 1369d021c344SAndy King goto out; 1370d021c344SAndy King } 1371d021c344SAndy King 1372b3a6dfe8SAsias He err = vsock_auto_bind(vsk); 1373b3a6dfe8SAsias He if (err) 1374d021c344SAndy King goto out; 1375d021c344SAndy King 13763b4477d2SStefan Hajnoczi sk->sk_state = TCP_SYN_SENT; 1377d021c344SAndy King 1378d021c344SAndy King err = transport->connect(vsk); 1379d021c344SAndy King if (err < 0) 1380d021c344SAndy King goto out; 1381d021c344SAndy King 1382d021c344SAndy King /* Mark sock as connecting and set the error code to in 1383d021c344SAndy King * progress in case this is a non-blocking connect. 1384d021c344SAndy King */ 1385d021c344SAndy King sock->state = SS_CONNECTING; 1386d021c344SAndy King err = -EINPROGRESS; 1387d021c344SAndy King } 1388d021c344SAndy King 1389d021c344SAndy King /* The receive path will handle all communication until we are able to 1390d021c344SAndy King * enter the connected state. Here we wait for the connection to be 1391d021c344SAndy King * completed or a notification of an error. 1392d021c344SAndy King */ 1393d021c344SAndy King timeout = vsk->connect_timeout; 1394d021c344SAndy King prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); 1395d021c344SAndy King 13963b4477d2SStefan Hajnoczi while (sk->sk_state != TCP_ESTABLISHED && sk->sk_err == 0) { 1397d021c344SAndy King if (flags & O_NONBLOCK) { 1398d021c344SAndy King /* If we're not going to block, we schedule a timeout 1399d021c344SAndy King * function to generate a timeout on the connection 1400d021c344SAndy King * attempt, in case the peer doesn't respond in a 1401d021c344SAndy King * timely manner. We hold on to the socket until the 1402d021c344SAndy King * timeout fires. 1403d021c344SAndy King */ 1404d021c344SAndy King sock_hold(sk); 14057e97cfedSPeilin Ye 14067e97cfedSPeilin Ye /* If the timeout function is already scheduled, 14077e97cfedSPeilin Ye * reschedule it, then ungrab the socket refcount to 14087e97cfedSPeilin Ye * keep it balanced. 14097e97cfedSPeilin Ye */ 14107e97cfedSPeilin Ye if (mod_delayed_work(system_wq, &vsk->connect_work, 14117e97cfedSPeilin Ye timeout)) 14127e97cfedSPeilin Ye sock_put(sk); 1413d021c344SAndy King 1414d021c344SAndy King /* Skip ahead to preserve error code set above. */ 1415d021c344SAndy King goto out_wait; 1416d021c344SAndy King } 1417d021c344SAndy King 1418d021c344SAndy King release_sock(sk); 1419d021c344SAndy King timeout = schedule_timeout(timeout); 1420d021c344SAndy King lock_sock(sk); 1421d021c344SAndy King 1422d021c344SAndy King if (signal_pending(current)) { 1423d021c344SAndy King err = sock_intr_errno(timeout); 1424c7ff9cffSLongpeng(Mike) sk->sk_state = sk->sk_state == TCP_ESTABLISHED ? TCP_CLOSING : TCP_CLOSE; 1425f7f9b5e7SClaudio Imbrenda sock->state = SS_UNCONNECTED; 1426380feae0SPeng Tao vsock_transport_cancel_pkt(vsk); 1427b9208492SSeth Forshee vsock_remove_connected(vsk); 1428f7f9b5e7SClaudio Imbrenda goto out_wait; 1429d021c344SAndy King } else if (timeout == 0) { 1430d021c344SAndy King err = -ETIMEDOUT; 14313b4477d2SStefan Hajnoczi sk->sk_state = TCP_CLOSE; 1432f7f9b5e7SClaudio Imbrenda sock->state = SS_UNCONNECTED; 1433380feae0SPeng Tao vsock_transport_cancel_pkt(vsk); 1434f7f9b5e7SClaudio Imbrenda goto out_wait; 1435d021c344SAndy King } 1436d021c344SAndy King 1437d021c344SAndy King prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); 1438d021c344SAndy King } 1439d021c344SAndy King 1440d021c344SAndy King if (sk->sk_err) { 1441d021c344SAndy King err = -sk->sk_err; 14423b4477d2SStefan Hajnoczi sk->sk_state = TCP_CLOSE; 1443f7f9b5e7SClaudio Imbrenda sock->state = SS_UNCONNECTED; 1444f7f9b5e7SClaudio Imbrenda } else { 1445d021c344SAndy King err = 0; 1446f7f9b5e7SClaudio Imbrenda } 1447d021c344SAndy King 1448d021c344SAndy King out_wait: 1449d021c344SAndy King finish_wait(sk_sleep(sk), &wait); 1450d021c344SAndy King out: 1451d021c344SAndy King release_sock(sk); 1452d021c344SAndy King return err; 1453d021c344SAndy King } 1454d021c344SAndy King 1455cdfbabfbSDavid Howells static int vsock_accept(struct socket *sock, struct socket *newsock, int flags, 1456cdfbabfbSDavid Howells bool kern) 1457d021c344SAndy King { 1458d021c344SAndy King struct sock *listener; 1459d021c344SAndy King int err; 1460d021c344SAndy King struct sock *connected; 1461d021c344SAndy King struct vsock_sock *vconnected; 1462d021c344SAndy King long timeout; 1463d021c344SAndy King DEFINE_WAIT(wait); 1464d021c344SAndy King 1465d021c344SAndy King err = 0; 1466d021c344SAndy King listener = sock->sk; 1467d021c344SAndy King 1468d021c344SAndy King lock_sock(listener); 1469d021c344SAndy King 1470a9e29e55SArseny Krasnov if (!sock_type_connectible(sock->type)) { 1471d021c344SAndy King err = -EOPNOTSUPP; 1472d021c344SAndy King goto out; 1473d021c344SAndy King } 1474d021c344SAndy King 14753b4477d2SStefan Hajnoczi if (listener->sk_state != TCP_LISTEN) { 1476d021c344SAndy King err = -EINVAL; 1477d021c344SAndy King goto out; 1478d021c344SAndy King } 1479d021c344SAndy King 1480d021c344SAndy King /* Wait for children sockets to appear; these are the new sockets 1481d021c344SAndy King * created upon connection establishment. 1482d021c344SAndy King */ 14837e0afbdfSStefano Garzarella timeout = sock_rcvtimeo(listener, flags & O_NONBLOCK); 1484d021c344SAndy King prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE); 1485d021c344SAndy King 1486d021c344SAndy King while ((connected = vsock_dequeue_accept(listener)) == NULL && 1487d021c344SAndy King listener->sk_err == 0) { 1488d021c344SAndy King release_sock(listener); 1489d021c344SAndy King timeout = schedule_timeout(timeout); 1490f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(listener), &wait); 1491d021c344SAndy King lock_sock(listener); 1492d021c344SAndy King 1493d021c344SAndy King if (signal_pending(current)) { 1494d021c344SAndy King err = sock_intr_errno(timeout); 1495f7f9b5e7SClaudio Imbrenda goto out; 1496d021c344SAndy King } else if (timeout == 0) { 1497d021c344SAndy King err = -EAGAIN; 1498f7f9b5e7SClaudio Imbrenda goto out; 1499d021c344SAndy King } 1500d021c344SAndy King 1501d021c344SAndy King prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE); 1502d021c344SAndy King } 1503f7f9b5e7SClaudio Imbrenda finish_wait(sk_sleep(listener), &wait); 1504d021c344SAndy King 1505d021c344SAndy King if (listener->sk_err) 1506d021c344SAndy King err = -listener->sk_err; 1507d021c344SAndy King 1508d021c344SAndy King if (connected) { 15097976a11bSEric Dumazet sk_acceptq_removed(listener); 1510d021c344SAndy King 15114192f672SStefan Hajnoczi lock_sock_nested(connected, SINGLE_DEPTH_NESTING); 1512d021c344SAndy King vconnected = vsock_sk(connected); 1513d021c344SAndy King 1514d021c344SAndy King /* If the listener socket has received an error, then we should 1515d021c344SAndy King * reject this socket and return. Note that we simply mark the 1516d021c344SAndy King * socket rejected, drop our reference, and let the cleanup 1517d021c344SAndy King * function handle the cleanup; the fact that we found it in 1518d021c344SAndy King * the listener's accept queue guarantees that the cleanup 1519d021c344SAndy King * function hasn't run yet. 1520d021c344SAndy King */ 1521d021c344SAndy King if (err) { 1522d021c344SAndy King vconnected->rejected = true; 1523f7f9b5e7SClaudio Imbrenda } else { 1524d021c344SAndy King newsock->state = SS_CONNECTED; 1525d021c344SAndy King sock_graft(connected, newsock); 1526f7f9b5e7SClaudio Imbrenda } 1527f7f9b5e7SClaudio Imbrenda 1528d021c344SAndy King release_sock(connected); 1529d021c344SAndy King sock_put(connected); 1530d021c344SAndy King } 1531d021c344SAndy King 1532d021c344SAndy King out: 1533d021c344SAndy King release_sock(listener); 1534d021c344SAndy King return err; 1535d021c344SAndy King } 1536d021c344SAndy King 1537d021c344SAndy King static int vsock_listen(struct socket *sock, int backlog) 1538d021c344SAndy King { 1539d021c344SAndy King int err; 1540d021c344SAndy King struct sock *sk; 1541d021c344SAndy King struct vsock_sock *vsk; 1542d021c344SAndy King 1543d021c344SAndy King sk = sock->sk; 1544d021c344SAndy King 1545d021c344SAndy King lock_sock(sk); 1546d021c344SAndy King 1547a9e29e55SArseny Krasnov if (!sock_type_connectible(sk->sk_type)) { 1548d021c344SAndy King err = -EOPNOTSUPP; 1549d021c344SAndy King goto out; 1550d021c344SAndy King } 1551d021c344SAndy King 1552d021c344SAndy King if (sock->state != SS_UNCONNECTED) { 1553d021c344SAndy King err = -EINVAL; 1554d021c344SAndy King goto out; 1555d021c344SAndy King } 1556d021c344SAndy King 1557d021c344SAndy King vsk = vsock_sk(sk); 1558d021c344SAndy King 1559d021c344SAndy King if (!vsock_addr_bound(&vsk->local_addr)) { 1560d021c344SAndy King err = -EINVAL; 1561d021c344SAndy King goto out; 1562d021c344SAndy King } 1563d021c344SAndy King 1564d021c344SAndy King sk->sk_max_ack_backlog = backlog; 15653b4477d2SStefan Hajnoczi sk->sk_state = TCP_LISTEN; 1566d021c344SAndy King 1567d021c344SAndy King err = 0; 1568d021c344SAndy King 1569d021c344SAndy King out: 1570d021c344SAndy King release_sock(sk); 1571d021c344SAndy King return err; 1572d021c344SAndy King } 1573d021c344SAndy King 1574b9f2b0ffSStefano Garzarella static void vsock_update_buffer_size(struct vsock_sock *vsk, 1575b9f2b0ffSStefano Garzarella const struct vsock_transport *transport, 1576b9f2b0ffSStefano Garzarella u64 val) 1577b9f2b0ffSStefano Garzarella { 1578b9f2b0ffSStefano Garzarella if (val > vsk->buffer_max_size) 1579b9f2b0ffSStefano Garzarella val = vsk->buffer_max_size; 1580b9f2b0ffSStefano Garzarella 1581b9f2b0ffSStefano Garzarella if (val < vsk->buffer_min_size) 1582b9f2b0ffSStefano Garzarella val = vsk->buffer_min_size; 1583b9f2b0ffSStefano Garzarella 1584b9f2b0ffSStefano Garzarella if (val != vsk->buffer_size && 1585b9f2b0ffSStefano Garzarella transport && transport->notify_buffer_size) 1586b9f2b0ffSStefano Garzarella transport->notify_buffer_size(vsk, &val); 1587b9f2b0ffSStefano Garzarella 1588b9f2b0ffSStefano Garzarella vsk->buffer_size = val; 1589b9f2b0ffSStefano Garzarella } 1590b9f2b0ffSStefano Garzarella 1591a9e29e55SArseny Krasnov static int vsock_connectible_setsockopt(struct socket *sock, 1592d021c344SAndy King int level, 1593d021c344SAndy King int optname, 1594a7b75c5aSChristoph Hellwig sockptr_t optval, 1595d021c344SAndy King unsigned int optlen) 1596d021c344SAndy King { 1597d021c344SAndy King int err; 1598d021c344SAndy King struct sock *sk; 1599d021c344SAndy King struct vsock_sock *vsk; 1600fe502c4aSStefano Garzarella const struct vsock_transport *transport; 1601d021c344SAndy King u64 val; 1602d021c344SAndy King 1603d021c344SAndy King if (level != AF_VSOCK) 1604d021c344SAndy King return -ENOPROTOOPT; 1605d021c344SAndy King 1606d021c344SAndy King #define COPY_IN(_v) \ 1607d021c344SAndy King do { \ 1608d021c344SAndy King if (optlen < sizeof(_v)) { \ 1609d021c344SAndy King err = -EINVAL; \ 1610d021c344SAndy King goto exit; \ 1611d021c344SAndy King } \ 1612a7b75c5aSChristoph Hellwig if (copy_from_sockptr(&_v, optval, sizeof(_v)) != 0) { \ 1613d021c344SAndy King err = -EFAULT; \ 1614d021c344SAndy King goto exit; \ 1615d021c344SAndy King } \ 1616d021c344SAndy King } while (0) 1617d021c344SAndy King 1618d021c344SAndy King err = 0; 1619d021c344SAndy King sk = sock->sk; 1620d021c344SAndy King vsk = vsock_sk(sk); 1621d021c344SAndy King 1622d021c344SAndy King lock_sock(sk); 1623d021c344SAndy King 1624c518adafSAlexander Popov transport = vsk->transport; 1625c518adafSAlexander Popov 1626d021c344SAndy King switch (optname) { 1627d021c344SAndy King case SO_VM_SOCKETS_BUFFER_SIZE: 1628d021c344SAndy King COPY_IN(val); 1629b9f2b0ffSStefano Garzarella vsock_update_buffer_size(vsk, transport, val); 1630d021c344SAndy King break; 1631d021c344SAndy King 1632d021c344SAndy King case SO_VM_SOCKETS_BUFFER_MAX_SIZE: 1633d021c344SAndy King COPY_IN(val); 1634b9f2b0ffSStefano Garzarella vsk->buffer_max_size = val; 1635b9f2b0ffSStefano Garzarella vsock_update_buffer_size(vsk, transport, vsk->buffer_size); 1636d021c344SAndy King break; 1637d021c344SAndy King 1638d021c344SAndy King case SO_VM_SOCKETS_BUFFER_MIN_SIZE: 1639d021c344SAndy King COPY_IN(val); 1640b9f2b0ffSStefano Garzarella vsk->buffer_min_size = val; 1641b9f2b0ffSStefano Garzarella vsock_update_buffer_size(vsk, transport, vsk->buffer_size); 1642d021c344SAndy King break; 1643d021c344SAndy King 16444c1e34c0SRichard Palethorpe case SO_VM_SOCKETS_CONNECT_TIMEOUT_NEW: 16454c1e34c0SRichard Palethorpe case SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD: { 16464c1e34c0SRichard Palethorpe struct __kernel_sock_timeval tv; 16474c1e34c0SRichard Palethorpe 16484c1e34c0SRichard Palethorpe err = sock_copy_user_timeval(&tv, optval, optlen, 16494c1e34c0SRichard Palethorpe optname == SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD); 16504c1e34c0SRichard Palethorpe if (err) 16514c1e34c0SRichard Palethorpe break; 1652d021c344SAndy King if (tv.tv_sec >= 0 && tv.tv_usec < USEC_PER_SEC && 1653d021c344SAndy King tv.tv_sec < (MAX_SCHEDULE_TIMEOUT / HZ - 1)) { 1654d021c344SAndy King vsk->connect_timeout = tv.tv_sec * HZ + 16554c1e34c0SRichard Palethorpe DIV_ROUND_UP((unsigned long)tv.tv_usec, (USEC_PER_SEC / HZ)); 1656d021c344SAndy King if (vsk->connect_timeout == 0) 1657d021c344SAndy King vsk->connect_timeout = 1658d021c344SAndy King VSOCK_DEFAULT_CONNECT_TIMEOUT; 1659d021c344SAndy King 1660d021c344SAndy King } else { 1661d021c344SAndy King err = -ERANGE; 1662d021c344SAndy King } 1663d021c344SAndy King break; 1664d021c344SAndy King } 1665d021c344SAndy King 1666d021c344SAndy King default: 1667d021c344SAndy King err = -ENOPROTOOPT; 1668d021c344SAndy King break; 1669d021c344SAndy King } 1670d021c344SAndy King 1671d021c344SAndy King #undef COPY_IN 1672d021c344SAndy King 1673d021c344SAndy King exit: 1674d021c344SAndy King release_sock(sk); 1675d021c344SAndy King return err; 1676d021c344SAndy King } 1677d021c344SAndy King 1678a9e29e55SArseny Krasnov static int vsock_connectible_getsockopt(struct socket *sock, 1679d021c344SAndy King int level, int optname, 1680d021c344SAndy King char __user *optval, 1681d021c344SAndy King int __user *optlen) 1682d021c344SAndy King { 1683685c3f2fSRichard Palethorpe struct sock *sk = sock->sk; 1684685c3f2fSRichard Palethorpe struct vsock_sock *vsk = vsock_sk(sk); 1685685c3f2fSRichard Palethorpe 1686685c3f2fSRichard Palethorpe union { 1687685c3f2fSRichard Palethorpe u64 val64; 16884c1e34c0SRichard Palethorpe struct old_timeval32 tm32; 1689685c3f2fSRichard Palethorpe struct __kernel_old_timeval tm; 16904c1e34c0SRichard Palethorpe struct __kernel_sock_timeval stm; 1691685c3f2fSRichard Palethorpe } v; 1692685c3f2fSRichard Palethorpe 1693685c3f2fSRichard Palethorpe int lv = sizeof(v.val64); 1694d021c344SAndy King int len; 1695d021c344SAndy King 1696d021c344SAndy King if (level != AF_VSOCK) 1697d021c344SAndy King return -ENOPROTOOPT; 1698d021c344SAndy King 1699685c3f2fSRichard Palethorpe if (get_user(len, optlen)) 1700685c3f2fSRichard Palethorpe return -EFAULT; 1701d021c344SAndy King 1702685c3f2fSRichard Palethorpe memset(&v, 0, sizeof(v)); 1703d021c344SAndy King 1704d021c344SAndy King switch (optname) { 1705d021c344SAndy King case SO_VM_SOCKETS_BUFFER_SIZE: 1706685c3f2fSRichard Palethorpe v.val64 = vsk->buffer_size; 1707d021c344SAndy King break; 1708d021c344SAndy King 1709d021c344SAndy King case SO_VM_SOCKETS_BUFFER_MAX_SIZE: 1710685c3f2fSRichard Palethorpe v.val64 = vsk->buffer_max_size; 1711d021c344SAndy King break; 1712d021c344SAndy King 1713d021c344SAndy King case SO_VM_SOCKETS_BUFFER_MIN_SIZE: 1714685c3f2fSRichard Palethorpe v.val64 = vsk->buffer_min_size; 1715d021c344SAndy King break; 1716d021c344SAndy King 17174c1e34c0SRichard Palethorpe case SO_VM_SOCKETS_CONNECT_TIMEOUT_NEW: 17184c1e34c0SRichard Palethorpe case SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD: 17194c1e34c0SRichard Palethorpe lv = sock_get_timeout(vsk->connect_timeout, &v, 17204c1e34c0SRichard Palethorpe optname == SO_VM_SOCKETS_CONNECT_TIMEOUT_OLD); 1721d021c344SAndy King break; 1722685c3f2fSRichard Palethorpe 1723d021c344SAndy King default: 1724d021c344SAndy King return -ENOPROTOOPT; 1725d021c344SAndy King } 1726d021c344SAndy King 1727685c3f2fSRichard Palethorpe if (len < lv) 1728685c3f2fSRichard Palethorpe return -EINVAL; 1729685c3f2fSRichard Palethorpe if (len > lv) 1730685c3f2fSRichard Palethorpe len = lv; 1731685c3f2fSRichard Palethorpe if (copy_to_user(optval, &v, len)) 1732d021c344SAndy King return -EFAULT; 1733d021c344SAndy King 1734685c3f2fSRichard Palethorpe if (put_user(len, optlen)) 1735685c3f2fSRichard Palethorpe return -EFAULT; 1736d021c344SAndy King 1737d021c344SAndy King return 0; 1738d021c344SAndy King } 1739d021c344SAndy King 1740a9e29e55SArseny Krasnov static int vsock_connectible_sendmsg(struct socket *sock, struct msghdr *msg, 17411b784140SYing Xue size_t len) 1742d021c344SAndy King { 1743d021c344SAndy King struct sock *sk; 1744d021c344SAndy King struct vsock_sock *vsk; 1745fe502c4aSStefano Garzarella const struct vsock_transport *transport; 1746d021c344SAndy King ssize_t total_written; 1747d021c344SAndy King long timeout; 1748d021c344SAndy King int err; 1749d021c344SAndy King struct vsock_transport_send_notify_data send_data; 1750499fde66SWANG Cong DEFINE_WAIT_FUNC(wait, woken_wake_function); 1751d021c344SAndy King 1752d021c344SAndy King sk = sock->sk; 1753d021c344SAndy King vsk = vsock_sk(sk); 1754d021c344SAndy King total_written = 0; 1755d021c344SAndy King err = 0; 1756d021c344SAndy King 1757d021c344SAndy King if (msg->msg_flags & MSG_OOB) 1758d021c344SAndy King return -EOPNOTSUPP; 1759d021c344SAndy King 1760d021c344SAndy King lock_sock(sk); 1761d021c344SAndy King 1762c518adafSAlexander Popov transport = vsk->transport; 1763c518adafSAlexander Popov 17648cb48554SArseny Krasnov /* Callers should not provide a destination with connection oriented 17658cb48554SArseny Krasnov * sockets. 17668cb48554SArseny Krasnov */ 1767d021c344SAndy King if (msg->msg_namelen) { 17683b4477d2SStefan Hajnoczi err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP; 1769d021c344SAndy King goto out; 1770d021c344SAndy King } 1771d021c344SAndy King 1772d021c344SAndy King /* Send data only if both sides are not shutdown in the direction. */ 1773d021c344SAndy King if (sk->sk_shutdown & SEND_SHUTDOWN || 1774d021c344SAndy King vsk->peer_shutdown & RCV_SHUTDOWN) { 1775d021c344SAndy King err = -EPIPE; 1776d021c344SAndy King goto out; 1777d021c344SAndy King } 1778d021c344SAndy King 1779c0cfa2d8SStefano Garzarella if (!transport || sk->sk_state != TCP_ESTABLISHED || 1780d021c344SAndy King !vsock_addr_bound(&vsk->local_addr)) { 1781d021c344SAndy King err = -ENOTCONN; 1782d021c344SAndy King goto out; 1783d021c344SAndy King } 1784d021c344SAndy King 1785d021c344SAndy King if (!vsock_addr_bound(&vsk->remote_addr)) { 1786d021c344SAndy King err = -EDESTADDRREQ; 1787d021c344SAndy King goto out; 1788d021c344SAndy King } 1789d021c344SAndy King 1790d021c344SAndy King /* Wait for room in the produce queue to enqueue our user's data. */ 1791d021c344SAndy King timeout = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT); 1792d021c344SAndy King 1793d021c344SAndy King err = transport->notify_send_init(vsk, &send_data); 1794d021c344SAndy King if (err < 0) 1795d021c344SAndy King goto out; 1796d021c344SAndy King 1797d021c344SAndy King while (total_written < len) { 1798d021c344SAndy King ssize_t written; 1799d021c344SAndy King 1800499fde66SWANG Cong add_wait_queue(sk_sleep(sk), &wait); 1801d021c344SAndy King while (vsock_stream_has_space(vsk) == 0 && 1802d021c344SAndy King sk->sk_err == 0 && 1803d021c344SAndy King !(sk->sk_shutdown & SEND_SHUTDOWN) && 1804d021c344SAndy King !(vsk->peer_shutdown & RCV_SHUTDOWN)) { 1805d021c344SAndy King 1806d021c344SAndy King /* Don't wait for non-blocking sockets. */ 1807d021c344SAndy King if (timeout == 0) { 1808d021c344SAndy King err = -EAGAIN; 1809499fde66SWANG Cong remove_wait_queue(sk_sleep(sk), &wait); 1810f7f9b5e7SClaudio Imbrenda goto out_err; 1811d021c344SAndy King } 1812d021c344SAndy King 1813d021c344SAndy King err = transport->notify_send_pre_block(vsk, &send_data); 1814f7f9b5e7SClaudio Imbrenda if (err < 0) { 1815499fde66SWANG Cong remove_wait_queue(sk_sleep(sk), &wait); 1816f7f9b5e7SClaudio Imbrenda goto out_err; 1817f7f9b5e7SClaudio Imbrenda } 1818d021c344SAndy King 1819d021c344SAndy King release_sock(sk); 1820499fde66SWANG Cong timeout = wait_woken(&wait, TASK_INTERRUPTIBLE, timeout); 1821d021c344SAndy King lock_sock(sk); 1822d021c344SAndy King if (signal_pending(current)) { 1823d021c344SAndy King err = sock_intr_errno(timeout); 1824499fde66SWANG Cong remove_wait_queue(sk_sleep(sk), &wait); 1825f7f9b5e7SClaudio Imbrenda goto out_err; 1826d021c344SAndy King } else if (timeout == 0) { 1827d021c344SAndy King err = -EAGAIN; 1828499fde66SWANG Cong remove_wait_queue(sk_sleep(sk), &wait); 1829f7f9b5e7SClaudio Imbrenda goto out_err; 1830d021c344SAndy King } 1831d021c344SAndy King } 1832499fde66SWANG Cong remove_wait_queue(sk_sleep(sk), &wait); 1833d021c344SAndy King 1834d021c344SAndy King /* These checks occur both as part of and after the loop 1835d021c344SAndy King * conditional since we need to check before and after 1836d021c344SAndy King * sleeping. 1837d021c344SAndy King */ 1838d021c344SAndy King if (sk->sk_err) { 1839d021c344SAndy King err = -sk->sk_err; 1840f7f9b5e7SClaudio Imbrenda goto out_err; 1841d021c344SAndy King } else if ((sk->sk_shutdown & SEND_SHUTDOWN) || 1842d021c344SAndy King (vsk->peer_shutdown & RCV_SHUTDOWN)) { 1843d021c344SAndy King err = -EPIPE; 1844f7f9b5e7SClaudio Imbrenda goto out_err; 1845d021c344SAndy King } 1846d021c344SAndy King 1847d021c344SAndy King err = transport->notify_send_pre_enqueue(vsk, &send_data); 1848d021c344SAndy King if (err < 0) 1849f7f9b5e7SClaudio Imbrenda goto out_err; 1850d021c344SAndy King 1851d021c344SAndy King /* Note that enqueue will only write as many bytes as are free 1852d021c344SAndy King * in the produce queue, so we don't need to ensure len is 1853d021c344SAndy King * smaller than the queue size. It is the caller's 1854d021c344SAndy King * responsibility to check how many bytes we were able to send. 1855d021c344SAndy King */ 1856d021c344SAndy King 1857fbe70c48SArseny Krasnov if (sk->sk_type == SOCK_SEQPACKET) { 1858fbe70c48SArseny Krasnov written = transport->seqpacket_enqueue(vsk, 1859fbe70c48SArseny Krasnov msg, len - total_written); 1860fbe70c48SArseny Krasnov } else { 1861fbe70c48SArseny Krasnov written = transport->stream_enqueue(vsk, 1862fbe70c48SArseny Krasnov msg, len - total_written); 1863fbe70c48SArseny Krasnov } 1864*c43170b7SBobby Eshleman 1865d021c344SAndy King if (written < 0) { 1866*c43170b7SBobby Eshleman err = written; 1867f7f9b5e7SClaudio Imbrenda goto out_err; 1868d021c344SAndy King } 1869d021c344SAndy King 1870d021c344SAndy King total_written += written; 1871d021c344SAndy King 1872d021c344SAndy King err = transport->notify_send_post_enqueue( 1873d021c344SAndy King vsk, written, &send_data); 1874d021c344SAndy King if (err < 0) 1875f7f9b5e7SClaudio Imbrenda goto out_err; 1876d021c344SAndy King 1877d021c344SAndy King } 1878d021c344SAndy King 1879f7f9b5e7SClaudio Imbrenda out_err: 1880fbe70c48SArseny Krasnov if (total_written > 0) { 1881fbe70c48SArseny Krasnov /* Return number of written bytes only if: 1882fbe70c48SArseny Krasnov * 1) SOCK_STREAM socket. 1883fbe70c48SArseny Krasnov * 2) SOCK_SEQPACKET socket when whole buffer is sent. 1884fbe70c48SArseny Krasnov */ 1885fbe70c48SArseny Krasnov if (sk->sk_type == SOCK_STREAM || total_written == len) 1886d021c344SAndy King err = total_written; 1887fbe70c48SArseny Krasnov } 1888d021c344SAndy King out: 1889d021c344SAndy King release_sock(sk); 1890d021c344SAndy King return err; 1891d021c344SAndy King } 1892d021c344SAndy King 18930de5b2e6SStefano Garzarella static int vsock_connectible_wait_data(struct sock *sk, 18940de5b2e6SStefano Garzarella struct wait_queue_entry *wait, 1895b3f7fd54SArseny Krasnov long timeout, 1896b3f7fd54SArseny Krasnov struct vsock_transport_recv_notify_data *recv_data, 1897b3f7fd54SArseny Krasnov size_t target) 1898b3f7fd54SArseny Krasnov { 1899b3f7fd54SArseny Krasnov const struct vsock_transport *transport; 1900b3f7fd54SArseny Krasnov struct vsock_sock *vsk; 1901b3f7fd54SArseny Krasnov s64 data; 1902b3f7fd54SArseny Krasnov int err; 1903b3f7fd54SArseny Krasnov 1904b3f7fd54SArseny Krasnov vsk = vsock_sk(sk); 1905b3f7fd54SArseny Krasnov err = 0; 1906b3f7fd54SArseny Krasnov transport = vsk->transport; 1907b3f7fd54SArseny Krasnov 1908466a8533SDexuan Cui while (1) { 1909b3f7fd54SArseny Krasnov prepare_to_wait(sk_sleep(sk), wait, TASK_INTERRUPTIBLE); 1910466a8533SDexuan Cui data = vsock_connectible_has_data(vsk); 1911466a8533SDexuan Cui if (data != 0) 1912466a8533SDexuan Cui break; 1913b3f7fd54SArseny Krasnov 1914b3f7fd54SArseny Krasnov if (sk->sk_err != 0 || 1915b3f7fd54SArseny Krasnov (sk->sk_shutdown & RCV_SHUTDOWN) || 1916b3f7fd54SArseny Krasnov (vsk->peer_shutdown & SEND_SHUTDOWN)) { 1917b3f7fd54SArseny Krasnov break; 1918b3f7fd54SArseny Krasnov } 1919b3f7fd54SArseny Krasnov 1920b3f7fd54SArseny Krasnov /* Don't wait for non-blocking sockets. */ 1921b3f7fd54SArseny Krasnov if (timeout == 0) { 1922b3f7fd54SArseny Krasnov err = -EAGAIN; 1923b3f7fd54SArseny Krasnov break; 1924b3f7fd54SArseny Krasnov } 1925b3f7fd54SArseny Krasnov 1926b3f7fd54SArseny Krasnov if (recv_data) { 1927b3f7fd54SArseny Krasnov err = transport->notify_recv_pre_block(vsk, target, recv_data); 1928b3f7fd54SArseny Krasnov if (err < 0) 1929b3f7fd54SArseny Krasnov break; 1930b3f7fd54SArseny Krasnov } 1931b3f7fd54SArseny Krasnov 1932b3f7fd54SArseny Krasnov release_sock(sk); 1933b3f7fd54SArseny Krasnov timeout = schedule_timeout(timeout); 1934b3f7fd54SArseny Krasnov lock_sock(sk); 1935b3f7fd54SArseny Krasnov 1936b3f7fd54SArseny Krasnov if (signal_pending(current)) { 1937b3f7fd54SArseny Krasnov err = sock_intr_errno(timeout); 1938b3f7fd54SArseny Krasnov break; 1939b3f7fd54SArseny Krasnov } else if (timeout == 0) { 1940b3f7fd54SArseny Krasnov err = -EAGAIN; 1941b3f7fd54SArseny Krasnov break; 1942b3f7fd54SArseny Krasnov } 1943b3f7fd54SArseny Krasnov } 1944b3f7fd54SArseny Krasnov 1945b3f7fd54SArseny Krasnov finish_wait(sk_sleep(sk), wait); 1946b3f7fd54SArseny Krasnov 1947b3f7fd54SArseny Krasnov if (err) 1948b3f7fd54SArseny Krasnov return err; 1949b3f7fd54SArseny Krasnov 1950b3f7fd54SArseny Krasnov /* Internal transport error when checking for available 1951b3f7fd54SArseny Krasnov * data. XXX This should be changed to a connection 1952b3f7fd54SArseny Krasnov * reset in a later change. 1953b3f7fd54SArseny Krasnov */ 1954b3f7fd54SArseny Krasnov if (data < 0) 1955b3f7fd54SArseny Krasnov return -ENOMEM; 1956b3f7fd54SArseny Krasnov 1957b3f7fd54SArseny Krasnov return data; 1958b3f7fd54SArseny Krasnov } 1959b3f7fd54SArseny Krasnov 196019c1b90eSArseny Krasnov static int __vsock_stream_recvmsg(struct sock *sk, struct msghdr *msg, 196119c1b90eSArseny Krasnov size_t len, int flags) 1962d021c344SAndy King { 1963d021c344SAndy King struct vsock_transport_recv_notify_data recv_data; 196419c1b90eSArseny Krasnov const struct vsock_transport *transport; 196519c1b90eSArseny Krasnov struct vsock_sock *vsk; 196619c1b90eSArseny Krasnov ssize_t copied; 196719c1b90eSArseny Krasnov size_t target; 196819c1b90eSArseny Krasnov long timeout; 196919c1b90eSArseny Krasnov int err; 1970d021c344SAndy King 1971d021c344SAndy King DEFINE_WAIT(wait); 1972d021c344SAndy King 1973d021c344SAndy King vsk = vsock_sk(sk); 1974c518adafSAlexander Popov transport = vsk->transport; 1975c518adafSAlexander Popov 1976d021c344SAndy King /* We must not copy less than target bytes into the user's buffer 1977d021c344SAndy King * before returning successfully, so we wait for the consume queue to 1978d021c344SAndy King * have that much data to consume before dequeueing. Note that this 1979d021c344SAndy King * makes it impossible to handle cases where target is greater than the 1980d021c344SAndy King * queue size. 1981d021c344SAndy King */ 1982d021c344SAndy King target = sock_rcvlowat(sk, flags & MSG_WAITALL, len); 1983d021c344SAndy King if (target >= transport->stream_rcvhiwat(vsk)) { 1984d021c344SAndy King err = -ENOMEM; 1985d021c344SAndy King goto out; 1986d021c344SAndy King } 1987d021c344SAndy King timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT); 1988d021c344SAndy King copied = 0; 1989d021c344SAndy King 1990d021c344SAndy King err = transport->notify_recv_init(vsk, target, &recv_data); 1991d021c344SAndy King if (err < 0) 1992d021c344SAndy King goto out; 1993d021c344SAndy King 1994d021c344SAndy King 1995d021c344SAndy King while (1) { 1996f7f9b5e7SClaudio Imbrenda ssize_t read; 1997f7f9b5e7SClaudio Imbrenda 19980de5b2e6SStefano Garzarella err = vsock_connectible_wait_data(sk, &wait, timeout, 19990de5b2e6SStefano Garzarella &recv_data, target); 2000b3f7fd54SArseny Krasnov if (err <= 0) 2001b3f7fd54SArseny Krasnov break; 2002d021c344SAndy King 2003b3f7fd54SArseny Krasnov err = transport->notify_recv_pre_dequeue(vsk, target, 2004b3f7fd54SArseny Krasnov &recv_data); 2005d021c344SAndy King if (err < 0) 2006d021c344SAndy King break; 2007d021c344SAndy King 2008b3f7fd54SArseny Krasnov read = transport->stream_dequeue(vsk, msg, len - copied, flags); 2009d021c344SAndy King if (read < 0) { 2010d021c344SAndy King err = -ENOMEM; 2011d021c344SAndy King break; 2012d021c344SAndy King } 2013d021c344SAndy King 2014d021c344SAndy King copied += read; 2015d021c344SAndy King 2016b3f7fd54SArseny Krasnov err = transport->notify_recv_post_dequeue(vsk, target, read, 2017d021c344SAndy King !(flags & MSG_PEEK), &recv_data); 2018d021c344SAndy King if (err < 0) 2019f7f9b5e7SClaudio Imbrenda goto out; 2020d021c344SAndy King 2021d021c344SAndy King if (read >= target || flags & MSG_PEEK) 2022d021c344SAndy King break; 2023d021c344SAndy King 2024d021c344SAndy King target -= read; 2025d021c344SAndy King } 2026d021c344SAndy King 2027d021c344SAndy King if (sk->sk_err) 2028d021c344SAndy King err = -sk->sk_err; 2029d021c344SAndy King else if (sk->sk_shutdown & RCV_SHUTDOWN) 2030d021c344SAndy King err = 0; 2031d021c344SAndy King 2032dedc58e0SIan Campbell if (copied > 0) 2033d021c344SAndy King err = copied; 2034d021c344SAndy King 2035d021c344SAndy King out: 203619c1b90eSArseny Krasnov return err; 203719c1b90eSArseny Krasnov } 203819c1b90eSArseny Krasnov 20399942c192SArseny Krasnov static int __vsock_seqpacket_recvmsg(struct sock *sk, struct msghdr *msg, 20409942c192SArseny Krasnov size_t len, int flags) 20419942c192SArseny Krasnov { 20429942c192SArseny Krasnov const struct vsock_transport *transport; 20439942c192SArseny Krasnov struct vsock_sock *vsk; 20448fc92b7cSArseny Krasnov ssize_t msg_len; 20459942c192SArseny Krasnov long timeout; 20469942c192SArseny Krasnov int err = 0; 20479942c192SArseny Krasnov DEFINE_WAIT(wait); 20489942c192SArseny Krasnov 20499942c192SArseny Krasnov vsk = vsock_sk(sk); 20509942c192SArseny Krasnov transport = vsk->transport; 20519942c192SArseny Krasnov 20529942c192SArseny Krasnov timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT); 20539942c192SArseny Krasnov 20540de5b2e6SStefano Garzarella err = vsock_connectible_wait_data(sk, &wait, timeout, NULL, 0); 20559942c192SArseny Krasnov if (err <= 0) 20569942c192SArseny Krasnov goto out; 20579942c192SArseny Krasnov 20588fc92b7cSArseny Krasnov msg_len = transport->seqpacket_dequeue(vsk, msg, flags); 20599942c192SArseny Krasnov 20608fc92b7cSArseny Krasnov if (msg_len < 0) { 20619942c192SArseny Krasnov err = -ENOMEM; 20629942c192SArseny Krasnov goto out; 20639942c192SArseny Krasnov } 20649942c192SArseny Krasnov 20659942c192SArseny Krasnov if (sk->sk_err) { 20669942c192SArseny Krasnov err = -sk->sk_err; 20679942c192SArseny Krasnov } else if (sk->sk_shutdown & RCV_SHUTDOWN) { 20689942c192SArseny Krasnov err = 0; 20699942c192SArseny Krasnov } else { 20709942c192SArseny Krasnov /* User sets MSG_TRUNC, so return real length of 20719942c192SArseny Krasnov * packet. 20729942c192SArseny Krasnov */ 20739942c192SArseny Krasnov if (flags & MSG_TRUNC) 20748fc92b7cSArseny Krasnov err = msg_len; 20759942c192SArseny Krasnov else 20769942c192SArseny Krasnov err = len - msg_data_left(msg); 20779942c192SArseny Krasnov 20789942c192SArseny Krasnov /* Always set MSG_TRUNC if real length of packet is 20799942c192SArseny Krasnov * bigger than user's buffer. 20809942c192SArseny Krasnov */ 20818fc92b7cSArseny Krasnov if (msg_len > len) 20829942c192SArseny Krasnov msg->msg_flags |= MSG_TRUNC; 20839942c192SArseny Krasnov } 20849942c192SArseny Krasnov 20859942c192SArseny Krasnov out: 20869942c192SArseny Krasnov return err; 20879942c192SArseny Krasnov } 20889942c192SArseny Krasnov 208919c1b90eSArseny Krasnov static int 209019c1b90eSArseny Krasnov vsock_connectible_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, 209119c1b90eSArseny Krasnov int flags) 209219c1b90eSArseny Krasnov { 209319c1b90eSArseny Krasnov struct sock *sk; 209419c1b90eSArseny Krasnov struct vsock_sock *vsk; 209519c1b90eSArseny Krasnov const struct vsock_transport *transport; 209619c1b90eSArseny Krasnov int err; 209719c1b90eSArseny Krasnov 209819c1b90eSArseny Krasnov sk = sock->sk; 209919c1b90eSArseny Krasnov vsk = vsock_sk(sk); 210019c1b90eSArseny Krasnov err = 0; 210119c1b90eSArseny Krasnov 210219c1b90eSArseny Krasnov lock_sock(sk); 210319c1b90eSArseny Krasnov 210419c1b90eSArseny Krasnov transport = vsk->transport; 210519c1b90eSArseny Krasnov 210619c1b90eSArseny Krasnov if (!transport || sk->sk_state != TCP_ESTABLISHED) { 210719c1b90eSArseny Krasnov /* Recvmsg is supposed to return 0 if a peer performs an 210819c1b90eSArseny Krasnov * orderly shutdown. Differentiate between that case and when a 210919c1b90eSArseny Krasnov * peer has not connected or a local shutdown occurred with the 211019c1b90eSArseny Krasnov * SOCK_DONE flag. 211119c1b90eSArseny Krasnov */ 211219c1b90eSArseny Krasnov if (sock_flag(sk, SOCK_DONE)) 211319c1b90eSArseny Krasnov err = 0; 211419c1b90eSArseny Krasnov else 211519c1b90eSArseny Krasnov err = -ENOTCONN; 211619c1b90eSArseny Krasnov 211719c1b90eSArseny Krasnov goto out; 211819c1b90eSArseny Krasnov } 211919c1b90eSArseny Krasnov 212019c1b90eSArseny Krasnov if (flags & MSG_OOB) { 212119c1b90eSArseny Krasnov err = -EOPNOTSUPP; 212219c1b90eSArseny Krasnov goto out; 212319c1b90eSArseny Krasnov } 212419c1b90eSArseny Krasnov 212519c1b90eSArseny Krasnov /* We don't check peer_shutdown flag here since peer may actually shut 212619c1b90eSArseny Krasnov * down, but there can be data in the queue that a local socket can 212719c1b90eSArseny Krasnov * receive. 212819c1b90eSArseny Krasnov */ 212919c1b90eSArseny Krasnov if (sk->sk_shutdown & RCV_SHUTDOWN) { 213019c1b90eSArseny Krasnov err = 0; 213119c1b90eSArseny Krasnov goto out; 213219c1b90eSArseny Krasnov } 213319c1b90eSArseny Krasnov 213419c1b90eSArseny Krasnov /* It is valid on Linux to pass in a zero-length receive buffer. This 213519c1b90eSArseny Krasnov * is not an error. We may as well bail out now. 213619c1b90eSArseny Krasnov */ 213719c1b90eSArseny Krasnov if (!len) { 213819c1b90eSArseny Krasnov err = 0; 213919c1b90eSArseny Krasnov goto out; 214019c1b90eSArseny Krasnov } 214119c1b90eSArseny Krasnov 21429942c192SArseny Krasnov if (sk->sk_type == SOCK_STREAM) 214319c1b90eSArseny Krasnov err = __vsock_stream_recvmsg(sk, msg, len, flags); 21449942c192SArseny Krasnov else 21459942c192SArseny Krasnov err = __vsock_seqpacket_recvmsg(sk, msg, len, flags); 214619c1b90eSArseny Krasnov 214719c1b90eSArseny Krasnov out: 2148d021c344SAndy King release_sock(sk); 2149d021c344SAndy King return err; 2150d021c344SAndy King } 2151d021c344SAndy King 2152e38f22c8SArseniy Krasnov static int vsock_set_rcvlowat(struct sock *sk, int val) 2153e38f22c8SArseniy Krasnov { 2154e38f22c8SArseniy Krasnov const struct vsock_transport *transport; 2155e38f22c8SArseniy Krasnov struct vsock_sock *vsk; 2156e38f22c8SArseniy Krasnov 2157e38f22c8SArseniy Krasnov vsk = vsock_sk(sk); 2158e38f22c8SArseniy Krasnov 2159e38f22c8SArseniy Krasnov if (val > vsk->buffer_size) 2160e38f22c8SArseniy Krasnov return -EINVAL; 2161e38f22c8SArseniy Krasnov 2162e38f22c8SArseniy Krasnov transport = vsk->transport; 2163e38f22c8SArseniy Krasnov 2164e38f22c8SArseniy Krasnov if (transport && transport->set_rcvlowat) 2165e38f22c8SArseniy Krasnov return transport->set_rcvlowat(vsk, val); 2166e38f22c8SArseniy Krasnov 2167e38f22c8SArseniy Krasnov WRITE_ONCE(sk->sk_rcvlowat, val ? : 1); 2168e38f22c8SArseniy Krasnov return 0; 2169e38f22c8SArseniy Krasnov } 2170e38f22c8SArseniy Krasnov 2171d021c344SAndy King static const struct proto_ops vsock_stream_ops = { 2172d021c344SAndy King .family = PF_VSOCK, 2173d021c344SAndy King .owner = THIS_MODULE, 2174d021c344SAndy King .release = vsock_release, 2175d021c344SAndy King .bind = vsock_bind, 2176a9e29e55SArseny Krasnov .connect = vsock_connect, 2177d021c344SAndy King .socketpair = sock_no_socketpair, 2178d021c344SAndy King .accept = vsock_accept, 2179d021c344SAndy King .getname = vsock_getname, 2180a11e1d43SLinus Torvalds .poll = vsock_poll, 2181d021c344SAndy King .ioctl = sock_no_ioctl, 2182d021c344SAndy King .listen = vsock_listen, 2183d021c344SAndy King .shutdown = vsock_shutdown, 2184a9e29e55SArseny Krasnov .setsockopt = vsock_connectible_setsockopt, 2185a9e29e55SArseny Krasnov .getsockopt = vsock_connectible_getsockopt, 2186a9e29e55SArseny Krasnov .sendmsg = vsock_connectible_sendmsg, 2187a9e29e55SArseny Krasnov .recvmsg = vsock_connectible_recvmsg, 2188d021c344SAndy King .mmap = sock_no_mmap, 2189d021c344SAndy King .sendpage = sock_no_sendpage, 2190e38f22c8SArseniy Krasnov .set_rcvlowat = vsock_set_rcvlowat, 2191d021c344SAndy King }; 2192d021c344SAndy King 21930798e78bSArseny Krasnov static const struct proto_ops vsock_seqpacket_ops = { 21940798e78bSArseny Krasnov .family = PF_VSOCK, 21950798e78bSArseny Krasnov .owner = THIS_MODULE, 21960798e78bSArseny Krasnov .release = vsock_release, 21970798e78bSArseny Krasnov .bind = vsock_bind, 21980798e78bSArseny Krasnov .connect = vsock_connect, 21990798e78bSArseny Krasnov .socketpair = sock_no_socketpair, 22000798e78bSArseny Krasnov .accept = vsock_accept, 22010798e78bSArseny Krasnov .getname = vsock_getname, 22020798e78bSArseny Krasnov .poll = vsock_poll, 22030798e78bSArseny Krasnov .ioctl = sock_no_ioctl, 22040798e78bSArseny Krasnov .listen = vsock_listen, 22050798e78bSArseny Krasnov .shutdown = vsock_shutdown, 22060798e78bSArseny Krasnov .setsockopt = vsock_connectible_setsockopt, 22070798e78bSArseny Krasnov .getsockopt = vsock_connectible_getsockopt, 22080798e78bSArseny Krasnov .sendmsg = vsock_connectible_sendmsg, 22090798e78bSArseny Krasnov .recvmsg = vsock_connectible_recvmsg, 22100798e78bSArseny Krasnov .mmap = sock_no_mmap, 22110798e78bSArseny Krasnov .sendpage = sock_no_sendpage, 22120798e78bSArseny Krasnov }; 22130798e78bSArseny Krasnov 2214d021c344SAndy King static int vsock_create(struct net *net, struct socket *sock, 2215d021c344SAndy King int protocol, int kern) 2216d021c344SAndy King { 2217c0cfa2d8SStefano Garzarella struct vsock_sock *vsk; 221855f3e149SStefano Garzarella struct sock *sk; 2219c0cfa2d8SStefano Garzarella int ret; 222055f3e149SStefano Garzarella 2221d021c344SAndy King if (!sock) 2222d021c344SAndy King return -EINVAL; 2223d021c344SAndy King 22246cf1c5fcSAndy King if (protocol && protocol != PF_VSOCK) 2225d021c344SAndy King return -EPROTONOSUPPORT; 2226d021c344SAndy King 2227d021c344SAndy King switch (sock->type) { 2228d021c344SAndy King case SOCK_DGRAM: 2229d021c344SAndy King sock->ops = &vsock_dgram_ops; 2230d021c344SAndy King break; 2231d021c344SAndy King case SOCK_STREAM: 2232d021c344SAndy King sock->ops = &vsock_stream_ops; 2233d021c344SAndy King break; 22340798e78bSArseny Krasnov case SOCK_SEQPACKET: 22350798e78bSArseny Krasnov sock->ops = &vsock_seqpacket_ops; 22360798e78bSArseny Krasnov break; 2237d021c344SAndy King default: 2238d021c344SAndy King return -ESOCKTNOSUPPORT; 2239d021c344SAndy King } 2240d021c344SAndy King 2241d021c344SAndy King sock->state = SS_UNCONNECTED; 2242d021c344SAndy King 224355f3e149SStefano Garzarella sk = __vsock_create(net, sock, NULL, GFP_KERNEL, 0, kern); 224455f3e149SStefano Garzarella if (!sk) 224555f3e149SStefano Garzarella return -ENOMEM; 224655f3e149SStefano Garzarella 2247c0cfa2d8SStefano Garzarella vsk = vsock_sk(sk); 2248c0cfa2d8SStefano Garzarella 2249c0cfa2d8SStefano Garzarella if (sock->type == SOCK_DGRAM) { 2250c0cfa2d8SStefano Garzarella ret = vsock_assign_transport(vsk, NULL); 2251c0cfa2d8SStefano Garzarella if (ret < 0) { 2252c0cfa2d8SStefano Garzarella sock_put(sk); 2253c0cfa2d8SStefano Garzarella return ret; 2254c0cfa2d8SStefano Garzarella } 2255c0cfa2d8SStefano Garzarella } 2256c0cfa2d8SStefano Garzarella 2257c0cfa2d8SStefano Garzarella vsock_insert_unbound(vsk); 225855f3e149SStefano Garzarella 225955f3e149SStefano Garzarella return 0; 2260d021c344SAndy King } 2261d021c344SAndy King 2262d021c344SAndy King static const struct net_proto_family vsock_family_ops = { 2263d021c344SAndy King .family = AF_VSOCK, 2264d021c344SAndy King .create = vsock_create, 2265d021c344SAndy King .owner = THIS_MODULE, 2266d021c344SAndy King }; 2267d021c344SAndy King 2268d021c344SAndy King static long vsock_dev_do_ioctl(struct file *filp, 2269d021c344SAndy King unsigned int cmd, void __user *ptr) 2270d021c344SAndy King { 2271d021c344SAndy King u32 __user *p = ptr; 2272c0cfa2d8SStefano Garzarella u32 cid = VMADDR_CID_ANY; 2273d021c344SAndy King int retval = 0; 2274d021c344SAndy King 2275d021c344SAndy King switch (cmd) { 2276d021c344SAndy King case IOCTL_VM_SOCKETS_GET_LOCAL_CID: 2277c0cfa2d8SStefano Garzarella /* To be compatible with the VMCI behavior, we prioritize the 2278c0cfa2d8SStefano Garzarella * guest CID instead of well-know host CID (VMADDR_CID_HOST). 2279c0cfa2d8SStefano Garzarella */ 2280c0cfa2d8SStefano Garzarella if (transport_g2h) 2281c0cfa2d8SStefano Garzarella cid = transport_g2h->get_local_cid(); 2282c0cfa2d8SStefano Garzarella else if (transport_h2g) 2283c0cfa2d8SStefano Garzarella cid = transport_h2g->get_local_cid(); 2284c0cfa2d8SStefano Garzarella 2285c0cfa2d8SStefano Garzarella if (put_user(cid, p) != 0) 2286d021c344SAndy King retval = -EFAULT; 2287d021c344SAndy King break; 2288d021c344SAndy King 2289d021c344SAndy King default: 2290c3e448cdSColin Ian King retval = -ENOIOCTLCMD; 2291d021c344SAndy King } 2292d021c344SAndy King 2293d021c344SAndy King return retval; 2294d021c344SAndy King } 2295d021c344SAndy King 2296d021c344SAndy King static long vsock_dev_ioctl(struct file *filp, 2297d021c344SAndy King unsigned int cmd, unsigned long arg) 2298d021c344SAndy King { 2299d021c344SAndy King return vsock_dev_do_ioctl(filp, cmd, (void __user *)arg); 2300d021c344SAndy King } 2301d021c344SAndy King 2302d021c344SAndy King #ifdef CONFIG_COMPAT 2303d021c344SAndy King static long vsock_dev_compat_ioctl(struct file *filp, 2304d021c344SAndy King unsigned int cmd, unsigned long arg) 2305d021c344SAndy King { 2306d021c344SAndy King return vsock_dev_do_ioctl(filp, cmd, compat_ptr(arg)); 2307d021c344SAndy King } 2308d021c344SAndy King #endif 2309d021c344SAndy King 2310d021c344SAndy King static const struct file_operations vsock_device_ops = { 2311d021c344SAndy King .owner = THIS_MODULE, 2312d021c344SAndy King .unlocked_ioctl = vsock_dev_ioctl, 2313d021c344SAndy King #ifdef CONFIG_COMPAT 2314d021c344SAndy King .compat_ioctl = vsock_dev_compat_ioctl, 2315d021c344SAndy King #endif 2316d021c344SAndy King .open = nonseekable_open, 2317d021c344SAndy King }; 2318d021c344SAndy King 2319d021c344SAndy King static struct miscdevice vsock_device = { 2320d021c344SAndy King .name = "vsock", 2321d021c344SAndy King .fops = &vsock_device_ops, 2322d021c344SAndy King }; 2323d021c344SAndy King 2324c0cfa2d8SStefano Garzarella static int __init vsock_init(void) 2325d021c344SAndy King { 2326c0cfa2d8SStefano Garzarella int err = 0; 23272c4a336eSAndy King 2328c0cfa2d8SStefano Garzarella vsock_init_tables(); 23292c4a336eSAndy King 2330c0cfa2d8SStefano Garzarella vsock_proto.owner = THIS_MODULE; 23316ad0b2f7SAsias He vsock_device.minor = MISC_DYNAMIC_MINOR; 2332d021c344SAndy King err = misc_register(&vsock_device); 2333d021c344SAndy King if (err) { 2334d021c344SAndy King pr_err("Failed to register misc device\n"); 2335f6a835bbSGao feng goto err_reset_transport; 2336d021c344SAndy King } 2337d021c344SAndy King 2338d021c344SAndy King err = proto_register(&vsock_proto, 1); /* we want our slab */ 2339d021c344SAndy King if (err) { 2340d021c344SAndy King pr_err("Cannot register vsock protocol\n"); 2341f6a835bbSGao feng goto err_deregister_misc; 2342d021c344SAndy King } 2343d021c344SAndy King 2344d021c344SAndy King err = sock_register(&vsock_family_ops); 2345d021c344SAndy King if (err) { 2346d021c344SAndy King pr_err("could not register af_vsock (%d) address family: %d\n", 2347d021c344SAndy King AF_VSOCK, err); 2348d021c344SAndy King goto err_unregister_proto; 2349d021c344SAndy King } 2350d021c344SAndy King 2351d021c344SAndy King return 0; 2352d021c344SAndy King 2353d021c344SAndy King err_unregister_proto: 2354d021c344SAndy King proto_unregister(&vsock_proto); 2355f6a835bbSGao feng err_deregister_misc: 2356d021c344SAndy King misc_deregister(&vsock_device); 2357f6a835bbSGao feng err_reset_transport: 2358d021c344SAndy King return err; 2359d021c344SAndy King } 2360d021c344SAndy King 2361c0cfa2d8SStefano Garzarella static void __exit vsock_exit(void) 2362d021c344SAndy King { 2363d021c344SAndy King misc_deregister(&vsock_device); 2364d021c344SAndy King sock_unregister(AF_VSOCK); 2365d021c344SAndy King proto_unregister(&vsock_proto); 2366d021c344SAndy King } 2367d021c344SAndy King 2368daabfbcaSStefano Garzarella const struct vsock_transport *vsock_core_get_transport(struct vsock_sock *vsk) 23690b01aeb3SStefan Hajnoczi { 2370daabfbcaSStefano Garzarella return vsk->transport; 23710b01aeb3SStefan Hajnoczi } 23720b01aeb3SStefan Hajnoczi EXPORT_SYMBOL_GPL(vsock_core_get_transport); 23730b01aeb3SStefan Hajnoczi 2374c0cfa2d8SStefano Garzarella int vsock_core_register(const struct vsock_transport *t, int features) 237505e489b1SStefan Hajnoczi { 23760e121905SStefano Garzarella const struct vsock_transport *t_h2g, *t_g2h, *t_dgram, *t_local; 2377c0cfa2d8SStefano Garzarella int err = mutex_lock_interruptible(&vsock_register_mutex); 2378c0cfa2d8SStefano Garzarella 2379c0cfa2d8SStefano Garzarella if (err) 2380c0cfa2d8SStefano Garzarella return err; 2381c0cfa2d8SStefano Garzarella 2382c0cfa2d8SStefano Garzarella t_h2g = transport_h2g; 2383c0cfa2d8SStefano Garzarella t_g2h = transport_g2h; 2384c0cfa2d8SStefano Garzarella t_dgram = transport_dgram; 23850e121905SStefano Garzarella t_local = transport_local; 2386c0cfa2d8SStefano Garzarella 2387c0cfa2d8SStefano Garzarella if (features & VSOCK_TRANSPORT_F_H2G) { 2388c0cfa2d8SStefano Garzarella if (t_h2g) { 2389c0cfa2d8SStefano Garzarella err = -EBUSY; 2390c0cfa2d8SStefano Garzarella goto err_busy; 2391c0cfa2d8SStefano Garzarella } 2392c0cfa2d8SStefano Garzarella t_h2g = t; 239305e489b1SStefan Hajnoczi } 239405e489b1SStefan Hajnoczi 2395c0cfa2d8SStefano Garzarella if (features & VSOCK_TRANSPORT_F_G2H) { 2396c0cfa2d8SStefano Garzarella if (t_g2h) { 2397c0cfa2d8SStefano Garzarella err = -EBUSY; 2398c0cfa2d8SStefano Garzarella goto err_busy; 2399c0cfa2d8SStefano Garzarella } 2400c0cfa2d8SStefano Garzarella t_g2h = t; 2401c0cfa2d8SStefano Garzarella } 2402c0cfa2d8SStefano Garzarella 2403c0cfa2d8SStefano Garzarella if (features & VSOCK_TRANSPORT_F_DGRAM) { 2404c0cfa2d8SStefano Garzarella if (t_dgram) { 2405c0cfa2d8SStefano Garzarella err = -EBUSY; 2406c0cfa2d8SStefano Garzarella goto err_busy; 2407c0cfa2d8SStefano Garzarella } 2408c0cfa2d8SStefano Garzarella t_dgram = t; 2409c0cfa2d8SStefano Garzarella } 2410c0cfa2d8SStefano Garzarella 24110e121905SStefano Garzarella if (features & VSOCK_TRANSPORT_F_LOCAL) { 24120e121905SStefano Garzarella if (t_local) { 24130e121905SStefano Garzarella err = -EBUSY; 24140e121905SStefano Garzarella goto err_busy; 24150e121905SStefano Garzarella } 24160e121905SStefano Garzarella t_local = t; 24170e121905SStefano Garzarella } 24180e121905SStefano Garzarella 2419c0cfa2d8SStefano Garzarella transport_h2g = t_h2g; 2420c0cfa2d8SStefano Garzarella transport_g2h = t_g2h; 2421c0cfa2d8SStefano Garzarella transport_dgram = t_dgram; 24220e121905SStefano Garzarella transport_local = t_local; 2423c0cfa2d8SStefano Garzarella 2424c0cfa2d8SStefano Garzarella err_busy: 2425c0cfa2d8SStefano Garzarella mutex_unlock(&vsock_register_mutex); 2426c0cfa2d8SStefano Garzarella return err; 2427c0cfa2d8SStefano Garzarella } 2428c0cfa2d8SStefano Garzarella EXPORT_SYMBOL_GPL(vsock_core_register); 2429c0cfa2d8SStefano Garzarella 2430c0cfa2d8SStefano Garzarella void vsock_core_unregister(const struct vsock_transport *t) 2431c0cfa2d8SStefano Garzarella { 2432c0cfa2d8SStefano Garzarella mutex_lock(&vsock_register_mutex); 2433c0cfa2d8SStefano Garzarella 2434c0cfa2d8SStefano Garzarella if (transport_h2g == t) 2435c0cfa2d8SStefano Garzarella transport_h2g = NULL; 2436c0cfa2d8SStefano Garzarella 2437c0cfa2d8SStefano Garzarella if (transport_g2h == t) 2438c0cfa2d8SStefano Garzarella transport_g2h = NULL; 2439c0cfa2d8SStefano Garzarella 2440c0cfa2d8SStefano Garzarella if (transport_dgram == t) 2441c0cfa2d8SStefano Garzarella transport_dgram = NULL; 2442c0cfa2d8SStefano Garzarella 24430e121905SStefano Garzarella if (transport_local == t) 24440e121905SStefano Garzarella transport_local = NULL; 24450e121905SStefano Garzarella 2446c0cfa2d8SStefano Garzarella mutex_unlock(&vsock_register_mutex); 2447c0cfa2d8SStefano Garzarella } 2448c0cfa2d8SStefano Garzarella EXPORT_SYMBOL_GPL(vsock_core_unregister); 2449c0cfa2d8SStefano Garzarella 2450c0cfa2d8SStefano Garzarella module_init(vsock_init); 245105e489b1SStefan Hajnoczi module_exit(vsock_exit); 2452c1eef220SCong Wang 2453d021c344SAndy King MODULE_AUTHOR("VMware, Inc."); 2454d021c344SAndy King MODULE_DESCRIPTION("VMware Virtual Socket Family"); 24551190cfdbSJorgen Hansen MODULE_VERSION("1.0.2.0-k"); 2456d021c344SAndy King MODULE_LICENSE("GPL v2"); 2457