1 // SPDX-License-Identifier: GPL-2.0-only 2 /* 3 * VMware vSockets Driver 4 * 5 * Copyright (C) 2007-2013 VMware, Inc. All rights reserved. 6 */ 7 8 /* Implementation notes: 9 * 10 * - There are two kinds of sockets: those created by user action (such as 11 * calling socket(2)) and those created by incoming connection request packets. 12 * 13 * - There are two "global" tables, one for bound sockets (sockets that have 14 * specified an address that they are responsible for) and one for connected 15 * sockets (sockets that have established a connection with another socket). 16 * These tables are "global" in that all sockets on the system are placed 17 * within them. - Note, though, that the bound table contains an extra entry 18 * for a list of unbound sockets and SOCK_DGRAM sockets will always remain in 19 * that list. The bound table is used solely for lookup of sockets when packets 20 * are received and that's not necessary for SOCK_DGRAM sockets since we create 21 * a datagram handle for each and need not perform a lookup. Keeping SOCK_DGRAM 22 * sockets out of the bound hash buckets will reduce the chance of collisions 23 * when looking for SOCK_STREAM sockets and prevents us from having to check the 24 * socket type in the hash table lookups. 25 * 26 * - Sockets created by user action will either be "client" sockets that 27 * initiate a connection or "server" sockets that listen for connections; we do 28 * not support simultaneous connects (two "client" sockets connecting). 29 * 30 * - "Server" sockets are referred to as listener sockets throughout this 31 * implementation because they are in the TCP_LISTEN state. When a 32 * connection request is received (the second kind of socket mentioned above), 33 * we create a new socket and refer to it as a pending socket. These pending 34 * sockets are placed on the pending connection list of the listener socket. 35 * When future packets are received for the address the listener socket is 36 * bound to, we check if the source of the packet is from one that has an 37 * existing pending connection. If it does, we process the packet for the 38 * pending socket. When that socket reaches the connected state, it is removed 39 * from the listener socket's pending list and enqueued in the listener 40 * socket's accept queue. Callers of accept(2) will accept connected sockets 41 * from the listener socket's accept queue. If the socket cannot be accepted 42 * for some reason then it is marked rejected. Once the connection is 43 * accepted, it is owned by the user process and the responsibility for cleanup 44 * falls with that user process. 45 * 46 * - It is possible that these pending sockets will never reach the connected 47 * state; in fact, we may never receive another packet after the connection 48 * request. Because of this, we must schedule a cleanup function to run in the 49 * future, after some amount of time passes where a connection should have been 50 * established. This function ensures that the socket is off all lists so it 51 * cannot be retrieved, then drops all references to the socket so it is cleaned 52 * up (sock_put() -> sk_free() -> our sk_destruct implementation). Note this 53 * function will also cleanup rejected sockets, those that reach the connected 54 * state but leave it before they have been accepted. 55 * 56 * - Lock ordering for pending or accept queue sockets is: 57 * 58 * lock_sock(listener); 59 * lock_sock_nested(pending, SINGLE_DEPTH_NESTING); 60 * 61 * Using explicit nested locking keeps lockdep happy since normally only one 62 * lock of a given class may be taken at a time. 63 * 64 * - Sockets created by user action will be cleaned up when the user process 65 * calls close(2), causing our release implementation to be called. Our release 66 * implementation will perform some cleanup then drop the last reference so our 67 * sk_destruct implementation is invoked. Our sk_destruct implementation will 68 * perform additional cleanup that's common for both types of sockets. 69 * 70 * - A socket's reference count is what ensures that the structure won't be 71 * freed. Each entry in a list (such as the "global" bound and connected tables 72 * and the listener socket's pending list and connected queue) ensures a 73 * reference. When we defer work until process context and pass a socket as our 74 * argument, we must ensure the reference count is increased to ensure the 75 * socket isn't freed before the function is run; the deferred function will 76 * then drop the reference. 77 * 78 * - sk->sk_state uses the TCP state constants because they are widely used by 79 * other address families and exposed to userspace tools like ss(8): 80 * 81 * TCP_CLOSE - unconnected 82 * TCP_SYN_SENT - connecting 83 * TCP_ESTABLISHED - connected 84 * TCP_CLOSING - disconnecting 85 * TCP_LISTEN - listening 86 */ 87 88 #include <linux/types.h> 89 #include <linux/bitops.h> 90 #include <linux/cred.h> 91 #include <linux/init.h> 92 #include <linux/io.h> 93 #include <linux/kernel.h> 94 #include <linux/sched/signal.h> 95 #include <linux/kmod.h> 96 #include <linux/list.h> 97 #include <linux/miscdevice.h> 98 #include <linux/module.h> 99 #include <linux/mutex.h> 100 #include <linux/net.h> 101 #include <linux/poll.h> 102 #include <linux/random.h> 103 #include <linux/skbuff.h> 104 #include <linux/smp.h> 105 #include <linux/socket.h> 106 #include <linux/stddef.h> 107 #include <linux/unistd.h> 108 #include <linux/wait.h> 109 #include <linux/workqueue.h> 110 #include <net/sock.h> 111 #include <net/af_vsock.h> 112 113 static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr); 114 static void vsock_sk_destruct(struct sock *sk); 115 static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb); 116 117 /* Protocol family. */ 118 static struct proto vsock_proto = { 119 .name = "AF_VSOCK", 120 .owner = THIS_MODULE, 121 .obj_size = sizeof(struct vsock_sock), 122 }; 123 124 /* The default peer timeout indicates how long we will wait for a peer response 125 * to a control message. 126 */ 127 #define VSOCK_DEFAULT_CONNECT_TIMEOUT (2 * HZ) 128 129 #define VSOCK_DEFAULT_BUFFER_SIZE (1024 * 256) 130 #define VSOCK_DEFAULT_BUFFER_MAX_SIZE (1024 * 256) 131 #define VSOCK_DEFAULT_BUFFER_MIN_SIZE 128 132 133 /* Transport used for host->guest communication */ 134 static const struct vsock_transport *transport_h2g; 135 /* Transport used for guest->host communication */ 136 static const struct vsock_transport *transport_g2h; 137 /* Transport used for DGRAM communication */ 138 static const struct vsock_transport *transport_dgram; 139 /* Transport used for local communication */ 140 static const struct vsock_transport *transport_local; 141 static DEFINE_MUTEX(vsock_register_mutex); 142 143 /**** UTILS ****/ 144 145 /* Each bound VSocket is stored in the bind hash table and each connected 146 * VSocket is stored in the connected hash table. 147 * 148 * Unbound sockets are all put on the same list attached to the end of the hash 149 * table (vsock_unbound_sockets). Bound sockets are added to the hash table in 150 * the bucket that their local address hashes to (vsock_bound_sockets(addr) 151 * represents the list that addr hashes to). 152 * 153 * Specifically, we initialize the vsock_bind_table array to a size of 154 * VSOCK_HASH_SIZE + 1 so that vsock_bind_table[0] through 155 * vsock_bind_table[VSOCK_HASH_SIZE - 1] are for bound sockets and 156 * vsock_bind_table[VSOCK_HASH_SIZE] is for unbound sockets. The hash function 157 * mods with VSOCK_HASH_SIZE to ensure this. 158 */ 159 #define MAX_PORT_RETRIES 24 160 161 #define VSOCK_HASH(addr) ((addr)->svm_port % VSOCK_HASH_SIZE) 162 #define vsock_bound_sockets(addr) (&vsock_bind_table[VSOCK_HASH(addr)]) 163 #define vsock_unbound_sockets (&vsock_bind_table[VSOCK_HASH_SIZE]) 164 165 /* XXX This can probably be implemented in a better way. */ 166 #define VSOCK_CONN_HASH(src, dst) \ 167 (((src)->svm_cid ^ (dst)->svm_port) % VSOCK_HASH_SIZE) 168 #define vsock_connected_sockets(src, dst) \ 169 (&vsock_connected_table[VSOCK_CONN_HASH(src, dst)]) 170 #define vsock_connected_sockets_vsk(vsk) \ 171 vsock_connected_sockets(&(vsk)->remote_addr, &(vsk)->local_addr) 172 173 struct list_head vsock_bind_table[VSOCK_HASH_SIZE + 1]; 174 EXPORT_SYMBOL_GPL(vsock_bind_table); 175 struct list_head vsock_connected_table[VSOCK_HASH_SIZE]; 176 EXPORT_SYMBOL_GPL(vsock_connected_table); 177 DEFINE_SPINLOCK(vsock_table_lock); 178 EXPORT_SYMBOL_GPL(vsock_table_lock); 179 180 /* Autobind this socket to the local address if necessary. */ 181 static int vsock_auto_bind(struct vsock_sock *vsk) 182 { 183 struct sock *sk = sk_vsock(vsk); 184 struct sockaddr_vm local_addr; 185 186 if (vsock_addr_bound(&vsk->local_addr)) 187 return 0; 188 vsock_addr_init(&local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 189 return __vsock_bind(sk, &local_addr); 190 } 191 192 static void vsock_init_tables(void) 193 { 194 int i; 195 196 for (i = 0; i < ARRAY_SIZE(vsock_bind_table); i++) 197 INIT_LIST_HEAD(&vsock_bind_table[i]); 198 199 for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) 200 INIT_LIST_HEAD(&vsock_connected_table[i]); 201 } 202 203 static void __vsock_insert_bound(struct list_head *list, 204 struct vsock_sock *vsk) 205 { 206 sock_hold(&vsk->sk); 207 list_add(&vsk->bound_table, list); 208 } 209 210 static void __vsock_insert_connected(struct list_head *list, 211 struct vsock_sock *vsk) 212 { 213 sock_hold(&vsk->sk); 214 list_add(&vsk->connected_table, list); 215 } 216 217 static void __vsock_remove_bound(struct vsock_sock *vsk) 218 { 219 list_del_init(&vsk->bound_table); 220 sock_put(&vsk->sk); 221 } 222 223 static void __vsock_remove_connected(struct vsock_sock *vsk) 224 { 225 list_del_init(&vsk->connected_table); 226 sock_put(&vsk->sk); 227 } 228 229 static struct sock *__vsock_find_bound_socket(struct sockaddr_vm *addr) 230 { 231 struct vsock_sock *vsk; 232 233 list_for_each_entry(vsk, vsock_bound_sockets(addr), bound_table) { 234 if (vsock_addr_equals_addr(addr, &vsk->local_addr)) 235 return sk_vsock(vsk); 236 237 if (addr->svm_port == vsk->local_addr.svm_port && 238 (vsk->local_addr.svm_cid == VMADDR_CID_ANY || 239 addr->svm_cid == VMADDR_CID_ANY)) 240 return sk_vsock(vsk); 241 } 242 243 return NULL; 244 } 245 246 static struct sock *__vsock_find_connected_socket(struct sockaddr_vm *src, 247 struct sockaddr_vm *dst) 248 { 249 struct vsock_sock *vsk; 250 251 list_for_each_entry(vsk, vsock_connected_sockets(src, dst), 252 connected_table) { 253 if (vsock_addr_equals_addr(src, &vsk->remote_addr) && 254 dst->svm_port == vsk->local_addr.svm_port) { 255 return sk_vsock(vsk); 256 } 257 } 258 259 return NULL; 260 } 261 262 static void vsock_insert_unbound(struct vsock_sock *vsk) 263 { 264 spin_lock_bh(&vsock_table_lock); 265 __vsock_insert_bound(vsock_unbound_sockets, vsk); 266 spin_unlock_bh(&vsock_table_lock); 267 } 268 269 void vsock_insert_connected(struct vsock_sock *vsk) 270 { 271 struct list_head *list = vsock_connected_sockets( 272 &vsk->remote_addr, &vsk->local_addr); 273 274 spin_lock_bh(&vsock_table_lock); 275 __vsock_insert_connected(list, vsk); 276 spin_unlock_bh(&vsock_table_lock); 277 } 278 EXPORT_SYMBOL_GPL(vsock_insert_connected); 279 280 void vsock_remove_bound(struct vsock_sock *vsk) 281 { 282 spin_lock_bh(&vsock_table_lock); 283 if (__vsock_in_bound_table(vsk)) 284 __vsock_remove_bound(vsk); 285 spin_unlock_bh(&vsock_table_lock); 286 } 287 EXPORT_SYMBOL_GPL(vsock_remove_bound); 288 289 void vsock_remove_connected(struct vsock_sock *vsk) 290 { 291 spin_lock_bh(&vsock_table_lock); 292 if (__vsock_in_connected_table(vsk)) 293 __vsock_remove_connected(vsk); 294 spin_unlock_bh(&vsock_table_lock); 295 } 296 EXPORT_SYMBOL_GPL(vsock_remove_connected); 297 298 struct sock *vsock_find_bound_socket(struct sockaddr_vm *addr) 299 { 300 struct sock *sk; 301 302 spin_lock_bh(&vsock_table_lock); 303 sk = __vsock_find_bound_socket(addr); 304 if (sk) 305 sock_hold(sk); 306 307 spin_unlock_bh(&vsock_table_lock); 308 309 return sk; 310 } 311 EXPORT_SYMBOL_GPL(vsock_find_bound_socket); 312 313 struct sock *vsock_find_connected_socket(struct sockaddr_vm *src, 314 struct sockaddr_vm *dst) 315 { 316 struct sock *sk; 317 318 spin_lock_bh(&vsock_table_lock); 319 sk = __vsock_find_connected_socket(src, dst); 320 if (sk) 321 sock_hold(sk); 322 323 spin_unlock_bh(&vsock_table_lock); 324 325 return sk; 326 } 327 EXPORT_SYMBOL_GPL(vsock_find_connected_socket); 328 329 void vsock_remove_sock(struct vsock_sock *vsk) 330 { 331 vsock_remove_bound(vsk); 332 vsock_remove_connected(vsk); 333 } 334 EXPORT_SYMBOL_GPL(vsock_remove_sock); 335 336 void vsock_for_each_connected_socket(void (*fn)(struct sock *sk)) 337 { 338 int i; 339 340 spin_lock_bh(&vsock_table_lock); 341 342 for (i = 0; i < ARRAY_SIZE(vsock_connected_table); i++) { 343 struct vsock_sock *vsk; 344 list_for_each_entry(vsk, &vsock_connected_table[i], 345 connected_table) 346 fn(sk_vsock(vsk)); 347 } 348 349 spin_unlock_bh(&vsock_table_lock); 350 } 351 EXPORT_SYMBOL_GPL(vsock_for_each_connected_socket); 352 353 void vsock_add_pending(struct sock *listener, struct sock *pending) 354 { 355 struct vsock_sock *vlistener; 356 struct vsock_sock *vpending; 357 358 vlistener = vsock_sk(listener); 359 vpending = vsock_sk(pending); 360 361 sock_hold(pending); 362 sock_hold(listener); 363 list_add_tail(&vpending->pending_links, &vlistener->pending_links); 364 } 365 EXPORT_SYMBOL_GPL(vsock_add_pending); 366 367 void vsock_remove_pending(struct sock *listener, struct sock *pending) 368 { 369 struct vsock_sock *vpending = vsock_sk(pending); 370 371 list_del_init(&vpending->pending_links); 372 sock_put(listener); 373 sock_put(pending); 374 } 375 EXPORT_SYMBOL_GPL(vsock_remove_pending); 376 377 void vsock_enqueue_accept(struct sock *listener, struct sock *connected) 378 { 379 struct vsock_sock *vlistener; 380 struct vsock_sock *vconnected; 381 382 vlistener = vsock_sk(listener); 383 vconnected = vsock_sk(connected); 384 385 sock_hold(connected); 386 sock_hold(listener); 387 list_add_tail(&vconnected->accept_queue, &vlistener->accept_queue); 388 } 389 EXPORT_SYMBOL_GPL(vsock_enqueue_accept); 390 391 static bool vsock_use_local_transport(unsigned int remote_cid) 392 { 393 if (!transport_local) 394 return false; 395 396 if (remote_cid == VMADDR_CID_LOCAL) 397 return true; 398 399 if (transport_g2h) { 400 return remote_cid == transport_g2h->get_local_cid(); 401 } else { 402 return remote_cid == VMADDR_CID_HOST; 403 } 404 } 405 406 static void vsock_deassign_transport(struct vsock_sock *vsk) 407 { 408 if (!vsk->transport) 409 return; 410 411 vsk->transport->destruct(vsk); 412 module_put(vsk->transport->module); 413 vsk->transport = NULL; 414 } 415 416 /* Assign a transport to a socket and call the .init transport callback. 417 * 418 * Note: for stream socket this must be called when vsk->remote_addr is set 419 * (e.g. during the connect() or when a connection request on a listener 420 * socket is received). 421 * The vsk->remote_addr is used to decide which transport to use: 422 * - remote CID == VMADDR_CID_LOCAL or g2h->local_cid or VMADDR_CID_HOST if 423 * g2h is not loaded, will use local transport; 424 * - remote CID <= VMADDR_CID_HOST will use guest->host transport; 425 * - remote CID > VMADDR_CID_HOST will use host->guest transport; 426 */ 427 int vsock_assign_transport(struct vsock_sock *vsk, struct vsock_sock *psk) 428 { 429 const struct vsock_transport *new_transport; 430 struct sock *sk = sk_vsock(vsk); 431 unsigned int remote_cid = vsk->remote_addr.svm_cid; 432 int ret; 433 434 /* If the packet is coming with the source and destination CIDs higher 435 * than VMADDR_CID_HOST, then a vsock channel where all the packets are 436 * forwarded to the host should be established. Then the host will 437 * need to forward the packets to the guest. 438 * 439 * The flag is set on the (listen) receive path (psk is not NULL). On 440 * the connect path the flag can be set by the user space application. 441 */ 442 if (psk && vsk->local_addr.svm_cid > VMADDR_CID_HOST && 443 vsk->remote_addr.svm_cid > VMADDR_CID_HOST) 444 vsk->remote_addr.svm_flags |= VMADDR_FLAG_TO_HOST; 445 446 switch (sk->sk_type) { 447 case SOCK_DGRAM: 448 new_transport = transport_dgram; 449 break; 450 case SOCK_STREAM: 451 if (vsock_use_local_transport(remote_cid)) 452 new_transport = transport_local; 453 else if (remote_cid <= VMADDR_CID_HOST || !transport_h2g) 454 new_transport = transport_g2h; 455 else 456 new_transport = transport_h2g; 457 break; 458 default: 459 return -ESOCKTNOSUPPORT; 460 } 461 462 if (vsk->transport) { 463 if (vsk->transport == new_transport) 464 return 0; 465 466 /* transport->release() must be called with sock lock acquired. 467 * This path can only be taken during vsock_stream_connect(), 468 * where we have already held the sock lock. 469 * In the other cases, this function is called on a new socket 470 * which is not assigned to any transport. 471 */ 472 vsk->transport->release(vsk); 473 vsock_deassign_transport(vsk); 474 } 475 476 /* We increase the module refcnt to prevent the transport unloading 477 * while there are open sockets assigned to it. 478 */ 479 if (!new_transport || !try_module_get(new_transport->module)) 480 return -ENODEV; 481 482 ret = new_transport->init(vsk, psk); 483 if (ret) { 484 module_put(new_transport->module); 485 return ret; 486 } 487 488 vsk->transport = new_transport; 489 490 return 0; 491 } 492 EXPORT_SYMBOL_GPL(vsock_assign_transport); 493 494 bool vsock_find_cid(unsigned int cid) 495 { 496 if (transport_g2h && cid == transport_g2h->get_local_cid()) 497 return true; 498 499 if (transport_h2g && cid == VMADDR_CID_HOST) 500 return true; 501 502 if (transport_local && cid == VMADDR_CID_LOCAL) 503 return true; 504 505 return false; 506 } 507 EXPORT_SYMBOL_GPL(vsock_find_cid); 508 509 static struct sock *vsock_dequeue_accept(struct sock *listener) 510 { 511 struct vsock_sock *vlistener; 512 struct vsock_sock *vconnected; 513 514 vlistener = vsock_sk(listener); 515 516 if (list_empty(&vlistener->accept_queue)) 517 return NULL; 518 519 vconnected = list_entry(vlistener->accept_queue.next, 520 struct vsock_sock, accept_queue); 521 522 list_del_init(&vconnected->accept_queue); 523 sock_put(listener); 524 /* The caller will need a reference on the connected socket so we let 525 * it call sock_put(). 526 */ 527 528 return sk_vsock(vconnected); 529 } 530 531 static bool vsock_is_accept_queue_empty(struct sock *sk) 532 { 533 struct vsock_sock *vsk = vsock_sk(sk); 534 return list_empty(&vsk->accept_queue); 535 } 536 537 static bool vsock_is_pending(struct sock *sk) 538 { 539 struct vsock_sock *vsk = vsock_sk(sk); 540 return !list_empty(&vsk->pending_links); 541 } 542 543 static int vsock_send_shutdown(struct sock *sk, int mode) 544 { 545 struct vsock_sock *vsk = vsock_sk(sk); 546 547 if (!vsk->transport) 548 return -ENODEV; 549 550 return vsk->transport->shutdown(vsk, mode); 551 } 552 553 static void vsock_pending_work(struct work_struct *work) 554 { 555 struct sock *sk; 556 struct sock *listener; 557 struct vsock_sock *vsk; 558 bool cleanup; 559 560 vsk = container_of(work, struct vsock_sock, pending_work.work); 561 sk = sk_vsock(vsk); 562 listener = vsk->listener; 563 cleanup = true; 564 565 lock_sock(listener); 566 lock_sock_nested(sk, SINGLE_DEPTH_NESTING); 567 568 if (vsock_is_pending(sk)) { 569 vsock_remove_pending(listener, sk); 570 571 sk_acceptq_removed(listener); 572 } else if (!vsk->rejected) { 573 /* We are not on the pending list and accept() did not reject 574 * us, so we must have been accepted by our user process. We 575 * just need to drop our references to the sockets and be on 576 * our way. 577 */ 578 cleanup = false; 579 goto out; 580 } 581 582 /* We need to remove ourself from the global connected sockets list so 583 * incoming packets can't find this socket, and to reduce the reference 584 * count. 585 */ 586 vsock_remove_connected(vsk); 587 588 sk->sk_state = TCP_CLOSE; 589 590 out: 591 release_sock(sk); 592 release_sock(listener); 593 if (cleanup) 594 sock_put(sk); 595 596 sock_put(sk); 597 sock_put(listener); 598 } 599 600 /**** SOCKET OPERATIONS ****/ 601 602 static int __vsock_bind_stream(struct vsock_sock *vsk, 603 struct sockaddr_vm *addr) 604 { 605 static u32 port; 606 struct sockaddr_vm new_addr; 607 608 if (!port) 609 port = LAST_RESERVED_PORT + 1 + 610 prandom_u32_max(U32_MAX - LAST_RESERVED_PORT); 611 612 vsock_addr_init(&new_addr, addr->svm_cid, addr->svm_port); 613 614 if (addr->svm_port == VMADDR_PORT_ANY) { 615 bool found = false; 616 unsigned int i; 617 618 for (i = 0; i < MAX_PORT_RETRIES; i++) { 619 if (port <= LAST_RESERVED_PORT) 620 port = LAST_RESERVED_PORT + 1; 621 622 new_addr.svm_port = port++; 623 624 if (!__vsock_find_bound_socket(&new_addr)) { 625 found = true; 626 break; 627 } 628 } 629 630 if (!found) 631 return -EADDRNOTAVAIL; 632 } else { 633 /* If port is in reserved range, ensure caller 634 * has necessary privileges. 635 */ 636 if (addr->svm_port <= LAST_RESERVED_PORT && 637 !capable(CAP_NET_BIND_SERVICE)) { 638 return -EACCES; 639 } 640 641 if (__vsock_find_bound_socket(&new_addr)) 642 return -EADDRINUSE; 643 } 644 645 vsock_addr_init(&vsk->local_addr, new_addr.svm_cid, new_addr.svm_port); 646 647 /* Remove stream sockets from the unbound list and add them to the hash 648 * table for easy lookup by its address. The unbound list is simply an 649 * extra entry at the end of the hash table, a trick used by AF_UNIX. 650 */ 651 __vsock_remove_bound(vsk); 652 __vsock_insert_bound(vsock_bound_sockets(&vsk->local_addr), vsk); 653 654 return 0; 655 } 656 657 static int __vsock_bind_dgram(struct vsock_sock *vsk, 658 struct sockaddr_vm *addr) 659 { 660 return vsk->transport->dgram_bind(vsk, addr); 661 } 662 663 static int __vsock_bind(struct sock *sk, struct sockaddr_vm *addr) 664 { 665 struct vsock_sock *vsk = vsock_sk(sk); 666 int retval; 667 668 /* First ensure this socket isn't already bound. */ 669 if (vsock_addr_bound(&vsk->local_addr)) 670 return -EINVAL; 671 672 /* Now bind to the provided address or select appropriate values if 673 * none are provided (VMADDR_CID_ANY and VMADDR_PORT_ANY). Note that 674 * like AF_INET prevents binding to a non-local IP address (in most 675 * cases), we only allow binding to a local CID. 676 */ 677 if (addr->svm_cid != VMADDR_CID_ANY && !vsock_find_cid(addr->svm_cid)) 678 return -EADDRNOTAVAIL; 679 680 switch (sk->sk_socket->type) { 681 case SOCK_STREAM: 682 spin_lock_bh(&vsock_table_lock); 683 retval = __vsock_bind_stream(vsk, addr); 684 spin_unlock_bh(&vsock_table_lock); 685 break; 686 687 case SOCK_DGRAM: 688 retval = __vsock_bind_dgram(vsk, addr); 689 break; 690 691 default: 692 retval = -EINVAL; 693 break; 694 } 695 696 return retval; 697 } 698 699 static void vsock_connect_timeout(struct work_struct *work); 700 701 static struct sock *__vsock_create(struct net *net, 702 struct socket *sock, 703 struct sock *parent, 704 gfp_t priority, 705 unsigned short type, 706 int kern) 707 { 708 struct sock *sk; 709 struct vsock_sock *psk; 710 struct vsock_sock *vsk; 711 712 sk = sk_alloc(net, AF_VSOCK, priority, &vsock_proto, kern); 713 if (!sk) 714 return NULL; 715 716 sock_init_data(sock, sk); 717 718 /* sk->sk_type is normally set in sock_init_data, but only if sock is 719 * non-NULL. We make sure that our sockets always have a type by 720 * setting it here if needed. 721 */ 722 if (!sock) 723 sk->sk_type = type; 724 725 vsk = vsock_sk(sk); 726 vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 727 vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 728 729 sk->sk_destruct = vsock_sk_destruct; 730 sk->sk_backlog_rcv = vsock_queue_rcv_skb; 731 sock_reset_flag(sk, SOCK_DONE); 732 733 INIT_LIST_HEAD(&vsk->bound_table); 734 INIT_LIST_HEAD(&vsk->connected_table); 735 vsk->listener = NULL; 736 INIT_LIST_HEAD(&vsk->pending_links); 737 INIT_LIST_HEAD(&vsk->accept_queue); 738 vsk->rejected = false; 739 vsk->sent_request = false; 740 vsk->ignore_connecting_rst = false; 741 vsk->peer_shutdown = 0; 742 INIT_DELAYED_WORK(&vsk->connect_work, vsock_connect_timeout); 743 INIT_DELAYED_WORK(&vsk->pending_work, vsock_pending_work); 744 745 psk = parent ? vsock_sk(parent) : NULL; 746 if (parent) { 747 vsk->trusted = psk->trusted; 748 vsk->owner = get_cred(psk->owner); 749 vsk->connect_timeout = psk->connect_timeout; 750 vsk->buffer_size = psk->buffer_size; 751 vsk->buffer_min_size = psk->buffer_min_size; 752 vsk->buffer_max_size = psk->buffer_max_size; 753 } else { 754 vsk->trusted = ns_capable_noaudit(&init_user_ns, CAP_NET_ADMIN); 755 vsk->owner = get_current_cred(); 756 vsk->connect_timeout = VSOCK_DEFAULT_CONNECT_TIMEOUT; 757 vsk->buffer_size = VSOCK_DEFAULT_BUFFER_SIZE; 758 vsk->buffer_min_size = VSOCK_DEFAULT_BUFFER_MIN_SIZE; 759 vsk->buffer_max_size = VSOCK_DEFAULT_BUFFER_MAX_SIZE; 760 } 761 762 return sk; 763 } 764 765 static void __vsock_release(struct sock *sk, int level) 766 { 767 if (sk) { 768 struct sock *pending; 769 struct vsock_sock *vsk; 770 771 vsk = vsock_sk(sk); 772 pending = NULL; /* Compiler warning. */ 773 774 /* When "level" is SINGLE_DEPTH_NESTING, use the nested 775 * version to avoid the warning "possible recursive locking 776 * detected". When "level" is 0, lock_sock_nested(sk, level) 777 * is the same as lock_sock(sk). 778 */ 779 lock_sock_nested(sk, level); 780 781 if (vsk->transport) 782 vsk->transport->release(vsk); 783 else if (sk->sk_type == SOCK_STREAM) 784 vsock_remove_sock(vsk); 785 786 sock_orphan(sk); 787 sk->sk_shutdown = SHUTDOWN_MASK; 788 789 skb_queue_purge(&sk->sk_receive_queue); 790 791 /* Clean up any sockets that never were accepted. */ 792 while ((pending = vsock_dequeue_accept(sk)) != NULL) { 793 __vsock_release(pending, SINGLE_DEPTH_NESTING); 794 sock_put(pending); 795 } 796 797 release_sock(sk); 798 sock_put(sk); 799 } 800 } 801 802 static void vsock_sk_destruct(struct sock *sk) 803 { 804 struct vsock_sock *vsk = vsock_sk(sk); 805 806 vsock_deassign_transport(vsk); 807 808 /* When clearing these addresses, there's no need to set the family and 809 * possibly register the address family with the kernel. 810 */ 811 vsock_addr_init(&vsk->local_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 812 vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, VMADDR_PORT_ANY); 813 814 put_cred(vsk->owner); 815 } 816 817 static int vsock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) 818 { 819 int err; 820 821 err = sock_queue_rcv_skb(sk, skb); 822 if (err) 823 kfree_skb(skb); 824 825 return err; 826 } 827 828 struct sock *vsock_create_connected(struct sock *parent) 829 { 830 return __vsock_create(sock_net(parent), NULL, parent, GFP_KERNEL, 831 parent->sk_type, 0); 832 } 833 EXPORT_SYMBOL_GPL(vsock_create_connected); 834 835 s64 vsock_stream_has_data(struct vsock_sock *vsk) 836 { 837 return vsk->transport->stream_has_data(vsk); 838 } 839 EXPORT_SYMBOL_GPL(vsock_stream_has_data); 840 841 s64 vsock_stream_has_space(struct vsock_sock *vsk) 842 { 843 return vsk->transport->stream_has_space(vsk); 844 } 845 EXPORT_SYMBOL_GPL(vsock_stream_has_space); 846 847 static int vsock_release(struct socket *sock) 848 { 849 __vsock_release(sock->sk, 0); 850 sock->sk = NULL; 851 sock->state = SS_FREE; 852 853 return 0; 854 } 855 856 static int 857 vsock_bind(struct socket *sock, struct sockaddr *addr, int addr_len) 858 { 859 int err; 860 struct sock *sk; 861 struct sockaddr_vm *vm_addr; 862 863 sk = sock->sk; 864 865 if (vsock_addr_cast(addr, addr_len, &vm_addr) != 0) 866 return -EINVAL; 867 868 lock_sock(sk); 869 err = __vsock_bind(sk, vm_addr); 870 release_sock(sk); 871 872 return err; 873 } 874 875 static int vsock_getname(struct socket *sock, 876 struct sockaddr *addr, int peer) 877 { 878 int err; 879 struct sock *sk; 880 struct vsock_sock *vsk; 881 struct sockaddr_vm *vm_addr; 882 883 sk = sock->sk; 884 vsk = vsock_sk(sk); 885 err = 0; 886 887 lock_sock(sk); 888 889 if (peer) { 890 if (sock->state != SS_CONNECTED) { 891 err = -ENOTCONN; 892 goto out; 893 } 894 vm_addr = &vsk->remote_addr; 895 } else { 896 vm_addr = &vsk->local_addr; 897 } 898 899 if (!vm_addr) { 900 err = -EINVAL; 901 goto out; 902 } 903 904 /* sys_getsockname() and sys_getpeername() pass us a 905 * MAX_SOCK_ADDR-sized buffer and don't set addr_len. Unfortunately 906 * that macro is defined in socket.c instead of .h, so we hardcode its 907 * value here. 908 */ 909 BUILD_BUG_ON(sizeof(*vm_addr) > 128); 910 memcpy(addr, vm_addr, sizeof(*vm_addr)); 911 err = sizeof(*vm_addr); 912 913 out: 914 release_sock(sk); 915 return err; 916 } 917 918 static int vsock_shutdown(struct socket *sock, int mode) 919 { 920 int err; 921 struct sock *sk; 922 923 /* User level uses SHUT_RD (0) and SHUT_WR (1), but the kernel uses 924 * RCV_SHUTDOWN (1) and SEND_SHUTDOWN (2), so we must increment mode 925 * here like the other address families do. Note also that the 926 * increment makes SHUT_RDWR (2) into RCV_SHUTDOWN | SEND_SHUTDOWN (3), 927 * which is what we want. 928 */ 929 mode++; 930 931 if ((mode & ~SHUTDOWN_MASK) || !mode) 932 return -EINVAL; 933 934 /* If this is a STREAM socket and it is not connected then bail out 935 * immediately. If it is a DGRAM socket then we must first kick the 936 * socket so that it wakes up from any sleeping calls, for example 937 * recv(), and then afterwards return the error. 938 */ 939 940 sk = sock->sk; 941 if (sock->state == SS_UNCONNECTED) { 942 err = -ENOTCONN; 943 if (sk->sk_type == SOCK_STREAM) 944 return err; 945 } else { 946 sock->state = SS_DISCONNECTING; 947 err = 0; 948 } 949 950 /* Receive and send shutdowns are treated alike. */ 951 mode = mode & (RCV_SHUTDOWN | SEND_SHUTDOWN); 952 if (mode) { 953 lock_sock(sk); 954 sk->sk_shutdown |= mode; 955 sk->sk_state_change(sk); 956 release_sock(sk); 957 958 if (sk->sk_type == SOCK_STREAM) { 959 sock_reset_flag(sk, SOCK_DONE); 960 vsock_send_shutdown(sk, mode); 961 } 962 } 963 964 return err; 965 } 966 967 static __poll_t vsock_poll(struct file *file, struct socket *sock, 968 poll_table *wait) 969 { 970 struct sock *sk; 971 __poll_t mask; 972 struct vsock_sock *vsk; 973 974 sk = sock->sk; 975 vsk = vsock_sk(sk); 976 977 poll_wait(file, sk_sleep(sk), wait); 978 mask = 0; 979 980 if (sk->sk_err) 981 /* Signify that there has been an error on this socket. */ 982 mask |= EPOLLERR; 983 984 /* INET sockets treat local write shutdown and peer write shutdown as a 985 * case of EPOLLHUP set. 986 */ 987 if ((sk->sk_shutdown == SHUTDOWN_MASK) || 988 ((sk->sk_shutdown & SEND_SHUTDOWN) && 989 (vsk->peer_shutdown & SEND_SHUTDOWN))) { 990 mask |= EPOLLHUP; 991 } 992 993 if (sk->sk_shutdown & RCV_SHUTDOWN || 994 vsk->peer_shutdown & SEND_SHUTDOWN) { 995 mask |= EPOLLRDHUP; 996 } 997 998 if (sock->type == SOCK_DGRAM) { 999 /* For datagram sockets we can read if there is something in 1000 * the queue and write as long as the socket isn't shutdown for 1001 * sending. 1002 */ 1003 if (!skb_queue_empty_lockless(&sk->sk_receive_queue) || 1004 (sk->sk_shutdown & RCV_SHUTDOWN)) { 1005 mask |= EPOLLIN | EPOLLRDNORM; 1006 } 1007 1008 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) 1009 mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND; 1010 1011 } else if (sock->type == SOCK_STREAM) { 1012 const struct vsock_transport *transport = vsk->transport; 1013 lock_sock(sk); 1014 1015 /* Listening sockets that have connections in their accept 1016 * queue can be read. 1017 */ 1018 if (sk->sk_state == TCP_LISTEN 1019 && !vsock_is_accept_queue_empty(sk)) 1020 mask |= EPOLLIN | EPOLLRDNORM; 1021 1022 /* If there is something in the queue then we can read. */ 1023 if (transport && transport->stream_is_active(vsk) && 1024 !(sk->sk_shutdown & RCV_SHUTDOWN)) { 1025 bool data_ready_now = false; 1026 int ret = transport->notify_poll_in( 1027 vsk, 1, &data_ready_now); 1028 if (ret < 0) { 1029 mask |= EPOLLERR; 1030 } else { 1031 if (data_ready_now) 1032 mask |= EPOLLIN | EPOLLRDNORM; 1033 1034 } 1035 } 1036 1037 /* Sockets whose connections have been closed, reset, or 1038 * terminated should also be considered read, and we check the 1039 * shutdown flag for that. 1040 */ 1041 if (sk->sk_shutdown & RCV_SHUTDOWN || 1042 vsk->peer_shutdown & SEND_SHUTDOWN) { 1043 mask |= EPOLLIN | EPOLLRDNORM; 1044 } 1045 1046 /* Connected sockets that can produce data can be written. */ 1047 if (transport && sk->sk_state == TCP_ESTABLISHED) { 1048 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) { 1049 bool space_avail_now = false; 1050 int ret = transport->notify_poll_out( 1051 vsk, 1, &space_avail_now); 1052 if (ret < 0) { 1053 mask |= EPOLLERR; 1054 } else { 1055 if (space_avail_now) 1056 /* Remove EPOLLWRBAND since INET 1057 * sockets are not setting it. 1058 */ 1059 mask |= EPOLLOUT | EPOLLWRNORM; 1060 1061 } 1062 } 1063 } 1064 1065 /* Simulate INET socket poll behaviors, which sets 1066 * EPOLLOUT|EPOLLWRNORM when peer is closed and nothing to read, 1067 * but local send is not shutdown. 1068 */ 1069 if (sk->sk_state == TCP_CLOSE || sk->sk_state == TCP_CLOSING) { 1070 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) 1071 mask |= EPOLLOUT | EPOLLWRNORM; 1072 1073 } 1074 1075 release_sock(sk); 1076 } 1077 1078 return mask; 1079 } 1080 1081 static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg, 1082 size_t len) 1083 { 1084 int err; 1085 struct sock *sk; 1086 struct vsock_sock *vsk; 1087 struct sockaddr_vm *remote_addr; 1088 const struct vsock_transport *transport; 1089 1090 if (msg->msg_flags & MSG_OOB) 1091 return -EOPNOTSUPP; 1092 1093 /* For now, MSG_DONTWAIT is always assumed... */ 1094 err = 0; 1095 sk = sock->sk; 1096 vsk = vsock_sk(sk); 1097 transport = vsk->transport; 1098 1099 lock_sock(sk); 1100 1101 err = vsock_auto_bind(vsk); 1102 if (err) 1103 goto out; 1104 1105 1106 /* If the provided message contains an address, use that. Otherwise 1107 * fall back on the socket's remote handle (if it has been connected). 1108 */ 1109 if (msg->msg_name && 1110 vsock_addr_cast(msg->msg_name, msg->msg_namelen, 1111 &remote_addr) == 0) { 1112 /* Ensure this address is of the right type and is a valid 1113 * destination. 1114 */ 1115 1116 if (remote_addr->svm_cid == VMADDR_CID_ANY) 1117 remote_addr->svm_cid = transport->get_local_cid(); 1118 1119 if (!vsock_addr_bound(remote_addr)) { 1120 err = -EINVAL; 1121 goto out; 1122 } 1123 } else if (sock->state == SS_CONNECTED) { 1124 remote_addr = &vsk->remote_addr; 1125 1126 if (remote_addr->svm_cid == VMADDR_CID_ANY) 1127 remote_addr->svm_cid = transport->get_local_cid(); 1128 1129 /* XXX Should connect() or this function ensure remote_addr is 1130 * bound? 1131 */ 1132 if (!vsock_addr_bound(&vsk->remote_addr)) { 1133 err = -EINVAL; 1134 goto out; 1135 } 1136 } else { 1137 err = -EINVAL; 1138 goto out; 1139 } 1140 1141 if (!transport->dgram_allow(remote_addr->svm_cid, 1142 remote_addr->svm_port)) { 1143 err = -EINVAL; 1144 goto out; 1145 } 1146 1147 err = transport->dgram_enqueue(vsk, remote_addr, msg, len); 1148 1149 out: 1150 release_sock(sk); 1151 return err; 1152 } 1153 1154 static int vsock_dgram_connect(struct socket *sock, 1155 struct sockaddr *addr, int addr_len, int flags) 1156 { 1157 int err; 1158 struct sock *sk; 1159 struct vsock_sock *vsk; 1160 struct sockaddr_vm *remote_addr; 1161 1162 sk = sock->sk; 1163 vsk = vsock_sk(sk); 1164 1165 err = vsock_addr_cast(addr, addr_len, &remote_addr); 1166 if (err == -EAFNOSUPPORT && remote_addr->svm_family == AF_UNSPEC) { 1167 lock_sock(sk); 1168 vsock_addr_init(&vsk->remote_addr, VMADDR_CID_ANY, 1169 VMADDR_PORT_ANY); 1170 sock->state = SS_UNCONNECTED; 1171 release_sock(sk); 1172 return 0; 1173 } else if (err != 0) 1174 return -EINVAL; 1175 1176 lock_sock(sk); 1177 1178 err = vsock_auto_bind(vsk); 1179 if (err) 1180 goto out; 1181 1182 if (!vsk->transport->dgram_allow(remote_addr->svm_cid, 1183 remote_addr->svm_port)) { 1184 err = -EINVAL; 1185 goto out; 1186 } 1187 1188 memcpy(&vsk->remote_addr, remote_addr, sizeof(vsk->remote_addr)); 1189 sock->state = SS_CONNECTED; 1190 1191 out: 1192 release_sock(sk); 1193 return err; 1194 } 1195 1196 static int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg, 1197 size_t len, int flags) 1198 { 1199 struct vsock_sock *vsk = vsock_sk(sock->sk); 1200 1201 return vsk->transport->dgram_dequeue(vsk, msg, len, flags); 1202 } 1203 1204 static const struct proto_ops vsock_dgram_ops = { 1205 .family = PF_VSOCK, 1206 .owner = THIS_MODULE, 1207 .release = vsock_release, 1208 .bind = vsock_bind, 1209 .connect = vsock_dgram_connect, 1210 .socketpair = sock_no_socketpair, 1211 .accept = sock_no_accept, 1212 .getname = vsock_getname, 1213 .poll = vsock_poll, 1214 .ioctl = sock_no_ioctl, 1215 .listen = sock_no_listen, 1216 .shutdown = vsock_shutdown, 1217 .sendmsg = vsock_dgram_sendmsg, 1218 .recvmsg = vsock_dgram_recvmsg, 1219 .mmap = sock_no_mmap, 1220 .sendpage = sock_no_sendpage, 1221 }; 1222 1223 static int vsock_transport_cancel_pkt(struct vsock_sock *vsk) 1224 { 1225 const struct vsock_transport *transport = vsk->transport; 1226 1227 if (!transport->cancel_pkt) 1228 return -EOPNOTSUPP; 1229 1230 return transport->cancel_pkt(vsk); 1231 } 1232 1233 static void vsock_connect_timeout(struct work_struct *work) 1234 { 1235 struct sock *sk; 1236 struct vsock_sock *vsk; 1237 int cancel = 0; 1238 1239 vsk = container_of(work, struct vsock_sock, connect_work.work); 1240 sk = sk_vsock(vsk); 1241 1242 lock_sock(sk); 1243 if (sk->sk_state == TCP_SYN_SENT && 1244 (sk->sk_shutdown != SHUTDOWN_MASK)) { 1245 sk->sk_state = TCP_CLOSE; 1246 sk->sk_err = ETIMEDOUT; 1247 sk->sk_error_report(sk); 1248 cancel = 1; 1249 } 1250 release_sock(sk); 1251 if (cancel) 1252 vsock_transport_cancel_pkt(vsk); 1253 1254 sock_put(sk); 1255 } 1256 1257 static int vsock_stream_connect(struct socket *sock, struct sockaddr *addr, 1258 int addr_len, int flags) 1259 { 1260 int err; 1261 struct sock *sk; 1262 struct vsock_sock *vsk; 1263 const struct vsock_transport *transport; 1264 struct sockaddr_vm *remote_addr; 1265 long timeout; 1266 DEFINE_WAIT(wait); 1267 1268 err = 0; 1269 sk = sock->sk; 1270 vsk = vsock_sk(sk); 1271 1272 lock_sock(sk); 1273 1274 /* XXX AF_UNSPEC should make us disconnect like AF_INET. */ 1275 switch (sock->state) { 1276 case SS_CONNECTED: 1277 err = -EISCONN; 1278 goto out; 1279 case SS_DISCONNECTING: 1280 err = -EINVAL; 1281 goto out; 1282 case SS_CONNECTING: 1283 /* This continues on so we can move sock into the SS_CONNECTED 1284 * state once the connection has completed (at which point err 1285 * will be set to zero also). Otherwise, we will either wait 1286 * for the connection or return -EALREADY should this be a 1287 * non-blocking call. 1288 */ 1289 err = -EALREADY; 1290 break; 1291 default: 1292 if ((sk->sk_state == TCP_LISTEN) || 1293 vsock_addr_cast(addr, addr_len, &remote_addr) != 0) { 1294 err = -EINVAL; 1295 goto out; 1296 } 1297 1298 /* Set the remote address that we are connecting to. */ 1299 memcpy(&vsk->remote_addr, remote_addr, 1300 sizeof(vsk->remote_addr)); 1301 1302 err = vsock_assign_transport(vsk, NULL); 1303 if (err) 1304 goto out; 1305 1306 transport = vsk->transport; 1307 1308 /* The hypervisor and well-known contexts do not have socket 1309 * endpoints. 1310 */ 1311 if (!transport || 1312 !transport->stream_allow(remote_addr->svm_cid, 1313 remote_addr->svm_port)) { 1314 err = -ENETUNREACH; 1315 goto out; 1316 } 1317 1318 err = vsock_auto_bind(vsk); 1319 if (err) 1320 goto out; 1321 1322 sk->sk_state = TCP_SYN_SENT; 1323 1324 err = transport->connect(vsk); 1325 if (err < 0) 1326 goto out; 1327 1328 /* Mark sock as connecting and set the error code to in 1329 * progress in case this is a non-blocking connect. 1330 */ 1331 sock->state = SS_CONNECTING; 1332 err = -EINPROGRESS; 1333 } 1334 1335 /* The receive path will handle all communication until we are able to 1336 * enter the connected state. Here we wait for the connection to be 1337 * completed or a notification of an error. 1338 */ 1339 timeout = vsk->connect_timeout; 1340 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); 1341 1342 while (sk->sk_state != TCP_ESTABLISHED && sk->sk_err == 0) { 1343 if (flags & O_NONBLOCK) { 1344 /* If we're not going to block, we schedule a timeout 1345 * function to generate a timeout on the connection 1346 * attempt, in case the peer doesn't respond in a 1347 * timely manner. We hold on to the socket until the 1348 * timeout fires. 1349 */ 1350 sock_hold(sk); 1351 schedule_delayed_work(&vsk->connect_work, timeout); 1352 1353 /* Skip ahead to preserve error code set above. */ 1354 goto out_wait; 1355 } 1356 1357 release_sock(sk); 1358 timeout = schedule_timeout(timeout); 1359 lock_sock(sk); 1360 1361 if (signal_pending(current)) { 1362 err = sock_intr_errno(timeout); 1363 sk->sk_state = TCP_CLOSE; 1364 sock->state = SS_UNCONNECTED; 1365 vsock_transport_cancel_pkt(vsk); 1366 goto out_wait; 1367 } else if (timeout == 0) { 1368 err = -ETIMEDOUT; 1369 sk->sk_state = TCP_CLOSE; 1370 sock->state = SS_UNCONNECTED; 1371 vsock_transport_cancel_pkt(vsk); 1372 goto out_wait; 1373 } 1374 1375 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); 1376 } 1377 1378 if (sk->sk_err) { 1379 err = -sk->sk_err; 1380 sk->sk_state = TCP_CLOSE; 1381 sock->state = SS_UNCONNECTED; 1382 } else { 1383 err = 0; 1384 } 1385 1386 out_wait: 1387 finish_wait(sk_sleep(sk), &wait); 1388 out: 1389 release_sock(sk); 1390 return err; 1391 } 1392 1393 static int vsock_accept(struct socket *sock, struct socket *newsock, int flags, 1394 bool kern) 1395 { 1396 struct sock *listener; 1397 int err; 1398 struct sock *connected; 1399 struct vsock_sock *vconnected; 1400 long timeout; 1401 DEFINE_WAIT(wait); 1402 1403 err = 0; 1404 listener = sock->sk; 1405 1406 lock_sock(listener); 1407 1408 if (sock->type != SOCK_STREAM) { 1409 err = -EOPNOTSUPP; 1410 goto out; 1411 } 1412 1413 if (listener->sk_state != TCP_LISTEN) { 1414 err = -EINVAL; 1415 goto out; 1416 } 1417 1418 /* Wait for children sockets to appear; these are the new sockets 1419 * created upon connection establishment. 1420 */ 1421 timeout = sock_rcvtimeo(listener, flags & O_NONBLOCK); 1422 prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE); 1423 1424 while ((connected = vsock_dequeue_accept(listener)) == NULL && 1425 listener->sk_err == 0) { 1426 release_sock(listener); 1427 timeout = schedule_timeout(timeout); 1428 finish_wait(sk_sleep(listener), &wait); 1429 lock_sock(listener); 1430 1431 if (signal_pending(current)) { 1432 err = sock_intr_errno(timeout); 1433 goto out; 1434 } else if (timeout == 0) { 1435 err = -EAGAIN; 1436 goto out; 1437 } 1438 1439 prepare_to_wait(sk_sleep(listener), &wait, TASK_INTERRUPTIBLE); 1440 } 1441 finish_wait(sk_sleep(listener), &wait); 1442 1443 if (listener->sk_err) 1444 err = -listener->sk_err; 1445 1446 if (connected) { 1447 sk_acceptq_removed(listener); 1448 1449 lock_sock_nested(connected, SINGLE_DEPTH_NESTING); 1450 vconnected = vsock_sk(connected); 1451 1452 /* If the listener socket has received an error, then we should 1453 * reject this socket and return. Note that we simply mark the 1454 * socket rejected, drop our reference, and let the cleanup 1455 * function handle the cleanup; the fact that we found it in 1456 * the listener's accept queue guarantees that the cleanup 1457 * function hasn't run yet. 1458 */ 1459 if (err) { 1460 vconnected->rejected = true; 1461 } else { 1462 newsock->state = SS_CONNECTED; 1463 sock_graft(connected, newsock); 1464 } 1465 1466 release_sock(connected); 1467 sock_put(connected); 1468 } 1469 1470 out: 1471 release_sock(listener); 1472 return err; 1473 } 1474 1475 static int vsock_listen(struct socket *sock, int backlog) 1476 { 1477 int err; 1478 struct sock *sk; 1479 struct vsock_sock *vsk; 1480 1481 sk = sock->sk; 1482 1483 lock_sock(sk); 1484 1485 if (sock->type != SOCK_STREAM) { 1486 err = -EOPNOTSUPP; 1487 goto out; 1488 } 1489 1490 if (sock->state != SS_UNCONNECTED) { 1491 err = -EINVAL; 1492 goto out; 1493 } 1494 1495 vsk = vsock_sk(sk); 1496 1497 if (!vsock_addr_bound(&vsk->local_addr)) { 1498 err = -EINVAL; 1499 goto out; 1500 } 1501 1502 sk->sk_max_ack_backlog = backlog; 1503 sk->sk_state = TCP_LISTEN; 1504 1505 err = 0; 1506 1507 out: 1508 release_sock(sk); 1509 return err; 1510 } 1511 1512 static void vsock_update_buffer_size(struct vsock_sock *vsk, 1513 const struct vsock_transport *transport, 1514 u64 val) 1515 { 1516 if (val > vsk->buffer_max_size) 1517 val = vsk->buffer_max_size; 1518 1519 if (val < vsk->buffer_min_size) 1520 val = vsk->buffer_min_size; 1521 1522 if (val != vsk->buffer_size && 1523 transport && transport->notify_buffer_size) 1524 transport->notify_buffer_size(vsk, &val); 1525 1526 vsk->buffer_size = val; 1527 } 1528 1529 static int vsock_stream_setsockopt(struct socket *sock, 1530 int level, 1531 int optname, 1532 sockptr_t optval, 1533 unsigned int optlen) 1534 { 1535 int err; 1536 struct sock *sk; 1537 struct vsock_sock *vsk; 1538 const struct vsock_transport *transport; 1539 u64 val; 1540 1541 if (level != AF_VSOCK) 1542 return -ENOPROTOOPT; 1543 1544 #define COPY_IN(_v) \ 1545 do { \ 1546 if (optlen < sizeof(_v)) { \ 1547 err = -EINVAL; \ 1548 goto exit; \ 1549 } \ 1550 if (copy_from_sockptr(&_v, optval, sizeof(_v)) != 0) { \ 1551 err = -EFAULT; \ 1552 goto exit; \ 1553 } \ 1554 } while (0) 1555 1556 err = 0; 1557 sk = sock->sk; 1558 vsk = vsock_sk(sk); 1559 transport = vsk->transport; 1560 1561 lock_sock(sk); 1562 1563 switch (optname) { 1564 case SO_VM_SOCKETS_BUFFER_SIZE: 1565 COPY_IN(val); 1566 vsock_update_buffer_size(vsk, transport, val); 1567 break; 1568 1569 case SO_VM_SOCKETS_BUFFER_MAX_SIZE: 1570 COPY_IN(val); 1571 vsk->buffer_max_size = val; 1572 vsock_update_buffer_size(vsk, transport, vsk->buffer_size); 1573 break; 1574 1575 case SO_VM_SOCKETS_BUFFER_MIN_SIZE: 1576 COPY_IN(val); 1577 vsk->buffer_min_size = val; 1578 vsock_update_buffer_size(vsk, transport, vsk->buffer_size); 1579 break; 1580 1581 case SO_VM_SOCKETS_CONNECT_TIMEOUT: { 1582 struct __kernel_old_timeval tv; 1583 COPY_IN(tv); 1584 if (tv.tv_sec >= 0 && tv.tv_usec < USEC_PER_SEC && 1585 tv.tv_sec < (MAX_SCHEDULE_TIMEOUT / HZ - 1)) { 1586 vsk->connect_timeout = tv.tv_sec * HZ + 1587 DIV_ROUND_UP(tv.tv_usec, (1000000 / HZ)); 1588 if (vsk->connect_timeout == 0) 1589 vsk->connect_timeout = 1590 VSOCK_DEFAULT_CONNECT_TIMEOUT; 1591 1592 } else { 1593 err = -ERANGE; 1594 } 1595 break; 1596 } 1597 1598 default: 1599 err = -ENOPROTOOPT; 1600 break; 1601 } 1602 1603 #undef COPY_IN 1604 1605 exit: 1606 release_sock(sk); 1607 return err; 1608 } 1609 1610 static int vsock_stream_getsockopt(struct socket *sock, 1611 int level, int optname, 1612 char __user *optval, 1613 int __user *optlen) 1614 { 1615 int err; 1616 int len; 1617 struct sock *sk; 1618 struct vsock_sock *vsk; 1619 u64 val; 1620 1621 if (level != AF_VSOCK) 1622 return -ENOPROTOOPT; 1623 1624 err = get_user(len, optlen); 1625 if (err != 0) 1626 return err; 1627 1628 #define COPY_OUT(_v) \ 1629 do { \ 1630 if (len < sizeof(_v)) \ 1631 return -EINVAL; \ 1632 \ 1633 len = sizeof(_v); \ 1634 if (copy_to_user(optval, &_v, len) != 0) \ 1635 return -EFAULT; \ 1636 \ 1637 } while (0) 1638 1639 err = 0; 1640 sk = sock->sk; 1641 vsk = vsock_sk(sk); 1642 1643 switch (optname) { 1644 case SO_VM_SOCKETS_BUFFER_SIZE: 1645 val = vsk->buffer_size; 1646 COPY_OUT(val); 1647 break; 1648 1649 case SO_VM_SOCKETS_BUFFER_MAX_SIZE: 1650 val = vsk->buffer_max_size; 1651 COPY_OUT(val); 1652 break; 1653 1654 case SO_VM_SOCKETS_BUFFER_MIN_SIZE: 1655 val = vsk->buffer_min_size; 1656 COPY_OUT(val); 1657 break; 1658 1659 case SO_VM_SOCKETS_CONNECT_TIMEOUT: { 1660 struct __kernel_old_timeval tv; 1661 tv.tv_sec = vsk->connect_timeout / HZ; 1662 tv.tv_usec = 1663 (vsk->connect_timeout - 1664 tv.tv_sec * HZ) * (1000000 / HZ); 1665 COPY_OUT(tv); 1666 break; 1667 } 1668 default: 1669 return -ENOPROTOOPT; 1670 } 1671 1672 err = put_user(len, optlen); 1673 if (err != 0) 1674 return -EFAULT; 1675 1676 #undef COPY_OUT 1677 1678 return 0; 1679 } 1680 1681 static int vsock_stream_sendmsg(struct socket *sock, struct msghdr *msg, 1682 size_t len) 1683 { 1684 struct sock *sk; 1685 struct vsock_sock *vsk; 1686 const struct vsock_transport *transport; 1687 ssize_t total_written; 1688 long timeout; 1689 int err; 1690 struct vsock_transport_send_notify_data send_data; 1691 DEFINE_WAIT_FUNC(wait, woken_wake_function); 1692 1693 sk = sock->sk; 1694 vsk = vsock_sk(sk); 1695 transport = vsk->transport; 1696 total_written = 0; 1697 err = 0; 1698 1699 if (msg->msg_flags & MSG_OOB) 1700 return -EOPNOTSUPP; 1701 1702 lock_sock(sk); 1703 1704 /* Callers should not provide a destination with stream sockets. */ 1705 if (msg->msg_namelen) { 1706 err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP; 1707 goto out; 1708 } 1709 1710 /* Send data only if both sides are not shutdown in the direction. */ 1711 if (sk->sk_shutdown & SEND_SHUTDOWN || 1712 vsk->peer_shutdown & RCV_SHUTDOWN) { 1713 err = -EPIPE; 1714 goto out; 1715 } 1716 1717 if (!transport || sk->sk_state != TCP_ESTABLISHED || 1718 !vsock_addr_bound(&vsk->local_addr)) { 1719 err = -ENOTCONN; 1720 goto out; 1721 } 1722 1723 if (!vsock_addr_bound(&vsk->remote_addr)) { 1724 err = -EDESTADDRREQ; 1725 goto out; 1726 } 1727 1728 /* Wait for room in the produce queue to enqueue our user's data. */ 1729 timeout = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT); 1730 1731 err = transport->notify_send_init(vsk, &send_data); 1732 if (err < 0) 1733 goto out; 1734 1735 while (total_written < len) { 1736 ssize_t written; 1737 1738 add_wait_queue(sk_sleep(sk), &wait); 1739 while (vsock_stream_has_space(vsk) == 0 && 1740 sk->sk_err == 0 && 1741 !(sk->sk_shutdown & SEND_SHUTDOWN) && 1742 !(vsk->peer_shutdown & RCV_SHUTDOWN)) { 1743 1744 /* Don't wait for non-blocking sockets. */ 1745 if (timeout == 0) { 1746 err = -EAGAIN; 1747 remove_wait_queue(sk_sleep(sk), &wait); 1748 goto out_err; 1749 } 1750 1751 err = transport->notify_send_pre_block(vsk, &send_data); 1752 if (err < 0) { 1753 remove_wait_queue(sk_sleep(sk), &wait); 1754 goto out_err; 1755 } 1756 1757 release_sock(sk); 1758 timeout = wait_woken(&wait, TASK_INTERRUPTIBLE, timeout); 1759 lock_sock(sk); 1760 if (signal_pending(current)) { 1761 err = sock_intr_errno(timeout); 1762 remove_wait_queue(sk_sleep(sk), &wait); 1763 goto out_err; 1764 } else if (timeout == 0) { 1765 err = -EAGAIN; 1766 remove_wait_queue(sk_sleep(sk), &wait); 1767 goto out_err; 1768 } 1769 } 1770 remove_wait_queue(sk_sleep(sk), &wait); 1771 1772 /* These checks occur both as part of and after the loop 1773 * conditional since we need to check before and after 1774 * sleeping. 1775 */ 1776 if (sk->sk_err) { 1777 err = -sk->sk_err; 1778 goto out_err; 1779 } else if ((sk->sk_shutdown & SEND_SHUTDOWN) || 1780 (vsk->peer_shutdown & RCV_SHUTDOWN)) { 1781 err = -EPIPE; 1782 goto out_err; 1783 } 1784 1785 err = transport->notify_send_pre_enqueue(vsk, &send_data); 1786 if (err < 0) 1787 goto out_err; 1788 1789 /* Note that enqueue will only write as many bytes as are free 1790 * in the produce queue, so we don't need to ensure len is 1791 * smaller than the queue size. It is the caller's 1792 * responsibility to check how many bytes we were able to send. 1793 */ 1794 1795 written = transport->stream_enqueue( 1796 vsk, msg, 1797 len - total_written); 1798 if (written < 0) { 1799 err = -ENOMEM; 1800 goto out_err; 1801 } 1802 1803 total_written += written; 1804 1805 err = transport->notify_send_post_enqueue( 1806 vsk, written, &send_data); 1807 if (err < 0) 1808 goto out_err; 1809 1810 } 1811 1812 out_err: 1813 if (total_written > 0) 1814 err = total_written; 1815 out: 1816 release_sock(sk); 1817 return err; 1818 } 1819 1820 1821 static int 1822 vsock_stream_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, 1823 int flags) 1824 { 1825 struct sock *sk; 1826 struct vsock_sock *vsk; 1827 const struct vsock_transport *transport; 1828 int err; 1829 size_t target; 1830 ssize_t copied; 1831 long timeout; 1832 struct vsock_transport_recv_notify_data recv_data; 1833 1834 DEFINE_WAIT(wait); 1835 1836 sk = sock->sk; 1837 vsk = vsock_sk(sk); 1838 transport = vsk->transport; 1839 err = 0; 1840 1841 lock_sock(sk); 1842 1843 if (!transport || sk->sk_state != TCP_ESTABLISHED) { 1844 /* Recvmsg is supposed to return 0 if a peer performs an 1845 * orderly shutdown. Differentiate between that case and when a 1846 * peer has not connected or a local shutdown occured with the 1847 * SOCK_DONE flag. 1848 */ 1849 if (sock_flag(sk, SOCK_DONE)) 1850 err = 0; 1851 else 1852 err = -ENOTCONN; 1853 1854 goto out; 1855 } 1856 1857 if (flags & MSG_OOB) { 1858 err = -EOPNOTSUPP; 1859 goto out; 1860 } 1861 1862 /* We don't check peer_shutdown flag here since peer may actually shut 1863 * down, but there can be data in the queue that a local socket can 1864 * receive. 1865 */ 1866 if (sk->sk_shutdown & RCV_SHUTDOWN) { 1867 err = 0; 1868 goto out; 1869 } 1870 1871 /* It is valid on Linux to pass in a zero-length receive buffer. This 1872 * is not an error. We may as well bail out now. 1873 */ 1874 if (!len) { 1875 err = 0; 1876 goto out; 1877 } 1878 1879 /* We must not copy less than target bytes into the user's buffer 1880 * before returning successfully, so we wait for the consume queue to 1881 * have that much data to consume before dequeueing. Note that this 1882 * makes it impossible to handle cases where target is greater than the 1883 * queue size. 1884 */ 1885 target = sock_rcvlowat(sk, flags & MSG_WAITALL, len); 1886 if (target >= transport->stream_rcvhiwat(vsk)) { 1887 err = -ENOMEM; 1888 goto out; 1889 } 1890 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT); 1891 copied = 0; 1892 1893 err = transport->notify_recv_init(vsk, target, &recv_data); 1894 if (err < 0) 1895 goto out; 1896 1897 1898 while (1) { 1899 s64 ready; 1900 1901 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); 1902 ready = vsock_stream_has_data(vsk); 1903 1904 if (ready == 0) { 1905 if (sk->sk_err != 0 || 1906 (sk->sk_shutdown & RCV_SHUTDOWN) || 1907 (vsk->peer_shutdown & SEND_SHUTDOWN)) { 1908 finish_wait(sk_sleep(sk), &wait); 1909 break; 1910 } 1911 /* Don't wait for non-blocking sockets. */ 1912 if (timeout == 0) { 1913 err = -EAGAIN; 1914 finish_wait(sk_sleep(sk), &wait); 1915 break; 1916 } 1917 1918 err = transport->notify_recv_pre_block( 1919 vsk, target, &recv_data); 1920 if (err < 0) { 1921 finish_wait(sk_sleep(sk), &wait); 1922 break; 1923 } 1924 release_sock(sk); 1925 timeout = schedule_timeout(timeout); 1926 lock_sock(sk); 1927 1928 if (signal_pending(current)) { 1929 err = sock_intr_errno(timeout); 1930 finish_wait(sk_sleep(sk), &wait); 1931 break; 1932 } else if (timeout == 0) { 1933 err = -EAGAIN; 1934 finish_wait(sk_sleep(sk), &wait); 1935 break; 1936 } 1937 } else { 1938 ssize_t read; 1939 1940 finish_wait(sk_sleep(sk), &wait); 1941 1942 if (ready < 0) { 1943 /* Invalid queue pair content. XXX This should 1944 * be changed to a connection reset in a later 1945 * change. 1946 */ 1947 1948 err = -ENOMEM; 1949 goto out; 1950 } 1951 1952 err = transport->notify_recv_pre_dequeue( 1953 vsk, target, &recv_data); 1954 if (err < 0) 1955 break; 1956 1957 read = transport->stream_dequeue( 1958 vsk, msg, 1959 len - copied, flags); 1960 if (read < 0) { 1961 err = -ENOMEM; 1962 break; 1963 } 1964 1965 copied += read; 1966 1967 err = transport->notify_recv_post_dequeue( 1968 vsk, target, read, 1969 !(flags & MSG_PEEK), &recv_data); 1970 if (err < 0) 1971 goto out; 1972 1973 if (read >= target || flags & MSG_PEEK) 1974 break; 1975 1976 target -= read; 1977 } 1978 } 1979 1980 if (sk->sk_err) 1981 err = -sk->sk_err; 1982 else if (sk->sk_shutdown & RCV_SHUTDOWN) 1983 err = 0; 1984 1985 if (copied > 0) 1986 err = copied; 1987 1988 out: 1989 release_sock(sk); 1990 return err; 1991 } 1992 1993 static const struct proto_ops vsock_stream_ops = { 1994 .family = PF_VSOCK, 1995 .owner = THIS_MODULE, 1996 .release = vsock_release, 1997 .bind = vsock_bind, 1998 .connect = vsock_stream_connect, 1999 .socketpair = sock_no_socketpair, 2000 .accept = vsock_accept, 2001 .getname = vsock_getname, 2002 .poll = vsock_poll, 2003 .ioctl = sock_no_ioctl, 2004 .listen = vsock_listen, 2005 .shutdown = vsock_shutdown, 2006 .setsockopt = vsock_stream_setsockopt, 2007 .getsockopt = vsock_stream_getsockopt, 2008 .sendmsg = vsock_stream_sendmsg, 2009 .recvmsg = vsock_stream_recvmsg, 2010 .mmap = sock_no_mmap, 2011 .sendpage = sock_no_sendpage, 2012 }; 2013 2014 static int vsock_create(struct net *net, struct socket *sock, 2015 int protocol, int kern) 2016 { 2017 struct vsock_sock *vsk; 2018 struct sock *sk; 2019 int ret; 2020 2021 if (!sock) 2022 return -EINVAL; 2023 2024 if (protocol && protocol != PF_VSOCK) 2025 return -EPROTONOSUPPORT; 2026 2027 switch (sock->type) { 2028 case SOCK_DGRAM: 2029 sock->ops = &vsock_dgram_ops; 2030 break; 2031 case SOCK_STREAM: 2032 sock->ops = &vsock_stream_ops; 2033 break; 2034 default: 2035 return -ESOCKTNOSUPPORT; 2036 } 2037 2038 sock->state = SS_UNCONNECTED; 2039 2040 sk = __vsock_create(net, sock, NULL, GFP_KERNEL, 0, kern); 2041 if (!sk) 2042 return -ENOMEM; 2043 2044 vsk = vsock_sk(sk); 2045 2046 if (sock->type == SOCK_DGRAM) { 2047 ret = vsock_assign_transport(vsk, NULL); 2048 if (ret < 0) { 2049 sock_put(sk); 2050 return ret; 2051 } 2052 } 2053 2054 vsock_insert_unbound(vsk); 2055 2056 return 0; 2057 } 2058 2059 static const struct net_proto_family vsock_family_ops = { 2060 .family = AF_VSOCK, 2061 .create = vsock_create, 2062 .owner = THIS_MODULE, 2063 }; 2064 2065 static long vsock_dev_do_ioctl(struct file *filp, 2066 unsigned int cmd, void __user *ptr) 2067 { 2068 u32 __user *p = ptr; 2069 u32 cid = VMADDR_CID_ANY; 2070 int retval = 0; 2071 2072 switch (cmd) { 2073 case IOCTL_VM_SOCKETS_GET_LOCAL_CID: 2074 /* To be compatible with the VMCI behavior, we prioritize the 2075 * guest CID instead of well-know host CID (VMADDR_CID_HOST). 2076 */ 2077 if (transport_g2h) 2078 cid = transport_g2h->get_local_cid(); 2079 else if (transport_h2g) 2080 cid = transport_h2g->get_local_cid(); 2081 2082 if (put_user(cid, p) != 0) 2083 retval = -EFAULT; 2084 break; 2085 2086 default: 2087 retval = -ENOIOCTLCMD; 2088 } 2089 2090 return retval; 2091 } 2092 2093 static long vsock_dev_ioctl(struct file *filp, 2094 unsigned int cmd, unsigned long arg) 2095 { 2096 return vsock_dev_do_ioctl(filp, cmd, (void __user *)arg); 2097 } 2098 2099 #ifdef CONFIG_COMPAT 2100 static long vsock_dev_compat_ioctl(struct file *filp, 2101 unsigned int cmd, unsigned long arg) 2102 { 2103 return vsock_dev_do_ioctl(filp, cmd, compat_ptr(arg)); 2104 } 2105 #endif 2106 2107 static const struct file_operations vsock_device_ops = { 2108 .owner = THIS_MODULE, 2109 .unlocked_ioctl = vsock_dev_ioctl, 2110 #ifdef CONFIG_COMPAT 2111 .compat_ioctl = vsock_dev_compat_ioctl, 2112 #endif 2113 .open = nonseekable_open, 2114 }; 2115 2116 static struct miscdevice vsock_device = { 2117 .name = "vsock", 2118 .fops = &vsock_device_ops, 2119 }; 2120 2121 static int __init vsock_init(void) 2122 { 2123 int err = 0; 2124 2125 vsock_init_tables(); 2126 2127 vsock_proto.owner = THIS_MODULE; 2128 vsock_device.minor = MISC_DYNAMIC_MINOR; 2129 err = misc_register(&vsock_device); 2130 if (err) { 2131 pr_err("Failed to register misc device\n"); 2132 goto err_reset_transport; 2133 } 2134 2135 err = proto_register(&vsock_proto, 1); /* we want our slab */ 2136 if (err) { 2137 pr_err("Cannot register vsock protocol\n"); 2138 goto err_deregister_misc; 2139 } 2140 2141 err = sock_register(&vsock_family_ops); 2142 if (err) { 2143 pr_err("could not register af_vsock (%d) address family: %d\n", 2144 AF_VSOCK, err); 2145 goto err_unregister_proto; 2146 } 2147 2148 return 0; 2149 2150 err_unregister_proto: 2151 proto_unregister(&vsock_proto); 2152 err_deregister_misc: 2153 misc_deregister(&vsock_device); 2154 err_reset_transport: 2155 return err; 2156 } 2157 2158 static void __exit vsock_exit(void) 2159 { 2160 misc_deregister(&vsock_device); 2161 sock_unregister(AF_VSOCK); 2162 proto_unregister(&vsock_proto); 2163 } 2164 2165 const struct vsock_transport *vsock_core_get_transport(struct vsock_sock *vsk) 2166 { 2167 return vsk->transport; 2168 } 2169 EXPORT_SYMBOL_GPL(vsock_core_get_transport); 2170 2171 int vsock_core_register(const struct vsock_transport *t, int features) 2172 { 2173 const struct vsock_transport *t_h2g, *t_g2h, *t_dgram, *t_local; 2174 int err = mutex_lock_interruptible(&vsock_register_mutex); 2175 2176 if (err) 2177 return err; 2178 2179 t_h2g = transport_h2g; 2180 t_g2h = transport_g2h; 2181 t_dgram = transport_dgram; 2182 t_local = transport_local; 2183 2184 if (features & VSOCK_TRANSPORT_F_H2G) { 2185 if (t_h2g) { 2186 err = -EBUSY; 2187 goto err_busy; 2188 } 2189 t_h2g = t; 2190 } 2191 2192 if (features & VSOCK_TRANSPORT_F_G2H) { 2193 if (t_g2h) { 2194 err = -EBUSY; 2195 goto err_busy; 2196 } 2197 t_g2h = t; 2198 } 2199 2200 if (features & VSOCK_TRANSPORT_F_DGRAM) { 2201 if (t_dgram) { 2202 err = -EBUSY; 2203 goto err_busy; 2204 } 2205 t_dgram = t; 2206 } 2207 2208 if (features & VSOCK_TRANSPORT_F_LOCAL) { 2209 if (t_local) { 2210 err = -EBUSY; 2211 goto err_busy; 2212 } 2213 t_local = t; 2214 } 2215 2216 transport_h2g = t_h2g; 2217 transport_g2h = t_g2h; 2218 transport_dgram = t_dgram; 2219 transport_local = t_local; 2220 2221 err_busy: 2222 mutex_unlock(&vsock_register_mutex); 2223 return err; 2224 } 2225 EXPORT_SYMBOL_GPL(vsock_core_register); 2226 2227 void vsock_core_unregister(const struct vsock_transport *t) 2228 { 2229 mutex_lock(&vsock_register_mutex); 2230 2231 if (transport_h2g == t) 2232 transport_h2g = NULL; 2233 2234 if (transport_g2h == t) 2235 transport_g2h = NULL; 2236 2237 if (transport_dgram == t) 2238 transport_dgram = NULL; 2239 2240 if (transport_local == t) 2241 transport_local = NULL; 2242 2243 mutex_unlock(&vsock_register_mutex); 2244 } 2245 EXPORT_SYMBOL_GPL(vsock_core_unregister); 2246 2247 module_init(vsock_init); 2248 module_exit(vsock_exit); 2249 2250 MODULE_AUTHOR("VMware, Inc."); 2251 MODULE_DESCRIPTION("VMware Virtual Socket Family"); 2252 MODULE_VERSION("1.0.2.0-k"); 2253 MODULE_LICENSE("GPL v2"); 2254