1 /* 2 * net/tipc/socket.c: TIPC socket API 3 * 4 * Copyright (c) 2001-2007, Ericsson AB 5 * Copyright (c) 2004-2008, Wind River Systems 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions are met: 10 * 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the names of the copyright holders nor the names of its 17 * contributors may be used to endorse or promote products derived from 18 * this software without specific prior written permission. 19 * 20 * Alternatively, this software may be distributed under the terms of the 21 * GNU General Public License ("GPL") version 2 as published by the Free 22 * Software Foundation. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 25 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE 28 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 31 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 32 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 33 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 34 * POSSIBILITY OF SUCH DAMAGE. 35 */ 36 37 #include <net/sock.h> 38 39 #include <linux/tipc.h> 40 #include <linux/tipc_config.h> 41 42 #include "core.h" 43 #include "port.h" 44 45 #define SS_LISTENING -1 /* socket is listening */ 46 #define SS_READY -2 /* socket is connectionless */ 47 48 #define OVERLOAD_LIMIT_BASE 5000 49 #define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */ 50 51 struct tipc_sock { 52 struct sock sk; 53 struct tipc_port *p; 54 struct tipc_portid peer_name; 55 long conn_timeout; 56 }; 57 58 #define tipc_sk(sk) ((struct tipc_sock *)(sk)) 59 #define tipc_sk_port(sk) ((struct tipc_port *)(tipc_sk(sk)->p)) 60 61 static int backlog_rcv(struct sock *sk, struct sk_buff *skb); 62 static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf); 63 static void wakeupdispatch(struct tipc_port *tport); 64 65 static const struct proto_ops packet_ops; 66 static const struct proto_ops stream_ops; 67 static const struct proto_ops msg_ops; 68 69 static struct proto tipc_proto; 70 71 static int sockets_enabled; 72 73 static atomic_t tipc_queue_size = ATOMIC_INIT(0); 74 75 /* 76 * Revised TIPC socket locking policy: 77 * 78 * Most socket operations take the standard socket lock when they start 79 * and hold it until they finish (or until they need to sleep). Acquiring 80 * this lock grants the owner exclusive access to the fields of the socket 81 * data structures, with the exception of the backlog queue. A few socket 82 * operations can be done without taking the socket lock because they only 83 * read socket information that never changes during the life of the socket. 84 * 85 * Socket operations may acquire the lock for the associated TIPC port if they 86 * need to perform an operation on the port. If any routine needs to acquire 87 * both the socket lock and the port lock it must take the socket lock first 88 * to avoid the risk of deadlock. 89 * 90 * The dispatcher handling incoming messages cannot grab the socket lock in 91 * the standard fashion, since invoked it runs at the BH level and cannot block. 92 * Instead, it checks to see if the socket lock is currently owned by someone, 93 * and either handles the message itself or adds it to the socket's backlog 94 * queue; in the latter case the queued message is processed once the process 95 * owning the socket lock releases it. 96 * 97 * NOTE: Releasing the socket lock while an operation is sleeping overcomes 98 * the problem of a blocked socket operation preventing any other operations 99 * from occurring. However, applications must be careful if they have 100 * multiple threads trying to send (or receive) on the same socket, as these 101 * operations might interfere with each other. For example, doing a connect 102 * and a receive at the same time might allow the receive to consume the 103 * ACK message meant for the connect. While additional work could be done 104 * to try and overcome this, it doesn't seem to be worthwhile at the present. 105 * 106 * NOTE: Releasing the socket lock while an operation is sleeping also ensures 107 * that another operation that must be performed in a non-blocking manner is 108 * not delayed for very long because the lock has already been taken. 109 * 110 * NOTE: This code assumes that certain fields of a port/socket pair are 111 * constant over its lifetime; such fields can be examined without taking 112 * the socket lock and/or port lock, and do not need to be re-read even 113 * after resuming processing after waiting. These fields include: 114 * - socket type 115 * - pointer to socket sk structure (aka tipc_sock structure) 116 * - pointer to port structure 117 * - port reference 118 */ 119 120 /** 121 * advance_rx_queue - discard first buffer in socket receive queue 122 * 123 * Caller must hold socket lock 124 */ 125 126 static void advance_rx_queue(struct sock *sk) 127 { 128 buf_discard(__skb_dequeue(&sk->sk_receive_queue)); 129 atomic_dec(&tipc_queue_size); 130 } 131 132 /** 133 * discard_rx_queue - discard all buffers in socket receive queue 134 * 135 * Caller must hold socket lock 136 */ 137 138 static void discard_rx_queue(struct sock *sk) 139 { 140 struct sk_buff *buf; 141 142 while ((buf = __skb_dequeue(&sk->sk_receive_queue))) { 143 atomic_dec(&tipc_queue_size); 144 buf_discard(buf); 145 } 146 } 147 148 /** 149 * reject_rx_queue - reject all buffers in socket receive queue 150 * 151 * Caller must hold socket lock 152 */ 153 154 static void reject_rx_queue(struct sock *sk) 155 { 156 struct sk_buff *buf; 157 158 while ((buf = __skb_dequeue(&sk->sk_receive_queue))) { 159 tipc_reject_msg(buf, TIPC_ERR_NO_PORT); 160 atomic_dec(&tipc_queue_size); 161 } 162 } 163 164 /** 165 * tipc_create - create a TIPC socket 166 * @net: network namespace (must be default network) 167 * @sock: pre-allocated socket structure 168 * @protocol: protocol indicator (must be 0) 169 * @kern: caused by kernel or by userspace? 170 * 171 * This routine creates additional data structures used by the TIPC socket, 172 * initializes them, and links them together. 173 * 174 * Returns 0 on success, errno otherwise 175 */ 176 177 static int tipc_create(struct net *net, struct socket *sock, int protocol, 178 int kern) 179 { 180 const struct proto_ops *ops; 181 socket_state state; 182 struct sock *sk; 183 struct tipc_port *tp_ptr; 184 185 /* Validate arguments */ 186 187 if (!net_eq(net, &init_net)) 188 return -EAFNOSUPPORT; 189 190 if (unlikely(protocol != 0)) 191 return -EPROTONOSUPPORT; 192 193 switch (sock->type) { 194 case SOCK_STREAM: 195 ops = &stream_ops; 196 state = SS_UNCONNECTED; 197 break; 198 case SOCK_SEQPACKET: 199 ops = &packet_ops; 200 state = SS_UNCONNECTED; 201 break; 202 case SOCK_DGRAM: 203 case SOCK_RDM: 204 ops = &msg_ops; 205 state = SS_READY; 206 break; 207 default: 208 return -EPROTOTYPE; 209 } 210 211 /* Allocate socket's protocol area */ 212 213 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto); 214 if (sk == NULL) 215 return -ENOMEM; 216 217 /* Allocate TIPC port for socket to use */ 218 219 tp_ptr = tipc_createport_raw(sk, &dispatch, &wakeupdispatch, 220 TIPC_LOW_IMPORTANCE); 221 if (unlikely(!tp_ptr)) { 222 sk_free(sk); 223 return -ENOMEM; 224 } 225 226 /* Finish initializing socket data structures */ 227 228 sock->ops = ops; 229 sock->state = state; 230 231 sock_init_data(sock, sk); 232 sk->sk_backlog_rcv = backlog_rcv; 233 tipc_sk(sk)->p = tp_ptr; 234 tipc_sk(sk)->conn_timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT); 235 236 spin_unlock_bh(tp_ptr->lock); 237 238 if (sock->state == SS_READY) { 239 tipc_set_portunreturnable(tp_ptr->ref, 1); 240 if (sock->type == SOCK_DGRAM) 241 tipc_set_portunreliable(tp_ptr->ref, 1); 242 } 243 244 atomic_inc(&tipc_user_count); 245 return 0; 246 } 247 248 /** 249 * release - destroy a TIPC socket 250 * @sock: socket to destroy 251 * 252 * This routine cleans up any messages that are still queued on the socket. 253 * For DGRAM and RDM socket types, all queued messages are rejected. 254 * For SEQPACKET and STREAM socket types, the first message is rejected 255 * and any others are discarded. (If the first message on a STREAM socket 256 * is partially-read, it is discarded and the next one is rejected instead.) 257 * 258 * NOTE: Rejected messages are not necessarily returned to the sender! They 259 * are returned or discarded according to the "destination droppable" setting 260 * specified for the message by the sender. 261 * 262 * Returns 0 on success, errno otherwise 263 */ 264 265 static int release(struct socket *sock) 266 { 267 struct sock *sk = sock->sk; 268 struct tipc_port *tport; 269 struct sk_buff *buf; 270 int res; 271 272 /* 273 * Exit if socket isn't fully initialized (occurs when a failed accept() 274 * releases a pre-allocated child socket that was never used) 275 */ 276 277 if (sk == NULL) 278 return 0; 279 280 tport = tipc_sk_port(sk); 281 lock_sock(sk); 282 283 /* 284 * Reject all unreceived messages, except on an active connection 285 * (which disconnects locally & sends a 'FIN+' to peer) 286 */ 287 288 while (sock->state != SS_DISCONNECTING) { 289 buf = __skb_dequeue(&sk->sk_receive_queue); 290 if (buf == NULL) 291 break; 292 atomic_dec(&tipc_queue_size); 293 if (TIPC_SKB_CB(buf)->handle != msg_data(buf_msg(buf))) 294 buf_discard(buf); 295 else { 296 if ((sock->state == SS_CONNECTING) || 297 (sock->state == SS_CONNECTED)) { 298 sock->state = SS_DISCONNECTING; 299 tipc_disconnect(tport->ref); 300 } 301 tipc_reject_msg(buf, TIPC_ERR_NO_PORT); 302 } 303 } 304 305 /* 306 * Delete TIPC port; this ensures no more messages are queued 307 * (also disconnects an active connection & sends a 'FIN-' to peer) 308 */ 309 310 res = tipc_deleteport(tport->ref); 311 312 /* Discard any remaining (connection-based) messages in receive queue */ 313 314 discard_rx_queue(sk); 315 316 /* Reject any messages that accumulated in backlog queue */ 317 318 sock->state = SS_DISCONNECTING; 319 release_sock(sk); 320 321 sock_put(sk); 322 sock->sk = NULL; 323 324 atomic_dec(&tipc_user_count); 325 return res; 326 } 327 328 /** 329 * bind - associate or disassocate TIPC name(s) with a socket 330 * @sock: socket structure 331 * @uaddr: socket address describing name(s) and desired operation 332 * @uaddr_len: size of socket address data structure 333 * 334 * Name and name sequence binding is indicated using a positive scope value; 335 * a negative scope value unbinds the specified name. Specifying no name 336 * (i.e. a socket address length of 0) unbinds all names from the socket. 337 * 338 * Returns 0 on success, errno otherwise 339 * 340 * NOTE: This routine doesn't need to take the socket lock since it doesn't 341 * access any non-constant socket information. 342 */ 343 344 static int bind(struct socket *sock, struct sockaddr *uaddr, int uaddr_len) 345 { 346 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr; 347 u32 portref = tipc_sk_port(sock->sk)->ref; 348 349 if (unlikely(!uaddr_len)) 350 return tipc_withdraw(portref, 0, NULL); 351 352 if (uaddr_len < sizeof(struct sockaddr_tipc)) 353 return -EINVAL; 354 if (addr->family != AF_TIPC) 355 return -EAFNOSUPPORT; 356 357 if (addr->addrtype == TIPC_ADDR_NAME) 358 addr->addr.nameseq.upper = addr->addr.nameseq.lower; 359 else if (addr->addrtype != TIPC_ADDR_NAMESEQ) 360 return -EAFNOSUPPORT; 361 362 return (addr->scope > 0) ? 363 tipc_publish(portref, addr->scope, &addr->addr.nameseq) : 364 tipc_withdraw(portref, -addr->scope, &addr->addr.nameseq); 365 } 366 367 /** 368 * get_name - get port ID of socket or peer socket 369 * @sock: socket structure 370 * @uaddr: area for returned socket address 371 * @uaddr_len: area for returned length of socket address 372 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID 373 * 374 * Returns 0 on success, errno otherwise 375 * 376 * NOTE: This routine doesn't need to take the socket lock since it only 377 * accesses socket information that is unchanging (or which changes in 378 * a completely predictable manner). 379 */ 380 381 static int get_name(struct socket *sock, struct sockaddr *uaddr, 382 int *uaddr_len, int peer) 383 { 384 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr; 385 struct tipc_sock *tsock = tipc_sk(sock->sk); 386 387 memset(addr, 0, sizeof(*addr)); 388 if (peer) { 389 if ((sock->state != SS_CONNECTED) && 390 ((peer != 2) || (sock->state != SS_DISCONNECTING))) 391 return -ENOTCONN; 392 addr->addr.id.ref = tsock->peer_name.ref; 393 addr->addr.id.node = tsock->peer_name.node; 394 } else { 395 addr->addr.id.ref = tsock->p->ref; 396 addr->addr.id.node = tipc_own_addr; 397 } 398 399 *uaddr_len = sizeof(*addr); 400 addr->addrtype = TIPC_ADDR_ID; 401 addr->family = AF_TIPC; 402 addr->scope = 0; 403 addr->addr.name.domain = 0; 404 405 return 0; 406 } 407 408 /** 409 * poll - read and possibly block on pollmask 410 * @file: file structure associated with the socket 411 * @sock: socket for which to calculate the poll bits 412 * @wait: ??? 413 * 414 * Returns pollmask value 415 * 416 * COMMENTARY: 417 * It appears that the usual socket locking mechanisms are not useful here 418 * since the pollmask info is potentially out-of-date the moment this routine 419 * exits. TCP and other protocols seem to rely on higher level poll routines 420 * to handle any preventable race conditions, so TIPC will do the same ... 421 * 422 * TIPC sets the returned events as follows: 423 * 424 * socket state flags set 425 * ------------ --------- 426 * unconnected no read flags 427 * no write flags 428 * 429 * connecting POLLIN/POLLRDNORM if ACK/NACK in rx queue 430 * no write flags 431 * 432 * connected POLLIN/POLLRDNORM if data in rx queue 433 * POLLOUT if port is not congested 434 * 435 * disconnecting POLLIN/POLLRDNORM/POLLHUP 436 * no write flags 437 * 438 * listening POLLIN if SYN in rx queue 439 * no write flags 440 * 441 * ready POLLIN/POLLRDNORM if data in rx queue 442 * [connectionless] POLLOUT (since port cannot be congested) 443 * 444 * IMPORTANT: The fact that a read or write operation is indicated does NOT 445 * imply that the operation will succeed, merely that it should be performed 446 * and will not block. 447 */ 448 449 static unsigned int poll(struct file *file, struct socket *sock, 450 poll_table *wait) 451 { 452 struct sock *sk = sock->sk; 453 u32 mask = 0; 454 455 poll_wait(file, sk_sleep(sk), wait); 456 457 switch ((int)sock->state) { 458 case SS_READY: 459 case SS_CONNECTED: 460 if (!tipc_sk_port(sk)->congested) 461 mask |= POLLOUT; 462 /* fall thru' */ 463 case SS_CONNECTING: 464 case SS_LISTENING: 465 if (!skb_queue_empty(&sk->sk_receive_queue)) 466 mask |= (POLLIN | POLLRDNORM); 467 break; 468 case SS_DISCONNECTING: 469 mask = (POLLIN | POLLRDNORM | POLLHUP); 470 break; 471 } 472 473 return mask; 474 } 475 476 /** 477 * dest_name_check - verify user is permitted to send to specified port name 478 * @dest: destination address 479 * @m: descriptor for message to be sent 480 * 481 * Prevents restricted configuration commands from being issued by 482 * unauthorized users. 483 * 484 * Returns 0 if permission is granted, otherwise errno 485 */ 486 487 static int dest_name_check(struct sockaddr_tipc *dest, struct msghdr *m) 488 { 489 struct tipc_cfg_msg_hdr hdr; 490 491 if (likely(dest->addr.name.name.type >= TIPC_RESERVED_TYPES)) 492 return 0; 493 if (likely(dest->addr.name.name.type == TIPC_TOP_SRV)) 494 return 0; 495 if (likely(dest->addr.name.name.type != TIPC_CFG_SRV)) 496 return -EACCES; 497 498 if (copy_from_user(&hdr, m->msg_iov[0].iov_base, sizeof(hdr))) 499 return -EFAULT; 500 if ((ntohs(hdr.tcm_type) & 0xC000) && (!capable(CAP_NET_ADMIN))) 501 return -EACCES; 502 503 return 0; 504 } 505 506 /** 507 * send_msg - send message in connectionless manner 508 * @iocb: if NULL, indicates that socket lock is already held 509 * @sock: socket structure 510 * @m: message to send 511 * @total_len: length of message 512 * 513 * Message must have an destination specified explicitly. 514 * Used for SOCK_RDM and SOCK_DGRAM messages, 515 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections. 516 * (Note: 'SYN+' is prohibited on SOCK_STREAM.) 517 * 518 * Returns the number of bytes sent on success, or errno otherwise 519 */ 520 521 static int send_msg(struct kiocb *iocb, struct socket *sock, 522 struct msghdr *m, size_t total_len) 523 { 524 struct sock *sk = sock->sk; 525 struct tipc_port *tport = tipc_sk_port(sk); 526 struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name; 527 int needs_conn; 528 int res = -EINVAL; 529 530 if (unlikely(!dest)) 531 return -EDESTADDRREQ; 532 if (unlikely((m->msg_namelen < sizeof(*dest)) || 533 (dest->family != AF_TIPC))) 534 return -EINVAL; 535 536 if (iocb) 537 lock_sock(sk); 538 539 needs_conn = (sock->state != SS_READY); 540 if (unlikely(needs_conn)) { 541 if (sock->state == SS_LISTENING) { 542 res = -EPIPE; 543 goto exit; 544 } 545 if (sock->state != SS_UNCONNECTED) { 546 res = -EISCONN; 547 goto exit; 548 } 549 if ((tport->published) || 550 ((sock->type == SOCK_STREAM) && (total_len != 0))) { 551 res = -EOPNOTSUPP; 552 goto exit; 553 } 554 if (dest->addrtype == TIPC_ADDR_NAME) { 555 tport->conn_type = dest->addr.name.name.type; 556 tport->conn_instance = dest->addr.name.name.instance; 557 } 558 559 /* Abort any pending connection attempts (very unlikely) */ 560 561 reject_rx_queue(sk); 562 } 563 564 do { 565 if (dest->addrtype == TIPC_ADDR_NAME) { 566 res = dest_name_check(dest, m); 567 if (res) 568 break; 569 res = tipc_send2name(tport->ref, 570 &dest->addr.name.name, 571 dest->addr.name.domain, 572 m->msg_iovlen, 573 m->msg_iov); 574 } else if (dest->addrtype == TIPC_ADDR_ID) { 575 res = tipc_send2port(tport->ref, 576 &dest->addr.id, 577 m->msg_iovlen, 578 m->msg_iov); 579 } else if (dest->addrtype == TIPC_ADDR_MCAST) { 580 if (needs_conn) { 581 res = -EOPNOTSUPP; 582 break; 583 } 584 res = dest_name_check(dest, m); 585 if (res) 586 break; 587 res = tipc_multicast(tport->ref, 588 &dest->addr.nameseq, 589 m->msg_iovlen, 590 m->msg_iov); 591 } 592 if (likely(res != -ELINKCONG)) { 593 if (needs_conn && (res >= 0)) 594 sock->state = SS_CONNECTING; 595 break; 596 } 597 if (m->msg_flags & MSG_DONTWAIT) { 598 res = -EWOULDBLOCK; 599 break; 600 } 601 release_sock(sk); 602 res = wait_event_interruptible(*sk_sleep(sk), 603 !tport->congested); 604 lock_sock(sk); 605 if (res) 606 break; 607 } while (1); 608 609 exit: 610 if (iocb) 611 release_sock(sk); 612 return res; 613 } 614 615 /** 616 * send_packet - send a connection-oriented message 617 * @iocb: if NULL, indicates that socket lock is already held 618 * @sock: socket structure 619 * @m: message to send 620 * @total_len: length of message 621 * 622 * Used for SOCK_SEQPACKET messages and SOCK_STREAM data. 623 * 624 * Returns the number of bytes sent on success, or errno otherwise 625 */ 626 627 static int send_packet(struct kiocb *iocb, struct socket *sock, 628 struct msghdr *m, size_t total_len) 629 { 630 struct sock *sk = sock->sk; 631 struct tipc_port *tport = tipc_sk_port(sk); 632 struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name; 633 int res; 634 635 /* Handle implied connection establishment */ 636 637 if (unlikely(dest)) 638 return send_msg(iocb, sock, m, total_len); 639 640 if (iocb) 641 lock_sock(sk); 642 643 do { 644 if (unlikely(sock->state != SS_CONNECTED)) { 645 if (sock->state == SS_DISCONNECTING) 646 res = -EPIPE; 647 else 648 res = -ENOTCONN; 649 break; 650 } 651 652 res = tipc_send(tport->ref, m->msg_iovlen, m->msg_iov); 653 if (likely(res != -ELINKCONG)) 654 break; 655 if (m->msg_flags & MSG_DONTWAIT) { 656 res = -EWOULDBLOCK; 657 break; 658 } 659 release_sock(sk); 660 res = wait_event_interruptible(*sk_sleep(sk), 661 (!tport->congested || !tport->connected)); 662 lock_sock(sk); 663 if (res) 664 break; 665 } while (1); 666 667 if (iocb) 668 release_sock(sk); 669 return res; 670 } 671 672 /** 673 * send_stream - send stream-oriented data 674 * @iocb: (unused) 675 * @sock: socket structure 676 * @m: data to send 677 * @total_len: total length of data to be sent 678 * 679 * Used for SOCK_STREAM data. 680 * 681 * Returns the number of bytes sent on success (or partial success), 682 * or errno if no data sent 683 */ 684 685 static int send_stream(struct kiocb *iocb, struct socket *sock, 686 struct msghdr *m, size_t total_len) 687 { 688 struct sock *sk = sock->sk; 689 struct tipc_port *tport = tipc_sk_port(sk); 690 struct msghdr my_msg; 691 struct iovec my_iov; 692 struct iovec *curr_iov; 693 int curr_iovlen; 694 char __user *curr_start; 695 u32 hdr_size; 696 int curr_left; 697 int bytes_to_send; 698 int bytes_sent; 699 int res; 700 701 lock_sock(sk); 702 703 /* Handle special cases where there is no connection */ 704 705 if (unlikely(sock->state != SS_CONNECTED)) { 706 if (sock->state == SS_UNCONNECTED) { 707 res = send_packet(NULL, sock, m, total_len); 708 goto exit; 709 } else if (sock->state == SS_DISCONNECTING) { 710 res = -EPIPE; 711 goto exit; 712 } else { 713 res = -ENOTCONN; 714 goto exit; 715 } 716 } 717 718 if (unlikely(m->msg_name)) { 719 res = -EISCONN; 720 goto exit; 721 } 722 723 /* 724 * Send each iovec entry using one or more messages 725 * 726 * Note: This algorithm is good for the most likely case 727 * (i.e. one large iovec entry), but could be improved to pass sets 728 * of small iovec entries into send_packet(). 729 */ 730 731 curr_iov = m->msg_iov; 732 curr_iovlen = m->msg_iovlen; 733 my_msg.msg_iov = &my_iov; 734 my_msg.msg_iovlen = 1; 735 my_msg.msg_flags = m->msg_flags; 736 my_msg.msg_name = NULL; 737 bytes_sent = 0; 738 739 hdr_size = msg_hdr_sz(&tport->phdr); 740 741 while (curr_iovlen--) { 742 curr_start = curr_iov->iov_base; 743 curr_left = curr_iov->iov_len; 744 745 while (curr_left) { 746 bytes_to_send = tport->max_pkt - hdr_size; 747 if (bytes_to_send > TIPC_MAX_USER_MSG_SIZE) 748 bytes_to_send = TIPC_MAX_USER_MSG_SIZE; 749 if (curr_left < bytes_to_send) 750 bytes_to_send = curr_left; 751 my_iov.iov_base = curr_start; 752 my_iov.iov_len = bytes_to_send; 753 res = send_packet(NULL, sock, &my_msg, 0); 754 if (res < 0) { 755 if (bytes_sent) 756 res = bytes_sent; 757 goto exit; 758 } 759 curr_left -= bytes_to_send; 760 curr_start += bytes_to_send; 761 bytes_sent += bytes_to_send; 762 } 763 764 curr_iov++; 765 } 766 res = bytes_sent; 767 exit: 768 release_sock(sk); 769 return res; 770 } 771 772 /** 773 * auto_connect - complete connection setup to a remote port 774 * @sock: socket structure 775 * @msg: peer's response message 776 * 777 * Returns 0 on success, errno otherwise 778 */ 779 780 static int auto_connect(struct socket *sock, struct tipc_msg *msg) 781 { 782 struct tipc_sock *tsock = tipc_sk(sock->sk); 783 784 if (msg_errcode(msg)) { 785 sock->state = SS_DISCONNECTING; 786 return -ECONNREFUSED; 787 } 788 789 tsock->peer_name.ref = msg_origport(msg); 790 tsock->peer_name.node = msg_orignode(msg); 791 tipc_connect2port(tsock->p->ref, &tsock->peer_name); 792 tipc_set_portimportance(tsock->p->ref, msg_importance(msg)); 793 sock->state = SS_CONNECTED; 794 return 0; 795 } 796 797 /** 798 * set_orig_addr - capture sender's address for received message 799 * @m: descriptor for message info 800 * @msg: received message header 801 * 802 * Note: Address is not captured if not requested by receiver. 803 */ 804 805 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg) 806 { 807 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)m->msg_name; 808 809 if (addr) { 810 addr->family = AF_TIPC; 811 addr->addrtype = TIPC_ADDR_ID; 812 addr->addr.id.ref = msg_origport(msg); 813 addr->addr.id.node = msg_orignode(msg); 814 addr->addr.name.domain = 0; /* could leave uninitialized */ 815 addr->scope = 0; /* could leave uninitialized */ 816 m->msg_namelen = sizeof(struct sockaddr_tipc); 817 } 818 } 819 820 /** 821 * anc_data_recv - optionally capture ancillary data for received message 822 * @m: descriptor for message info 823 * @msg: received message header 824 * @tport: TIPC port associated with message 825 * 826 * Note: Ancillary data is not captured if not requested by receiver. 827 * 828 * Returns 0 if successful, otherwise errno 829 */ 830 831 static int anc_data_recv(struct msghdr *m, struct tipc_msg *msg, 832 struct tipc_port *tport) 833 { 834 u32 anc_data[3]; 835 u32 err; 836 u32 dest_type; 837 int has_name; 838 int res; 839 840 if (likely(m->msg_controllen == 0)) 841 return 0; 842 843 /* Optionally capture errored message object(s) */ 844 845 err = msg ? msg_errcode(msg) : 0; 846 if (unlikely(err)) { 847 anc_data[0] = err; 848 anc_data[1] = msg_data_sz(msg); 849 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data); 850 if (res) 851 return res; 852 if (anc_data[1]) { 853 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1], 854 msg_data(msg)); 855 if (res) 856 return res; 857 } 858 } 859 860 /* Optionally capture message destination object */ 861 862 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG; 863 switch (dest_type) { 864 case TIPC_NAMED_MSG: 865 has_name = 1; 866 anc_data[0] = msg_nametype(msg); 867 anc_data[1] = msg_namelower(msg); 868 anc_data[2] = msg_namelower(msg); 869 break; 870 case TIPC_MCAST_MSG: 871 has_name = 1; 872 anc_data[0] = msg_nametype(msg); 873 anc_data[1] = msg_namelower(msg); 874 anc_data[2] = msg_nameupper(msg); 875 break; 876 case TIPC_CONN_MSG: 877 has_name = (tport->conn_type != 0); 878 anc_data[0] = tport->conn_type; 879 anc_data[1] = tport->conn_instance; 880 anc_data[2] = tport->conn_instance; 881 break; 882 default: 883 has_name = 0; 884 } 885 if (has_name) { 886 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data); 887 if (res) 888 return res; 889 } 890 891 return 0; 892 } 893 894 /** 895 * recv_msg - receive packet-oriented message 896 * @iocb: (unused) 897 * @m: descriptor for message info 898 * @buf_len: total size of user buffer area 899 * @flags: receive flags 900 * 901 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages. 902 * If the complete message doesn't fit in user area, truncate it. 903 * 904 * Returns size of returned message data, errno otherwise 905 */ 906 907 static int recv_msg(struct kiocb *iocb, struct socket *sock, 908 struct msghdr *m, size_t buf_len, int flags) 909 { 910 struct sock *sk = sock->sk; 911 struct tipc_port *tport = tipc_sk_port(sk); 912 struct sk_buff *buf; 913 struct tipc_msg *msg; 914 unsigned int sz; 915 u32 err; 916 int res; 917 918 /* Catch invalid receive requests */ 919 920 if (m->msg_iovlen != 1) 921 return -EOPNOTSUPP; /* Don't do multiple iovec entries yet */ 922 923 if (unlikely(!buf_len)) 924 return -EINVAL; 925 926 lock_sock(sk); 927 928 if (unlikely(sock->state == SS_UNCONNECTED)) { 929 res = -ENOTCONN; 930 goto exit; 931 } 932 933 restart: 934 935 /* Look for a message in receive queue; wait if necessary */ 936 937 while (skb_queue_empty(&sk->sk_receive_queue)) { 938 if (sock->state == SS_DISCONNECTING) { 939 res = -ENOTCONN; 940 goto exit; 941 } 942 if (flags & MSG_DONTWAIT) { 943 res = -EWOULDBLOCK; 944 goto exit; 945 } 946 release_sock(sk); 947 res = wait_event_interruptible(*sk_sleep(sk), 948 (!skb_queue_empty(&sk->sk_receive_queue) || 949 (sock->state == SS_DISCONNECTING))); 950 lock_sock(sk); 951 if (res) 952 goto exit; 953 } 954 955 /* Look at first message in receive queue */ 956 957 buf = skb_peek(&sk->sk_receive_queue); 958 msg = buf_msg(buf); 959 sz = msg_data_sz(msg); 960 err = msg_errcode(msg); 961 962 /* Complete connection setup for an implied connect */ 963 964 if (unlikely(sock->state == SS_CONNECTING)) { 965 res = auto_connect(sock, msg); 966 if (res) 967 goto exit; 968 } 969 970 /* Discard an empty non-errored message & try again */ 971 972 if ((!sz) && (!err)) { 973 advance_rx_queue(sk); 974 goto restart; 975 } 976 977 /* Capture sender's address (optional) */ 978 979 set_orig_addr(m, msg); 980 981 /* Capture ancillary data (optional) */ 982 983 res = anc_data_recv(m, msg, tport); 984 if (res) 985 goto exit; 986 987 /* Capture message data (if valid) & compute return value (always) */ 988 989 if (!err) { 990 if (unlikely(buf_len < sz)) { 991 sz = buf_len; 992 m->msg_flags |= MSG_TRUNC; 993 } 994 if (unlikely(copy_to_user(m->msg_iov->iov_base, msg_data(msg), 995 sz))) { 996 res = -EFAULT; 997 goto exit; 998 } 999 res = sz; 1000 } else { 1001 if ((sock->state == SS_READY) || 1002 ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)) 1003 res = 0; 1004 else 1005 res = -ECONNRESET; 1006 } 1007 1008 /* Consume received message (optional) */ 1009 1010 if (likely(!(flags & MSG_PEEK))) { 1011 if ((sock->state != SS_READY) && 1012 (++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN)) 1013 tipc_acknowledge(tport->ref, tport->conn_unacked); 1014 advance_rx_queue(sk); 1015 } 1016 exit: 1017 release_sock(sk); 1018 return res; 1019 } 1020 1021 /** 1022 * recv_stream - receive stream-oriented data 1023 * @iocb: (unused) 1024 * @m: descriptor for message info 1025 * @buf_len: total size of user buffer area 1026 * @flags: receive flags 1027 * 1028 * Used for SOCK_STREAM messages only. If not enough data is available 1029 * will optionally wait for more; never truncates data. 1030 * 1031 * Returns size of returned message data, errno otherwise 1032 */ 1033 1034 static int recv_stream(struct kiocb *iocb, struct socket *sock, 1035 struct msghdr *m, size_t buf_len, int flags) 1036 { 1037 struct sock *sk = sock->sk; 1038 struct tipc_port *tport = tipc_sk_port(sk); 1039 struct sk_buff *buf; 1040 struct tipc_msg *msg; 1041 unsigned int sz; 1042 int sz_to_copy, target, needed; 1043 int sz_copied = 0; 1044 char __user *crs = m->msg_iov->iov_base; 1045 unsigned char *buf_crs; 1046 u32 err; 1047 int res = 0; 1048 1049 /* Catch invalid receive attempts */ 1050 1051 if (m->msg_iovlen != 1) 1052 return -EOPNOTSUPP; /* Don't do multiple iovec entries yet */ 1053 1054 if (unlikely(!buf_len)) 1055 return -EINVAL; 1056 1057 lock_sock(sk); 1058 1059 if (unlikely((sock->state == SS_UNCONNECTED) || 1060 (sock->state == SS_CONNECTING))) { 1061 res = -ENOTCONN; 1062 goto exit; 1063 } 1064 1065 target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len); 1066 1067 restart: 1068 1069 /* Look for a message in receive queue; wait if necessary */ 1070 1071 while (skb_queue_empty(&sk->sk_receive_queue)) { 1072 if (sock->state == SS_DISCONNECTING) { 1073 res = -ENOTCONN; 1074 goto exit; 1075 } 1076 if (flags & MSG_DONTWAIT) { 1077 res = -EWOULDBLOCK; 1078 goto exit; 1079 } 1080 release_sock(sk); 1081 res = wait_event_interruptible(*sk_sleep(sk), 1082 (!skb_queue_empty(&sk->sk_receive_queue) || 1083 (sock->state == SS_DISCONNECTING))); 1084 lock_sock(sk); 1085 if (res) 1086 goto exit; 1087 } 1088 1089 /* Look at first message in receive queue */ 1090 1091 buf = skb_peek(&sk->sk_receive_queue); 1092 msg = buf_msg(buf); 1093 sz = msg_data_sz(msg); 1094 err = msg_errcode(msg); 1095 1096 /* Discard an empty non-errored message & try again */ 1097 1098 if ((!sz) && (!err)) { 1099 advance_rx_queue(sk); 1100 goto restart; 1101 } 1102 1103 /* Optionally capture sender's address & ancillary data of first msg */ 1104 1105 if (sz_copied == 0) { 1106 set_orig_addr(m, msg); 1107 res = anc_data_recv(m, msg, tport); 1108 if (res) 1109 goto exit; 1110 } 1111 1112 /* Capture message data (if valid) & compute return value (always) */ 1113 1114 if (!err) { 1115 buf_crs = (unsigned char *)(TIPC_SKB_CB(buf)->handle); 1116 sz = (unsigned char *)msg + msg_size(msg) - buf_crs; 1117 1118 needed = (buf_len - sz_copied); 1119 sz_to_copy = (sz <= needed) ? sz : needed; 1120 if (unlikely(copy_to_user(crs, buf_crs, sz_to_copy))) { 1121 res = -EFAULT; 1122 goto exit; 1123 } 1124 sz_copied += sz_to_copy; 1125 1126 if (sz_to_copy < sz) { 1127 if (!(flags & MSG_PEEK)) 1128 TIPC_SKB_CB(buf)->handle = buf_crs + sz_to_copy; 1129 goto exit; 1130 } 1131 1132 crs += sz_to_copy; 1133 } else { 1134 if (sz_copied != 0) 1135 goto exit; /* can't add error msg to valid data */ 1136 1137 if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control) 1138 res = 0; 1139 else 1140 res = -ECONNRESET; 1141 } 1142 1143 /* Consume received message (optional) */ 1144 1145 if (likely(!(flags & MSG_PEEK))) { 1146 if (unlikely(++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN)) 1147 tipc_acknowledge(tport->ref, tport->conn_unacked); 1148 advance_rx_queue(sk); 1149 } 1150 1151 /* Loop around if more data is required */ 1152 1153 if ((sz_copied < buf_len) && /* didn't get all requested data */ 1154 (!skb_queue_empty(&sk->sk_receive_queue) || 1155 (sz_copied < target)) && /* and more is ready or required */ 1156 (!(flags & MSG_PEEK)) && /* and aren't just peeking at data */ 1157 (!err)) /* and haven't reached a FIN */ 1158 goto restart; 1159 1160 exit: 1161 release_sock(sk); 1162 return sz_copied ? sz_copied : res; 1163 } 1164 1165 /** 1166 * rx_queue_full - determine if receive queue can accept another message 1167 * @msg: message to be added to queue 1168 * @queue_size: current size of queue 1169 * @base: nominal maximum size of queue 1170 * 1171 * Returns 1 if queue is unable to accept message, 0 otherwise 1172 */ 1173 1174 static int rx_queue_full(struct tipc_msg *msg, u32 queue_size, u32 base) 1175 { 1176 u32 threshold; 1177 u32 imp = msg_importance(msg); 1178 1179 if (imp == TIPC_LOW_IMPORTANCE) 1180 threshold = base; 1181 else if (imp == TIPC_MEDIUM_IMPORTANCE) 1182 threshold = base * 2; 1183 else if (imp == TIPC_HIGH_IMPORTANCE) 1184 threshold = base * 100; 1185 else 1186 return 0; 1187 1188 if (msg_connected(msg)) 1189 threshold *= 4; 1190 1191 return queue_size >= threshold; 1192 } 1193 1194 /** 1195 * filter_rcv - validate incoming message 1196 * @sk: socket 1197 * @buf: message 1198 * 1199 * Enqueues message on receive queue if acceptable; optionally handles 1200 * disconnect indication for a connected socket. 1201 * 1202 * Called with socket lock already taken; port lock may also be taken. 1203 * 1204 * Returns TIPC error status code (TIPC_OK if message is not to be rejected) 1205 */ 1206 1207 static u32 filter_rcv(struct sock *sk, struct sk_buff *buf) 1208 { 1209 struct socket *sock = sk->sk_socket; 1210 struct tipc_msg *msg = buf_msg(buf); 1211 u32 recv_q_len; 1212 1213 /* Reject message if it is wrong sort of message for socket */ 1214 1215 /* 1216 * WOULD IT BE BETTER TO JUST DISCARD THESE MESSAGES INSTEAD? 1217 * "NO PORT" ISN'T REALLY THE RIGHT ERROR CODE, AND THERE MAY 1218 * BE SECURITY IMPLICATIONS INHERENT IN REJECTING INVALID TRAFFIC 1219 */ 1220 1221 if (sock->state == SS_READY) { 1222 if (msg_connected(msg)) 1223 return TIPC_ERR_NO_PORT; 1224 } else { 1225 if (msg_mcast(msg)) 1226 return TIPC_ERR_NO_PORT; 1227 if (sock->state == SS_CONNECTED) { 1228 if (!msg_connected(msg)) 1229 return TIPC_ERR_NO_PORT; 1230 } else if (sock->state == SS_CONNECTING) { 1231 if (!msg_connected(msg) && (msg_errcode(msg) == 0)) 1232 return TIPC_ERR_NO_PORT; 1233 } else if (sock->state == SS_LISTENING) { 1234 if (msg_connected(msg) || msg_errcode(msg)) 1235 return TIPC_ERR_NO_PORT; 1236 } else if (sock->state == SS_DISCONNECTING) { 1237 return TIPC_ERR_NO_PORT; 1238 } else /* (sock->state == SS_UNCONNECTED) */ { 1239 if (msg_connected(msg) || msg_errcode(msg)) 1240 return TIPC_ERR_NO_PORT; 1241 } 1242 } 1243 1244 /* Reject message if there isn't room to queue it */ 1245 1246 recv_q_len = (u32)atomic_read(&tipc_queue_size); 1247 if (unlikely(recv_q_len >= OVERLOAD_LIMIT_BASE)) { 1248 if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE)) 1249 return TIPC_ERR_OVERLOAD; 1250 } 1251 recv_q_len = skb_queue_len(&sk->sk_receive_queue); 1252 if (unlikely(recv_q_len >= (OVERLOAD_LIMIT_BASE / 2))) { 1253 if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE / 2)) 1254 return TIPC_ERR_OVERLOAD; 1255 } 1256 1257 /* Enqueue message (finally!) */ 1258 1259 TIPC_SKB_CB(buf)->handle = msg_data(msg); 1260 atomic_inc(&tipc_queue_size); 1261 __skb_queue_tail(&sk->sk_receive_queue, buf); 1262 1263 /* Initiate connection termination for an incoming 'FIN' */ 1264 1265 if (unlikely(msg_errcode(msg) && (sock->state == SS_CONNECTED))) { 1266 sock->state = SS_DISCONNECTING; 1267 tipc_disconnect_port(tipc_sk_port(sk)); 1268 } 1269 1270 if (waitqueue_active(sk_sleep(sk))) 1271 wake_up_interruptible(sk_sleep(sk)); 1272 return TIPC_OK; 1273 } 1274 1275 /** 1276 * backlog_rcv - handle incoming message from backlog queue 1277 * @sk: socket 1278 * @buf: message 1279 * 1280 * Caller must hold socket lock, but not port lock. 1281 * 1282 * Returns 0 1283 */ 1284 1285 static int backlog_rcv(struct sock *sk, struct sk_buff *buf) 1286 { 1287 u32 res; 1288 1289 res = filter_rcv(sk, buf); 1290 if (res) 1291 tipc_reject_msg(buf, res); 1292 return 0; 1293 } 1294 1295 /** 1296 * dispatch - handle incoming message 1297 * @tport: TIPC port that received message 1298 * @buf: message 1299 * 1300 * Called with port lock already taken. 1301 * 1302 * Returns TIPC error status code (TIPC_OK if message is not to be rejected) 1303 */ 1304 1305 static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf) 1306 { 1307 struct sock *sk = (struct sock *)tport->usr_handle; 1308 u32 res; 1309 1310 /* 1311 * Process message if socket is unlocked; otherwise add to backlog queue 1312 * 1313 * This code is based on sk_receive_skb(), but must be distinct from it 1314 * since a TIPC-specific filter/reject mechanism is utilized 1315 */ 1316 1317 bh_lock_sock(sk); 1318 if (!sock_owned_by_user(sk)) { 1319 res = filter_rcv(sk, buf); 1320 } else { 1321 if (sk_add_backlog(sk, buf)) 1322 res = TIPC_ERR_OVERLOAD; 1323 else 1324 res = TIPC_OK; 1325 } 1326 bh_unlock_sock(sk); 1327 1328 return res; 1329 } 1330 1331 /** 1332 * wakeupdispatch - wake up port after congestion 1333 * @tport: port to wakeup 1334 * 1335 * Called with port lock already taken. 1336 */ 1337 1338 static void wakeupdispatch(struct tipc_port *tport) 1339 { 1340 struct sock *sk = (struct sock *)tport->usr_handle; 1341 1342 if (waitqueue_active(sk_sleep(sk))) 1343 wake_up_interruptible(sk_sleep(sk)); 1344 } 1345 1346 /** 1347 * connect - establish a connection to another TIPC port 1348 * @sock: socket structure 1349 * @dest: socket address for destination port 1350 * @destlen: size of socket address data structure 1351 * @flags: file-related flags associated with socket 1352 * 1353 * Returns 0 on success, errno otherwise 1354 */ 1355 1356 static int connect(struct socket *sock, struct sockaddr *dest, int destlen, 1357 int flags) 1358 { 1359 struct sock *sk = sock->sk; 1360 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest; 1361 struct msghdr m = {NULL,}; 1362 struct sk_buff *buf; 1363 struct tipc_msg *msg; 1364 long timeout; 1365 int res; 1366 1367 lock_sock(sk); 1368 1369 /* For now, TIPC does not allow use of connect() with DGRAM/RDM types */ 1370 1371 if (sock->state == SS_READY) { 1372 res = -EOPNOTSUPP; 1373 goto exit; 1374 } 1375 1376 /* For now, TIPC does not support the non-blocking form of connect() */ 1377 1378 if (flags & O_NONBLOCK) { 1379 res = -EOPNOTSUPP; 1380 goto exit; 1381 } 1382 1383 /* Issue Posix-compliant error code if socket is in the wrong state */ 1384 1385 if (sock->state == SS_LISTENING) { 1386 res = -EOPNOTSUPP; 1387 goto exit; 1388 } 1389 if (sock->state == SS_CONNECTING) { 1390 res = -EALREADY; 1391 goto exit; 1392 } 1393 if (sock->state != SS_UNCONNECTED) { 1394 res = -EISCONN; 1395 goto exit; 1396 } 1397 1398 /* 1399 * Reject connection attempt using multicast address 1400 * 1401 * Note: send_msg() validates the rest of the address fields, 1402 * so there's no need to do it here 1403 */ 1404 1405 if (dst->addrtype == TIPC_ADDR_MCAST) { 1406 res = -EINVAL; 1407 goto exit; 1408 } 1409 1410 /* Reject any messages already in receive queue (very unlikely) */ 1411 1412 reject_rx_queue(sk); 1413 1414 /* Send a 'SYN-' to destination */ 1415 1416 m.msg_name = dest; 1417 m.msg_namelen = destlen; 1418 res = send_msg(NULL, sock, &m, 0); 1419 if (res < 0) 1420 goto exit; 1421 1422 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */ 1423 1424 timeout = tipc_sk(sk)->conn_timeout; 1425 release_sock(sk); 1426 res = wait_event_interruptible_timeout(*sk_sleep(sk), 1427 (!skb_queue_empty(&sk->sk_receive_queue) || 1428 (sock->state != SS_CONNECTING)), 1429 timeout ? timeout : MAX_SCHEDULE_TIMEOUT); 1430 lock_sock(sk); 1431 1432 if (res > 0) { 1433 buf = skb_peek(&sk->sk_receive_queue); 1434 if (buf != NULL) { 1435 msg = buf_msg(buf); 1436 res = auto_connect(sock, msg); 1437 if (!res) { 1438 if (!msg_data_sz(msg)) 1439 advance_rx_queue(sk); 1440 } 1441 } else { 1442 if (sock->state == SS_CONNECTED) 1443 res = -EISCONN; 1444 else 1445 res = -ECONNREFUSED; 1446 } 1447 } else { 1448 if (res == 0) 1449 res = -ETIMEDOUT; 1450 else 1451 ; /* leave "res" unchanged */ 1452 sock->state = SS_DISCONNECTING; 1453 } 1454 1455 exit: 1456 release_sock(sk); 1457 return res; 1458 } 1459 1460 /** 1461 * listen - allow socket to listen for incoming connections 1462 * @sock: socket structure 1463 * @len: (unused) 1464 * 1465 * Returns 0 on success, errno otherwise 1466 */ 1467 1468 static int listen(struct socket *sock, int len) 1469 { 1470 struct sock *sk = sock->sk; 1471 int res; 1472 1473 lock_sock(sk); 1474 1475 if (sock->state == SS_READY) 1476 res = -EOPNOTSUPP; 1477 else if (sock->state != SS_UNCONNECTED) 1478 res = -EINVAL; 1479 else { 1480 sock->state = SS_LISTENING; 1481 res = 0; 1482 } 1483 1484 release_sock(sk); 1485 return res; 1486 } 1487 1488 /** 1489 * accept - wait for connection request 1490 * @sock: listening socket 1491 * @newsock: new socket that is to be connected 1492 * @flags: file-related flags associated with socket 1493 * 1494 * Returns 0 on success, errno otherwise 1495 */ 1496 1497 static int accept(struct socket *sock, struct socket *new_sock, int flags) 1498 { 1499 struct sock *sk = sock->sk; 1500 struct sk_buff *buf; 1501 int res; 1502 1503 lock_sock(sk); 1504 1505 if (sock->state == SS_READY) { 1506 res = -EOPNOTSUPP; 1507 goto exit; 1508 } 1509 if (sock->state != SS_LISTENING) { 1510 res = -EINVAL; 1511 goto exit; 1512 } 1513 1514 while (skb_queue_empty(&sk->sk_receive_queue)) { 1515 if (flags & O_NONBLOCK) { 1516 res = -EWOULDBLOCK; 1517 goto exit; 1518 } 1519 release_sock(sk); 1520 res = wait_event_interruptible(*sk_sleep(sk), 1521 (!skb_queue_empty(&sk->sk_receive_queue))); 1522 lock_sock(sk); 1523 if (res) 1524 goto exit; 1525 } 1526 1527 buf = skb_peek(&sk->sk_receive_queue); 1528 1529 res = tipc_create(sock_net(sock->sk), new_sock, 0, 0); 1530 if (!res) { 1531 struct sock *new_sk = new_sock->sk; 1532 struct tipc_sock *new_tsock = tipc_sk(new_sk); 1533 struct tipc_port *new_tport = new_tsock->p; 1534 u32 new_ref = new_tport->ref; 1535 struct tipc_msg *msg = buf_msg(buf); 1536 1537 lock_sock(new_sk); 1538 1539 /* 1540 * Reject any stray messages received by new socket 1541 * before the socket lock was taken (very, very unlikely) 1542 */ 1543 1544 reject_rx_queue(new_sk); 1545 1546 /* Connect new socket to it's peer */ 1547 1548 new_tsock->peer_name.ref = msg_origport(msg); 1549 new_tsock->peer_name.node = msg_orignode(msg); 1550 tipc_connect2port(new_ref, &new_tsock->peer_name); 1551 new_sock->state = SS_CONNECTED; 1552 1553 tipc_set_portimportance(new_ref, msg_importance(msg)); 1554 if (msg_named(msg)) { 1555 new_tport->conn_type = msg_nametype(msg); 1556 new_tport->conn_instance = msg_nameinst(msg); 1557 } 1558 1559 /* 1560 * Respond to 'SYN-' by discarding it & returning 'ACK'-. 1561 * Respond to 'SYN+' by queuing it on new socket. 1562 */ 1563 1564 if (!msg_data_sz(msg)) { 1565 struct msghdr m = {NULL,}; 1566 1567 advance_rx_queue(sk); 1568 send_packet(NULL, new_sock, &m, 0); 1569 } else { 1570 __skb_dequeue(&sk->sk_receive_queue); 1571 __skb_queue_head(&new_sk->sk_receive_queue, buf); 1572 } 1573 release_sock(new_sk); 1574 } 1575 exit: 1576 release_sock(sk); 1577 return res; 1578 } 1579 1580 /** 1581 * shutdown - shutdown socket connection 1582 * @sock: socket structure 1583 * @how: direction to close (must be SHUT_RDWR) 1584 * 1585 * Terminates connection (if necessary), then purges socket's receive queue. 1586 * 1587 * Returns 0 on success, errno otherwise 1588 */ 1589 1590 static int shutdown(struct socket *sock, int how) 1591 { 1592 struct sock *sk = sock->sk; 1593 struct tipc_port *tport = tipc_sk_port(sk); 1594 struct sk_buff *buf; 1595 int res; 1596 1597 if (how != SHUT_RDWR) 1598 return -EINVAL; 1599 1600 lock_sock(sk); 1601 1602 switch (sock->state) { 1603 case SS_CONNECTING: 1604 case SS_CONNECTED: 1605 1606 /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */ 1607 restart: 1608 buf = __skb_dequeue(&sk->sk_receive_queue); 1609 if (buf) { 1610 atomic_dec(&tipc_queue_size); 1611 if (TIPC_SKB_CB(buf)->handle != msg_data(buf_msg(buf))) { 1612 buf_discard(buf); 1613 goto restart; 1614 } 1615 tipc_disconnect(tport->ref); 1616 tipc_reject_msg(buf, TIPC_CONN_SHUTDOWN); 1617 } else { 1618 tipc_shutdown(tport->ref); 1619 } 1620 1621 sock->state = SS_DISCONNECTING; 1622 1623 /* fall through */ 1624 1625 case SS_DISCONNECTING: 1626 1627 /* Discard any unreceived messages; wake up sleeping tasks */ 1628 1629 discard_rx_queue(sk); 1630 if (waitqueue_active(sk_sleep(sk))) 1631 wake_up_interruptible(sk_sleep(sk)); 1632 res = 0; 1633 break; 1634 1635 default: 1636 res = -ENOTCONN; 1637 } 1638 1639 release_sock(sk); 1640 return res; 1641 } 1642 1643 /** 1644 * setsockopt - set socket option 1645 * @sock: socket structure 1646 * @lvl: option level 1647 * @opt: option identifier 1648 * @ov: pointer to new option value 1649 * @ol: length of option value 1650 * 1651 * For stream sockets only, accepts and ignores all IPPROTO_TCP options 1652 * (to ease compatibility). 1653 * 1654 * Returns 0 on success, errno otherwise 1655 */ 1656 1657 static int setsockopt(struct socket *sock, 1658 int lvl, int opt, char __user *ov, unsigned int ol) 1659 { 1660 struct sock *sk = sock->sk; 1661 struct tipc_port *tport = tipc_sk_port(sk); 1662 u32 value; 1663 int res; 1664 1665 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM)) 1666 return 0; 1667 if (lvl != SOL_TIPC) 1668 return -ENOPROTOOPT; 1669 if (ol < sizeof(value)) 1670 return -EINVAL; 1671 res = get_user(value, (u32 __user *)ov); 1672 if (res) 1673 return res; 1674 1675 lock_sock(sk); 1676 1677 switch (opt) { 1678 case TIPC_IMPORTANCE: 1679 res = tipc_set_portimportance(tport->ref, value); 1680 break; 1681 case TIPC_SRC_DROPPABLE: 1682 if (sock->type != SOCK_STREAM) 1683 res = tipc_set_portunreliable(tport->ref, value); 1684 else 1685 res = -ENOPROTOOPT; 1686 break; 1687 case TIPC_DEST_DROPPABLE: 1688 res = tipc_set_portunreturnable(tport->ref, value); 1689 break; 1690 case TIPC_CONN_TIMEOUT: 1691 tipc_sk(sk)->conn_timeout = msecs_to_jiffies(value); 1692 /* no need to set "res", since already 0 at this point */ 1693 break; 1694 default: 1695 res = -EINVAL; 1696 } 1697 1698 release_sock(sk); 1699 1700 return res; 1701 } 1702 1703 /** 1704 * getsockopt - get socket option 1705 * @sock: socket structure 1706 * @lvl: option level 1707 * @opt: option identifier 1708 * @ov: receptacle for option value 1709 * @ol: receptacle for length of option value 1710 * 1711 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options 1712 * (to ease compatibility). 1713 * 1714 * Returns 0 on success, errno otherwise 1715 */ 1716 1717 static int getsockopt(struct socket *sock, 1718 int lvl, int opt, char __user *ov, int __user *ol) 1719 { 1720 struct sock *sk = sock->sk; 1721 struct tipc_port *tport = tipc_sk_port(sk); 1722 int len; 1723 u32 value; 1724 int res; 1725 1726 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM)) 1727 return put_user(0, ol); 1728 if (lvl != SOL_TIPC) 1729 return -ENOPROTOOPT; 1730 res = get_user(len, ol); 1731 if (res) 1732 return res; 1733 1734 lock_sock(sk); 1735 1736 switch (opt) { 1737 case TIPC_IMPORTANCE: 1738 res = tipc_portimportance(tport->ref, &value); 1739 break; 1740 case TIPC_SRC_DROPPABLE: 1741 res = tipc_portunreliable(tport->ref, &value); 1742 break; 1743 case TIPC_DEST_DROPPABLE: 1744 res = tipc_portunreturnable(tport->ref, &value); 1745 break; 1746 case TIPC_CONN_TIMEOUT: 1747 value = jiffies_to_msecs(tipc_sk(sk)->conn_timeout); 1748 /* no need to set "res", since already 0 at this point */ 1749 break; 1750 case TIPC_NODE_RECVQ_DEPTH: 1751 value = (u32)atomic_read(&tipc_queue_size); 1752 break; 1753 case TIPC_SOCK_RECVQ_DEPTH: 1754 value = skb_queue_len(&sk->sk_receive_queue); 1755 break; 1756 default: 1757 res = -EINVAL; 1758 } 1759 1760 release_sock(sk); 1761 1762 if (res) 1763 return res; /* "get" failed */ 1764 1765 if (len < sizeof(value)) 1766 return -EINVAL; 1767 1768 if (copy_to_user(ov, &value, sizeof(value))) 1769 return -EFAULT; 1770 1771 return put_user(sizeof(value), ol); 1772 } 1773 1774 /** 1775 * Protocol switches for the various types of TIPC sockets 1776 */ 1777 1778 static const struct proto_ops msg_ops = { 1779 .owner = THIS_MODULE, 1780 .family = AF_TIPC, 1781 .release = release, 1782 .bind = bind, 1783 .connect = connect, 1784 .socketpair = sock_no_socketpair, 1785 .accept = accept, 1786 .getname = get_name, 1787 .poll = poll, 1788 .ioctl = sock_no_ioctl, 1789 .listen = listen, 1790 .shutdown = shutdown, 1791 .setsockopt = setsockopt, 1792 .getsockopt = getsockopt, 1793 .sendmsg = send_msg, 1794 .recvmsg = recv_msg, 1795 .mmap = sock_no_mmap, 1796 .sendpage = sock_no_sendpage 1797 }; 1798 1799 static const struct proto_ops packet_ops = { 1800 .owner = THIS_MODULE, 1801 .family = AF_TIPC, 1802 .release = release, 1803 .bind = bind, 1804 .connect = connect, 1805 .socketpair = sock_no_socketpair, 1806 .accept = accept, 1807 .getname = get_name, 1808 .poll = poll, 1809 .ioctl = sock_no_ioctl, 1810 .listen = listen, 1811 .shutdown = shutdown, 1812 .setsockopt = setsockopt, 1813 .getsockopt = getsockopt, 1814 .sendmsg = send_packet, 1815 .recvmsg = recv_msg, 1816 .mmap = sock_no_mmap, 1817 .sendpage = sock_no_sendpage 1818 }; 1819 1820 static const struct proto_ops stream_ops = { 1821 .owner = THIS_MODULE, 1822 .family = AF_TIPC, 1823 .release = release, 1824 .bind = bind, 1825 .connect = connect, 1826 .socketpair = sock_no_socketpair, 1827 .accept = accept, 1828 .getname = get_name, 1829 .poll = poll, 1830 .ioctl = sock_no_ioctl, 1831 .listen = listen, 1832 .shutdown = shutdown, 1833 .setsockopt = setsockopt, 1834 .getsockopt = getsockopt, 1835 .sendmsg = send_stream, 1836 .recvmsg = recv_stream, 1837 .mmap = sock_no_mmap, 1838 .sendpage = sock_no_sendpage 1839 }; 1840 1841 static const struct net_proto_family tipc_family_ops = { 1842 .owner = THIS_MODULE, 1843 .family = AF_TIPC, 1844 .create = tipc_create 1845 }; 1846 1847 static struct proto tipc_proto = { 1848 .name = "TIPC", 1849 .owner = THIS_MODULE, 1850 .obj_size = sizeof(struct tipc_sock) 1851 }; 1852 1853 /** 1854 * tipc_socket_init - initialize TIPC socket interface 1855 * 1856 * Returns 0 on success, errno otherwise 1857 */ 1858 int tipc_socket_init(void) 1859 { 1860 int res; 1861 1862 res = proto_register(&tipc_proto, 1); 1863 if (res) { 1864 err("Failed to register TIPC protocol type\n"); 1865 goto out; 1866 } 1867 1868 res = sock_register(&tipc_family_ops); 1869 if (res) { 1870 err("Failed to register TIPC socket type\n"); 1871 proto_unregister(&tipc_proto); 1872 goto out; 1873 } 1874 1875 sockets_enabled = 1; 1876 out: 1877 return res; 1878 } 1879 1880 /** 1881 * tipc_socket_stop - stop TIPC socket interface 1882 */ 1883 1884 void tipc_socket_stop(void) 1885 { 1886 if (!sockets_enabled) 1887 return; 1888 1889 sockets_enabled = 0; 1890 sock_unregister(tipc_family_ops.family); 1891 proto_unregister(&tipc_proto); 1892 } 1893 1894