| 1 | /* |
| 2 | * net/tipc/socket.c: TIPC socket API |
| 3 | * |
| 4 | * Copyright (c) 2001-2007, 2012-2014, Ericsson AB |
| 5 | * Copyright (c) 2004-2008, 2010-2013, 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 "core.h" |
| 38 | #include "port.h" |
| 39 | #include "name_table.h" |
| 40 | #include "node.h" |
| 41 | #include "link.h" |
| 42 | #include <linux/export.h> |
| 43 | #include "link.h" |
| 44 | |
| 45 | #define SS_LISTENING -1 /* socket is listening */ |
| 46 | #define SS_READY -2 /* socket is connectionless */ |
| 47 | |
| 48 | #define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */ |
| 49 | |
| 50 | static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb); |
| 51 | static void tipc_data_ready(struct sock *sk); |
| 52 | static void tipc_write_space(struct sock *sk); |
| 53 | static int tipc_release(struct socket *sock); |
| 54 | static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags); |
| 55 | |
| 56 | static const struct proto_ops packet_ops; |
| 57 | static const struct proto_ops stream_ops; |
| 58 | static const struct proto_ops msg_ops; |
| 59 | |
| 60 | static struct proto tipc_proto; |
| 61 | static struct proto tipc_proto_kern; |
| 62 | |
| 63 | /* |
| 64 | * Revised TIPC socket locking policy: |
| 65 | * |
| 66 | * Most socket operations take the standard socket lock when they start |
| 67 | * and hold it until they finish (or until they need to sleep). Acquiring |
| 68 | * this lock grants the owner exclusive access to the fields of the socket |
| 69 | * data structures, with the exception of the backlog queue. A few socket |
| 70 | * operations can be done without taking the socket lock because they only |
| 71 | * read socket information that never changes during the life of the socket. |
| 72 | * |
| 73 | * Socket operations may acquire the lock for the associated TIPC port if they |
| 74 | * need to perform an operation on the port. If any routine needs to acquire |
| 75 | * both the socket lock and the port lock it must take the socket lock first |
| 76 | * to avoid the risk of deadlock. |
| 77 | * |
| 78 | * The dispatcher handling incoming messages cannot grab the socket lock in |
| 79 | * the standard fashion, since invoked it runs at the BH level and cannot block. |
| 80 | * Instead, it checks to see if the socket lock is currently owned by someone, |
| 81 | * and either handles the message itself or adds it to the socket's backlog |
| 82 | * queue; in the latter case the queued message is processed once the process |
| 83 | * owning the socket lock releases it. |
| 84 | * |
| 85 | * NOTE: Releasing the socket lock while an operation is sleeping overcomes |
| 86 | * the problem of a blocked socket operation preventing any other operations |
| 87 | * from occurring. However, applications must be careful if they have |
| 88 | * multiple threads trying to send (or receive) on the same socket, as these |
| 89 | * operations might interfere with each other. For example, doing a connect |
| 90 | * and a receive at the same time might allow the receive to consume the |
| 91 | * ACK message meant for the connect. While additional work could be done |
| 92 | * to try and overcome this, it doesn't seem to be worthwhile at the present. |
| 93 | * |
| 94 | * NOTE: Releasing the socket lock while an operation is sleeping also ensures |
| 95 | * that another operation that must be performed in a non-blocking manner is |
| 96 | * not delayed for very long because the lock has already been taken. |
| 97 | * |
| 98 | * NOTE: This code assumes that certain fields of a port/socket pair are |
| 99 | * constant over its lifetime; such fields can be examined without taking |
| 100 | * the socket lock and/or port lock, and do not need to be re-read even |
| 101 | * after resuming processing after waiting. These fields include: |
| 102 | * - socket type |
| 103 | * - pointer to socket sk structure (aka tipc_sock structure) |
| 104 | * - pointer to port structure |
| 105 | * - port reference |
| 106 | */ |
| 107 | |
| 108 | #include "socket.h" |
| 109 | |
| 110 | /** |
| 111 | * advance_rx_queue - discard first buffer in socket receive queue |
| 112 | * |
| 113 | * Caller must hold socket lock |
| 114 | */ |
| 115 | static void advance_rx_queue(struct sock *sk) |
| 116 | { |
| 117 | kfree_skb(__skb_dequeue(&sk->sk_receive_queue)); |
| 118 | } |
| 119 | |
| 120 | /** |
| 121 | * reject_rx_queue - reject all buffers in socket receive queue |
| 122 | * |
| 123 | * Caller must hold socket lock |
| 124 | */ |
| 125 | static void reject_rx_queue(struct sock *sk) |
| 126 | { |
| 127 | struct sk_buff *buf; |
| 128 | u32 dnode; |
| 129 | |
| 130 | while ((buf = __skb_dequeue(&sk->sk_receive_queue))) { |
| 131 | if (tipc_msg_reverse(buf, &dnode, TIPC_ERR_NO_PORT)) |
| 132 | tipc_link_xmit2(buf, dnode, 0); |
| 133 | } |
| 134 | } |
| 135 | |
| 136 | /** |
| 137 | * tipc_sk_create - create a TIPC socket |
| 138 | * @net: network namespace (must be default network) |
| 139 | * @sock: pre-allocated socket structure |
| 140 | * @protocol: protocol indicator (must be 0) |
| 141 | * @kern: caused by kernel or by userspace? |
| 142 | * |
| 143 | * This routine creates additional data structures used by the TIPC socket, |
| 144 | * initializes them, and links them together. |
| 145 | * |
| 146 | * Returns 0 on success, errno otherwise |
| 147 | */ |
| 148 | static int tipc_sk_create(struct net *net, struct socket *sock, |
| 149 | int protocol, int kern) |
| 150 | { |
| 151 | const struct proto_ops *ops; |
| 152 | socket_state state; |
| 153 | struct sock *sk; |
| 154 | struct tipc_sock *tsk; |
| 155 | struct tipc_port *port; |
| 156 | u32 ref; |
| 157 | |
| 158 | /* Validate arguments */ |
| 159 | if (unlikely(protocol != 0)) |
| 160 | return -EPROTONOSUPPORT; |
| 161 | |
| 162 | switch (sock->type) { |
| 163 | case SOCK_STREAM: |
| 164 | ops = &stream_ops; |
| 165 | state = SS_UNCONNECTED; |
| 166 | break; |
| 167 | case SOCK_SEQPACKET: |
| 168 | ops = &packet_ops; |
| 169 | state = SS_UNCONNECTED; |
| 170 | break; |
| 171 | case SOCK_DGRAM: |
| 172 | case SOCK_RDM: |
| 173 | ops = &msg_ops; |
| 174 | state = SS_READY; |
| 175 | break; |
| 176 | default: |
| 177 | return -EPROTOTYPE; |
| 178 | } |
| 179 | |
| 180 | /* Allocate socket's protocol area */ |
| 181 | if (!kern) |
| 182 | sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto); |
| 183 | else |
| 184 | sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto_kern); |
| 185 | |
| 186 | if (sk == NULL) |
| 187 | return -ENOMEM; |
| 188 | |
| 189 | tsk = tipc_sk(sk); |
| 190 | port = &tsk->port; |
| 191 | |
| 192 | ref = tipc_port_init(port, TIPC_LOW_IMPORTANCE); |
| 193 | if (!ref) { |
| 194 | pr_warn("Socket registration failed, ref. table exhausted\n"); |
| 195 | sk_free(sk); |
| 196 | return -ENOMEM; |
| 197 | } |
| 198 | |
| 199 | /* Finish initializing socket data structures */ |
| 200 | sock->ops = ops; |
| 201 | sock->state = state; |
| 202 | |
| 203 | sock_init_data(sock, sk); |
| 204 | sk->sk_backlog_rcv = tipc_backlog_rcv; |
| 205 | sk->sk_rcvbuf = sysctl_tipc_rmem[1]; |
| 206 | sk->sk_data_ready = tipc_data_ready; |
| 207 | sk->sk_write_space = tipc_write_space; |
| 208 | tsk->conn_timeout = CONN_TIMEOUT_DEFAULT; |
| 209 | tsk->port.sent = 0; |
| 210 | atomic_set(&tsk->dupl_rcvcnt, 0); |
| 211 | tipc_port_unlock(port); |
| 212 | |
| 213 | if (sock->state == SS_READY) { |
| 214 | tipc_port_set_unreturnable(port, true); |
| 215 | if (sock->type == SOCK_DGRAM) |
| 216 | tipc_port_set_unreliable(port, true); |
| 217 | } |
| 218 | return 0; |
| 219 | } |
| 220 | |
| 221 | /** |
| 222 | * tipc_sock_create_local - create TIPC socket from inside TIPC module |
| 223 | * @type: socket type - SOCK_RDM or SOCK_SEQPACKET |
| 224 | * |
| 225 | * We cannot use sock_creat_kern here because it bumps module user count. |
| 226 | * Since socket owner and creator is the same module we must make sure |
| 227 | * that module count remains zero for module local sockets, otherwise |
| 228 | * we cannot do rmmod. |
| 229 | * |
| 230 | * Returns 0 on success, errno otherwise |
| 231 | */ |
| 232 | int tipc_sock_create_local(int type, struct socket **res) |
| 233 | { |
| 234 | int rc; |
| 235 | |
| 236 | rc = sock_create_lite(AF_TIPC, type, 0, res); |
| 237 | if (rc < 0) { |
| 238 | pr_err("Failed to create kernel socket\n"); |
| 239 | return rc; |
| 240 | } |
| 241 | tipc_sk_create(&init_net, *res, 0, 1); |
| 242 | |
| 243 | return 0; |
| 244 | } |
| 245 | |
| 246 | /** |
| 247 | * tipc_sock_release_local - release socket created by tipc_sock_create_local |
| 248 | * @sock: the socket to be released. |
| 249 | * |
| 250 | * Module reference count is not incremented when such sockets are created, |
| 251 | * so we must keep it from being decremented when they are released. |
| 252 | */ |
| 253 | void tipc_sock_release_local(struct socket *sock) |
| 254 | { |
| 255 | tipc_release(sock); |
| 256 | sock->ops = NULL; |
| 257 | sock_release(sock); |
| 258 | } |
| 259 | |
| 260 | /** |
| 261 | * tipc_sock_accept_local - accept a connection on a socket created |
| 262 | * with tipc_sock_create_local. Use this function to avoid that |
| 263 | * module reference count is inadvertently incremented. |
| 264 | * |
| 265 | * @sock: the accepting socket |
| 266 | * @newsock: reference to the new socket to be created |
| 267 | * @flags: socket flags |
| 268 | */ |
| 269 | |
| 270 | int tipc_sock_accept_local(struct socket *sock, struct socket **newsock, |
| 271 | int flags) |
| 272 | { |
| 273 | struct sock *sk = sock->sk; |
| 274 | int ret; |
| 275 | |
| 276 | ret = sock_create_lite(sk->sk_family, sk->sk_type, |
| 277 | sk->sk_protocol, newsock); |
| 278 | if (ret < 0) |
| 279 | return ret; |
| 280 | |
| 281 | ret = tipc_accept(sock, *newsock, flags); |
| 282 | if (ret < 0) { |
| 283 | sock_release(*newsock); |
| 284 | return ret; |
| 285 | } |
| 286 | (*newsock)->ops = sock->ops; |
| 287 | return ret; |
| 288 | } |
| 289 | |
| 290 | /** |
| 291 | * tipc_release - destroy a TIPC socket |
| 292 | * @sock: socket to destroy |
| 293 | * |
| 294 | * This routine cleans up any messages that are still queued on the socket. |
| 295 | * For DGRAM and RDM socket types, all queued messages are rejected. |
| 296 | * For SEQPACKET and STREAM socket types, the first message is rejected |
| 297 | * and any others are discarded. (If the first message on a STREAM socket |
| 298 | * is partially-read, it is discarded and the next one is rejected instead.) |
| 299 | * |
| 300 | * NOTE: Rejected messages are not necessarily returned to the sender! They |
| 301 | * are returned or discarded according to the "destination droppable" setting |
| 302 | * specified for the message by the sender. |
| 303 | * |
| 304 | * Returns 0 on success, errno otherwise |
| 305 | */ |
| 306 | static int tipc_release(struct socket *sock) |
| 307 | { |
| 308 | struct sock *sk = sock->sk; |
| 309 | struct tipc_sock *tsk; |
| 310 | struct tipc_port *port; |
| 311 | struct sk_buff *buf; |
| 312 | u32 dnode; |
| 313 | |
| 314 | /* |
| 315 | * Exit if socket isn't fully initialized (occurs when a failed accept() |
| 316 | * releases a pre-allocated child socket that was never used) |
| 317 | */ |
| 318 | if (sk == NULL) |
| 319 | return 0; |
| 320 | |
| 321 | tsk = tipc_sk(sk); |
| 322 | port = &tsk->port; |
| 323 | lock_sock(sk); |
| 324 | |
| 325 | /* |
| 326 | * Reject all unreceived messages, except on an active connection |
| 327 | * (which disconnects locally & sends a 'FIN+' to peer) |
| 328 | */ |
| 329 | while (sock->state != SS_DISCONNECTING) { |
| 330 | buf = __skb_dequeue(&sk->sk_receive_queue); |
| 331 | if (buf == NULL) |
| 332 | break; |
| 333 | if (TIPC_SKB_CB(buf)->handle != NULL) |
| 334 | kfree_skb(buf); |
| 335 | else { |
| 336 | if ((sock->state == SS_CONNECTING) || |
| 337 | (sock->state == SS_CONNECTED)) { |
| 338 | sock->state = SS_DISCONNECTING; |
| 339 | tipc_port_disconnect(port->ref); |
| 340 | } |
| 341 | if (tipc_msg_reverse(buf, &dnode, TIPC_ERR_NO_PORT)) |
| 342 | tipc_link_xmit2(buf, dnode, 0); |
| 343 | } |
| 344 | } |
| 345 | |
| 346 | /* Destroy TIPC port; also disconnects an active connection and |
| 347 | * sends a 'FIN-' to peer. |
| 348 | */ |
| 349 | tipc_port_destroy(port); |
| 350 | |
| 351 | /* Discard any remaining (connection-based) messages in receive queue */ |
| 352 | __skb_queue_purge(&sk->sk_receive_queue); |
| 353 | |
| 354 | /* Reject any messages that accumulated in backlog queue */ |
| 355 | sock->state = SS_DISCONNECTING; |
| 356 | release_sock(sk); |
| 357 | |
| 358 | sock_put(sk); |
| 359 | sock->sk = NULL; |
| 360 | |
| 361 | return 0; |
| 362 | } |
| 363 | |
| 364 | /** |
| 365 | * tipc_bind - associate or disassocate TIPC name(s) with a socket |
| 366 | * @sock: socket structure |
| 367 | * @uaddr: socket address describing name(s) and desired operation |
| 368 | * @uaddr_len: size of socket address data structure |
| 369 | * |
| 370 | * Name and name sequence binding is indicated using a positive scope value; |
| 371 | * a negative scope value unbinds the specified name. Specifying no name |
| 372 | * (i.e. a socket address length of 0) unbinds all names from the socket. |
| 373 | * |
| 374 | * Returns 0 on success, errno otherwise |
| 375 | * |
| 376 | * NOTE: This routine doesn't need to take the socket lock since it doesn't |
| 377 | * access any non-constant socket information. |
| 378 | */ |
| 379 | static int tipc_bind(struct socket *sock, struct sockaddr *uaddr, |
| 380 | int uaddr_len) |
| 381 | { |
| 382 | struct sock *sk = sock->sk; |
| 383 | struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr; |
| 384 | struct tipc_sock *tsk = tipc_sk(sk); |
| 385 | int res = -EINVAL; |
| 386 | |
| 387 | lock_sock(sk); |
| 388 | if (unlikely(!uaddr_len)) { |
| 389 | res = tipc_withdraw(&tsk->port, 0, NULL); |
| 390 | goto exit; |
| 391 | } |
| 392 | |
| 393 | if (uaddr_len < sizeof(struct sockaddr_tipc)) { |
| 394 | res = -EINVAL; |
| 395 | goto exit; |
| 396 | } |
| 397 | if (addr->family != AF_TIPC) { |
| 398 | res = -EAFNOSUPPORT; |
| 399 | goto exit; |
| 400 | } |
| 401 | |
| 402 | if (addr->addrtype == TIPC_ADDR_NAME) |
| 403 | addr->addr.nameseq.upper = addr->addr.nameseq.lower; |
| 404 | else if (addr->addrtype != TIPC_ADDR_NAMESEQ) { |
| 405 | res = -EAFNOSUPPORT; |
| 406 | goto exit; |
| 407 | } |
| 408 | |
| 409 | if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) && |
| 410 | (addr->addr.nameseq.type != TIPC_TOP_SRV) && |
| 411 | (addr->addr.nameseq.type != TIPC_CFG_SRV)) { |
| 412 | res = -EACCES; |
| 413 | goto exit; |
| 414 | } |
| 415 | |
| 416 | res = (addr->scope > 0) ? |
| 417 | tipc_publish(&tsk->port, addr->scope, &addr->addr.nameseq) : |
| 418 | tipc_withdraw(&tsk->port, -addr->scope, &addr->addr.nameseq); |
| 419 | exit: |
| 420 | release_sock(sk); |
| 421 | return res; |
| 422 | } |
| 423 | |
| 424 | /** |
| 425 | * tipc_getname - get port ID of socket or peer socket |
| 426 | * @sock: socket structure |
| 427 | * @uaddr: area for returned socket address |
| 428 | * @uaddr_len: area for returned length of socket address |
| 429 | * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID |
| 430 | * |
| 431 | * Returns 0 on success, errno otherwise |
| 432 | * |
| 433 | * NOTE: This routine doesn't need to take the socket lock since it only |
| 434 | * accesses socket information that is unchanging (or which changes in |
| 435 | * a completely predictable manner). |
| 436 | */ |
| 437 | static int tipc_getname(struct socket *sock, struct sockaddr *uaddr, |
| 438 | int *uaddr_len, int peer) |
| 439 | { |
| 440 | struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr; |
| 441 | struct tipc_sock *tsk = tipc_sk(sock->sk); |
| 442 | |
| 443 | memset(addr, 0, sizeof(*addr)); |
| 444 | if (peer) { |
| 445 | if ((sock->state != SS_CONNECTED) && |
| 446 | ((peer != 2) || (sock->state != SS_DISCONNECTING))) |
| 447 | return -ENOTCONN; |
| 448 | addr->addr.id.ref = tipc_port_peerport(&tsk->port); |
| 449 | addr->addr.id.node = tipc_port_peernode(&tsk->port); |
| 450 | } else { |
| 451 | addr->addr.id.ref = tsk->port.ref; |
| 452 | addr->addr.id.node = tipc_own_addr; |
| 453 | } |
| 454 | |
| 455 | *uaddr_len = sizeof(*addr); |
| 456 | addr->addrtype = TIPC_ADDR_ID; |
| 457 | addr->family = AF_TIPC; |
| 458 | addr->scope = 0; |
| 459 | addr->addr.name.domain = 0; |
| 460 | |
| 461 | return 0; |
| 462 | } |
| 463 | |
| 464 | /** |
| 465 | * tipc_poll - read and possibly block on pollmask |
| 466 | * @file: file structure associated with the socket |
| 467 | * @sock: socket for which to calculate the poll bits |
| 468 | * @wait: ??? |
| 469 | * |
| 470 | * Returns pollmask value |
| 471 | * |
| 472 | * COMMENTARY: |
| 473 | * It appears that the usual socket locking mechanisms are not useful here |
| 474 | * since the pollmask info is potentially out-of-date the moment this routine |
| 475 | * exits. TCP and other protocols seem to rely on higher level poll routines |
| 476 | * to handle any preventable race conditions, so TIPC will do the same ... |
| 477 | * |
| 478 | * TIPC sets the returned events as follows: |
| 479 | * |
| 480 | * socket state flags set |
| 481 | * ------------ --------- |
| 482 | * unconnected no read flags |
| 483 | * POLLOUT if port is not congested |
| 484 | * |
| 485 | * connecting POLLIN/POLLRDNORM if ACK/NACK in rx queue |
| 486 | * no write flags |
| 487 | * |
| 488 | * connected POLLIN/POLLRDNORM if data in rx queue |
| 489 | * POLLOUT if port is not congested |
| 490 | * |
| 491 | * disconnecting POLLIN/POLLRDNORM/POLLHUP |
| 492 | * no write flags |
| 493 | * |
| 494 | * listening POLLIN if SYN in rx queue |
| 495 | * no write flags |
| 496 | * |
| 497 | * ready POLLIN/POLLRDNORM if data in rx queue |
| 498 | * [connectionless] POLLOUT (since port cannot be congested) |
| 499 | * |
| 500 | * IMPORTANT: The fact that a read or write operation is indicated does NOT |
| 501 | * imply that the operation will succeed, merely that it should be performed |
| 502 | * and will not block. |
| 503 | */ |
| 504 | static unsigned int tipc_poll(struct file *file, struct socket *sock, |
| 505 | poll_table *wait) |
| 506 | { |
| 507 | struct sock *sk = sock->sk; |
| 508 | struct tipc_sock *tsk = tipc_sk(sk); |
| 509 | u32 mask = 0; |
| 510 | |
| 511 | sock_poll_wait(file, sk_sleep(sk), wait); |
| 512 | |
| 513 | switch ((int)sock->state) { |
| 514 | case SS_UNCONNECTED: |
| 515 | if (!tsk->port.congested) |
| 516 | mask |= POLLOUT; |
| 517 | break; |
| 518 | case SS_READY: |
| 519 | case SS_CONNECTED: |
| 520 | if (!tsk->port.congested) |
| 521 | mask |= POLLOUT; |
| 522 | /* fall thru' */ |
| 523 | case SS_CONNECTING: |
| 524 | case SS_LISTENING: |
| 525 | if (!skb_queue_empty(&sk->sk_receive_queue)) |
| 526 | mask |= (POLLIN | POLLRDNORM); |
| 527 | break; |
| 528 | case SS_DISCONNECTING: |
| 529 | mask = (POLLIN | POLLRDNORM | POLLHUP); |
| 530 | break; |
| 531 | } |
| 532 | |
| 533 | return mask; |
| 534 | } |
| 535 | |
| 536 | /** |
| 537 | * dest_name_check - verify user is permitted to send to specified port name |
| 538 | * @dest: destination address |
| 539 | * @m: descriptor for message to be sent |
| 540 | * |
| 541 | * Prevents restricted configuration commands from being issued by |
| 542 | * unauthorized users. |
| 543 | * |
| 544 | * Returns 0 if permission is granted, otherwise errno |
| 545 | */ |
| 546 | static int dest_name_check(struct sockaddr_tipc *dest, struct msghdr *m) |
| 547 | { |
| 548 | struct tipc_cfg_msg_hdr hdr; |
| 549 | |
| 550 | if (unlikely(dest->addrtype == TIPC_ADDR_ID)) |
| 551 | return 0; |
| 552 | if (likely(dest->addr.name.name.type >= TIPC_RESERVED_TYPES)) |
| 553 | return 0; |
| 554 | if (likely(dest->addr.name.name.type == TIPC_TOP_SRV)) |
| 555 | return 0; |
| 556 | if (likely(dest->addr.name.name.type != TIPC_CFG_SRV)) |
| 557 | return -EACCES; |
| 558 | |
| 559 | if (!m->msg_iovlen || (m->msg_iov[0].iov_len < sizeof(hdr))) |
| 560 | return -EMSGSIZE; |
| 561 | if (copy_from_user(&hdr, m->msg_iov[0].iov_base, sizeof(hdr))) |
| 562 | return -EFAULT; |
| 563 | if ((ntohs(hdr.tcm_type) & 0xC000) && (!capable(CAP_NET_ADMIN))) |
| 564 | return -EACCES; |
| 565 | |
| 566 | return 0; |
| 567 | } |
| 568 | |
| 569 | static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p) |
| 570 | { |
| 571 | struct sock *sk = sock->sk; |
| 572 | struct tipc_sock *tsk = tipc_sk(sk); |
| 573 | DEFINE_WAIT(wait); |
| 574 | int done; |
| 575 | |
| 576 | do { |
| 577 | int err = sock_error(sk); |
| 578 | if (err) |
| 579 | return err; |
| 580 | if (sock->state == SS_DISCONNECTING) |
| 581 | return -EPIPE; |
| 582 | if (!*timeo_p) |
| 583 | return -EAGAIN; |
| 584 | if (signal_pending(current)) |
| 585 | return sock_intr_errno(*timeo_p); |
| 586 | |
| 587 | prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); |
| 588 | done = sk_wait_event(sk, timeo_p, !tsk->port.congested); |
| 589 | finish_wait(sk_sleep(sk), &wait); |
| 590 | } while (!done); |
| 591 | return 0; |
| 592 | } |
| 593 | |
| 594 | /** |
| 595 | * tipc_sendmcast - send multicast message |
| 596 | * @sock: socket structure |
| 597 | * @seq: destination address |
| 598 | * @iov: message data to send |
| 599 | * @dsz: total length of message data |
| 600 | * @timeo: timeout to wait for wakeup |
| 601 | * |
| 602 | * Called from function tipc_sendmsg(), which has done all sanity checks |
| 603 | * Returns the number of bytes sent on success, or errno |
| 604 | */ |
| 605 | static int tipc_sendmcast(struct socket *sock, struct tipc_name_seq *seq, |
| 606 | struct iovec *iov, size_t dsz, long timeo) |
| 607 | { |
| 608 | struct sock *sk = sock->sk; |
| 609 | struct tipc_sock *tsk = tipc_sk(sk); |
| 610 | int rc; |
| 611 | |
| 612 | do { |
| 613 | if (sock->state != SS_READY) { |
| 614 | rc = -EOPNOTSUPP; |
| 615 | break; |
| 616 | } |
| 617 | rc = tipc_port_mcast_xmit(&tsk->port, seq, iov, dsz); |
| 618 | if (likely(rc >= 0)) { |
| 619 | if (sock->state != SS_READY) |
| 620 | sock->state = SS_CONNECTING; |
| 621 | break; |
| 622 | } |
| 623 | if (rc != -ELINKCONG) |
| 624 | break; |
| 625 | rc = tipc_wait_for_sndmsg(sock, &timeo); |
| 626 | } while (!rc); |
| 627 | |
| 628 | return rc; |
| 629 | } |
| 630 | |
| 631 | /** |
| 632 | * tipc_sendmsg - send message in connectionless manner |
| 633 | * @iocb: if NULL, indicates that socket lock is already held |
| 634 | * @sock: socket structure |
| 635 | * @m: message to send |
| 636 | * @dsz: amount of user data to be sent |
| 637 | * |
| 638 | * Message must have an destination specified explicitly. |
| 639 | * Used for SOCK_RDM and SOCK_DGRAM messages, |
| 640 | * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections. |
| 641 | * (Note: 'SYN+' is prohibited on SOCK_STREAM.) |
| 642 | * |
| 643 | * Returns the number of bytes sent on success, or errno otherwise |
| 644 | */ |
| 645 | static int tipc_sendmsg(struct kiocb *iocb, struct socket *sock, |
| 646 | struct msghdr *m, size_t dsz) |
| 647 | { |
| 648 | DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name); |
| 649 | struct sock *sk = sock->sk; |
| 650 | struct tipc_sock *tsk = tipc_sk(sk); |
| 651 | struct tipc_port *port = &tsk->port; |
| 652 | struct tipc_msg *mhdr = &port->phdr; |
| 653 | struct iovec *iov = m->msg_iov; |
| 654 | u32 dnode, dport; |
| 655 | struct sk_buff *buf; |
| 656 | struct tipc_name_seq *seq = &dest->addr.nameseq; |
| 657 | u32 mtu; |
| 658 | long timeo; |
| 659 | int rc = -EINVAL; |
| 660 | |
| 661 | if (unlikely(!dest)) |
| 662 | return -EDESTADDRREQ; |
| 663 | |
| 664 | if (unlikely((m->msg_namelen < sizeof(*dest)) || |
| 665 | (dest->family != AF_TIPC))) |
| 666 | return -EINVAL; |
| 667 | |
| 668 | if (dsz > TIPC_MAX_USER_MSG_SIZE) |
| 669 | return -EMSGSIZE; |
| 670 | |
| 671 | if (iocb) |
| 672 | lock_sock(sk); |
| 673 | |
| 674 | if (unlikely(sock->state != SS_READY)) { |
| 675 | if (sock->state == SS_LISTENING) { |
| 676 | rc = -EPIPE; |
| 677 | goto exit; |
| 678 | } |
| 679 | if (sock->state != SS_UNCONNECTED) { |
| 680 | rc = -EISCONN; |
| 681 | goto exit; |
| 682 | } |
| 683 | if (tsk->port.published) { |
| 684 | rc = -EOPNOTSUPP; |
| 685 | goto exit; |
| 686 | } |
| 687 | if (dest->addrtype == TIPC_ADDR_NAME) { |
| 688 | tsk->port.conn_type = dest->addr.name.name.type; |
| 689 | tsk->port.conn_instance = dest->addr.name.name.instance; |
| 690 | } |
| 691 | } |
| 692 | rc = dest_name_check(dest, m); |
| 693 | if (rc) |
| 694 | goto exit; |
| 695 | |
| 696 | timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT); |
| 697 | |
| 698 | if (dest->addrtype == TIPC_ADDR_MCAST) { |
| 699 | rc = tipc_sendmcast(sock, seq, iov, dsz, timeo); |
| 700 | goto exit; |
| 701 | } else if (dest->addrtype == TIPC_ADDR_NAME) { |
| 702 | u32 type = dest->addr.name.name.type; |
| 703 | u32 inst = dest->addr.name.name.instance; |
| 704 | u32 domain = dest->addr.name.domain; |
| 705 | |
| 706 | dnode = domain; |
| 707 | msg_set_type(mhdr, TIPC_NAMED_MSG); |
| 708 | msg_set_hdr_sz(mhdr, NAMED_H_SIZE); |
| 709 | msg_set_nametype(mhdr, type); |
| 710 | msg_set_nameinst(mhdr, inst); |
| 711 | msg_set_lookup_scope(mhdr, tipc_addr_scope(domain)); |
| 712 | dport = tipc_nametbl_translate(type, inst, &dnode); |
| 713 | msg_set_destnode(mhdr, dnode); |
| 714 | msg_set_destport(mhdr, dport); |
| 715 | if (unlikely(!dport && !dnode)) { |
| 716 | rc = -EHOSTUNREACH; |
| 717 | goto exit; |
| 718 | } |
| 719 | } else if (dest->addrtype == TIPC_ADDR_ID) { |
| 720 | dnode = dest->addr.id.node; |
| 721 | msg_set_type(mhdr, TIPC_DIRECT_MSG); |
| 722 | msg_set_lookup_scope(mhdr, 0); |
| 723 | msg_set_destnode(mhdr, dnode); |
| 724 | msg_set_destport(mhdr, dest->addr.id.ref); |
| 725 | msg_set_hdr_sz(mhdr, BASIC_H_SIZE); |
| 726 | } |
| 727 | |
| 728 | new_mtu: |
| 729 | mtu = tipc_node_get_mtu(dnode, tsk->port.ref); |
| 730 | rc = tipc_msg_build2(mhdr, iov, 0, dsz, mtu, &buf); |
| 731 | if (rc < 0) |
| 732 | goto exit; |
| 733 | |
| 734 | do { |
| 735 | rc = tipc_link_xmit2(buf, dnode, tsk->port.ref); |
| 736 | if (likely(rc >= 0)) { |
| 737 | if (sock->state != SS_READY) |
| 738 | sock->state = SS_CONNECTING; |
| 739 | rc = dsz; |
| 740 | break; |
| 741 | } |
| 742 | if (rc == -EMSGSIZE) |
| 743 | goto new_mtu; |
| 744 | |
| 745 | if (rc != -ELINKCONG) |
| 746 | break; |
| 747 | |
| 748 | rc = tipc_wait_for_sndmsg(sock, &timeo); |
| 749 | } while (!rc); |
| 750 | |
| 751 | exit: |
| 752 | if (iocb) |
| 753 | release_sock(sk); |
| 754 | |
| 755 | return rc; |
| 756 | } |
| 757 | |
| 758 | static int tipc_wait_for_sndpkt(struct socket *sock, long *timeo_p) |
| 759 | { |
| 760 | struct sock *sk = sock->sk; |
| 761 | struct tipc_sock *tsk = tipc_sk(sk); |
| 762 | struct tipc_port *port = &tsk->port; |
| 763 | DEFINE_WAIT(wait); |
| 764 | int done; |
| 765 | |
| 766 | do { |
| 767 | int err = sock_error(sk); |
| 768 | if (err) |
| 769 | return err; |
| 770 | if (sock->state == SS_DISCONNECTING) |
| 771 | return -EPIPE; |
| 772 | else if (sock->state != SS_CONNECTED) |
| 773 | return -ENOTCONN; |
| 774 | if (!*timeo_p) |
| 775 | return -EAGAIN; |
| 776 | if (signal_pending(current)) |
| 777 | return sock_intr_errno(*timeo_p); |
| 778 | |
| 779 | prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); |
| 780 | done = sk_wait_event(sk, timeo_p, |
| 781 | (!port->congested || !port->connected)); |
| 782 | finish_wait(sk_sleep(sk), &wait); |
| 783 | } while (!done); |
| 784 | return 0; |
| 785 | } |
| 786 | |
| 787 | /** |
| 788 | * tipc_send_stream - send stream-oriented data |
| 789 | * @iocb: (unused) |
| 790 | * @sock: socket structure |
| 791 | * @m: data to send |
| 792 | * @dsz: total length of data to be transmitted |
| 793 | * |
| 794 | * Used for SOCK_STREAM data. |
| 795 | * |
| 796 | * Returns the number of bytes sent on success (or partial success), |
| 797 | * or errno if no data sent |
| 798 | */ |
| 799 | static int tipc_send_stream(struct kiocb *iocb, struct socket *sock, |
| 800 | struct msghdr *m, size_t dsz) |
| 801 | { |
| 802 | struct sock *sk = sock->sk; |
| 803 | struct tipc_sock *tsk = tipc_sk(sk); |
| 804 | struct tipc_port *port = &tsk->port; |
| 805 | struct tipc_msg *mhdr = &port->phdr; |
| 806 | struct sk_buff *buf; |
| 807 | DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name); |
| 808 | u32 ref = port->ref; |
| 809 | int rc = -EINVAL; |
| 810 | long timeo; |
| 811 | u32 dnode; |
| 812 | uint mtu, send, sent = 0; |
| 813 | |
| 814 | /* Handle implied connection establishment */ |
| 815 | if (unlikely(dest)) { |
| 816 | rc = tipc_sendmsg(iocb, sock, m, dsz); |
| 817 | if (dsz && (dsz == rc)) |
| 818 | tsk->port.sent = 1; |
| 819 | return rc; |
| 820 | } |
| 821 | if (dsz > (uint)INT_MAX) |
| 822 | return -EMSGSIZE; |
| 823 | |
| 824 | if (iocb) |
| 825 | lock_sock(sk); |
| 826 | |
| 827 | if (unlikely(sock->state != SS_CONNECTED)) { |
| 828 | if (sock->state == SS_DISCONNECTING) |
| 829 | rc = -EPIPE; |
| 830 | else |
| 831 | rc = -ENOTCONN; |
| 832 | goto exit; |
| 833 | } |
| 834 | |
| 835 | timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT); |
| 836 | dnode = tipc_port_peernode(port); |
| 837 | port->congested = 1; |
| 838 | |
| 839 | next: |
| 840 | mtu = port->max_pkt; |
| 841 | send = min_t(uint, dsz - sent, TIPC_MAX_USER_MSG_SIZE); |
| 842 | rc = tipc_msg_build2(mhdr, m->msg_iov, sent, send, mtu, &buf); |
| 843 | if (unlikely(rc < 0)) |
| 844 | goto exit; |
| 845 | do { |
| 846 | port->congested = 1; |
| 847 | if (likely(!tipc_port_congested(port))) { |
| 848 | rc = tipc_link_xmit2(buf, dnode, ref); |
| 849 | if (likely(!rc)) { |
| 850 | port->sent++; |
| 851 | sent += send; |
| 852 | if (sent == dsz) |
| 853 | break; |
| 854 | goto next; |
| 855 | } |
| 856 | if (rc == -EMSGSIZE) { |
| 857 | port->max_pkt = tipc_node_get_mtu(dnode, ref); |
| 858 | goto next; |
| 859 | } |
| 860 | if (rc != -ELINKCONG) |
| 861 | break; |
| 862 | } |
| 863 | rc = tipc_wait_for_sndpkt(sock, &timeo); |
| 864 | } while (!rc); |
| 865 | |
| 866 | port->congested = 0; |
| 867 | exit: |
| 868 | if (iocb) |
| 869 | release_sock(sk); |
| 870 | return sent ? sent : rc; |
| 871 | } |
| 872 | |
| 873 | /** |
| 874 | * tipc_send_packet - send a connection-oriented message |
| 875 | * @iocb: if NULL, indicates that socket lock is already held |
| 876 | * @sock: socket structure |
| 877 | * @m: message to send |
| 878 | * @dsz: length of data to be transmitted |
| 879 | * |
| 880 | * Used for SOCK_SEQPACKET messages. |
| 881 | * |
| 882 | * Returns the number of bytes sent on success, or errno otherwise |
| 883 | */ |
| 884 | static int tipc_send_packet(struct kiocb *iocb, struct socket *sock, |
| 885 | struct msghdr *m, size_t dsz) |
| 886 | { |
| 887 | if (dsz > TIPC_MAX_USER_MSG_SIZE) |
| 888 | return -EMSGSIZE; |
| 889 | |
| 890 | return tipc_send_stream(iocb, sock, m, dsz); |
| 891 | } |
| 892 | |
| 893 | /** |
| 894 | * auto_connect - complete connection setup to a remote port |
| 895 | * @tsk: tipc socket structure |
| 896 | * @msg: peer's response message |
| 897 | * |
| 898 | * Returns 0 on success, errno otherwise |
| 899 | */ |
| 900 | static int auto_connect(struct tipc_sock *tsk, struct tipc_msg *msg) |
| 901 | { |
| 902 | struct tipc_port *port = &tsk->port; |
| 903 | struct socket *sock = tsk->sk.sk_socket; |
| 904 | struct tipc_portid peer; |
| 905 | |
| 906 | peer.ref = msg_origport(msg); |
| 907 | peer.node = msg_orignode(msg); |
| 908 | |
| 909 | __tipc_port_connect(port->ref, port, &peer); |
| 910 | |
| 911 | if (msg_importance(msg) > TIPC_CRITICAL_IMPORTANCE) |
| 912 | return -EINVAL; |
| 913 | msg_set_importance(&port->phdr, (u32)msg_importance(msg)); |
| 914 | sock->state = SS_CONNECTED; |
| 915 | return 0; |
| 916 | } |
| 917 | |
| 918 | /** |
| 919 | * set_orig_addr - capture sender's address for received message |
| 920 | * @m: descriptor for message info |
| 921 | * @msg: received message header |
| 922 | * |
| 923 | * Note: Address is not captured if not requested by receiver. |
| 924 | */ |
| 925 | static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg) |
| 926 | { |
| 927 | DECLARE_SOCKADDR(struct sockaddr_tipc *, addr, m->msg_name); |
| 928 | |
| 929 | if (addr) { |
| 930 | addr->family = AF_TIPC; |
| 931 | addr->addrtype = TIPC_ADDR_ID; |
| 932 | memset(&addr->addr, 0, sizeof(addr->addr)); |
| 933 | addr->addr.id.ref = msg_origport(msg); |
| 934 | addr->addr.id.node = msg_orignode(msg); |
| 935 | addr->addr.name.domain = 0; /* could leave uninitialized */ |
| 936 | addr->scope = 0; /* could leave uninitialized */ |
| 937 | m->msg_namelen = sizeof(struct sockaddr_tipc); |
| 938 | } |
| 939 | } |
| 940 | |
| 941 | /** |
| 942 | * anc_data_recv - optionally capture ancillary data for received message |
| 943 | * @m: descriptor for message info |
| 944 | * @msg: received message header |
| 945 | * @tport: TIPC port associated with message |
| 946 | * |
| 947 | * Note: Ancillary data is not captured if not requested by receiver. |
| 948 | * |
| 949 | * Returns 0 if successful, otherwise errno |
| 950 | */ |
| 951 | static int anc_data_recv(struct msghdr *m, struct tipc_msg *msg, |
| 952 | struct tipc_port *tport) |
| 953 | { |
| 954 | u32 anc_data[3]; |
| 955 | u32 err; |
| 956 | u32 dest_type; |
| 957 | int has_name; |
| 958 | int res; |
| 959 | |
| 960 | if (likely(m->msg_controllen == 0)) |
| 961 | return 0; |
| 962 | |
| 963 | /* Optionally capture errored message object(s) */ |
| 964 | err = msg ? msg_errcode(msg) : 0; |
| 965 | if (unlikely(err)) { |
| 966 | anc_data[0] = err; |
| 967 | anc_data[1] = msg_data_sz(msg); |
| 968 | res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data); |
| 969 | if (res) |
| 970 | return res; |
| 971 | if (anc_data[1]) { |
| 972 | res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1], |
| 973 | msg_data(msg)); |
| 974 | if (res) |
| 975 | return res; |
| 976 | } |
| 977 | } |
| 978 | |
| 979 | /* Optionally capture message destination object */ |
| 980 | dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG; |
| 981 | switch (dest_type) { |
| 982 | case TIPC_NAMED_MSG: |
| 983 | has_name = 1; |
| 984 | anc_data[0] = msg_nametype(msg); |
| 985 | anc_data[1] = msg_namelower(msg); |
| 986 | anc_data[2] = msg_namelower(msg); |
| 987 | break; |
| 988 | case TIPC_MCAST_MSG: |
| 989 | has_name = 1; |
| 990 | anc_data[0] = msg_nametype(msg); |
| 991 | anc_data[1] = msg_namelower(msg); |
| 992 | anc_data[2] = msg_nameupper(msg); |
| 993 | break; |
| 994 | case TIPC_CONN_MSG: |
| 995 | has_name = (tport->conn_type != 0); |
| 996 | anc_data[0] = tport->conn_type; |
| 997 | anc_data[1] = tport->conn_instance; |
| 998 | anc_data[2] = tport->conn_instance; |
| 999 | break; |
| 1000 | default: |
| 1001 | has_name = 0; |
| 1002 | } |
| 1003 | if (has_name) { |
| 1004 | res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data); |
| 1005 | if (res) |
| 1006 | return res; |
| 1007 | } |
| 1008 | |
| 1009 | return 0; |
| 1010 | } |
| 1011 | |
| 1012 | static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop) |
| 1013 | { |
| 1014 | struct sock *sk = sock->sk; |
| 1015 | DEFINE_WAIT(wait); |
| 1016 | long timeo = *timeop; |
| 1017 | int err; |
| 1018 | |
| 1019 | for (;;) { |
| 1020 | prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); |
| 1021 | if (timeo && skb_queue_empty(&sk->sk_receive_queue)) { |
| 1022 | if (sock->state == SS_DISCONNECTING) { |
| 1023 | err = -ENOTCONN; |
| 1024 | break; |
| 1025 | } |
| 1026 | release_sock(sk); |
| 1027 | timeo = schedule_timeout(timeo); |
| 1028 | lock_sock(sk); |
| 1029 | } |
| 1030 | err = 0; |
| 1031 | if (!skb_queue_empty(&sk->sk_receive_queue)) |
| 1032 | break; |
| 1033 | err = sock_intr_errno(timeo); |
| 1034 | if (signal_pending(current)) |
| 1035 | break; |
| 1036 | err = -EAGAIN; |
| 1037 | if (!timeo) |
| 1038 | break; |
| 1039 | } |
| 1040 | finish_wait(sk_sleep(sk), &wait); |
| 1041 | *timeop = timeo; |
| 1042 | return err; |
| 1043 | } |
| 1044 | |
| 1045 | /** |
| 1046 | * tipc_recvmsg - receive packet-oriented message |
| 1047 | * @iocb: (unused) |
| 1048 | * @m: descriptor for message info |
| 1049 | * @buf_len: total size of user buffer area |
| 1050 | * @flags: receive flags |
| 1051 | * |
| 1052 | * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages. |
| 1053 | * If the complete message doesn't fit in user area, truncate it. |
| 1054 | * |
| 1055 | * Returns size of returned message data, errno otherwise |
| 1056 | */ |
| 1057 | static int tipc_recvmsg(struct kiocb *iocb, struct socket *sock, |
| 1058 | struct msghdr *m, size_t buf_len, int flags) |
| 1059 | { |
| 1060 | struct sock *sk = sock->sk; |
| 1061 | struct tipc_sock *tsk = tipc_sk(sk); |
| 1062 | struct tipc_port *port = &tsk->port; |
| 1063 | struct sk_buff *buf; |
| 1064 | struct tipc_msg *msg; |
| 1065 | long timeo; |
| 1066 | unsigned int sz; |
| 1067 | u32 err; |
| 1068 | int res; |
| 1069 | |
| 1070 | /* Catch invalid receive requests */ |
| 1071 | if (unlikely(!buf_len)) |
| 1072 | return -EINVAL; |
| 1073 | |
| 1074 | lock_sock(sk); |
| 1075 | |
| 1076 | if (unlikely(sock->state == SS_UNCONNECTED)) { |
| 1077 | res = -ENOTCONN; |
| 1078 | goto exit; |
| 1079 | } |
| 1080 | |
| 1081 | timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT); |
| 1082 | restart: |
| 1083 | |
| 1084 | /* Look for a message in receive queue; wait if necessary */ |
| 1085 | res = tipc_wait_for_rcvmsg(sock, &timeo); |
| 1086 | if (res) |
| 1087 | goto exit; |
| 1088 | |
| 1089 | /* Look at first message in receive queue */ |
| 1090 | buf = skb_peek(&sk->sk_receive_queue); |
| 1091 | msg = buf_msg(buf); |
| 1092 | sz = msg_data_sz(msg); |
| 1093 | err = msg_errcode(msg); |
| 1094 | |
| 1095 | /* Discard an empty non-errored message & try again */ |
| 1096 | if ((!sz) && (!err)) { |
| 1097 | advance_rx_queue(sk); |
| 1098 | goto restart; |
| 1099 | } |
| 1100 | |
| 1101 | /* Capture sender's address (optional) */ |
| 1102 | set_orig_addr(m, msg); |
| 1103 | |
| 1104 | /* Capture ancillary data (optional) */ |
| 1105 | res = anc_data_recv(m, msg, port); |
| 1106 | if (res) |
| 1107 | goto exit; |
| 1108 | |
| 1109 | /* Capture message data (if valid) & compute return value (always) */ |
| 1110 | if (!err) { |
| 1111 | if (unlikely(buf_len < sz)) { |
| 1112 | sz = buf_len; |
| 1113 | m->msg_flags |= MSG_TRUNC; |
| 1114 | } |
| 1115 | res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg), |
| 1116 | m->msg_iov, sz); |
| 1117 | if (res) |
| 1118 | goto exit; |
| 1119 | res = sz; |
| 1120 | } else { |
| 1121 | if ((sock->state == SS_READY) || |
| 1122 | ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)) |
| 1123 | res = 0; |
| 1124 | else |
| 1125 | res = -ECONNRESET; |
| 1126 | } |
| 1127 | |
| 1128 | /* Consume received message (optional) */ |
| 1129 | if (likely(!(flags & MSG_PEEK))) { |
| 1130 | if ((sock->state != SS_READY) && |
| 1131 | (++port->conn_unacked >= TIPC_CONNACK_INTV)) |
| 1132 | tipc_acknowledge(port->ref, port->conn_unacked); |
| 1133 | advance_rx_queue(sk); |
| 1134 | } |
| 1135 | exit: |
| 1136 | release_sock(sk); |
| 1137 | return res; |
| 1138 | } |
| 1139 | |
| 1140 | /** |
| 1141 | * tipc_recv_stream - receive stream-oriented data |
| 1142 | * @iocb: (unused) |
| 1143 | * @m: descriptor for message info |
| 1144 | * @buf_len: total size of user buffer area |
| 1145 | * @flags: receive flags |
| 1146 | * |
| 1147 | * Used for SOCK_STREAM messages only. If not enough data is available |
| 1148 | * will optionally wait for more; never truncates data. |
| 1149 | * |
| 1150 | * Returns size of returned message data, errno otherwise |
| 1151 | */ |
| 1152 | static int tipc_recv_stream(struct kiocb *iocb, struct socket *sock, |
| 1153 | struct msghdr *m, size_t buf_len, int flags) |
| 1154 | { |
| 1155 | struct sock *sk = sock->sk; |
| 1156 | struct tipc_sock *tsk = tipc_sk(sk); |
| 1157 | struct tipc_port *port = &tsk->port; |
| 1158 | struct sk_buff *buf; |
| 1159 | struct tipc_msg *msg; |
| 1160 | long timeo; |
| 1161 | unsigned int sz; |
| 1162 | int sz_to_copy, target, needed; |
| 1163 | int sz_copied = 0; |
| 1164 | u32 err; |
| 1165 | int res = 0; |
| 1166 | |
| 1167 | /* Catch invalid receive attempts */ |
| 1168 | if (unlikely(!buf_len)) |
| 1169 | return -EINVAL; |
| 1170 | |
| 1171 | lock_sock(sk); |
| 1172 | |
| 1173 | if (unlikely(sock->state == SS_UNCONNECTED)) { |
| 1174 | res = -ENOTCONN; |
| 1175 | goto exit; |
| 1176 | } |
| 1177 | |
| 1178 | target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len); |
| 1179 | timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT); |
| 1180 | |
| 1181 | restart: |
| 1182 | /* Look for a message in receive queue; wait if necessary */ |
| 1183 | res = tipc_wait_for_rcvmsg(sock, &timeo); |
| 1184 | if (res) |
| 1185 | goto exit; |
| 1186 | |
| 1187 | /* Look at first message in receive queue */ |
| 1188 | buf = skb_peek(&sk->sk_receive_queue); |
| 1189 | msg = buf_msg(buf); |
| 1190 | sz = msg_data_sz(msg); |
| 1191 | err = msg_errcode(msg); |
| 1192 | |
| 1193 | /* Discard an empty non-errored message & try again */ |
| 1194 | if ((!sz) && (!err)) { |
| 1195 | advance_rx_queue(sk); |
| 1196 | goto restart; |
| 1197 | } |
| 1198 | |
| 1199 | /* Optionally capture sender's address & ancillary data of first msg */ |
| 1200 | if (sz_copied == 0) { |
| 1201 | set_orig_addr(m, msg); |
| 1202 | res = anc_data_recv(m, msg, port); |
| 1203 | if (res) |
| 1204 | goto exit; |
| 1205 | } |
| 1206 | |
| 1207 | /* Capture message data (if valid) & compute return value (always) */ |
| 1208 | if (!err) { |
| 1209 | u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle); |
| 1210 | |
| 1211 | sz -= offset; |
| 1212 | needed = (buf_len - sz_copied); |
| 1213 | sz_to_copy = (sz <= needed) ? sz : needed; |
| 1214 | |
| 1215 | res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg) + offset, |
| 1216 | m->msg_iov, sz_to_copy); |
| 1217 | if (res) |
| 1218 | goto exit; |
| 1219 | |
| 1220 | sz_copied += sz_to_copy; |
| 1221 | |
| 1222 | if (sz_to_copy < sz) { |
| 1223 | if (!(flags & MSG_PEEK)) |
| 1224 | TIPC_SKB_CB(buf)->handle = |
| 1225 | (void *)(unsigned long)(offset + sz_to_copy); |
| 1226 | goto exit; |
| 1227 | } |
| 1228 | } else { |
| 1229 | if (sz_copied != 0) |
| 1230 | goto exit; /* can't add error msg to valid data */ |
| 1231 | |
| 1232 | if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control) |
| 1233 | res = 0; |
| 1234 | else |
| 1235 | res = -ECONNRESET; |
| 1236 | } |
| 1237 | |
| 1238 | /* Consume received message (optional) */ |
| 1239 | if (likely(!(flags & MSG_PEEK))) { |
| 1240 | if (unlikely(++port->conn_unacked >= TIPC_CONNACK_INTV)) |
| 1241 | tipc_acknowledge(port->ref, port->conn_unacked); |
| 1242 | advance_rx_queue(sk); |
| 1243 | } |
| 1244 | |
| 1245 | /* Loop around if more data is required */ |
| 1246 | if ((sz_copied < buf_len) && /* didn't get all requested data */ |
| 1247 | (!skb_queue_empty(&sk->sk_receive_queue) || |
| 1248 | (sz_copied < target)) && /* and more is ready or required */ |
| 1249 | (!(flags & MSG_PEEK)) && /* and aren't just peeking at data */ |
| 1250 | (!err)) /* and haven't reached a FIN */ |
| 1251 | goto restart; |
| 1252 | |
| 1253 | exit: |
| 1254 | release_sock(sk); |
| 1255 | return sz_copied ? sz_copied : res; |
| 1256 | } |
| 1257 | |
| 1258 | /** |
| 1259 | * tipc_write_space - wake up thread if port congestion is released |
| 1260 | * @sk: socket |
| 1261 | */ |
| 1262 | static void tipc_write_space(struct sock *sk) |
| 1263 | { |
| 1264 | struct socket_wq *wq; |
| 1265 | |
| 1266 | rcu_read_lock(); |
| 1267 | wq = rcu_dereference(sk->sk_wq); |
| 1268 | if (wq_has_sleeper(wq)) |
| 1269 | wake_up_interruptible_sync_poll(&wq->wait, POLLOUT | |
| 1270 | POLLWRNORM | POLLWRBAND); |
| 1271 | rcu_read_unlock(); |
| 1272 | } |
| 1273 | |
| 1274 | /** |
| 1275 | * tipc_data_ready - wake up threads to indicate messages have been received |
| 1276 | * @sk: socket |
| 1277 | * @len: the length of messages |
| 1278 | */ |
| 1279 | static void tipc_data_ready(struct sock *sk) |
| 1280 | { |
| 1281 | struct socket_wq *wq; |
| 1282 | |
| 1283 | rcu_read_lock(); |
| 1284 | wq = rcu_dereference(sk->sk_wq); |
| 1285 | if (wq_has_sleeper(wq)) |
| 1286 | wake_up_interruptible_sync_poll(&wq->wait, POLLIN | |
| 1287 | POLLRDNORM | POLLRDBAND); |
| 1288 | rcu_read_unlock(); |
| 1289 | } |
| 1290 | |
| 1291 | /** |
| 1292 | * filter_connect - Handle all incoming messages for a connection-based socket |
| 1293 | * @tsk: TIPC socket |
| 1294 | * @msg: message |
| 1295 | * |
| 1296 | * Returns 0 (TIPC_OK) if everyting ok, -TIPC_ERR_NO_PORT otherwise |
| 1297 | */ |
| 1298 | static int filter_connect(struct tipc_sock *tsk, struct sk_buff **buf) |
| 1299 | { |
| 1300 | struct sock *sk = &tsk->sk; |
| 1301 | struct tipc_port *port = &tsk->port; |
| 1302 | struct socket *sock = sk->sk_socket; |
| 1303 | struct tipc_msg *msg = buf_msg(*buf); |
| 1304 | |
| 1305 | int retval = -TIPC_ERR_NO_PORT; |
| 1306 | int res; |
| 1307 | |
| 1308 | if (msg_mcast(msg)) |
| 1309 | return retval; |
| 1310 | |
| 1311 | switch ((int)sock->state) { |
| 1312 | case SS_CONNECTED: |
| 1313 | /* Accept only connection-based messages sent by peer */ |
| 1314 | if (msg_connected(msg) && tipc_port_peer_msg(port, msg)) { |
| 1315 | if (unlikely(msg_errcode(msg))) { |
| 1316 | sock->state = SS_DISCONNECTING; |
| 1317 | __tipc_port_disconnect(port); |
| 1318 | } |
| 1319 | retval = TIPC_OK; |
| 1320 | } |
| 1321 | break; |
| 1322 | case SS_CONNECTING: |
| 1323 | /* Accept only ACK or NACK message */ |
| 1324 | if (unlikely(msg_errcode(msg))) { |
| 1325 | sock->state = SS_DISCONNECTING; |
| 1326 | sk->sk_err = ECONNREFUSED; |
| 1327 | retval = TIPC_OK; |
| 1328 | break; |
| 1329 | } |
| 1330 | |
| 1331 | if (unlikely(!msg_connected(msg))) |
| 1332 | break; |
| 1333 | |
| 1334 | res = auto_connect(tsk, msg); |
| 1335 | if (res) { |
| 1336 | sock->state = SS_DISCONNECTING; |
| 1337 | sk->sk_err = -res; |
| 1338 | retval = TIPC_OK; |
| 1339 | break; |
| 1340 | } |
| 1341 | |
| 1342 | /* If an incoming message is an 'ACK-', it should be |
| 1343 | * discarded here because it doesn't contain useful |
| 1344 | * data. In addition, we should try to wake up |
| 1345 | * connect() routine if sleeping. |
| 1346 | */ |
| 1347 | if (msg_data_sz(msg) == 0) { |
| 1348 | kfree_skb(*buf); |
| 1349 | *buf = NULL; |
| 1350 | if (waitqueue_active(sk_sleep(sk))) |
| 1351 | wake_up_interruptible(sk_sleep(sk)); |
| 1352 | } |
| 1353 | retval = TIPC_OK; |
| 1354 | break; |
| 1355 | case SS_LISTENING: |
| 1356 | case SS_UNCONNECTED: |
| 1357 | /* Accept only SYN message */ |
| 1358 | if (!msg_connected(msg) && !(msg_errcode(msg))) |
| 1359 | retval = TIPC_OK; |
| 1360 | break; |
| 1361 | case SS_DISCONNECTING: |
| 1362 | break; |
| 1363 | default: |
| 1364 | pr_err("Unknown socket state %u\n", sock->state); |
| 1365 | } |
| 1366 | return retval; |
| 1367 | } |
| 1368 | |
| 1369 | /** |
| 1370 | * rcvbuf_limit - get proper overload limit of socket receive queue |
| 1371 | * @sk: socket |
| 1372 | * @buf: message |
| 1373 | * |
| 1374 | * For all connection oriented messages, irrespective of importance, |
| 1375 | * the default overload value (i.e. 67MB) is set as limit. |
| 1376 | * |
| 1377 | * For all connectionless messages, by default new queue limits are |
| 1378 | * as belows: |
| 1379 | * |
| 1380 | * TIPC_LOW_IMPORTANCE (4 MB) |
| 1381 | * TIPC_MEDIUM_IMPORTANCE (8 MB) |
| 1382 | * TIPC_HIGH_IMPORTANCE (16 MB) |
| 1383 | * TIPC_CRITICAL_IMPORTANCE (32 MB) |
| 1384 | * |
| 1385 | * Returns overload limit according to corresponding message importance |
| 1386 | */ |
| 1387 | static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *buf) |
| 1388 | { |
| 1389 | struct tipc_msg *msg = buf_msg(buf); |
| 1390 | |
| 1391 | if (msg_connected(msg)) |
| 1392 | return sysctl_tipc_rmem[2]; |
| 1393 | |
| 1394 | return sk->sk_rcvbuf >> TIPC_CRITICAL_IMPORTANCE << |
| 1395 | msg_importance(msg); |
| 1396 | } |
| 1397 | |
| 1398 | /** |
| 1399 | * filter_rcv - validate incoming message |
| 1400 | * @sk: socket |
| 1401 | * @buf: message |
| 1402 | * |
| 1403 | * Enqueues message on receive queue if acceptable; optionally handles |
| 1404 | * disconnect indication for a connected socket. |
| 1405 | * |
| 1406 | * Called with socket lock already taken; port lock may also be taken. |
| 1407 | * |
| 1408 | * Returns 0 (TIPC_OK) if message was consumed, -TIPC error code if message |
| 1409 | * to be rejected. |
| 1410 | */ |
| 1411 | static int filter_rcv(struct sock *sk, struct sk_buff *buf) |
| 1412 | { |
| 1413 | struct socket *sock = sk->sk_socket; |
| 1414 | struct tipc_sock *tsk = tipc_sk(sk); |
| 1415 | struct tipc_msg *msg = buf_msg(buf); |
| 1416 | unsigned int limit = rcvbuf_limit(sk, buf); |
| 1417 | int rc = TIPC_OK; |
| 1418 | |
| 1419 | if (unlikely(msg_user(msg) == CONN_MANAGER)) { |
| 1420 | tipc_port_proto_rcv(&tsk->port, buf); |
| 1421 | return TIPC_OK; |
| 1422 | } |
| 1423 | |
| 1424 | /* Reject message if it is wrong sort of message for socket */ |
| 1425 | if (msg_type(msg) > TIPC_DIRECT_MSG) |
| 1426 | return -TIPC_ERR_NO_PORT; |
| 1427 | |
| 1428 | if (sock->state == SS_READY) { |
| 1429 | if (msg_connected(msg)) |
| 1430 | return -TIPC_ERR_NO_PORT; |
| 1431 | } else { |
| 1432 | rc = filter_connect(tsk, &buf); |
| 1433 | if (rc != TIPC_OK || buf == NULL) |
| 1434 | return rc; |
| 1435 | } |
| 1436 | |
| 1437 | /* Reject message if there isn't room to queue it */ |
| 1438 | if (sk_rmem_alloc_get(sk) + buf->truesize >= limit) |
| 1439 | return -TIPC_ERR_OVERLOAD; |
| 1440 | |
| 1441 | /* Enqueue message */ |
| 1442 | TIPC_SKB_CB(buf)->handle = NULL; |
| 1443 | __skb_queue_tail(&sk->sk_receive_queue, buf); |
| 1444 | skb_set_owner_r(buf, sk); |
| 1445 | |
| 1446 | sk->sk_data_ready(sk); |
| 1447 | return TIPC_OK; |
| 1448 | } |
| 1449 | |
| 1450 | /** |
| 1451 | * tipc_backlog_rcv - handle incoming message from backlog queue |
| 1452 | * @sk: socket |
| 1453 | * @buf: message |
| 1454 | * |
| 1455 | * Caller must hold socket lock, but not port lock. |
| 1456 | * |
| 1457 | * Returns 0 |
| 1458 | */ |
| 1459 | static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *buf) |
| 1460 | { |
| 1461 | int rc; |
| 1462 | u32 onode; |
| 1463 | struct tipc_sock *tsk = tipc_sk(sk); |
| 1464 | uint truesize = buf->truesize; |
| 1465 | |
| 1466 | rc = filter_rcv(sk, buf); |
| 1467 | |
| 1468 | if (unlikely(rc && tipc_msg_reverse(buf, &onode, -rc))) |
| 1469 | tipc_link_xmit2(buf, onode, 0); |
| 1470 | else if (atomic_read(&tsk->dupl_rcvcnt) < TIPC_CONN_OVERLOAD_LIMIT) |
| 1471 | atomic_add(truesize, &tsk->dupl_rcvcnt); |
| 1472 | |
| 1473 | return 0; |
| 1474 | } |
| 1475 | |
| 1476 | /** |
| 1477 | * tipc_sk_rcv - handle incoming message |
| 1478 | * @buf: buffer containing arriving message |
| 1479 | * Consumes buffer |
| 1480 | * Returns 0 if success, or errno: -EHOSTUNREACH |
| 1481 | */ |
| 1482 | int tipc_sk_rcv(struct sk_buff *buf) |
| 1483 | { |
| 1484 | struct tipc_sock *tsk; |
| 1485 | struct tipc_port *port; |
| 1486 | struct sock *sk; |
| 1487 | u32 dport = msg_destport(buf_msg(buf)); |
| 1488 | int rc = TIPC_OK; |
| 1489 | uint limit; |
| 1490 | u32 dnode; |
| 1491 | |
| 1492 | /* Validate destination and message */ |
| 1493 | port = tipc_port_lock(dport); |
| 1494 | if (unlikely(!port)) { |
| 1495 | rc = tipc_msg_eval(buf, &dnode); |
| 1496 | goto exit; |
| 1497 | } |
| 1498 | |
| 1499 | tsk = tipc_port_to_sock(port); |
| 1500 | sk = &tsk->sk; |
| 1501 | |
| 1502 | /* Queue message */ |
| 1503 | bh_lock_sock(sk); |
| 1504 | |
| 1505 | if (!sock_owned_by_user(sk)) { |
| 1506 | rc = filter_rcv(sk, buf); |
| 1507 | } else { |
| 1508 | if (sk->sk_backlog.len == 0) |
| 1509 | atomic_set(&tsk->dupl_rcvcnt, 0); |
| 1510 | limit = rcvbuf_limit(sk, buf) + atomic_read(&tsk->dupl_rcvcnt); |
| 1511 | if (sk_add_backlog(sk, buf, limit)) |
| 1512 | rc = -TIPC_ERR_OVERLOAD; |
| 1513 | } |
| 1514 | bh_unlock_sock(sk); |
| 1515 | tipc_port_unlock(port); |
| 1516 | |
| 1517 | if (likely(!rc)) |
| 1518 | return 0; |
| 1519 | exit: |
| 1520 | if ((rc < 0) && !tipc_msg_reverse(buf, &dnode, -rc)) |
| 1521 | return -EHOSTUNREACH; |
| 1522 | |
| 1523 | tipc_link_xmit2(buf, dnode, 0); |
| 1524 | return (rc < 0) ? -EHOSTUNREACH : 0; |
| 1525 | } |
| 1526 | |
| 1527 | static int tipc_wait_for_connect(struct socket *sock, long *timeo_p) |
| 1528 | { |
| 1529 | struct sock *sk = sock->sk; |
| 1530 | DEFINE_WAIT(wait); |
| 1531 | int done; |
| 1532 | |
| 1533 | do { |
| 1534 | int err = sock_error(sk); |
| 1535 | if (err) |
| 1536 | return err; |
| 1537 | if (!*timeo_p) |
| 1538 | return -ETIMEDOUT; |
| 1539 | if (signal_pending(current)) |
| 1540 | return sock_intr_errno(*timeo_p); |
| 1541 | |
| 1542 | prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); |
| 1543 | done = sk_wait_event(sk, timeo_p, sock->state != SS_CONNECTING); |
| 1544 | finish_wait(sk_sleep(sk), &wait); |
| 1545 | } while (!done); |
| 1546 | return 0; |
| 1547 | } |
| 1548 | |
| 1549 | /** |
| 1550 | * tipc_connect - establish a connection to another TIPC port |
| 1551 | * @sock: socket structure |
| 1552 | * @dest: socket address for destination port |
| 1553 | * @destlen: size of socket address data structure |
| 1554 | * @flags: file-related flags associated with socket |
| 1555 | * |
| 1556 | * Returns 0 on success, errno otherwise |
| 1557 | */ |
| 1558 | static int tipc_connect(struct socket *sock, struct sockaddr *dest, |
| 1559 | int destlen, int flags) |
| 1560 | { |
| 1561 | struct sock *sk = sock->sk; |
| 1562 | struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest; |
| 1563 | struct msghdr m = {NULL,}; |
| 1564 | long timeout = (flags & O_NONBLOCK) ? 0 : tipc_sk(sk)->conn_timeout; |
| 1565 | socket_state previous; |
| 1566 | int res; |
| 1567 | |
| 1568 | lock_sock(sk); |
| 1569 | |
| 1570 | /* For now, TIPC does not allow use of connect() with DGRAM/RDM types */ |
| 1571 | if (sock->state == SS_READY) { |
| 1572 | res = -EOPNOTSUPP; |
| 1573 | goto exit; |
| 1574 | } |
| 1575 | |
| 1576 | /* |
| 1577 | * Reject connection attempt using multicast address |
| 1578 | * |
| 1579 | * Note: send_msg() validates the rest of the address fields, |
| 1580 | * so there's no need to do it here |
| 1581 | */ |
| 1582 | if (dst->addrtype == TIPC_ADDR_MCAST) { |
| 1583 | res = -EINVAL; |
| 1584 | goto exit; |
| 1585 | } |
| 1586 | |
| 1587 | previous = sock->state; |
| 1588 | switch (sock->state) { |
| 1589 | case SS_UNCONNECTED: |
| 1590 | /* Send a 'SYN-' to destination */ |
| 1591 | m.msg_name = dest; |
| 1592 | m.msg_namelen = destlen; |
| 1593 | |
| 1594 | /* If connect is in non-blocking case, set MSG_DONTWAIT to |
| 1595 | * indicate send_msg() is never blocked. |
| 1596 | */ |
| 1597 | if (!timeout) |
| 1598 | m.msg_flags = MSG_DONTWAIT; |
| 1599 | |
| 1600 | res = tipc_sendmsg(NULL, sock, &m, 0); |
| 1601 | if ((res < 0) && (res != -EWOULDBLOCK)) |
| 1602 | goto exit; |
| 1603 | |
| 1604 | /* Just entered SS_CONNECTING state; the only |
| 1605 | * difference is that return value in non-blocking |
| 1606 | * case is EINPROGRESS, rather than EALREADY. |
| 1607 | */ |
| 1608 | res = -EINPROGRESS; |
| 1609 | case SS_CONNECTING: |
| 1610 | if (previous == SS_CONNECTING) |
| 1611 | res = -EALREADY; |
| 1612 | if (!timeout) |
| 1613 | goto exit; |
| 1614 | timeout = msecs_to_jiffies(timeout); |
| 1615 | /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */ |
| 1616 | res = tipc_wait_for_connect(sock, &timeout); |
| 1617 | break; |
| 1618 | case SS_CONNECTED: |
| 1619 | res = -EISCONN; |
| 1620 | break; |
| 1621 | default: |
| 1622 | res = -EINVAL; |
| 1623 | break; |
| 1624 | } |
| 1625 | exit: |
| 1626 | release_sock(sk); |
| 1627 | return res; |
| 1628 | } |
| 1629 | |
| 1630 | /** |
| 1631 | * tipc_listen - allow socket to listen for incoming connections |
| 1632 | * @sock: socket structure |
| 1633 | * @len: (unused) |
| 1634 | * |
| 1635 | * Returns 0 on success, errno otherwise |
| 1636 | */ |
| 1637 | static int tipc_listen(struct socket *sock, int len) |
| 1638 | { |
| 1639 | struct sock *sk = sock->sk; |
| 1640 | int res; |
| 1641 | |
| 1642 | lock_sock(sk); |
| 1643 | |
| 1644 | if (sock->state != SS_UNCONNECTED) |
| 1645 | res = -EINVAL; |
| 1646 | else { |
| 1647 | sock->state = SS_LISTENING; |
| 1648 | res = 0; |
| 1649 | } |
| 1650 | |
| 1651 | release_sock(sk); |
| 1652 | return res; |
| 1653 | } |
| 1654 | |
| 1655 | static int tipc_wait_for_accept(struct socket *sock, long timeo) |
| 1656 | { |
| 1657 | struct sock *sk = sock->sk; |
| 1658 | DEFINE_WAIT(wait); |
| 1659 | int err; |
| 1660 | |
| 1661 | /* True wake-one mechanism for incoming connections: only |
| 1662 | * one process gets woken up, not the 'whole herd'. |
| 1663 | * Since we do not 'race & poll' for established sockets |
| 1664 | * anymore, the common case will execute the loop only once. |
| 1665 | */ |
| 1666 | for (;;) { |
| 1667 | prepare_to_wait_exclusive(sk_sleep(sk), &wait, |
| 1668 | TASK_INTERRUPTIBLE); |
| 1669 | if (timeo && skb_queue_empty(&sk->sk_receive_queue)) { |
| 1670 | release_sock(sk); |
| 1671 | timeo = schedule_timeout(timeo); |
| 1672 | lock_sock(sk); |
| 1673 | } |
| 1674 | err = 0; |
| 1675 | if (!skb_queue_empty(&sk->sk_receive_queue)) |
| 1676 | break; |
| 1677 | err = -EINVAL; |
| 1678 | if (sock->state != SS_LISTENING) |
| 1679 | break; |
| 1680 | err = sock_intr_errno(timeo); |
| 1681 | if (signal_pending(current)) |
| 1682 | break; |
| 1683 | err = -EAGAIN; |
| 1684 | if (!timeo) |
| 1685 | break; |
| 1686 | } |
| 1687 | finish_wait(sk_sleep(sk), &wait); |
| 1688 | return err; |
| 1689 | } |
| 1690 | |
| 1691 | /** |
| 1692 | * tipc_accept - wait for connection request |
| 1693 | * @sock: listening socket |
| 1694 | * @newsock: new socket that is to be connected |
| 1695 | * @flags: file-related flags associated with socket |
| 1696 | * |
| 1697 | * Returns 0 on success, errno otherwise |
| 1698 | */ |
| 1699 | static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags) |
| 1700 | { |
| 1701 | struct sock *new_sk, *sk = sock->sk; |
| 1702 | struct sk_buff *buf; |
| 1703 | struct tipc_port *new_port; |
| 1704 | struct tipc_msg *msg; |
| 1705 | struct tipc_portid peer; |
| 1706 | u32 new_ref; |
| 1707 | long timeo; |
| 1708 | int res; |
| 1709 | |
| 1710 | lock_sock(sk); |
| 1711 | |
| 1712 | if (sock->state != SS_LISTENING) { |
| 1713 | res = -EINVAL; |
| 1714 | goto exit; |
| 1715 | } |
| 1716 | timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK); |
| 1717 | res = tipc_wait_for_accept(sock, timeo); |
| 1718 | if (res) |
| 1719 | goto exit; |
| 1720 | |
| 1721 | buf = skb_peek(&sk->sk_receive_queue); |
| 1722 | |
| 1723 | res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, 1); |
| 1724 | if (res) |
| 1725 | goto exit; |
| 1726 | |
| 1727 | new_sk = new_sock->sk; |
| 1728 | new_port = &tipc_sk(new_sk)->port; |
| 1729 | new_ref = new_port->ref; |
| 1730 | msg = buf_msg(buf); |
| 1731 | |
| 1732 | /* we lock on new_sk; but lockdep sees the lock on sk */ |
| 1733 | lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING); |
| 1734 | |
| 1735 | /* |
| 1736 | * Reject any stray messages received by new socket |
| 1737 | * before the socket lock was taken (very, very unlikely) |
| 1738 | */ |
| 1739 | reject_rx_queue(new_sk); |
| 1740 | |
| 1741 | /* Connect new socket to it's peer */ |
| 1742 | peer.ref = msg_origport(msg); |
| 1743 | peer.node = msg_orignode(msg); |
| 1744 | tipc_port_connect(new_ref, &peer); |
| 1745 | new_sock->state = SS_CONNECTED; |
| 1746 | |
| 1747 | tipc_port_set_importance(new_port, msg_importance(msg)); |
| 1748 | if (msg_named(msg)) { |
| 1749 | new_port->conn_type = msg_nametype(msg); |
| 1750 | new_port->conn_instance = msg_nameinst(msg); |
| 1751 | } |
| 1752 | |
| 1753 | /* |
| 1754 | * Respond to 'SYN-' by discarding it & returning 'ACK'-. |
| 1755 | * Respond to 'SYN+' by queuing it on new socket. |
| 1756 | */ |
| 1757 | if (!msg_data_sz(msg)) { |
| 1758 | struct msghdr m = {NULL,}; |
| 1759 | |
| 1760 | advance_rx_queue(sk); |
| 1761 | tipc_send_packet(NULL, new_sock, &m, 0); |
| 1762 | } else { |
| 1763 | __skb_dequeue(&sk->sk_receive_queue); |
| 1764 | __skb_queue_head(&new_sk->sk_receive_queue, buf); |
| 1765 | skb_set_owner_r(buf, new_sk); |
| 1766 | } |
| 1767 | release_sock(new_sk); |
| 1768 | exit: |
| 1769 | release_sock(sk); |
| 1770 | return res; |
| 1771 | } |
| 1772 | |
| 1773 | /** |
| 1774 | * tipc_shutdown - shutdown socket connection |
| 1775 | * @sock: socket structure |
| 1776 | * @how: direction to close (must be SHUT_RDWR) |
| 1777 | * |
| 1778 | * Terminates connection (if necessary), then purges socket's receive queue. |
| 1779 | * |
| 1780 | * Returns 0 on success, errno otherwise |
| 1781 | */ |
| 1782 | static int tipc_shutdown(struct socket *sock, int how) |
| 1783 | { |
| 1784 | struct sock *sk = sock->sk; |
| 1785 | struct tipc_sock *tsk = tipc_sk(sk); |
| 1786 | struct tipc_port *port = &tsk->port; |
| 1787 | struct sk_buff *buf; |
| 1788 | u32 peer; |
| 1789 | int res; |
| 1790 | |
| 1791 | if (how != SHUT_RDWR) |
| 1792 | return -EINVAL; |
| 1793 | |
| 1794 | lock_sock(sk); |
| 1795 | |
| 1796 | switch (sock->state) { |
| 1797 | case SS_CONNECTING: |
| 1798 | case SS_CONNECTED: |
| 1799 | |
| 1800 | restart: |
| 1801 | /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */ |
| 1802 | buf = __skb_dequeue(&sk->sk_receive_queue); |
| 1803 | if (buf) { |
| 1804 | if (TIPC_SKB_CB(buf)->handle != NULL) { |
| 1805 | kfree_skb(buf); |
| 1806 | goto restart; |
| 1807 | } |
| 1808 | tipc_port_disconnect(port->ref); |
| 1809 | if (tipc_msg_reverse(buf, &peer, TIPC_CONN_SHUTDOWN)) |
| 1810 | tipc_link_xmit2(buf, peer, 0); |
| 1811 | } else { |
| 1812 | tipc_port_shutdown(port->ref); |
| 1813 | } |
| 1814 | |
| 1815 | sock->state = SS_DISCONNECTING; |
| 1816 | |
| 1817 | /* fall through */ |
| 1818 | |
| 1819 | case SS_DISCONNECTING: |
| 1820 | |
| 1821 | /* Discard any unreceived messages */ |
| 1822 | __skb_queue_purge(&sk->sk_receive_queue); |
| 1823 | |
| 1824 | /* Wake up anyone sleeping in poll */ |
| 1825 | sk->sk_state_change(sk); |
| 1826 | res = 0; |
| 1827 | break; |
| 1828 | |
| 1829 | default: |
| 1830 | res = -ENOTCONN; |
| 1831 | } |
| 1832 | |
| 1833 | release_sock(sk); |
| 1834 | return res; |
| 1835 | } |
| 1836 | |
| 1837 | /** |
| 1838 | * tipc_setsockopt - set socket option |
| 1839 | * @sock: socket structure |
| 1840 | * @lvl: option level |
| 1841 | * @opt: option identifier |
| 1842 | * @ov: pointer to new option value |
| 1843 | * @ol: length of option value |
| 1844 | * |
| 1845 | * For stream sockets only, accepts and ignores all IPPROTO_TCP options |
| 1846 | * (to ease compatibility). |
| 1847 | * |
| 1848 | * Returns 0 on success, errno otherwise |
| 1849 | */ |
| 1850 | static int tipc_setsockopt(struct socket *sock, int lvl, int opt, |
| 1851 | char __user *ov, unsigned int ol) |
| 1852 | { |
| 1853 | struct sock *sk = sock->sk; |
| 1854 | struct tipc_sock *tsk = tipc_sk(sk); |
| 1855 | struct tipc_port *port = &tsk->port; |
| 1856 | u32 value; |
| 1857 | int res; |
| 1858 | |
| 1859 | if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM)) |
| 1860 | return 0; |
| 1861 | if (lvl != SOL_TIPC) |
| 1862 | return -ENOPROTOOPT; |
| 1863 | if (ol < sizeof(value)) |
| 1864 | return -EINVAL; |
| 1865 | res = get_user(value, (u32 __user *)ov); |
| 1866 | if (res) |
| 1867 | return res; |
| 1868 | |
| 1869 | lock_sock(sk); |
| 1870 | |
| 1871 | switch (opt) { |
| 1872 | case TIPC_IMPORTANCE: |
| 1873 | tipc_port_set_importance(port, value); |
| 1874 | break; |
| 1875 | case TIPC_SRC_DROPPABLE: |
| 1876 | if (sock->type != SOCK_STREAM) |
| 1877 | tipc_port_set_unreliable(port, value); |
| 1878 | else |
| 1879 | res = -ENOPROTOOPT; |
| 1880 | break; |
| 1881 | case TIPC_DEST_DROPPABLE: |
| 1882 | tipc_port_set_unreturnable(port, value); |
| 1883 | break; |
| 1884 | case TIPC_CONN_TIMEOUT: |
| 1885 | tipc_sk(sk)->conn_timeout = value; |
| 1886 | /* no need to set "res", since already 0 at this point */ |
| 1887 | break; |
| 1888 | default: |
| 1889 | res = -EINVAL; |
| 1890 | } |
| 1891 | |
| 1892 | release_sock(sk); |
| 1893 | |
| 1894 | return res; |
| 1895 | } |
| 1896 | |
| 1897 | /** |
| 1898 | * tipc_getsockopt - get socket option |
| 1899 | * @sock: socket structure |
| 1900 | * @lvl: option level |
| 1901 | * @opt: option identifier |
| 1902 | * @ov: receptacle for option value |
| 1903 | * @ol: receptacle for length of option value |
| 1904 | * |
| 1905 | * For stream sockets only, returns 0 length result for all IPPROTO_TCP options |
| 1906 | * (to ease compatibility). |
| 1907 | * |
| 1908 | * Returns 0 on success, errno otherwise |
| 1909 | */ |
| 1910 | static int tipc_getsockopt(struct socket *sock, int lvl, int opt, |
| 1911 | char __user *ov, int __user *ol) |
| 1912 | { |
| 1913 | struct sock *sk = sock->sk; |
| 1914 | struct tipc_sock *tsk = tipc_sk(sk); |
| 1915 | struct tipc_port *port = &tsk->port; |
| 1916 | int len; |
| 1917 | u32 value; |
| 1918 | int res; |
| 1919 | |
| 1920 | if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM)) |
| 1921 | return put_user(0, ol); |
| 1922 | if (lvl != SOL_TIPC) |
| 1923 | return -ENOPROTOOPT; |
| 1924 | res = get_user(len, ol); |
| 1925 | if (res) |
| 1926 | return res; |
| 1927 | |
| 1928 | lock_sock(sk); |
| 1929 | |
| 1930 | switch (opt) { |
| 1931 | case TIPC_IMPORTANCE: |
| 1932 | value = tipc_port_importance(port); |
| 1933 | break; |
| 1934 | case TIPC_SRC_DROPPABLE: |
| 1935 | value = tipc_port_unreliable(port); |
| 1936 | break; |
| 1937 | case TIPC_DEST_DROPPABLE: |
| 1938 | value = tipc_port_unreturnable(port); |
| 1939 | break; |
| 1940 | case TIPC_CONN_TIMEOUT: |
| 1941 | value = tipc_sk(sk)->conn_timeout; |
| 1942 | /* no need to set "res", since already 0 at this point */ |
| 1943 | break; |
| 1944 | case TIPC_NODE_RECVQ_DEPTH: |
| 1945 | value = 0; /* was tipc_queue_size, now obsolete */ |
| 1946 | break; |
| 1947 | case TIPC_SOCK_RECVQ_DEPTH: |
| 1948 | value = skb_queue_len(&sk->sk_receive_queue); |
| 1949 | break; |
| 1950 | default: |
| 1951 | res = -EINVAL; |
| 1952 | } |
| 1953 | |
| 1954 | release_sock(sk); |
| 1955 | |
| 1956 | if (res) |
| 1957 | return res; /* "get" failed */ |
| 1958 | |
| 1959 | if (len < sizeof(value)) |
| 1960 | return -EINVAL; |
| 1961 | |
| 1962 | if (copy_to_user(ov, &value, sizeof(value))) |
| 1963 | return -EFAULT; |
| 1964 | |
| 1965 | return put_user(sizeof(value), ol); |
| 1966 | } |
| 1967 | |
| 1968 | int tipc_ioctl(struct socket *sk, unsigned int cmd, unsigned long arg) |
| 1969 | { |
| 1970 | struct tipc_sioc_ln_req lnr; |
| 1971 | void __user *argp = (void __user *)arg; |
| 1972 | |
| 1973 | switch (cmd) { |
| 1974 | case SIOCGETLINKNAME: |
| 1975 | if (copy_from_user(&lnr, argp, sizeof(lnr))) |
| 1976 | return -EFAULT; |
| 1977 | if (!tipc_node_get_linkname(lnr.bearer_id, lnr.peer, |
| 1978 | lnr.linkname, TIPC_MAX_LINK_NAME)) { |
| 1979 | if (copy_to_user(argp, &lnr, sizeof(lnr))) |
| 1980 | return -EFAULT; |
| 1981 | return 0; |
| 1982 | } |
| 1983 | return -EADDRNOTAVAIL; |
| 1984 | break; |
| 1985 | default: |
| 1986 | return -ENOIOCTLCMD; |
| 1987 | } |
| 1988 | } |
| 1989 | |
| 1990 | /* Protocol switches for the various types of TIPC sockets */ |
| 1991 | |
| 1992 | static const struct proto_ops msg_ops = { |
| 1993 | .owner = THIS_MODULE, |
| 1994 | .family = AF_TIPC, |
| 1995 | .release = tipc_release, |
| 1996 | .bind = tipc_bind, |
| 1997 | .connect = tipc_connect, |
| 1998 | .socketpair = sock_no_socketpair, |
| 1999 | .accept = sock_no_accept, |
| 2000 | .getname = tipc_getname, |
| 2001 | .poll = tipc_poll, |
| 2002 | .ioctl = tipc_ioctl, |
| 2003 | .listen = sock_no_listen, |
| 2004 | .shutdown = tipc_shutdown, |
| 2005 | .setsockopt = tipc_setsockopt, |
| 2006 | .getsockopt = tipc_getsockopt, |
| 2007 | .sendmsg = tipc_sendmsg, |
| 2008 | .recvmsg = tipc_recvmsg, |
| 2009 | .mmap = sock_no_mmap, |
| 2010 | .sendpage = sock_no_sendpage |
| 2011 | }; |
| 2012 | |
| 2013 | static const struct proto_ops packet_ops = { |
| 2014 | .owner = THIS_MODULE, |
| 2015 | .family = AF_TIPC, |
| 2016 | .release = tipc_release, |
| 2017 | .bind = tipc_bind, |
| 2018 | .connect = tipc_connect, |
| 2019 | .socketpair = sock_no_socketpair, |
| 2020 | .accept = tipc_accept, |
| 2021 | .getname = tipc_getname, |
| 2022 | .poll = tipc_poll, |
| 2023 | .ioctl = tipc_ioctl, |
| 2024 | .listen = tipc_listen, |
| 2025 | .shutdown = tipc_shutdown, |
| 2026 | .setsockopt = tipc_setsockopt, |
| 2027 | .getsockopt = tipc_getsockopt, |
| 2028 | .sendmsg = tipc_send_packet, |
| 2029 | .recvmsg = tipc_recvmsg, |
| 2030 | .mmap = sock_no_mmap, |
| 2031 | .sendpage = sock_no_sendpage |
| 2032 | }; |
| 2033 | |
| 2034 | static const struct proto_ops stream_ops = { |
| 2035 | .owner = THIS_MODULE, |
| 2036 | .family = AF_TIPC, |
| 2037 | .release = tipc_release, |
| 2038 | .bind = tipc_bind, |
| 2039 | .connect = tipc_connect, |
| 2040 | .socketpair = sock_no_socketpair, |
| 2041 | .accept = tipc_accept, |
| 2042 | .getname = tipc_getname, |
| 2043 | .poll = tipc_poll, |
| 2044 | .ioctl = tipc_ioctl, |
| 2045 | .listen = tipc_listen, |
| 2046 | .shutdown = tipc_shutdown, |
| 2047 | .setsockopt = tipc_setsockopt, |
| 2048 | .getsockopt = tipc_getsockopt, |
| 2049 | .sendmsg = tipc_send_stream, |
| 2050 | .recvmsg = tipc_recv_stream, |
| 2051 | .mmap = sock_no_mmap, |
| 2052 | .sendpage = sock_no_sendpage |
| 2053 | }; |
| 2054 | |
| 2055 | static const struct net_proto_family tipc_family_ops = { |
| 2056 | .owner = THIS_MODULE, |
| 2057 | .family = AF_TIPC, |
| 2058 | .create = tipc_sk_create |
| 2059 | }; |
| 2060 | |
| 2061 | static struct proto tipc_proto = { |
| 2062 | .name = "TIPC", |
| 2063 | .owner = THIS_MODULE, |
| 2064 | .obj_size = sizeof(struct tipc_sock), |
| 2065 | .sysctl_rmem = sysctl_tipc_rmem |
| 2066 | }; |
| 2067 | |
| 2068 | static struct proto tipc_proto_kern = { |
| 2069 | .name = "TIPC", |
| 2070 | .obj_size = sizeof(struct tipc_sock), |
| 2071 | .sysctl_rmem = sysctl_tipc_rmem |
| 2072 | }; |
| 2073 | |
| 2074 | /** |
| 2075 | * tipc_socket_init - initialize TIPC socket interface |
| 2076 | * |
| 2077 | * Returns 0 on success, errno otherwise |
| 2078 | */ |
| 2079 | int tipc_socket_init(void) |
| 2080 | { |
| 2081 | int res; |
| 2082 | |
| 2083 | res = proto_register(&tipc_proto, 1); |
| 2084 | if (res) { |
| 2085 | pr_err("Failed to register TIPC protocol type\n"); |
| 2086 | goto out; |
| 2087 | } |
| 2088 | |
| 2089 | res = sock_register(&tipc_family_ops); |
| 2090 | if (res) { |
| 2091 | pr_err("Failed to register TIPC socket type\n"); |
| 2092 | proto_unregister(&tipc_proto); |
| 2093 | goto out; |
| 2094 | } |
| 2095 | out: |
| 2096 | return res; |
| 2097 | } |
| 2098 | |
| 2099 | /** |
| 2100 | * tipc_socket_stop - stop TIPC socket interface |
| 2101 | */ |
| 2102 | void tipc_socket_stop(void) |
| 2103 | { |
| 2104 | sock_unregister(tipc_family_ops.family); |
| 2105 | proto_unregister(&tipc_proto); |
| 2106 | } |