| 1 | /* |
| 2 | * net/tipc/socket.c: TIPC socket API |
| 3 | * |
| 4 | * Copyright (c) 2001-2007, 2012 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 | |
| 40 | #include <linux/export.h> |
| 41 | #include <net/sock.h> |
| 42 | |
| 43 | #define SS_LISTENING -1 /* socket is listening */ |
| 44 | #define SS_READY -2 /* socket is connectionless */ |
| 45 | |
| 46 | #define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */ |
| 47 | |
| 48 | struct tipc_sock { |
| 49 | struct sock sk; |
| 50 | struct tipc_port *p; |
| 51 | struct tipc_portid peer_name; |
| 52 | unsigned int conn_timeout; |
| 53 | }; |
| 54 | |
| 55 | #define tipc_sk(sk) ((struct tipc_sock *)(sk)) |
| 56 | #define tipc_sk_port(sk) (tipc_sk(sk)->p) |
| 57 | |
| 58 | #define tipc_rx_ready(sock) (!skb_queue_empty(&sock->sk->sk_receive_queue) || \ |
| 59 | (sock->state == SS_DISCONNECTING)) |
| 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 | static void tipc_data_ready(struct sock *sk, int len); |
| 65 | static void tipc_write_space(struct sock *sk); |
| 66 | static int release(struct socket *sock); |
| 67 | static int accept(struct socket *sock, struct socket *new_sock, int flags); |
| 68 | |
| 69 | static const struct proto_ops packet_ops; |
| 70 | static const struct proto_ops stream_ops; |
| 71 | static const struct proto_ops msg_ops; |
| 72 | |
| 73 | static struct proto tipc_proto; |
| 74 | static struct proto tipc_proto_kern; |
| 75 | |
| 76 | static int sockets_enabled; |
| 77 | |
| 78 | /* |
| 79 | * Revised TIPC socket locking policy: |
| 80 | * |
| 81 | * Most socket operations take the standard socket lock when they start |
| 82 | * and hold it until they finish (or until they need to sleep). Acquiring |
| 83 | * this lock grants the owner exclusive access to the fields of the socket |
| 84 | * data structures, with the exception of the backlog queue. A few socket |
| 85 | * operations can be done without taking the socket lock because they only |
| 86 | * read socket information that never changes during the life of the socket. |
| 87 | * |
| 88 | * Socket operations may acquire the lock for the associated TIPC port if they |
| 89 | * need to perform an operation on the port. If any routine needs to acquire |
| 90 | * both the socket lock and the port lock it must take the socket lock first |
| 91 | * to avoid the risk of deadlock. |
| 92 | * |
| 93 | * The dispatcher handling incoming messages cannot grab the socket lock in |
| 94 | * the standard fashion, since invoked it runs at the BH level and cannot block. |
| 95 | * Instead, it checks to see if the socket lock is currently owned by someone, |
| 96 | * and either handles the message itself or adds it to the socket's backlog |
| 97 | * queue; in the latter case the queued message is processed once the process |
| 98 | * owning the socket lock releases it. |
| 99 | * |
| 100 | * NOTE: Releasing the socket lock while an operation is sleeping overcomes |
| 101 | * the problem of a blocked socket operation preventing any other operations |
| 102 | * from occurring. However, applications must be careful if they have |
| 103 | * multiple threads trying to send (or receive) on the same socket, as these |
| 104 | * operations might interfere with each other. For example, doing a connect |
| 105 | * and a receive at the same time might allow the receive to consume the |
| 106 | * ACK message meant for the connect. While additional work could be done |
| 107 | * to try and overcome this, it doesn't seem to be worthwhile at the present. |
| 108 | * |
| 109 | * NOTE: Releasing the socket lock while an operation is sleeping also ensures |
| 110 | * that another operation that must be performed in a non-blocking manner is |
| 111 | * not delayed for very long because the lock has already been taken. |
| 112 | * |
| 113 | * NOTE: This code assumes that certain fields of a port/socket pair are |
| 114 | * constant over its lifetime; such fields can be examined without taking |
| 115 | * the socket lock and/or port lock, and do not need to be re-read even |
| 116 | * after resuming processing after waiting. These fields include: |
| 117 | * - socket type |
| 118 | * - pointer to socket sk structure (aka tipc_sock structure) |
| 119 | * - pointer to port structure |
| 120 | * - port reference |
| 121 | */ |
| 122 | |
| 123 | /** |
| 124 | * advance_rx_queue - discard first buffer in socket receive queue |
| 125 | * |
| 126 | * Caller must hold socket lock |
| 127 | */ |
| 128 | static void advance_rx_queue(struct sock *sk) |
| 129 | { |
| 130 | kfree_skb(__skb_dequeue(&sk->sk_receive_queue)); |
| 131 | } |
| 132 | |
| 133 | /** |
| 134 | * reject_rx_queue - reject all buffers in socket receive queue |
| 135 | * |
| 136 | * Caller must hold socket lock |
| 137 | */ |
| 138 | static void reject_rx_queue(struct sock *sk) |
| 139 | { |
| 140 | struct sk_buff *buf; |
| 141 | |
| 142 | while ((buf = __skb_dequeue(&sk->sk_receive_queue))) |
| 143 | tipc_reject_msg(buf, TIPC_ERR_NO_PORT); |
| 144 | } |
| 145 | |
| 146 | /** |
| 147 | * tipc_sk_create - create a TIPC socket |
| 148 | * @net: network namespace (must be default network) |
| 149 | * @sock: pre-allocated socket structure |
| 150 | * @protocol: protocol indicator (must be 0) |
| 151 | * @kern: caused by kernel or by userspace? |
| 152 | * |
| 153 | * This routine creates additional data structures used by the TIPC socket, |
| 154 | * initializes them, and links them together. |
| 155 | * |
| 156 | * Returns 0 on success, errno otherwise |
| 157 | */ |
| 158 | static int tipc_sk_create(struct net *net, struct socket *sock, int protocol, |
| 159 | int kern) |
| 160 | { |
| 161 | const struct proto_ops *ops; |
| 162 | socket_state state; |
| 163 | struct sock *sk; |
| 164 | struct tipc_port *tp_ptr; |
| 165 | |
| 166 | /* Validate arguments */ |
| 167 | if (unlikely(protocol != 0)) |
| 168 | return -EPROTONOSUPPORT; |
| 169 | |
| 170 | switch (sock->type) { |
| 171 | case SOCK_STREAM: |
| 172 | ops = &stream_ops; |
| 173 | state = SS_UNCONNECTED; |
| 174 | break; |
| 175 | case SOCK_SEQPACKET: |
| 176 | ops = &packet_ops; |
| 177 | state = SS_UNCONNECTED; |
| 178 | break; |
| 179 | case SOCK_DGRAM: |
| 180 | case SOCK_RDM: |
| 181 | ops = &msg_ops; |
| 182 | state = SS_READY; |
| 183 | break; |
| 184 | default: |
| 185 | return -EPROTOTYPE; |
| 186 | } |
| 187 | |
| 188 | /* Allocate socket's protocol area */ |
| 189 | if (!kern) |
| 190 | sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto); |
| 191 | else |
| 192 | sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto_kern); |
| 193 | |
| 194 | if (sk == NULL) |
| 195 | return -ENOMEM; |
| 196 | |
| 197 | /* Allocate TIPC port for socket to use */ |
| 198 | tp_ptr = tipc_createport_raw(sk, &dispatch, &wakeupdispatch, |
| 199 | TIPC_LOW_IMPORTANCE); |
| 200 | if (unlikely(!tp_ptr)) { |
| 201 | sk_free(sk); |
| 202 | return -ENOMEM; |
| 203 | } |
| 204 | |
| 205 | /* Finish initializing socket data structures */ |
| 206 | sock->ops = ops; |
| 207 | sock->state = state; |
| 208 | |
| 209 | sock_init_data(sock, sk); |
| 210 | sk->sk_backlog_rcv = backlog_rcv; |
| 211 | sk->sk_rcvbuf = sysctl_tipc_rmem[1]; |
| 212 | sk->sk_data_ready = tipc_data_ready; |
| 213 | sk->sk_write_space = tipc_write_space; |
| 214 | tipc_sk(sk)->p = tp_ptr; |
| 215 | tipc_sk(sk)->conn_timeout = CONN_TIMEOUT_DEFAULT; |
| 216 | |
| 217 | spin_unlock_bh(tp_ptr->lock); |
| 218 | |
| 219 | if (sock->state == SS_READY) { |
| 220 | tipc_set_portunreturnable(tp_ptr->ref, 1); |
| 221 | if (sock->type == SOCK_DGRAM) |
| 222 | tipc_set_portunreliable(tp_ptr->ref, 1); |
| 223 | } |
| 224 | |
| 225 | return 0; |
| 226 | } |
| 227 | |
| 228 | /** |
| 229 | * tipc_sock_create_local - create TIPC socket from inside TIPC module |
| 230 | * @type: socket type - SOCK_RDM or SOCK_SEQPACKET |
| 231 | * |
| 232 | * We cannot use sock_creat_kern here because it bumps module user count. |
| 233 | * Since socket owner and creator is the same module we must make sure |
| 234 | * that module count remains zero for module local sockets, otherwise |
| 235 | * we cannot do rmmod. |
| 236 | * |
| 237 | * Returns 0 on success, errno otherwise |
| 238 | */ |
| 239 | int tipc_sock_create_local(int type, struct socket **res) |
| 240 | { |
| 241 | int rc; |
| 242 | struct sock *sk; |
| 243 | |
| 244 | rc = sock_create_lite(AF_TIPC, type, 0, res); |
| 245 | if (rc < 0) { |
| 246 | pr_err("Failed to create kernel socket\n"); |
| 247 | return rc; |
| 248 | } |
| 249 | tipc_sk_create(&init_net, *res, 0, 1); |
| 250 | |
| 251 | sk = (*res)->sk; |
| 252 | |
| 253 | return 0; |
| 254 | } |
| 255 | |
| 256 | /** |
| 257 | * tipc_sock_release_local - release socket created by tipc_sock_create_local |
| 258 | * @sock: the socket to be released. |
| 259 | * |
| 260 | * Module reference count is not incremented when such sockets are created, |
| 261 | * so we must keep it from being decremented when they are released. |
| 262 | */ |
| 263 | void tipc_sock_release_local(struct socket *sock) |
| 264 | { |
| 265 | release(sock); |
| 266 | sock->ops = NULL; |
| 267 | sock_release(sock); |
| 268 | } |
| 269 | |
| 270 | /** |
| 271 | * tipc_sock_accept_local - accept a connection on a socket created |
| 272 | * with tipc_sock_create_local. Use this function to avoid that |
| 273 | * module reference count is inadvertently incremented. |
| 274 | * |
| 275 | * @sock: the accepting socket |
| 276 | * @newsock: reference to the new socket to be created |
| 277 | * @flags: socket flags |
| 278 | */ |
| 279 | |
| 280 | int tipc_sock_accept_local(struct socket *sock, struct socket **newsock, |
| 281 | int flags) |
| 282 | { |
| 283 | struct sock *sk = sock->sk; |
| 284 | int ret; |
| 285 | |
| 286 | ret = sock_create_lite(sk->sk_family, sk->sk_type, |
| 287 | sk->sk_protocol, newsock); |
| 288 | if (ret < 0) |
| 289 | return ret; |
| 290 | |
| 291 | ret = accept(sock, *newsock, flags); |
| 292 | if (ret < 0) { |
| 293 | sock_release(*newsock); |
| 294 | return ret; |
| 295 | } |
| 296 | (*newsock)->ops = sock->ops; |
| 297 | return ret; |
| 298 | } |
| 299 | |
| 300 | /** |
| 301 | * release - destroy a TIPC socket |
| 302 | * @sock: socket to destroy |
| 303 | * |
| 304 | * This routine cleans up any messages that are still queued on the socket. |
| 305 | * For DGRAM and RDM socket types, all queued messages are rejected. |
| 306 | * For SEQPACKET and STREAM socket types, the first message is rejected |
| 307 | * and any others are discarded. (If the first message on a STREAM socket |
| 308 | * is partially-read, it is discarded and the next one is rejected instead.) |
| 309 | * |
| 310 | * NOTE: Rejected messages are not necessarily returned to the sender! They |
| 311 | * are returned or discarded according to the "destination droppable" setting |
| 312 | * specified for the message by the sender. |
| 313 | * |
| 314 | * Returns 0 on success, errno otherwise |
| 315 | */ |
| 316 | static int release(struct socket *sock) |
| 317 | { |
| 318 | struct sock *sk = sock->sk; |
| 319 | struct tipc_port *tport; |
| 320 | struct sk_buff *buf; |
| 321 | int res; |
| 322 | |
| 323 | /* |
| 324 | * Exit if socket isn't fully initialized (occurs when a failed accept() |
| 325 | * releases a pre-allocated child socket that was never used) |
| 326 | */ |
| 327 | if (sk == NULL) |
| 328 | return 0; |
| 329 | |
| 330 | tport = tipc_sk_port(sk); |
| 331 | lock_sock(sk); |
| 332 | |
| 333 | /* |
| 334 | * Reject all unreceived messages, except on an active connection |
| 335 | * (which disconnects locally & sends a 'FIN+' to peer) |
| 336 | */ |
| 337 | while (sock->state != SS_DISCONNECTING) { |
| 338 | buf = __skb_dequeue(&sk->sk_receive_queue); |
| 339 | if (buf == NULL) |
| 340 | break; |
| 341 | if (TIPC_SKB_CB(buf)->handle != 0) |
| 342 | kfree_skb(buf); |
| 343 | else { |
| 344 | if ((sock->state == SS_CONNECTING) || |
| 345 | (sock->state == SS_CONNECTED)) { |
| 346 | sock->state = SS_DISCONNECTING; |
| 347 | tipc_disconnect(tport->ref); |
| 348 | } |
| 349 | tipc_reject_msg(buf, TIPC_ERR_NO_PORT); |
| 350 | } |
| 351 | } |
| 352 | |
| 353 | /* |
| 354 | * Delete TIPC port; this ensures no more messages are queued |
| 355 | * (also disconnects an active connection & sends a 'FIN-' to peer) |
| 356 | */ |
| 357 | res = tipc_deleteport(tport->ref); |
| 358 | |
| 359 | /* Discard any remaining (connection-based) messages in receive queue */ |
| 360 | __skb_queue_purge(&sk->sk_receive_queue); |
| 361 | |
| 362 | /* Reject any messages that accumulated in backlog queue */ |
| 363 | sock->state = SS_DISCONNECTING; |
| 364 | release_sock(sk); |
| 365 | |
| 366 | sock_put(sk); |
| 367 | sock->sk = NULL; |
| 368 | |
| 369 | return res; |
| 370 | } |
| 371 | |
| 372 | /** |
| 373 | * bind - associate or disassocate TIPC name(s) with a socket |
| 374 | * @sock: socket structure |
| 375 | * @uaddr: socket address describing name(s) and desired operation |
| 376 | * @uaddr_len: size of socket address data structure |
| 377 | * |
| 378 | * Name and name sequence binding is indicated using a positive scope value; |
| 379 | * a negative scope value unbinds the specified name. Specifying no name |
| 380 | * (i.e. a socket address length of 0) unbinds all names from the socket. |
| 381 | * |
| 382 | * Returns 0 on success, errno otherwise |
| 383 | * |
| 384 | * NOTE: This routine doesn't need to take the socket lock since it doesn't |
| 385 | * access any non-constant socket information. |
| 386 | */ |
| 387 | static int bind(struct socket *sock, struct sockaddr *uaddr, int uaddr_len) |
| 388 | { |
| 389 | struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr; |
| 390 | u32 portref = tipc_sk_port(sock->sk)->ref; |
| 391 | |
| 392 | if (unlikely(!uaddr_len)) |
| 393 | return tipc_withdraw(portref, 0, NULL); |
| 394 | |
| 395 | if (uaddr_len < sizeof(struct sockaddr_tipc)) |
| 396 | return -EINVAL; |
| 397 | if (addr->family != AF_TIPC) |
| 398 | return -EAFNOSUPPORT; |
| 399 | |
| 400 | if (addr->addrtype == TIPC_ADDR_NAME) |
| 401 | addr->addr.nameseq.upper = addr->addr.nameseq.lower; |
| 402 | else if (addr->addrtype != TIPC_ADDR_NAMESEQ) |
| 403 | return -EAFNOSUPPORT; |
| 404 | |
| 405 | if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) && |
| 406 | (addr->addr.nameseq.type != TIPC_TOP_SRV)) |
| 407 | return -EACCES; |
| 408 | |
| 409 | return (addr->scope > 0) ? |
| 410 | tipc_publish(portref, addr->scope, &addr->addr.nameseq) : |
| 411 | tipc_withdraw(portref, -addr->scope, &addr->addr.nameseq); |
| 412 | } |
| 413 | |
| 414 | /** |
| 415 | * get_name - get port ID of socket or peer socket |
| 416 | * @sock: socket structure |
| 417 | * @uaddr: area for returned socket address |
| 418 | * @uaddr_len: area for returned length of socket address |
| 419 | * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID |
| 420 | * |
| 421 | * Returns 0 on success, errno otherwise |
| 422 | * |
| 423 | * NOTE: This routine doesn't need to take the socket lock since it only |
| 424 | * accesses socket information that is unchanging (or which changes in |
| 425 | * a completely predictable manner). |
| 426 | */ |
| 427 | static int get_name(struct socket *sock, struct sockaddr *uaddr, |
| 428 | int *uaddr_len, int peer) |
| 429 | { |
| 430 | struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr; |
| 431 | struct tipc_sock *tsock = tipc_sk(sock->sk); |
| 432 | |
| 433 | memset(addr, 0, sizeof(*addr)); |
| 434 | if (peer) { |
| 435 | if ((sock->state != SS_CONNECTED) && |
| 436 | ((peer != 2) || (sock->state != SS_DISCONNECTING))) |
| 437 | return -ENOTCONN; |
| 438 | addr->addr.id.ref = tsock->peer_name.ref; |
| 439 | addr->addr.id.node = tsock->peer_name.node; |
| 440 | } else { |
| 441 | addr->addr.id.ref = tsock->p->ref; |
| 442 | addr->addr.id.node = tipc_own_addr; |
| 443 | } |
| 444 | |
| 445 | *uaddr_len = sizeof(*addr); |
| 446 | addr->addrtype = TIPC_ADDR_ID; |
| 447 | addr->family = AF_TIPC; |
| 448 | addr->scope = 0; |
| 449 | addr->addr.name.domain = 0; |
| 450 | |
| 451 | return 0; |
| 452 | } |
| 453 | |
| 454 | /** |
| 455 | * poll - read and possibly block on pollmask |
| 456 | * @file: file structure associated with the socket |
| 457 | * @sock: socket for which to calculate the poll bits |
| 458 | * @wait: ??? |
| 459 | * |
| 460 | * Returns pollmask value |
| 461 | * |
| 462 | * COMMENTARY: |
| 463 | * It appears that the usual socket locking mechanisms are not useful here |
| 464 | * since the pollmask info is potentially out-of-date the moment this routine |
| 465 | * exits. TCP and other protocols seem to rely on higher level poll routines |
| 466 | * to handle any preventable race conditions, so TIPC will do the same ... |
| 467 | * |
| 468 | * TIPC sets the returned events as follows: |
| 469 | * |
| 470 | * socket state flags set |
| 471 | * ------------ --------- |
| 472 | * unconnected no read flags |
| 473 | * POLLOUT if port is not congested |
| 474 | * |
| 475 | * connecting POLLIN/POLLRDNORM if ACK/NACK in rx queue |
| 476 | * no write flags |
| 477 | * |
| 478 | * connected POLLIN/POLLRDNORM if data in rx queue |
| 479 | * POLLOUT if port is not congested |
| 480 | * |
| 481 | * disconnecting POLLIN/POLLRDNORM/POLLHUP |
| 482 | * no write flags |
| 483 | * |
| 484 | * listening POLLIN if SYN in rx queue |
| 485 | * no write flags |
| 486 | * |
| 487 | * ready POLLIN/POLLRDNORM if data in rx queue |
| 488 | * [connectionless] POLLOUT (since port cannot be congested) |
| 489 | * |
| 490 | * IMPORTANT: The fact that a read or write operation is indicated does NOT |
| 491 | * imply that the operation will succeed, merely that it should be performed |
| 492 | * and will not block. |
| 493 | */ |
| 494 | static unsigned int poll(struct file *file, struct socket *sock, |
| 495 | poll_table *wait) |
| 496 | { |
| 497 | struct sock *sk = sock->sk; |
| 498 | u32 mask = 0; |
| 499 | |
| 500 | sock_poll_wait(file, sk_sleep(sk), wait); |
| 501 | |
| 502 | switch ((int)sock->state) { |
| 503 | case SS_UNCONNECTED: |
| 504 | if (!tipc_sk_port(sk)->congested) |
| 505 | mask |= POLLOUT; |
| 506 | break; |
| 507 | case SS_READY: |
| 508 | case SS_CONNECTED: |
| 509 | if (!tipc_sk_port(sk)->congested) |
| 510 | mask |= POLLOUT; |
| 511 | /* fall thru' */ |
| 512 | case SS_CONNECTING: |
| 513 | case SS_LISTENING: |
| 514 | if (!skb_queue_empty(&sk->sk_receive_queue)) |
| 515 | mask |= (POLLIN | POLLRDNORM); |
| 516 | break; |
| 517 | case SS_DISCONNECTING: |
| 518 | mask = (POLLIN | POLLRDNORM | POLLHUP); |
| 519 | break; |
| 520 | } |
| 521 | |
| 522 | return mask; |
| 523 | } |
| 524 | |
| 525 | /** |
| 526 | * dest_name_check - verify user is permitted to send to specified port name |
| 527 | * @dest: destination address |
| 528 | * @m: descriptor for message to be sent |
| 529 | * |
| 530 | * Prevents restricted configuration commands from being issued by |
| 531 | * unauthorized users. |
| 532 | * |
| 533 | * Returns 0 if permission is granted, otherwise errno |
| 534 | */ |
| 535 | static int dest_name_check(struct sockaddr_tipc *dest, struct msghdr *m) |
| 536 | { |
| 537 | struct tipc_cfg_msg_hdr hdr; |
| 538 | |
| 539 | if (likely(dest->addr.name.name.type >= TIPC_RESERVED_TYPES)) |
| 540 | return 0; |
| 541 | if (likely(dest->addr.name.name.type == TIPC_TOP_SRV)) |
| 542 | return 0; |
| 543 | if (likely(dest->addr.name.name.type != TIPC_CFG_SRV)) |
| 544 | return -EACCES; |
| 545 | |
| 546 | if (!m->msg_iovlen || (m->msg_iov[0].iov_len < sizeof(hdr))) |
| 547 | return -EMSGSIZE; |
| 548 | if (copy_from_user(&hdr, m->msg_iov[0].iov_base, sizeof(hdr))) |
| 549 | return -EFAULT; |
| 550 | if ((ntohs(hdr.tcm_type) & 0xC000) && (!capable(CAP_NET_ADMIN))) |
| 551 | return -EACCES; |
| 552 | |
| 553 | return 0; |
| 554 | } |
| 555 | |
| 556 | /** |
| 557 | * send_msg - send message in connectionless manner |
| 558 | * @iocb: if NULL, indicates that socket lock is already held |
| 559 | * @sock: socket structure |
| 560 | * @m: message to send |
| 561 | * @total_len: length of message |
| 562 | * |
| 563 | * Message must have an destination specified explicitly. |
| 564 | * Used for SOCK_RDM and SOCK_DGRAM messages, |
| 565 | * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections. |
| 566 | * (Note: 'SYN+' is prohibited on SOCK_STREAM.) |
| 567 | * |
| 568 | * Returns the number of bytes sent on success, or errno otherwise |
| 569 | */ |
| 570 | static int send_msg(struct kiocb *iocb, struct socket *sock, |
| 571 | struct msghdr *m, size_t total_len) |
| 572 | { |
| 573 | struct sock *sk = sock->sk; |
| 574 | struct tipc_port *tport = tipc_sk_port(sk); |
| 575 | struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name; |
| 576 | int needs_conn; |
| 577 | long timeout_val; |
| 578 | int res = -EINVAL; |
| 579 | |
| 580 | if (unlikely(!dest)) |
| 581 | return -EDESTADDRREQ; |
| 582 | if (unlikely((m->msg_namelen < sizeof(*dest)) || |
| 583 | (dest->family != AF_TIPC))) |
| 584 | return -EINVAL; |
| 585 | if (total_len > TIPC_MAX_USER_MSG_SIZE) |
| 586 | return -EMSGSIZE; |
| 587 | |
| 588 | if (iocb) |
| 589 | lock_sock(sk); |
| 590 | |
| 591 | needs_conn = (sock->state != SS_READY); |
| 592 | if (unlikely(needs_conn)) { |
| 593 | if (sock->state == SS_LISTENING) { |
| 594 | res = -EPIPE; |
| 595 | goto exit; |
| 596 | } |
| 597 | if (sock->state != SS_UNCONNECTED) { |
| 598 | res = -EISCONN; |
| 599 | goto exit; |
| 600 | } |
| 601 | if (tport->published) { |
| 602 | res = -EOPNOTSUPP; |
| 603 | goto exit; |
| 604 | } |
| 605 | if (dest->addrtype == TIPC_ADDR_NAME) { |
| 606 | tport->conn_type = dest->addr.name.name.type; |
| 607 | tport->conn_instance = dest->addr.name.name.instance; |
| 608 | } |
| 609 | |
| 610 | /* Abort any pending connection attempts (very unlikely) */ |
| 611 | reject_rx_queue(sk); |
| 612 | } |
| 613 | |
| 614 | timeout_val = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT); |
| 615 | |
| 616 | do { |
| 617 | if (dest->addrtype == TIPC_ADDR_NAME) { |
| 618 | res = dest_name_check(dest, m); |
| 619 | if (res) |
| 620 | break; |
| 621 | res = tipc_send2name(tport->ref, |
| 622 | &dest->addr.name.name, |
| 623 | dest->addr.name.domain, |
| 624 | m->msg_iovlen, |
| 625 | m->msg_iov, |
| 626 | total_len); |
| 627 | } else if (dest->addrtype == TIPC_ADDR_ID) { |
| 628 | res = tipc_send2port(tport->ref, |
| 629 | &dest->addr.id, |
| 630 | m->msg_iovlen, |
| 631 | m->msg_iov, |
| 632 | total_len); |
| 633 | } else if (dest->addrtype == TIPC_ADDR_MCAST) { |
| 634 | if (needs_conn) { |
| 635 | res = -EOPNOTSUPP; |
| 636 | break; |
| 637 | } |
| 638 | res = dest_name_check(dest, m); |
| 639 | if (res) |
| 640 | break; |
| 641 | res = tipc_multicast(tport->ref, |
| 642 | &dest->addr.nameseq, |
| 643 | m->msg_iovlen, |
| 644 | m->msg_iov, |
| 645 | total_len); |
| 646 | } |
| 647 | if (likely(res != -ELINKCONG)) { |
| 648 | if (needs_conn && (res >= 0)) |
| 649 | sock->state = SS_CONNECTING; |
| 650 | break; |
| 651 | } |
| 652 | if (timeout_val <= 0L) { |
| 653 | res = timeout_val ? timeout_val : -EWOULDBLOCK; |
| 654 | break; |
| 655 | } |
| 656 | release_sock(sk); |
| 657 | timeout_val = wait_event_interruptible_timeout(*sk_sleep(sk), |
| 658 | !tport->congested, timeout_val); |
| 659 | lock_sock(sk); |
| 660 | } while (1); |
| 661 | |
| 662 | exit: |
| 663 | if (iocb) |
| 664 | release_sock(sk); |
| 665 | return res; |
| 666 | } |
| 667 | |
| 668 | /** |
| 669 | * send_packet - send a connection-oriented message |
| 670 | * @iocb: if NULL, indicates that socket lock is already held |
| 671 | * @sock: socket structure |
| 672 | * @m: message to send |
| 673 | * @total_len: length of message |
| 674 | * |
| 675 | * Used for SOCK_SEQPACKET messages and SOCK_STREAM data. |
| 676 | * |
| 677 | * Returns the number of bytes sent on success, or errno otherwise |
| 678 | */ |
| 679 | static int send_packet(struct kiocb *iocb, struct socket *sock, |
| 680 | struct msghdr *m, size_t total_len) |
| 681 | { |
| 682 | struct sock *sk = sock->sk; |
| 683 | struct tipc_port *tport = tipc_sk_port(sk); |
| 684 | struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name; |
| 685 | long timeout_val; |
| 686 | int res; |
| 687 | |
| 688 | /* Handle implied connection establishment */ |
| 689 | if (unlikely(dest)) |
| 690 | return send_msg(iocb, sock, m, total_len); |
| 691 | |
| 692 | if (total_len > TIPC_MAX_USER_MSG_SIZE) |
| 693 | return -EMSGSIZE; |
| 694 | |
| 695 | if (iocb) |
| 696 | lock_sock(sk); |
| 697 | |
| 698 | timeout_val = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT); |
| 699 | |
| 700 | do { |
| 701 | if (unlikely(sock->state != SS_CONNECTED)) { |
| 702 | if (sock->state == SS_DISCONNECTING) |
| 703 | res = -EPIPE; |
| 704 | else |
| 705 | res = -ENOTCONN; |
| 706 | break; |
| 707 | } |
| 708 | |
| 709 | res = tipc_send(tport->ref, m->msg_iovlen, m->msg_iov, |
| 710 | total_len); |
| 711 | if (likely(res != -ELINKCONG)) |
| 712 | break; |
| 713 | if (timeout_val <= 0L) { |
| 714 | res = timeout_val ? timeout_val : -EWOULDBLOCK; |
| 715 | break; |
| 716 | } |
| 717 | release_sock(sk); |
| 718 | timeout_val = wait_event_interruptible_timeout(*sk_sleep(sk), |
| 719 | (!tport->congested || !tport->connected), timeout_val); |
| 720 | lock_sock(sk); |
| 721 | } while (1); |
| 722 | |
| 723 | if (iocb) |
| 724 | release_sock(sk); |
| 725 | return res; |
| 726 | } |
| 727 | |
| 728 | /** |
| 729 | * send_stream - send stream-oriented data |
| 730 | * @iocb: (unused) |
| 731 | * @sock: socket structure |
| 732 | * @m: data to send |
| 733 | * @total_len: total length of data to be sent |
| 734 | * |
| 735 | * Used for SOCK_STREAM data. |
| 736 | * |
| 737 | * Returns the number of bytes sent on success (or partial success), |
| 738 | * or errno if no data sent |
| 739 | */ |
| 740 | static int send_stream(struct kiocb *iocb, struct socket *sock, |
| 741 | struct msghdr *m, size_t total_len) |
| 742 | { |
| 743 | struct sock *sk = sock->sk; |
| 744 | struct tipc_port *tport = tipc_sk_port(sk); |
| 745 | struct msghdr my_msg; |
| 746 | struct iovec my_iov; |
| 747 | struct iovec *curr_iov; |
| 748 | int curr_iovlen; |
| 749 | char __user *curr_start; |
| 750 | u32 hdr_size; |
| 751 | int curr_left; |
| 752 | int bytes_to_send; |
| 753 | int bytes_sent; |
| 754 | int res; |
| 755 | |
| 756 | lock_sock(sk); |
| 757 | |
| 758 | /* Handle special cases where there is no connection */ |
| 759 | if (unlikely(sock->state != SS_CONNECTED)) { |
| 760 | if (sock->state == SS_UNCONNECTED) { |
| 761 | res = send_packet(NULL, sock, m, total_len); |
| 762 | goto exit; |
| 763 | } else if (sock->state == SS_DISCONNECTING) { |
| 764 | res = -EPIPE; |
| 765 | goto exit; |
| 766 | } else { |
| 767 | res = -ENOTCONN; |
| 768 | goto exit; |
| 769 | } |
| 770 | } |
| 771 | |
| 772 | if (unlikely(m->msg_name)) { |
| 773 | res = -EISCONN; |
| 774 | goto exit; |
| 775 | } |
| 776 | |
| 777 | if (total_len > (unsigned int)INT_MAX) { |
| 778 | res = -EMSGSIZE; |
| 779 | goto exit; |
| 780 | } |
| 781 | |
| 782 | /* |
| 783 | * Send each iovec entry using one or more messages |
| 784 | * |
| 785 | * Note: This algorithm is good for the most likely case |
| 786 | * (i.e. one large iovec entry), but could be improved to pass sets |
| 787 | * of small iovec entries into send_packet(). |
| 788 | */ |
| 789 | curr_iov = m->msg_iov; |
| 790 | curr_iovlen = m->msg_iovlen; |
| 791 | my_msg.msg_iov = &my_iov; |
| 792 | my_msg.msg_iovlen = 1; |
| 793 | my_msg.msg_flags = m->msg_flags; |
| 794 | my_msg.msg_name = NULL; |
| 795 | bytes_sent = 0; |
| 796 | |
| 797 | hdr_size = msg_hdr_sz(&tport->phdr); |
| 798 | |
| 799 | while (curr_iovlen--) { |
| 800 | curr_start = curr_iov->iov_base; |
| 801 | curr_left = curr_iov->iov_len; |
| 802 | |
| 803 | while (curr_left) { |
| 804 | bytes_to_send = tport->max_pkt - hdr_size; |
| 805 | if (bytes_to_send > TIPC_MAX_USER_MSG_SIZE) |
| 806 | bytes_to_send = TIPC_MAX_USER_MSG_SIZE; |
| 807 | if (curr_left < bytes_to_send) |
| 808 | bytes_to_send = curr_left; |
| 809 | my_iov.iov_base = curr_start; |
| 810 | my_iov.iov_len = bytes_to_send; |
| 811 | res = send_packet(NULL, sock, &my_msg, bytes_to_send); |
| 812 | if (res < 0) { |
| 813 | if (bytes_sent) |
| 814 | res = bytes_sent; |
| 815 | goto exit; |
| 816 | } |
| 817 | curr_left -= bytes_to_send; |
| 818 | curr_start += bytes_to_send; |
| 819 | bytes_sent += bytes_to_send; |
| 820 | } |
| 821 | |
| 822 | curr_iov++; |
| 823 | } |
| 824 | res = bytes_sent; |
| 825 | exit: |
| 826 | release_sock(sk); |
| 827 | return res; |
| 828 | } |
| 829 | |
| 830 | /** |
| 831 | * auto_connect - complete connection setup to a remote port |
| 832 | * @sock: socket structure |
| 833 | * @msg: peer's response message |
| 834 | * |
| 835 | * Returns 0 on success, errno otherwise |
| 836 | */ |
| 837 | static int auto_connect(struct socket *sock, struct tipc_msg *msg) |
| 838 | { |
| 839 | struct tipc_sock *tsock = tipc_sk(sock->sk); |
| 840 | struct tipc_port *p_ptr; |
| 841 | |
| 842 | tsock->peer_name.ref = msg_origport(msg); |
| 843 | tsock->peer_name.node = msg_orignode(msg); |
| 844 | p_ptr = tipc_port_deref(tsock->p->ref); |
| 845 | if (!p_ptr) |
| 846 | return -EINVAL; |
| 847 | |
| 848 | __tipc_connect(tsock->p->ref, p_ptr, &tsock->peer_name); |
| 849 | |
| 850 | if (msg_importance(msg) > TIPC_CRITICAL_IMPORTANCE) |
| 851 | return -EINVAL; |
| 852 | msg_set_importance(&p_ptr->phdr, (u32)msg_importance(msg)); |
| 853 | sock->state = SS_CONNECTED; |
| 854 | return 0; |
| 855 | } |
| 856 | |
| 857 | /** |
| 858 | * set_orig_addr - capture sender's address for received message |
| 859 | * @m: descriptor for message info |
| 860 | * @msg: received message header |
| 861 | * |
| 862 | * Note: Address is not captured if not requested by receiver. |
| 863 | */ |
| 864 | static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg) |
| 865 | { |
| 866 | struct sockaddr_tipc *addr = (struct sockaddr_tipc *)m->msg_name; |
| 867 | |
| 868 | if (addr) { |
| 869 | addr->family = AF_TIPC; |
| 870 | addr->addrtype = TIPC_ADDR_ID; |
| 871 | memset(&addr->addr, 0, sizeof(addr->addr)); |
| 872 | addr->addr.id.ref = msg_origport(msg); |
| 873 | addr->addr.id.node = msg_orignode(msg); |
| 874 | addr->addr.name.domain = 0; /* could leave uninitialized */ |
| 875 | addr->scope = 0; /* could leave uninitialized */ |
| 876 | m->msg_namelen = sizeof(struct sockaddr_tipc); |
| 877 | } |
| 878 | } |
| 879 | |
| 880 | /** |
| 881 | * anc_data_recv - optionally capture ancillary data for received message |
| 882 | * @m: descriptor for message info |
| 883 | * @msg: received message header |
| 884 | * @tport: TIPC port associated with message |
| 885 | * |
| 886 | * Note: Ancillary data is not captured if not requested by receiver. |
| 887 | * |
| 888 | * Returns 0 if successful, otherwise errno |
| 889 | */ |
| 890 | static int anc_data_recv(struct msghdr *m, struct tipc_msg *msg, |
| 891 | struct tipc_port *tport) |
| 892 | { |
| 893 | u32 anc_data[3]; |
| 894 | u32 err; |
| 895 | u32 dest_type; |
| 896 | int has_name; |
| 897 | int res; |
| 898 | |
| 899 | if (likely(m->msg_controllen == 0)) |
| 900 | return 0; |
| 901 | |
| 902 | /* Optionally capture errored message object(s) */ |
| 903 | err = msg ? msg_errcode(msg) : 0; |
| 904 | if (unlikely(err)) { |
| 905 | anc_data[0] = err; |
| 906 | anc_data[1] = msg_data_sz(msg); |
| 907 | res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data); |
| 908 | if (res) |
| 909 | return res; |
| 910 | if (anc_data[1]) { |
| 911 | res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1], |
| 912 | msg_data(msg)); |
| 913 | if (res) |
| 914 | return res; |
| 915 | } |
| 916 | } |
| 917 | |
| 918 | /* Optionally capture message destination object */ |
| 919 | dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG; |
| 920 | switch (dest_type) { |
| 921 | case TIPC_NAMED_MSG: |
| 922 | has_name = 1; |
| 923 | anc_data[0] = msg_nametype(msg); |
| 924 | anc_data[1] = msg_namelower(msg); |
| 925 | anc_data[2] = msg_namelower(msg); |
| 926 | break; |
| 927 | case TIPC_MCAST_MSG: |
| 928 | has_name = 1; |
| 929 | anc_data[0] = msg_nametype(msg); |
| 930 | anc_data[1] = msg_namelower(msg); |
| 931 | anc_data[2] = msg_nameupper(msg); |
| 932 | break; |
| 933 | case TIPC_CONN_MSG: |
| 934 | has_name = (tport->conn_type != 0); |
| 935 | anc_data[0] = tport->conn_type; |
| 936 | anc_data[1] = tport->conn_instance; |
| 937 | anc_data[2] = tport->conn_instance; |
| 938 | break; |
| 939 | default: |
| 940 | has_name = 0; |
| 941 | } |
| 942 | if (has_name) { |
| 943 | res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data); |
| 944 | if (res) |
| 945 | return res; |
| 946 | } |
| 947 | |
| 948 | return 0; |
| 949 | } |
| 950 | |
| 951 | /** |
| 952 | * recv_msg - receive packet-oriented message |
| 953 | * @iocb: (unused) |
| 954 | * @m: descriptor for message info |
| 955 | * @buf_len: total size of user buffer area |
| 956 | * @flags: receive flags |
| 957 | * |
| 958 | * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages. |
| 959 | * If the complete message doesn't fit in user area, truncate it. |
| 960 | * |
| 961 | * Returns size of returned message data, errno otherwise |
| 962 | */ |
| 963 | static int recv_msg(struct kiocb *iocb, struct socket *sock, |
| 964 | struct msghdr *m, size_t buf_len, int flags) |
| 965 | { |
| 966 | struct sock *sk = sock->sk; |
| 967 | struct tipc_port *tport = tipc_sk_port(sk); |
| 968 | struct sk_buff *buf; |
| 969 | struct tipc_msg *msg; |
| 970 | long timeout; |
| 971 | unsigned int sz; |
| 972 | u32 err; |
| 973 | int res; |
| 974 | |
| 975 | /* Catch invalid receive requests */ |
| 976 | if (unlikely(!buf_len)) |
| 977 | return -EINVAL; |
| 978 | |
| 979 | lock_sock(sk); |
| 980 | |
| 981 | if (unlikely(sock->state == SS_UNCONNECTED)) { |
| 982 | res = -ENOTCONN; |
| 983 | goto exit; |
| 984 | } |
| 985 | |
| 986 | /* will be updated in set_orig_addr() if needed */ |
| 987 | m->msg_namelen = 0; |
| 988 | |
| 989 | timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT); |
| 990 | restart: |
| 991 | |
| 992 | /* Look for a message in receive queue; wait if necessary */ |
| 993 | while (skb_queue_empty(&sk->sk_receive_queue)) { |
| 994 | if (sock->state == SS_DISCONNECTING) { |
| 995 | res = -ENOTCONN; |
| 996 | goto exit; |
| 997 | } |
| 998 | if (timeout <= 0L) { |
| 999 | res = timeout ? timeout : -EWOULDBLOCK; |
| 1000 | goto exit; |
| 1001 | } |
| 1002 | release_sock(sk); |
| 1003 | timeout = wait_event_interruptible_timeout(*sk_sleep(sk), |
| 1004 | tipc_rx_ready(sock), |
| 1005 | timeout); |
| 1006 | lock_sock(sk); |
| 1007 | } |
| 1008 | |
| 1009 | /* Look at first message in receive queue */ |
| 1010 | buf = skb_peek(&sk->sk_receive_queue); |
| 1011 | msg = buf_msg(buf); |
| 1012 | sz = msg_data_sz(msg); |
| 1013 | err = msg_errcode(msg); |
| 1014 | |
| 1015 | /* Discard an empty non-errored message & try again */ |
| 1016 | if ((!sz) && (!err)) { |
| 1017 | advance_rx_queue(sk); |
| 1018 | goto restart; |
| 1019 | } |
| 1020 | |
| 1021 | /* Capture sender's address (optional) */ |
| 1022 | set_orig_addr(m, msg); |
| 1023 | |
| 1024 | /* Capture ancillary data (optional) */ |
| 1025 | res = anc_data_recv(m, msg, tport); |
| 1026 | if (res) |
| 1027 | goto exit; |
| 1028 | |
| 1029 | /* Capture message data (if valid) & compute return value (always) */ |
| 1030 | if (!err) { |
| 1031 | if (unlikely(buf_len < sz)) { |
| 1032 | sz = buf_len; |
| 1033 | m->msg_flags |= MSG_TRUNC; |
| 1034 | } |
| 1035 | res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg), |
| 1036 | m->msg_iov, sz); |
| 1037 | if (res) |
| 1038 | goto exit; |
| 1039 | res = sz; |
| 1040 | } else { |
| 1041 | if ((sock->state == SS_READY) || |
| 1042 | ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)) |
| 1043 | res = 0; |
| 1044 | else |
| 1045 | res = -ECONNRESET; |
| 1046 | } |
| 1047 | |
| 1048 | /* Consume received message (optional) */ |
| 1049 | if (likely(!(flags & MSG_PEEK))) { |
| 1050 | if ((sock->state != SS_READY) && |
| 1051 | (++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN)) |
| 1052 | tipc_acknowledge(tport->ref, tport->conn_unacked); |
| 1053 | advance_rx_queue(sk); |
| 1054 | } |
| 1055 | exit: |
| 1056 | release_sock(sk); |
| 1057 | return res; |
| 1058 | } |
| 1059 | |
| 1060 | /** |
| 1061 | * recv_stream - receive stream-oriented data |
| 1062 | * @iocb: (unused) |
| 1063 | * @m: descriptor for message info |
| 1064 | * @buf_len: total size of user buffer area |
| 1065 | * @flags: receive flags |
| 1066 | * |
| 1067 | * Used for SOCK_STREAM messages only. If not enough data is available |
| 1068 | * will optionally wait for more; never truncates data. |
| 1069 | * |
| 1070 | * Returns size of returned message data, errno otherwise |
| 1071 | */ |
| 1072 | static int recv_stream(struct kiocb *iocb, struct socket *sock, |
| 1073 | struct msghdr *m, size_t buf_len, int flags) |
| 1074 | { |
| 1075 | struct sock *sk = sock->sk; |
| 1076 | struct tipc_port *tport = tipc_sk_port(sk); |
| 1077 | struct sk_buff *buf; |
| 1078 | struct tipc_msg *msg; |
| 1079 | long timeout; |
| 1080 | unsigned int sz; |
| 1081 | int sz_to_copy, target, needed; |
| 1082 | int sz_copied = 0; |
| 1083 | u32 err; |
| 1084 | int res = 0; |
| 1085 | |
| 1086 | /* Catch invalid receive attempts */ |
| 1087 | if (unlikely(!buf_len)) |
| 1088 | return -EINVAL; |
| 1089 | |
| 1090 | lock_sock(sk); |
| 1091 | |
| 1092 | if (unlikely((sock->state == SS_UNCONNECTED))) { |
| 1093 | res = -ENOTCONN; |
| 1094 | goto exit; |
| 1095 | } |
| 1096 | |
| 1097 | /* will be updated in set_orig_addr() if needed */ |
| 1098 | m->msg_namelen = 0; |
| 1099 | |
| 1100 | target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len); |
| 1101 | timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT); |
| 1102 | |
| 1103 | restart: |
| 1104 | /* Look for a message in receive queue; wait if necessary */ |
| 1105 | while (skb_queue_empty(&sk->sk_receive_queue)) { |
| 1106 | if (sock->state == SS_DISCONNECTING) { |
| 1107 | res = -ENOTCONN; |
| 1108 | goto exit; |
| 1109 | } |
| 1110 | if (timeout <= 0L) { |
| 1111 | res = timeout ? timeout : -EWOULDBLOCK; |
| 1112 | goto exit; |
| 1113 | } |
| 1114 | release_sock(sk); |
| 1115 | timeout = wait_event_interruptible_timeout(*sk_sleep(sk), |
| 1116 | tipc_rx_ready(sock), |
| 1117 | timeout); |
| 1118 | lock_sock(sk); |
| 1119 | } |
| 1120 | |
| 1121 | /* Look at first message in receive queue */ |
| 1122 | buf = skb_peek(&sk->sk_receive_queue); |
| 1123 | msg = buf_msg(buf); |
| 1124 | sz = msg_data_sz(msg); |
| 1125 | err = msg_errcode(msg); |
| 1126 | |
| 1127 | /* Discard an empty non-errored message & try again */ |
| 1128 | if ((!sz) && (!err)) { |
| 1129 | advance_rx_queue(sk); |
| 1130 | goto restart; |
| 1131 | } |
| 1132 | |
| 1133 | /* Optionally capture sender's address & ancillary data of first msg */ |
| 1134 | if (sz_copied == 0) { |
| 1135 | set_orig_addr(m, msg); |
| 1136 | res = anc_data_recv(m, msg, tport); |
| 1137 | if (res) |
| 1138 | goto exit; |
| 1139 | } |
| 1140 | |
| 1141 | /* Capture message data (if valid) & compute return value (always) */ |
| 1142 | if (!err) { |
| 1143 | u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle); |
| 1144 | |
| 1145 | sz -= offset; |
| 1146 | needed = (buf_len - sz_copied); |
| 1147 | sz_to_copy = (sz <= needed) ? sz : needed; |
| 1148 | |
| 1149 | res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg) + offset, |
| 1150 | m->msg_iov, sz_to_copy); |
| 1151 | if (res) |
| 1152 | goto exit; |
| 1153 | |
| 1154 | sz_copied += sz_to_copy; |
| 1155 | |
| 1156 | if (sz_to_copy < sz) { |
| 1157 | if (!(flags & MSG_PEEK)) |
| 1158 | TIPC_SKB_CB(buf)->handle = |
| 1159 | (void *)(unsigned long)(offset + sz_to_copy); |
| 1160 | goto exit; |
| 1161 | } |
| 1162 | } else { |
| 1163 | if (sz_copied != 0) |
| 1164 | goto exit; /* can't add error msg to valid data */ |
| 1165 | |
| 1166 | if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control) |
| 1167 | res = 0; |
| 1168 | else |
| 1169 | res = -ECONNRESET; |
| 1170 | } |
| 1171 | |
| 1172 | /* Consume received message (optional) */ |
| 1173 | if (likely(!(flags & MSG_PEEK))) { |
| 1174 | if (unlikely(++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN)) |
| 1175 | tipc_acknowledge(tport->ref, tport->conn_unacked); |
| 1176 | advance_rx_queue(sk); |
| 1177 | } |
| 1178 | |
| 1179 | /* Loop around if more data is required */ |
| 1180 | if ((sz_copied < buf_len) && /* didn't get all requested data */ |
| 1181 | (!skb_queue_empty(&sk->sk_receive_queue) || |
| 1182 | (sz_copied < target)) && /* and more is ready or required */ |
| 1183 | (!(flags & MSG_PEEK)) && /* and aren't just peeking at data */ |
| 1184 | (!err)) /* and haven't reached a FIN */ |
| 1185 | goto restart; |
| 1186 | |
| 1187 | exit: |
| 1188 | release_sock(sk); |
| 1189 | return sz_copied ? sz_copied : res; |
| 1190 | } |
| 1191 | |
| 1192 | /** |
| 1193 | * tipc_write_space - wake up thread if port congestion is released |
| 1194 | * @sk: socket |
| 1195 | */ |
| 1196 | static void tipc_write_space(struct sock *sk) |
| 1197 | { |
| 1198 | struct socket_wq *wq; |
| 1199 | |
| 1200 | rcu_read_lock(); |
| 1201 | wq = rcu_dereference(sk->sk_wq); |
| 1202 | if (wq_has_sleeper(wq)) |
| 1203 | wake_up_interruptible_sync_poll(&wq->wait, POLLOUT | |
| 1204 | POLLWRNORM | POLLWRBAND); |
| 1205 | rcu_read_unlock(); |
| 1206 | } |
| 1207 | |
| 1208 | /** |
| 1209 | * tipc_data_ready - wake up threads to indicate messages have been received |
| 1210 | * @sk: socket |
| 1211 | * @len: the length of messages |
| 1212 | */ |
| 1213 | static void tipc_data_ready(struct sock *sk, int len) |
| 1214 | { |
| 1215 | struct socket_wq *wq; |
| 1216 | |
| 1217 | rcu_read_lock(); |
| 1218 | wq = rcu_dereference(sk->sk_wq); |
| 1219 | if (wq_has_sleeper(wq)) |
| 1220 | wake_up_interruptible_sync_poll(&wq->wait, POLLIN | |
| 1221 | POLLRDNORM | POLLRDBAND); |
| 1222 | rcu_read_unlock(); |
| 1223 | } |
| 1224 | |
| 1225 | /** |
| 1226 | * filter_connect - Handle all incoming messages for a connection-based socket |
| 1227 | * @tsock: TIPC socket |
| 1228 | * @msg: message |
| 1229 | * |
| 1230 | * Returns TIPC error status code and socket error status code |
| 1231 | * once it encounters some errors |
| 1232 | */ |
| 1233 | static u32 filter_connect(struct tipc_sock *tsock, struct sk_buff **buf) |
| 1234 | { |
| 1235 | struct socket *sock = tsock->sk.sk_socket; |
| 1236 | struct tipc_msg *msg = buf_msg(*buf); |
| 1237 | struct sock *sk = &tsock->sk; |
| 1238 | u32 retval = TIPC_ERR_NO_PORT; |
| 1239 | int res; |
| 1240 | |
| 1241 | if (msg_mcast(msg)) |
| 1242 | return retval; |
| 1243 | |
| 1244 | switch ((int)sock->state) { |
| 1245 | case SS_CONNECTED: |
| 1246 | /* Accept only connection-based messages sent by peer */ |
| 1247 | if (msg_connected(msg) && tipc_port_peer_msg(tsock->p, msg)) { |
| 1248 | if (unlikely(msg_errcode(msg))) { |
| 1249 | sock->state = SS_DISCONNECTING; |
| 1250 | __tipc_disconnect(tsock->p); |
| 1251 | } |
| 1252 | retval = TIPC_OK; |
| 1253 | } |
| 1254 | break; |
| 1255 | case SS_CONNECTING: |
| 1256 | /* Accept only ACK or NACK message */ |
| 1257 | if (unlikely(msg_errcode(msg))) { |
| 1258 | sock->state = SS_DISCONNECTING; |
| 1259 | sk->sk_err = -ECONNREFUSED; |
| 1260 | retval = TIPC_OK; |
| 1261 | break; |
| 1262 | } |
| 1263 | |
| 1264 | if (unlikely(!msg_connected(msg))) |
| 1265 | break; |
| 1266 | |
| 1267 | res = auto_connect(sock, msg); |
| 1268 | if (res) { |
| 1269 | sock->state = SS_DISCONNECTING; |
| 1270 | sk->sk_err = res; |
| 1271 | retval = TIPC_OK; |
| 1272 | break; |
| 1273 | } |
| 1274 | |
| 1275 | /* If an incoming message is an 'ACK-', it should be |
| 1276 | * discarded here because it doesn't contain useful |
| 1277 | * data. In addition, we should try to wake up |
| 1278 | * connect() routine if sleeping. |
| 1279 | */ |
| 1280 | if (msg_data_sz(msg) == 0) { |
| 1281 | kfree_skb(*buf); |
| 1282 | *buf = NULL; |
| 1283 | if (waitqueue_active(sk_sleep(sk))) |
| 1284 | wake_up_interruptible(sk_sleep(sk)); |
| 1285 | } |
| 1286 | retval = TIPC_OK; |
| 1287 | break; |
| 1288 | case SS_LISTENING: |
| 1289 | case SS_UNCONNECTED: |
| 1290 | /* Accept only SYN message */ |
| 1291 | if (!msg_connected(msg) && !(msg_errcode(msg))) |
| 1292 | retval = TIPC_OK; |
| 1293 | break; |
| 1294 | case SS_DISCONNECTING: |
| 1295 | break; |
| 1296 | default: |
| 1297 | pr_err("Unknown socket state %u\n", sock->state); |
| 1298 | } |
| 1299 | return retval; |
| 1300 | } |
| 1301 | |
| 1302 | /** |
| 1303 | * rcvbuf_limit - get proper overload limit of socket receive queue |
| 1304 | * @sk: socket |
| 1305 | * @buf: message |
| 1306 | * |
| 1307 | * For all connection oriented messages, irrespective of importance, |
| 1308 | * the default overload value (i.e. 67MB) is set as limit. |
| 1309 | * |
| 1310 | * For all connectionless messages, by default new queue limits are |
| 1311 | * as belows: |
| 1312 | * |
| 1313 | * TIPC_LOW_IMPORTANCE (4 MB) |
| 1314 | * TIPC_MEDIUM_IMPORTANCE (8 MB) |
| 1315 | * TIPC_HIGH_IMPORTANCE (16 MB) |
| 1316 | * TIPC_CRITICAL_IMPORTANCE (32 MB) |
| 1317 | * |
| 1318 | * Returns overload limit according to corresponding message importance |
| 1319 | */ |
| 1320 | static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *buf) |
| 1321 | { |
| 1322 | struct tipc_msg *msg = buf_msg(buf); |
| 1323 | unsigned int limit; |
| 1324 | |
| 1325 | if (msg_connected(msg)) |
| 1326 | limit = sysctl_tipc_rmem[2]; |
| 1327 | else |
| 1328 | limit = sk->sk_rcvbuf >> TIPC_CRITICAL_IMPORTANCE << |
| 1329 | msg_importance(msg); |
| 1330 | return limit; |
| 1331 | } |
| 1332 | |
| 1333 | /** |
| 1334 | * filter_rcv - validate incoming message |
| 1335 | * @sk: socket |
| 1336 | * @buf: message |
| 1337 | * |
| 1338 | * Enqueues message on receive queue if acceptable; optionally handles |
| 1339 | * disconnect indication for a connected socket. |
| 1340 | * |
| 1341 | * Called with socket lock already taken; port lock may also be taken. |
| 1342 | * |
| 1343 | * Returns TIPC error status code (TIPC_OK if message is not to be rejected) |
| 1344 | */ |
| 1345 | static u32 filter_rcv(struct sock *sk, struct sk_buff *buf) |
| 1346 | { |
| 1347 | struct socket *sock = sk->sk_socket; |
| 1348 | struct tipc_msg *msg = buf_msg(buf); |
| 1349 | unsigned int limit = rcvbuf_limit(sk, buf); |
| 1350 | u32 res = TIPC_OK; |
| 1351 | |
| 1352 | /* Reject message if it is wrong sort of message for socket */ |
| 1353 | if (msg_type(msg) > TIPC_DIRECT_MSG) |
| 1354 | return TIPC_ERR_NO_PORT; |
| 1355 | |
| 1356 | if (sock->state == SS_READY) { |
| 1357 | if (msg_connected(msg)) |
| 1358 | return TIPC_ERR_NO_PORT; |
| 1359 | } else { |
| 1360 | res = filter_connect(tipc_sk(sk), &buf); |
| 1361 | if (res != TIPC_OK || buf == NULL) |
| 1362 | return res; |
| 1363 | } |
| 1364 | |
| 1365 | /* Reject message if there isn't room to queue it */ |
| 1366 | if (sk_rmem_alloc_get(sk) + buf->truesize >= limit) |
| 1367 | return TIPC_ERR_OVERLOAD; |
| 1368 | |
| 1369 | /* Enqueue message */ |
| 1370 | TIPC_SKB_CB(buf)->handle = 0; |
| 1371 | __skb_queue_tail(&sk->sk_receive_queue, buf); |
| 1372 | skb_set_owner_r(buf, sk); |
| 1373 | |
| 1374 | sk->sk_data_ready(sk, 0); |
| 1375 | return TIPC_OK; |
| 1376 | } |
| 1377 | |
| 1378 | /** |
| 1379 | * backlog_rcv - handle incoming message from backlog queue |
| 1380 | * @sk: socket |
| 1381 | * @buf: message |
| 1382 | * |
| 1383 | * Caller must hold socket lock, but not port lock. |
| 1384 | * |
| 1385 | * Returns 0 |
| 1386 | */ |
| 1387 | static int backlog_rcv(struct sock *sk, struct sk_buff *buf) |
| 1388 | { |
| 1389 | u32 res; |
| 1390 | |
| 1391 | res = filter_rcv(sk, buf); |
| 1392 | if (res) |
| 1393 | tipc_reject_msg(buf, res); |
| 1394 | return 0; |
| 1395 | } |
| 1396 | |
| 1397 | /** |
| 1398 | * dispatch - handle incoming message |
| 1399 | * @tport: TIPC port that received message |
| 1400 | * @buf: message |
| 1401 | * |
| 1402 | * Called with port lock already taken. |
| 1403 | * |
| 1404 | * Returns TIPC error status code (TIPC_OK if message is not to be rejected) |
| 1405 | */ |
| 1406 | static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf) |
| 1407 | { |
| 1408 | struct sock *sk = (struct sock *)tport->usr_handle; |
| 1409 | u32 res; |
| 1410 | |
| 1411 | /* |
| 1412 | * Process message if socket is unlocked; otherwise add to backlog queue |
| 1413 | * |
| 1414 | * This code is based on sk_receive_skb(), but must be distinct from it |
| 1415 | * since a TIPC-specific filter/reject mechanism is utilized |
| 1416 | */ |
| 1417 | bh_lock_sock(sk); |
| 1418 | if (!sock_owned_by_user(sk)) { |
| 1419 | res = filter_rcv(sk, buf); |
| 1420 | } else { |
| 1421 | if (sk_add_backlog(sk, buf, rcvbuf_limit(sk, buf))) |
| 1422 | res = TIPC_ERR_OVERLOAD; |
| 1423 | else |
| 1424 | res = TIPC_OK; |
| 1425 | } |
| 1426 | bh_unlock_sock(sk); |
| 1427 | |
| 1428 | return res; |
| 1429 | } |
| 1430 | |
| 1431 | /** |
| 1432 | * wakeupdispatch - wake up port after congestion |
| 1433 | * @tport: port to wakeup |
| 1434 | * |
| 1435 | * Called with port lock already taken. |
| 1436 | */ |
| 1437 | static void wakeupdispatch(struct tipc_port *tport) |
| 1438 | { |
| 1439 | struct sock *sk = (struct sock *)tport->usr_handle; |
| 1440 | |
| 1441 | sk->sk_write_space(sk); |
| 1442 | } |
| 1443 | |
| 1444 | /** |
| 1445 | * connect - establish a connection to another TIPC port |
| 1446 | * @sock: socket structure |
| 1447 | * @dest: socket address for destination port |
| 1448 | * @destlen: size of socket address data structure |
| 1449 | * @flags: file-related flags associated with socket |
| 1450 | * |
| 1451 | * Returns 0 on success, errno otherwise |
| 1452 | */ |
| 1453 | static int connect(struct socket *sock, struct sockaddr *dest, int destlen, |
| 1454 | int flags) |
| 1455 | { |
| 1456 | struct sock *sk = sock->sk; |
| 1457 | struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest; |
| 1458 | struct msghdr m = {NULL,}; |
| 1459 | unsigned int timeout; |
| 1460 | int res; |
| 1461 | |
| 1462 | lock_sock(sk); |
| 1463 | |
| 1464 | /* For now, TIPC does not allow use of connect() with DGRAM/RDM types */ |
| 1465 | if (sock->state == SS_READY) { |
| 1466 | res = -EOPNOTSUPP; |
| 1467 | goto exit; |
| 1468 | } |
| 1469 | |
| 1470 | /* |
| 1471 | * Reject connection attempt using multicast address |
| 1472 | * |
| 1473 | * Note: send_msg() validates the rest of the address fields, |
| 1474 | * so there's no need to do it here |
| 1475 | */ |
| 1476 | if (dst->addrtype == TIPC_ADDR_MCAST) { |
| 1477 | res = -EINVAL; |
| 1478 | goto exit; |
| 1479 | } |
| 1480 | |
| 1481 | timeout = (flags & O_NONBLOCK) ? 0 : tipc_sk(sk)->conn_timeout; |
| 1482 | |
| 1483 | switch (sock->state) { |
| 1484 | case SS_UNCONNECTED: |
| 1485 | /* Send a 'SYN-' to destination */ |
| 1486 | m.msg_name = dest; |
| 1487 | m.msg_namelen = destlen; |
| 1488 | |
| 1489 | /* If connect is in non-blocking case, set MSG_DONTWAIT to |
| 1490 | * indicate send_msg() is never blocked. |
| 1491 | */ |
| 1492 | if (!timeout) |
| 1493 | m.msg_flags = MSG_DONTWAIT; |
| 1494 | |
| 1495 | res = send_msg(NULL, sock, &m, 0); |
| 1496 | if ((res < 0) && (res != -EWOULDBLOCK)) |
| 1497 | goto exit; |
| 1498 | |
| 1499 | /* Just entered SS_CONNECTING state; the only |
| 1500 | * difference is that return value in non-blocking |
| 1501 | * case is EINPROGRESS, rather than EALREADY. |
| 1502 | */ |
| 1503 | res = -EINPROGRESS; |
| 1504 | break; |
| 1505 | case SS_CONNECTING: |
| 1506 | res = -EALREADY; |
| 1507 | break; |
| 1508 | case SS_CONNECTED: |
| 1509 | res = -EISCONN; |
| 1510 | break; |
| 1511 | default: |
| 1512 | res = -EINVAL; |
| 1513 | goto exit; |
| 1514 | } |
| 1515 | |
| 1516 | if (sock->state == SS_CONNECTING) { |
| 1517 | if (!timeout) |
| 1518 | goto exit; |
| 1519 | |
| 1520 | /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */ |
| 1521 | release_sock(sk); |
| 1522 | res = wait_event_interruptible_timeout(*sk_sleep(sk), |
| 1523 | sock->state != SS_CONNECTING, |
| 1524 | timeout ? (long)msecs_to_jiffies(timeout) |
| 1525 | : MAX_SCHEDULE_TIMEOUT); |
| 1526 | lock_sock(sk); |
| 1527 | if (res <= 0) { |
| 1528 | if (res == 0) |
| 1529 | res = -ETIMEDOUT; |
| 1530 | else |
| 1531 | ; /* leave "res" unchanged */ |
| 1532 | goto exit; |
| 1533 | } |
| 1534 | } |
| 1535 | |
| 1536 | if (unlikely(sock->state == SS_DISCONNECTING)) |
| 1537 | res = sock_error(sk); |
| 1538 | else |
| 1539 | res = 0; |
| 1540 | |
| 1541 | exit: |
| 1542 | release_sock(sk); |
| 1543 | return res; |
| 1544 | } |
| 1545 | |
| 1546 | /** |
| 1547 | * listen - allow socket to listen for incoming connections |
| 1548 | * @sock: socket structure |
| 1549 | * @len: (unused) |
| 1550 | * |
| 1551 | * Returns 0 on success, errno otherwise |
| 1552 | */ |
| 1553 | static int listen(struct socket *sock, int len) |
| 1554 | { |
| 1555 | struct sock *sk = sock->sk; |
| 1556 | int res; |
| 1557 | |
| 1558 | lock_sock(sk); |
| 1559 | |
| 1560 | if (sock->state != SS_UNCONNECTED) |
| 1561 | res = -EINVAL; |
| 1562 | else { |
| 1563 | sock->state = SS_LISTENING; |
| 1564 | res = 0; |
| 1565 | } |
| 1566 | |
| 1567 | release_sock(sk); |
| 1568 | return res; |
| 1569 | } |
| 1570 | |
| 1571 | /** |
| 1572 | * accept - wait for connection request |
| 1573 | * @sock: listening socket |
| 1574 | * @newsock: new socket that is to be connected |
| 1575 | * @flags: file-related flags associated with socket |
| 1576 | * |
| 1577 | * Returns 0 on success, errno otherwise |
| 1578 | */ |
| 1579 | static int accept(struct socket *sock, struct socket *new_sock, int flags) |
| 1580 | { |
| 1581 | struct sock *new_sk, *sk = sock->sk; |
| 1582 | struct sk_buff *buf; |
| 1583 | struct tipc_sock *new_tsock; |
| 1584 | struct tipc_port *new_tport; |
| 1585 | struct tipc_msg *msg; |
| 1586 | u32 new_ref; |
| 1587 | |
| 1588 | int res; |
| 1589 | |
| 1590 | lock_sock(sk); |
| 1591 | |
| 1592 | if (sock->state != SS_LISTENING) { |
| 1593 | res = -EINVAL; |
| 1594 | goto exit; |
| 1595 | } |
| 1596 | |
| 1597 | while (skb_queue_empty(&sk->sk_receive_queue)) { |
| 1598 | if (flags & O_NONBLOCK) { |
| 1599 | res = -EWOULDBLOCK; |
| 1600 | goto exit; |
| 1601 | } |
| 1602 | release_sock(sk); |
| 1603 | res = wait_event_interruptible(*sk_sleep(sk), |
| 1604 | (!skb_queue_empty(&sk->sk_receive_queue))); |
| 1605 | lock_sock(sk); |
| 1606 | if (res) |
| 1607 | goto exit; |
| 1608 | } |
| 1609 | |
| 1610 | buf = skb_peek(&sk->sk_receive_queue); |
| 1611 | |
| 1612 | res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, 1); |
| 1613 | if (res) |
| 1614 | goto exit; |
| 1615 | |
| 1616 | new_sk = new_sock->sk; |
| 1617 | new_tsock = tipc_sk(new_sk); |
| 1618 | new_tport = new_tsock->p; |
| 1619 | new_ref = new_tport->ref; |
| 1620 | msg = buf_msg(buf); |
| 1621 | |
| 1622 | /* we lock on new_sk; but lockdep sees the lock on sk */ |
| 1623 | lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING); |
| 1624 | |
| 1625 | /* |
| 1626 | * Reject any stray messages received by new socket |
| 1627 | * before the socket lock was taken (very, very unlikely) |
| 1628 | */ |
| 1629 | reject_rx_queue(new_sk); |
| 1630 | |
| 1631 | /* Connect new socket to it's peer */ |
| 1632 | new_tsock->peer_name.ref = msg_origport(msg); |
| 1633 | new_tsock->peer_name.node = msg_orignode(msg); |
| 1634 | tipc_connect(new_ref, &new_tsock->peer_name); |
| 1635 | new_sock->state = SS_CONNECTED; |
| 1636 | |
| 1637 | tipc_set_portimportance(new_ref, msg_importance(msg)); |
| 1638 | if (msg_named(msg)) { |
| 1639 | new_tport->conn_type = msg_nametype(msg); |
| 1640 | new_tport->conn_instance = msg_nameinst(msg); |
| 1641 | } |
| 1642 | |
| 1643 | /* |
| 1644 | * Respond to 'SYN-' by discarding it & returning 'ACK'-. |
| 1645 | * Respond to 'SYN+' by queuing it on new socket. |
| 1646 | */ |
| 1647 | if (!msg_data_sz(msg)) { |
| 1648 | struct msghdr m = {NULL,}; |
| 1649 | |
| 1650 | advance_rx_queue(sk); |
| 1651 | send_packet(NULL, new_sock, &m, 0); |
| 1652 | } else { |
| 1653 | __skb_dequeue(&sk->sk_receive_queue); |
| 1654 | __skb_queue_head(&new_sk->sk_receive_queue, buf); |
| 1655 | skb_set_owner_r(buf, new_sk); |
| 1656 | } |
| 1657 | release_sock(new_sk); |
| 1658 | |
| 1659 | exit: |
| 1660 | release_sock(sk); |
| 1661 | return res; |
| 1662 | } |
| 1663 | |
| 1664 | /** |
| 1665 | * shutdown - shutdown socket connection |
| 1666 | * @sock: socket structure |
| 1667 | * @how: direction to close (must be SHUT_RDWR) |
| 1668 | * |
| 1669 | * Terminates connection (if necessary), then purges socket's receive queue. |
| 1670 | * |
| 1671 | * Returns 0 on success, errno otherwise |
| 1672 | */ |
| 1673 | static int shutdown(struct socket *sock, int how) |
| 1674 | { |
| 1675 | struct sock *sk = sock->sk; |
| 1676 | struct tipc_port *tport = tipc_sk_port(sk); |
| 1677 | struct sk_buff *buf; |
| 1678 | int res; |
| 1679 | |
| 1680 | if (how != SHUT_RDWR) |
| 1681 | return -EINVAL; |
| 1682 | |
| 1683 | lock_sock(sk); |
| 1684 | |
| 1685 | switch (sock->state) { |
| 1686 | case SS_CONNECTING: |
| 1687 | case SS_CONNECTED: |
| 1688 | |
| 1689 | restart: |
| 1690 | /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */ |
| 1691 | buf = __skb_dequeue(&sk->sk_receive_queue); |
| 1692 | if (buf) { |
| 1693 | if (TIPC_SKB_CB(buf)->handle != 0) { |
| 1694 | kfree_skb(buf); |
| 1695 | goto restart; |
| 1696 | } |
| 1697 | tipc_disconnect(tport->ref); |
| 1698 | tipc_reject_msg(buf, TIPC_CONN_SHUTDOWN); |
| 1699 | } else { |
| 1700 | tipc_shutdown(tport->ref); |
| 1701 | } |
| 1702 | |
| 1703 | sock->state = SS_DISCONNECTING; |
| 1704 | |
| 1705 | /* fall through */ |
| 1706 | |
| 1707 | case SS_DISCONNECTING: |
| 1708 | |
| 1709 | /* Discard any unreceived messages */ |
| 1710 | __skb_queue_purge(&sk->sk_receive_queue); |
| 1711 | |
| 1712 | /* Wake up anyone sleeping in poll */ |
| 1713 | sk->sk_state_change(sk); |
| 1714 | res = 0; |
| 1715 | break; |
| 1716 | |
| 1717 | default: |
| 1718 | res = -ENOTCONN; |
| 1719 | } |
| 1720 | |
| 1721 | release_sock(sk); |
| 1722 | return res; |
| 1723 | } |
| 1724 | |
| 1725 | /** |
| 1726 | * setsockopt - set socket option |
| 1727 | * @sock: socket structure |
| 1728 | * @lvl: option level |
| 1729 | * @opt: option identifier |
| 1730 | * @ov: pointer to new option value |
| 1731 | * @ol: length of option value |
| 1732 | * |
| 1733 | * For stream sockets only, accepts and ignores all IPPROTO_TCP options |
| 1734 | * (to ease compatibility). |
| 1735 | * |
| 1736 | * Returns 0 on success, errno otherwise |
| 1737 | */ |
| 1738 | static int setsockopt(struct socket *sock, |
| 1739 | int lvl, int opt, char __user *ov, unsigned int ol) |
| 1740 | { |
| 1741 | struct sock *sk = sock->sk; |
| 1742 | struct tipc_port *tport = tipc_sk_port(sk); |
| 1743 | u32 value; |
| 1744 | int res; |
| 1745 | |
| 1746 | if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM)) |
| 1747 | return 0; |
| 1748 | if (lvl != SOL_TIPC) |
| 1749 | return -ENOPROTOOPT; |
| 1750 | if (ol < sizeof(value)) |
| 1751 | return -EINVAL; |
| 1752 | res = get_user(value, (u32 __user *)ov); |
| 1753 | if (res) |
| 1754 | return res; |
| 1755 | |
| 1756 | lock_sock(sk); |
| 1757 | |
| 1758 | switch (opt) { |
| 1759 | case TIPC_IMPORTANCE: |
| 1760 | res = tipc_set_portimportance(tport->ref, value); |
| 1761 | break; |
| 1762 | case TIPC_SRC_DROPPABLE: |
| 1763 | if (sock->type != SOCK_STREAM) |
| 1764 | res = tipc_set_portunreliable(tport->ref, value); |
| 1765 | else |
| 1766 | res = -ENOPROTOOPT; |
| 1767 | break; |
| 1768 | case TIPC_DEST_DROPPABLE: |
| 1769 | res = tipc_set_portunreturnable(tport->ref, value); |
| 1770 | break; |
| 1771 | case TIPC_CONN_TIMEOUT: |
| 1772 | tipc_sk(sk)->conn_timeout = value; |
| 1773 | /* no need to set "res", since already 0 at this point */ |
| 1774 | break; |
| 1775 | default: |
| 1776 | res = -EINVAL; |
| 1777 | } |
| 1778 | |
| 1779 | release_sock(sk); |
| 1780 | |
| 1781 | return res; |
| 1782 | } |
| 1783 | |
| 1784 | /** |
| 1785 | * getsockopt - get socket option |
| 1786 | * @sock: socket structure |
| 1787 | * @lvl: option level |
| 1788 | * @opt: option identifier |
| 1789 | * @ov: receptacle for option value |
| 1790 | * @ol: receptacle for length of option value |
| 1791 | * |
| 1792 | * For stream sockets only, returns 0 length result for all IPPROTO_TCP options |
| 1793 | * (to ease compatibility). |
| 1794 | * |
| 1795 | * Returns 0 on success, errno otherwise |
| 1796 | */ |
| 1797 | static int getsockopt(struct socket *sock, |
| 1798 | int lvl, int opt, char __user *ov, int __user *ol) |
| 1799 | { |
| 1800 | struct sock *sk = sock->sk; |
| 1801 | struct tipc_port *tport = tipc_sk_port(sk); |
| 1802 | int len; |
| 1803 | u32 value; |
| 1804 | int res; |
| 1805 | |
| 1806 | if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM)) |
| 1807 | return put_user(0, ol); |
| 1808 | if (lvl != SOL_TIPC) |
| 1809 | return -ENOPROTOOPT; |
| 1810 | res = get_user(len, ol); |
| 1811 | if (res) |
| 1812 | return res; |
| 1813 | |
| 1814 | lock_sock(sk); |
| 1815 | |
| 1816 | switch (opt) { |
| 1817 | case TIPC_IMPORTANCE: |
| 1818 | res = tipc_portimportance(tport->ref, &value); |
| 1819 | break; |
| 1820 | case TIPC_SRC_DROPPABLE: |
| 1821 | res = tipc_portunreliable(tport->ref, &value); |
| 1822 | break; |
| 1823 | case TIPC_DEST_DROPPABLE: |
| 1824 | res = tipc_portunreturnable(tport->ref, &value); |
| 1825 | break; |
| 1826 | case TIPC_CONN_TIMEOUT: |
| 1827 | value = tipc_sk(sk)->conn_timeout; |
| 1828 | /* no need to set "res", since already 0 at this point */ |
| 1829 | break; |
| 1830 | case TIPC_NODE_RECVQ_DEPTH: |
| 1831 | value = 0; /* was tipc_queue_size, now obsolete */ |
| 1832 | break; |
| 1833 | case TIPC_SOCK_RECVQ_DEPTH: |
| 1834 | value = skb_queue_len(&sk->sk_receive_queue); |
| 1835 | break; |
| 1836 | default: |
| 1837 | res = -EINVAL; |
| 1838 | } |
| 1839 | |
| 1840 | release_sock(sk); |
| 1841 | |
| 1842 | if (res) |
| 1843 | return res; /* "get" failed */ |
| 1844 | |
| 1845 | if (len < sizeof(value)) |
| 1846 | return -EINVAL; |
| 1847 | |
| 1848 | if (copy_to_user(ov, &value, sizeof(value))) |
| 1849 | return -EFAULT; |
| 1850 | |
| 1851 | return put_user(sizeof(value), ol); |
| 1852 | } |
| 1853 | |
| 1854 | /* Protocol switches for the various types of TIPC sockets */ |
| 1855 | |
| 1856 | static const struct proto_ops msg_ops = { |
| 1857 | .owner = THIS_MODULE, |
| 1858 | .family = AF_TIPC, |
| 1859 | .release = release, |
| 1860 | .bind = bind, |
| 1861 | .connect = connect, |
| 1862 | .socketpair = sock_no_socketpair, |
| 1863 | .accept = sock_no_accept, |
| 1864 | .getname = get_name, |
| 1865 | .poll = poll, |
| 1866 | .ioctl = sock_no_ioctl, |
| 1867 | .listen = sock_no_listen, |
| 1868 | .shutdown = shutdown, |
| 1869 | .setsockopt = setsockopt, |
| 1870 | .getsockopt = getsockopt, |
| 1871 | .sendmsg = send_msg, |
| 1872 | .recvmsg = recv_msg, |
| 1873 | .mmap = sock_no_mmap, |
| 1874 | .sendpage = sock_no_sendpage |
| 1875 | }; |
| 1876 | |
| 1877 | static const struct proto_ops packet_ops = { |
| 1878 | .owner = THIS_MODULE, |
| 1879 | .family = AF_TIPC, |
| 1880 | .release = release, |
| 1881 | .bind = bind, |
| 1882 | .connect = connect, |
| 1883 | .socketpair = sock_no_socketpair, |
| 1884 | .accept = accept, |
| 1885 | .getname = get_name, |
| 1886 | .poll = poll, |
| 1887 | .ioctl = sock_no_ioctl, |
| 1888 | .listen = listen, |
| 1889 | .shutdown = shutdown, |
| 1890 | .setsockopt = setsockopt, |
| 1891 | .getsockopt = getsockopt, |
| 1892 | .sendmsg = send_packet, |
| 1893 | .recvmsg = recv_msg, |
| 1894 | .mmap = sock_no_mmap, |
| 1895 | .sendpage = sock_no_sendpage |
| 1896 | }; |
| 1897 | |
| 1898 | static const struct proto_ops stream_ops = { |
| 1899 | .owner = THIS_MODULE, |
| 1900 | .family = AF_TIPC, |
| 1901 | .release = release, |
| 1902 | .bind = bind, |
| 1903 | .connect = connect, |
| 1904 | .socketpair = sock_no_socketpair, |
| 1905 | .accept = accept, |
| 1906 | .getname = get_name, |
| 1907 | .poll = poll, |
| 1908 | .ioctl = sock_no_ioctl, |
| 1909 | .listen = listen, |
| 1910 | .shutdown = shutdown, |
| 1911 | .setsockopt = setsockopt, |
| 1912 | .getsockopt = getsockopt, |
| 1913 | .sendmsg = send_stream, |
| 1914 | .recvmsg = recv_stream, |
| 1915 | .mmap = sock_no_mmap, |
| 1916 | .sendpage = sock_no_sendpage |
| 1917 | }; |
| 1918 | |
| 1919 | static const struct net_proto_family tipc_family_ops = { |
| 1920 | .owner = THIS_MODULE, |
| 1921 | .family = AF_TIPC, |
| 1922 | .create = tipc_sk_create |
| 1923 | }; |
| 1924 | |
| 1925 | static struct proto tipc_proto = { |
| 1926 | .name = "TIPC", |
| 1927 | .owner = THIS_MODULE, |
| 1928 | .obj_size = sizeof(struct tipc_sock), |
| 1929 | .sysctl_rmem = sysctl_tipc_rmem |
| 1930 | }; |
| 1931 | |
| 1932 | static struct proto tipc_proto_kern = { |
| 1933 | .name = "TIPC", |
| 1934 | .obj_size = sizeof(struct tipc_sock), |
| 1935 | .sysctl_rmem = sysctl_tipc_rmem |
| 1936 | }; |
| 1937 | |
| 1938 | /** |
| 1939 | * tipc_socket_init - initialize TIPC socket interface |
| 1940 | * |
| 1941 | * Returns 0 on success, errno otherwise |
| 1942 | */ |
| 1943 | int tipc_socket_init(void) |
| 1944 | { |
| 1945 | int res; |
| 1946 | |
| 1947 | res = proto_register(&tipc_proto, 1); |
| 1948 | if (res) { |
| 1949 | pr_err("Failed to register TIPC protocol type\n"); |
| 1950 | goto out; |
| 1951 | } |
| 1952 | |
| 1953 | res = sock_register(&tipc_family_ops); |
| 1954 | if (res) { |
| 1955 | pr_err("Failed to register TIPC socket type\n"); |
| 1956 | proto_unregister(&tipc_proto); |
| 1957 | goto out; |
| 1958 | } |
| 1959 | |
| 1960 | sockets_enabled = 1; |
| 1961 | out: |
| 1962 | return res; |
| 1963 | } |
| 1964 | |
| 1965 | /** |
| 1966 | * tipc_socket_stop - stop TIPC socket interface |
| 1967 | */ |
| 1968 | void tipc_socket_stop(void) |
| 1969 | { |
| 1970 | if (!sockets_enabled) |
| 1971 | return; |
| 1972 | |
| 1973 | sockets_enabled = 0; |
| 1974 | sock_unregister(tipc_family_ops.family); |
| 1975 | proto_unregister(&tipc_proto); |
| 1976 | } |