1 /* 2 * net/tipc/socket.c: TIPC socket API 3 * 4 * Copyright (c) 2001-2007, 2012-2017, 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 <linux/rhashtable.h> 38 #include <linux/sched/signal.h> 39 40 #include "core.h" 41 #include "name_table.h" 42 #include "node.h" 43 #include "link.h" 44 #include "name_distr.h" 45 #include "socket.h" 46 #include "bcast.h" 47 #include "netlink.h" 48 #include "group.h" 49 50 #define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */ 51 #define CONN_PROBING_INTV msecs_to_jiffies(3600000) /* [ms] => 1 h */ 52 #define TIPC_FWD_MSG 1 53 #define TIPC_MAX_PORT 0xffffffff 54 #define TIPC_MIN_PORT 1 55 #define TIPC_ACK_RATE 4 /* ACK at 1/4 of of rcv window size */ 56 57 enum { 58 TIPC_LISTEN = TCP_LISTEN, 59 TIPC_ESTABLISHED = TCP_ESTABLISHED, 60 TIPC_OPEN = TCP_CLOSE, 61 TIPC_DISCONNECTING = TCP_CLOSE_WAIT, 62 TIPC_CONNECTING = TCP_SYN_SENT, 63 }; 64 65 struct sockaddr_pair { 66 struct sockaddr_tipc sock; 67 struct sockaddr_tipc member; 68 }; 69 70 /** 71 * struct tipc_sock - TIPC socket structure 72 * @sk: socket - interacts with 'port' and with user via the socket API 73 * @conn_type: TIPC type used when connection was established 74 * @conn_instance: TIPC instance used when connection was established 75 * @published: non-zero if port has one or more associated names 76 * @max_pkt: maximum packet size "hint" used when building messages sent by port 77 * @portid: unique port identity in TIPC socket hash table 78 * @phdr: preformatted message header used when sending messages 79 * #cong_links: list of congested links 80 * @publications: list of publications for port 81 * @blocking_link: address of the congested link we are currently sleeping on 82 * @pub_count: total # of publications port has made during its lifetime 83 * @probing_state: 84 * @conn_timeout: the time we can wait for an unresponded setup request 85 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue 86 * @cong_link_cnt: number of congested links 87 * @snt_unacked: # messages sent by socket, and not yet acked by peer 88 * @rcv_unacked: # messages read by user, but not yet acked back to peer 89 * @peer: 'connected' peer for dgram/rdm 90 * @node: hash table node 91 * @mc_method: cookie for use between socket and broadcast layer 92 * @rcu: rcu struct for tipc_sock 93 */ 94 struct tipc_sock { 95 struct sock sk; 96 u32 conn_type; 97 u32 conn_instance; 98 int published; 99 u32 max_pkt; 100 u32 portid; 101 struct tipc_msg phdr; 102 struct list_head cong_links; 103 struct list_head publications; 104 u32 pub_count; 105 uint conn_timeout; 106 atomic_t dupl_rcvcnt; 107 bool probe_unacked; 108 u16 cong_link_cnt; 109 u16 snt_unacked; 110 u16 snd_win; 111 u16 peer_caps; 112 u16 rcv_unacked; 113 u16 rcv_win; 114 struct sockaddr_tipc peer; 115 struct rhash_head node; 116 struct tipc_mc_method mc_method; 117 struct rcu_head rcu; 118 struct tipc_group *group; 119 bool group_is_open; 120 }; 121 122 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb); 123 static void tipc_data_ready(struct sock *sk); 124 static void tipc_write_space(struct sock *sk); 125 static void tipc_sock_destruct(struct sock *sk); 126 static int tipc_release(struct socket *sock); 127 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags, 128 bool kern); 129 static void tipc_sk_timeout(struct timer_list *t); 130 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope, 131 struct tipc_name_seq const *seq); 132 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope, 133 struct tipc_name_seq const *seq); 134 static int tipc_sk_leave(struct tipc_sock *tsk); 135 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid); 136 static int tipc_sk_insert(struct tipc_sock *tsk); 137 static void tipc_sk_remove(struct tipc_sock *tsk); 138 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz); 139 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz); 140 141 static const struct proto_ops packet_ops; 142 static const struct proto_ops stream_ops; 143 static const struct proto_ops msg_ops; 144 static struct proto tipc_proto; 145 static const struct rhashtable_params tsk_rht_params; 146 147 static u32 tsk_own_node(struct tipc_sock *tsk) 148 { 149 return msg_prevnode(&tsk->phdr); 150 } 151 152 static u32 tsk_peer_node(struct tipc_sock *tsk) 153 { 154 return msg_destnode(&tsk->phdr); 155 } 156 157 static u32 tsk_peer_port(struct tipc_sock *tsk) 158 { 159 return msg_destport(&tsk->phdr); 160 } 161 162 static bool tsk_unreliable(struct tipc_sock *tsk) 163 { 164 return msg_src_droppable(&tsk->phdr) != 0; 165 } 166 167 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable) 168 { 169 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0); 170 } 171 172 static bool tsk_unreturnable(struct tipc_sock *tsk) 173 { 174 return msg_dest_droppable(&tsk->phdr) != 0; 175 } 176 177 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable) 178 { 179 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0); 180 } 181 182 static int tsk_importance(struct tipc_sock *tsk) 183 { 184 return msg_importance(&tsk->phdr); 185 } 186 187 static int tsk_set_importance(struct tipc_sock *tsk, int imp) 188 { 189 if (imp > TIPC_CRITICAL_IMPORTANCE) 190 return -EINVAL; 191 msg_set_importance(&tsk->phdr, (u32)imp); 192 return 0; 193 } 194 195 static struct tipc_sock *tipc_sk(const struct sock *sk) 196 { 197 return container_of(sk, struct tipc_sock, sk); 198 } 199 200 static bool tsk_conn_cong(struct tipc_sock *tsk) 201 { 202 return tsk->snt_unacked > tsk->snd_win; 203 } 204 205 static u16 tsk_blocks(int len) 206 { 207 return ((len / FLOWCTL_BLK_SZ) + 1); 208 } 209 210 /* tsk_blocks(): translate a buffer size in bytes to number of 211 * advertisable blocks, taking into account the ratio truesize(len)/len 212 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ 213 */ 214 static u16 tsk_adv_blocks(int len) 215 { 216 return len / FLOWCTL_BLK_SZ / 4; 217 } 218 219 /* tsk_inc(): increment counter for sent or received data 220 * - If block based flow control is not supported by peer we 221 * fall back to message based ditto, incrementing the counter 222 */ 223 static u16 tsk_inc(struct tipc_sock *tsk, int msglen) 224 { 225 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL)) 226 return ((msglen / FLOWCTL_BLK_SZ) + 1); 227 return 1; 228 } 229 230 /** 231 * tsk_advance_rx_queue - discard first buffer in socket receive queue 232 * 233 * Caller must hold socket lock 234 */ 235 static void tsk_advance_rx_queue(struct sock *sk) 236 { 237 kfree_skb(__skb_dequeue(&sk->sk_receive_queue)); 238 } 239 240 /* tipc_sk_respond() : send response message back to sender 241 */ 242 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err) 243 { 244 u32 selector; 245 u32 dnode; 246 u32 onode = tipc_own_addr(sock_net(sk)); 247 248 if (!tipc_msg_reverse(onode, &skb, err)) 249 return; 250 251 dnode = msg_destnode(buf_msg(skb)); 252 selector = msg_origport(buf_msg(skb)); 253 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector); 254 } 255 256 /** 257 * tsk_rej_rx_queue - reject all buffers in socket receive queue 258 * 259 * Caller must hold socket lock 260 */ 261 static void tsk_rej_rx_queue(struct sock *sk) 262 { 263 struct sk_buff *skb; 264 265 while ((skb = __skb_dequeue(&sk->sk_receive_queue))) 266 tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT); 267 } 268 269 static bool tipc_sk_connected(struct sock *sk) 270 { 271 return sk->sk_state == TIPC_ESTABLISHED; 272 } 273 274 /* tipc_sk_type_connectionless - check if the socket is datagram socket 275 * @sk: socket 276 * 277 * Returns true if connection less, false otherwise 278 */ 279 static bool tipc_sk_type_connectionless(struct sock *sk) 280 { 281 return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM; 282 } 283 284 /* tsk_peer_msg - verify if message was sent by connected port's peer 285 * 286 * Handles cases where the node's network address has changed from 287 * the default of <0.0.0> to its configured setting. 288 */ 289 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg) 290 { 291 struct sock *sk = &tsk->sk; 292 u32 self = tipc_own_addr(sock_net(sk)); 293 u32 peer_port = tsk_peer_port(tsk); 294 u32 orig_node, peer_node; 295 296 if (unlikely(!tipc_sk_connected(sk))) 297 return false; 298 299 if (unlikely(msg_origport(msg) != peer_port)) 300 return false; 301 302 orig_node = msg_orignode(msg); 303 peer_node = tsk_peer_node(tsk); 304 305 if (likely(orig_node == peer_node)) 306 return true; 307 308 if (!orig_node && peer_node == self) 309 return true; 310 311 if (!peer_node && orig_node == self) 312 return true; 313 314 return false; 315 } 316 317 /* tipc_set_sk_state - set the sk_state of the socket 318 * @sk: socket 319 * 320 * Caller must hold socket lock 321 * 322 * Returns 0 on success, errno otherwise 323 */ 324 static int tipc_set_sk_state(struct sock *sk, int state) 325 { 326 int oldsk_state = sk->sk_state; 327 int res = -EINVAL; 328 329 switch (state) { 330 case TIPC_OPEN: 331 res = 0; 332 break; 333 case TIPC_LISTEN: 334 case TIPC_CONNECTING: 335 if (oldsk_state == TIPC_OPEN) 336 res = 0; 337 break; 338 case TIPC_ESTABLISHED: 339 if (oldsk_state == TIPC_CONNECTING || 340 oldsk_state == TIPC_OPEN) 341 res = 0; 342 break; 343 case TIPC_DISCONNECTING: 344 if (oldsk_state == TIPC_CONNECTING || 345 oldsk_state == TIPC_ESTABLISHED) 346 res = 0; 347 break; 348 } 349 350 if (!res) 351 sk->sk_state = state; 352 353 return res; 354 } 355 356 static int tipc_sk_sock_err(struct socket *sock, long *timeout) 357 { 358 struct sock *sk = sock->sk; 359 int err = sock_error(sk); 360 int typ = sock->type; 361 362 if (err) 363 return err; 364 if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) { 365 if (sk->sk_state == TIPC_DISCONNECTING) 366 return -EPIPE; 367 else if (!tipc_sk_connected(sk)) 368 return -ENOTCONN; 369 } 370 if (!*timeout) 371 return -EAGAIN; 372 if (signal_pending(current)) 373 return sock_intr_errno(*timeout); 374 375 return 0; 376 } 377 378 #define tipc_wait_for_cond(sock_, timeo_, condition_) \ 379 ({ \ 380 struct sock *sk_; \ 381 int rc_; \ 382 \ 383 while ((rc_ = !(condition_))) { \ 384 DEFINE_WAIT_FUNC(wait_, woken_wake_function); \ 385 sk_ = (sock_)->sk; \ 386 rc_ = tipc_sk_sock_err((sock_), timeo_); \ 387 if (rc_) \ 388 break; \ 389 prepare_to_wait(sk_sleep(sk_), &wait_, TASK_INTERRUPTIBLE); \ 390 release_sock(sk_); \ 391 *(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \ 392 sched_annotate_sleep(); \ 393 lock_sock(sk_); \ 394 remove_wait_queue(sk_sleep(sk_), &wait_); \ 395 } \ 396 rc_; \ 397 }) 398 399 /** 400 * tipc_sk_create - create a TIPC socket 401 * @net: network namespace (must be default network) 402 * @sock: pre-allocated socket structure 403 * @protocol: protocol indicator (must be 0) 404 * @kern: caused by kernel or by userspace? 405 * 406 * This routine creates additional data structures used by the TIPC socket, 407 * initializes them, and links them together. 408 * 409 * Returns 0 on success, errno otherwise 410 */ 411 static int tipc_sk_create(struct net *net, struct socket *sock, 412 int protocol, int kern) 413 { 414 const struct proto_ops *ops; 415 struct sock *sk; 416 struct tipc_sock *tsk; 417 struct tipc_msg *msg; 418 419 /* Validate arguments */ 420 if (unlikely(protocol != 0)) 421 return -EPROTONOSUPPORT; 422 423 switch (sock->type) { 424 case SOCK_STREAM: 425 ops = &stream_ops; 426 break; 427 case SOCK_SEQPACKET: 428 ops = &packet_ops; 429 break; 430 case SOCK_DGRAM: 431 case SOCK_RDM: 432 ops = &msg_ops; 433 break; 434 default: 435 return -EPROTOTYPE; 436 } 437 438 /* Allocate socket's protocol area */ 439 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern); 440 if (sk == NULL) 441 return -ENOMEM; 442 443 tsk = tipc_sk(sk); 444 tsk->max_pkt = MAX_PKT_DEFAULT; 445 INIT_LIST_HEAD(&tsk->publications); 446 INIT_LIST_HEAD(&tsk->cong_links); 447 msg = &tsk->phdr; 448 449 /* Finish initializing socket data structures */ 450 sock->ops = ops; 451 sock_init_data(sock, sk); 452 tipc_set_sk_state(sk, TIPC_OPEN); 453 if (tipc_sk_insert(tsk)) { 454 pr_warn("Socket create failed; port number exhausted\n"); 455 return -EINVAL; 456 } 457 458 /* Ensure tsk is visible before we read own_addr. */ 459 smp_mb(); 460 461 tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE, 462 TIPC_NAMED_MSG, NAMED_H_SIZE, 0); 463 464 msg_set_origport(msg, tsk->portid); 465 timer_setup(&sk->sk_timer, tipc_sk_timeout, 0); 466 sk->sk_shutdown = 0; 467 sk->sk_backlog_rcv = tipc_sk_backlog_rcv; 468 sk->sk_rcvbuf = sysctl_tipc_rmem[1]; 469 sk->sk_data_ready = tipc_data_ready; 470 sk->sk_write_space = tipc_write_space; 471 sk->sk_destruct = tipc_sock_destruct; 472 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT; 473 tsk->group_is_open = true; 474 atomic_set(&tsk->dupl_rcvcnt, 0); 475 476 /* Start out with safe limits until we receive an advertised window */ 477 tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN); 478 tsk->rcv_win = tsk->snd_win; 479 480 if (tipc_sk_type_connectionless(sk)) { 481 tsk_set_unreturnable(tsk, true); 482 if (sock->type == SOCK_DGRAM) 483 tsk_set_unreliable(tsk, true); 484 } 485 486 return 0; 487 } 488 489 static void tipc_sk_callback(struct rcu_head *head) 490 { 491 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu); 492 493 sock_put(&tsk->sk); 494 } 495 496 /* Caller should hold socket lock for the socket. */ 497 static void __tipc_shutdown(struct socket *sock, int error) 498 { 499 struct sock *sk = sock->sk; 500 struct tipc_sock *tsk = tipc_sk(sk); 501 struct net *net = sock_net(sk); 502 long timeout = CONN_TIMEOUT_DEFAULT; 503 u32 dnode = tsk_peer_node(tsk); 504 struct sk_buff *skb; 505 506 /* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */ 507 tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt && 508 !tsk_conn_cong(tsk))); 509 510 /* Reject all unreceived messages, except on an active connection 511 * (which disconnects locally & sends a 'FIN+' to peer). 512 */ 513 while ((skb = __skb_dequeue(&sk->sk_receive_queue)) != NULL) { 514 if (TIPC_SKB_CB(skb)->bytes_read) { 515 kfree_skb(skb); 516 continue; 517 } 518 if (!tipc_sk_type_connectionless(sk) && 519 sk->sk_state != TIPC_DISCONNECTING) { 520 tipc_set_sk_state(sk, TIPC_DISCONNECTING); 521 tipc_node_remove_conn(net, dnode, tsk->portid); 522 } 523 tipc_sk_respond(sk, skb, error); 524 } 525 526 if (tipc_sk_type_connectionless(sk)) 527 return; 528 529 if (sk->sk_state != TIPC_DISCONNECTING) { 530 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE, 531 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode, 532 tsk_own_node(tsk), tsk_peer_port(tsk), 533 tsk->portid, error); 534 if (skb) 535 tipc_node_xmit_skb(net, skb, dnode, tsk->portid); 536 tipc_node_remove_conn(net, dnode, tsk->portid); 537 tipc_set_sk_state(sk, TIPC_DISCONNECTING); 538 } 539 } 540 541 /** 542 * tipc_release - destroy a TIPC socket 543 * @sock: socket to destroy 544 * 545 * This routine cleans up any messages that are still queued on the socket. 546 * For DGRAM and RDM socket types, all queued messages are rejected. 547 * For SEQPACKET and STREAM socket types, the first message is rejected 548 * and any others are discarded. (If the first message on a STREAM socket 549 * is partially-read, it is discarded and the next one is rejected instead.) 550 * 551 * NOTE: Rejected messages are not necessarily returned to the sender! They 552 * are returned or discarded according to the "destination droppable" setting 553 * specified for the message by the sender. 554 * 555 * Returns 0 on success, errno otherwise 556 */ 557 static int tipc_release(struct socket *sock) 558 { 559 struct sock *sk = sock->sk; 560 struct tipc_sock *tsk; 561 562 /* 563 * Exit if socket isn't fully initialized (occurs when a failed accept() 564 * releases a pre-allocated child socket that was never used) 565 */ 566 if (sk == NULL) 567 return 0; 568 569 tsk = tipc_sk(sk); 570 lock_sock(sk); 571 572 __tipc_shutdown(sock, TIPC_ERR_NO_PORT); 573 sk->sk_shutdown = SHUTDOWN_MASK; 574 tipc_sk_leave(tsk); 575 tipc_sk_withdraw(tsk, 0, NULL); 576 sk_stop_timer(sk, &sk->sk_timer); 577 tipc_sk_remove(tsk); 578 579 sock_orphan(sk); 580 /* Reject any messages that accumulated in backlog queue */ 581 release_sock(sk); 582 tipc_dest_list_purge(&tsk->cong_links); 583 tsk->cong_link_cnt = 0; 584 call_rcu(&tsk->rcu, tipc_sk_callback); 585 sock->sk = NULL; 586 587 return 0; 588 } 589 590 /** 591 * tipc_bind - associate or disassocate TIPC name(s) with a socket 592 * @sock: socket structure 593 * @uaddr: socket address describing name(s) and desired operation 594 * @uaddr_len: size of socket address data structure 595 * 596 * Name and name sequence binding is indicated using a positive scope value; 597 * a negative scope value unbinds the specified name. Specifying no name 598 * (i.e. a socket address length of 0) unbinds all names from the socket. 599 * 600 * Returns 0 on success, errno otherwise 601 * 602 * NOTE: This routine doesn't need to take the socket lock since it doesn't 603 * access any non-constant socket information. 604 */ 605 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr, 606 int uaddr_len) 607 { 608 struct sock *sk = sock->sk; 609 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr; 610 struct tipc_sock *tsk = tipc_sk(sk); 611 int res = -EINVAL; 612 613 lock_sock(sk); 614 if (unlikely(!uaddr_len)) { 615 res = tipc_sk_withdraw(tsk, 0, NULL); 616 goto exit; 617 } 618 if (tsk->group) { 619 res = -EACCES; 620 goto exit; 621 } 622 if (uaddr_len < sizeof(struct sockaddr_tipc)) { 623 res = -EINVAL; 624 goto exit; 625 } 626 if (addr->family != AF_TIPC) { 627 res = -EAFNOSUPPORT; 628 goto exit; 629 } 630 631 if (addr->addrtype == TIPC_ADDR_NAME) 632 addr->addr.nameseq.upper = addr->addr.nameseq.lower; 633 else if (addr->addrtype != TIPC_ADDR_NAMESEQ) { 634 res = -EAFNOSUPPORT; 635 goto exit; 636 } 637 638 if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) && 639 (addr->addr.nameseq.type != TIPC_TOP_SRV) && 640 (addr->addr.nameseq.type != TIPC_CFG_SRV)) { 641 res = -EACCES; 642 goto exit; 643 } 644 645 res = (addr->scope >= 0) ? 646 tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) : 647 tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq); 648 exit: 649 release_sock(sk); 650 return res; 651 } 652 653 /** 654 * tipc_getname - get port ID of socket or peer socket 655 * @sock: socket structure 656 * @uaddr: area for returned socket address 657 * @uaddr_len: area for returned length of socket address 658 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID 659 * 660 * Returns 0 on success, errno otherwise 661 * 662 * NOTE: This routine doesn't need to take the socket lock since it only 663 * accesses socket information that is unchanging (or which changes in 664 * a completely predictable manner). 665 */ 666 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr, 667 int peer) 668 { 669 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr; 670 struct sock *sk = sock->sk; 671 struct tipc_sock *tsk = tipc_sk(sk); 672 673 memset(addr, 0, sizeof(*addr)); 674 if (peer) { 675 if ((!tipc_sk_connected(sk)) && 676 ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING))) 677 return -ENOTCONN; 678 addr->addr.id.ref = tsk_peer_port(tsk); 679 addr->addr.id.node = tsk_peer_node(tsk); 680 } else { 681 addr->addr.id.ref = tsk->portid; 682 addr->addr.id.node = tipc_own_addr(sock_net(sk)); 683 } 684 685 addr->addrtype = TIPC_ADDR_ID; 686 addr->family = AF_TIPC; 687 addr->scope = 0; 688 addr->addr.name.domain = 0; 689 690 return sizeof(*addr); 691 } 692 693 /** 694 * tipc_poll - read and possibly block on pollmask 695 * @file: file structure associated with the socket 696 * @sock: socket for which to calculate the poll bits 697 * @wait: ??? 698 * 699 * Returns pollmask value 700 * 701 * COMMENTARY: 702 * It appears that the usual socket locking mechanisms are not useful here 703 * since the pollmask info is potentially out-of-date the moment this routine 704 * exits. TCP and other protocols seem to rely on higher level poll routines 705 * to handle any preventable race conditions, so TIPC will do the same ... 706 * 707 * IMPORTANT: The fact that a read or write operation is indicated does NOT 708 * imply that the operation will succeed, merely that it should be performed 709 * and will not block. 710 */ 711 static __poll_t tipc_poll(struct file *file, struct socket *sock, 712 poll_table *wait) 713 { 714 struct sock *sk = sock->sk; 715 struct tipc_sock *tsk = tipc_sk(sk); 716 __poll_t revents = 0; 717 718 sock_poll_wait(file, wait); 719 720 if (sk->sk_shutdown & RCV_SHUTDOWN) 721 revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM; 722 if (sk->sk_shutdown == SHUTDOWN_MASK) 723 revents |= EPOLLHUP; 724 725 switch (sk->sk_state) { 726 case TIPC_ESTABLISHED: 727 case TIPC_CONNECTING: 728 if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk)) 729 revents |= EPOLLOUT; 730 /* fall thru' */ 731 case TIPC_LISTEN: 732 if (!skb_queue_empty(&sk->sk_receive_queue)) 733 revents |= EPOLLIN | EPOLLRDNORM; 734 break; 735 case TIPC_OPEN: 736 if (tsk->group_is_open && !tsk->cong_link_cnt) 737 revents |= EPOLLOUT; 738 if (!tipc_sk_type_connectionless(sk)) 739 break; 740 if (skb_queue_empty(&sk->sk_receive_queue)) 741 break; 742 revents |= EPOLLIN | EPOLLRDNORM; 743 break; 744 case TIPC_DISCONNECTING: 745 revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP; 746 break; 747 } 748 return revents; 749 } 750 751 /** 752 * tipc_sendmcast - send multicast message 753 * @sock: socket structure 754 * @seq: destination address 755 * @msg: message to send 756 * @dlen: length of data to send 757 * @timeout: timeout to wait for wakeup 758 * 759 * Called from function tipc_sendmsg(), which has done all sanity checks 760 * Returns the number of bytes sent on success, or errno 761 */ 762 static int tipc_sendmcast(struct socket *sock, struct tipc_name_seq *seq, 763 struct msghdr *msg, size_t dlen, long timeout) 764 { 765 struct sock *sk = sock->sk; 766 struct tipc_sock *tsk = tipc_sk(sk); 767 struct tipc_msg *hdr = &tsk->phdr; 768 struct net *net = sock_net(sk); 769 int mtu = tipc_bcast_get_mtu(net); 770 struct tipc_mc_method *method = &tsk->mc_method; 771 struct sk_buff_head pkts; 772 struct tipc_nlist dsts; 773 int rc; 774 775 if (tsk->group) 776 return -EACCES; 777 778 /* Block or return if any destination link is congested */ 779 rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt); 780 if (unlikely(rc)) 781 return rc; 782 783 /* Lookup destination nodes */ 784 tipc_nlist_init(&dsts, tipc_own_addr(net)); 785 tipc_nametbl_lookup_dst_nodes(net, seq->type, seq->lower, 786 seq->upper, &dsts); 787 if (!dsts.local && !dsts.remote) 788 return -EHOSTUNREACH; 789 790 /* Build message header */ 791 msg_set_type(hdr, TIPC_MCAST_MSG); 792 msg_set_hdr_sz(hdr, MCAST_H_SIZE); 793 msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE); 794 msg_set_destport(hdr, 0); 795 msg_set_destnode(hdr, 0); 796 msg_set_nametype(hdr, seq->type); 797 msg_set_namelower(hdr, seq->lower); 798 msg_set_nameupper(hdr, seq->upper); 799 800 /* Build message as chain of buffers */ 801 skb_queue_head_init(&pkts); 802 rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts); 803 804 /* Send message if build was successful */ 805 if (unlikely(rc == dlen)) 806 rc = tipc_mcast_xmit(net, &pkts, method, &dsts, 807 &tsk->cong_link_cnt); 808 809 tipc_nlist_purge(&dsts); 810 811 return rc ? rc : dlen; 812 } 813 814 /** 815 * tipc_send_group_msg - send a message to a member in the group 816 * @net: network namespace 817 * @m: message to send 818 * @mb: group member 819 * @dnode: destination node 820 * @dport: destination port 821 * @dlen: total length of message data 822 */ 823 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk, 824 struct msghdr *m, struct tipc_member *mb, 825 u32 dnode, u32 dport, int dlen) 826 { 827 u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group); 828 struct tipc_mc_method *method = &tsk->mc_method; 829 int blks = tsk_blocks(GROUP_H_SIZE + dlen); 830 struct tipc_msg *hdr = &tsk->phdr; 831 struct sk_buff_head pkts; 832 int mtu, rc; 833 834 /* Complete message header */ 835 msg_set_type(hdr, TIPC_GRP_UCAST_MSG); 836 msg_set_hdr_sz(hdr, GROUP_H_SIZE); 837 msg_set_destport(hdr, dport); 838 msg_set_destnode(hdr, dnode); 839 msg_set_grp_bc_seqno(hdr, bc_snd_nxt); 840 841 /* Build message as chain of buffers */ 842 skb_queue_head_init(&pkts); 843 mtu = tipc_node_get_mtu(net, dnode, tsk->portid); 844 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts); 845 if (unlikely(rc != dlen)) 846 return rc; 847 848 /* Send message */ 849 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid); 850 if (unlikely(rc == -ELINKCONG)) { 851 tipc_dest_push(&tsk->cong_links, dnode, 0); 852 tsk->cong_link_cnt++; 853 } 854 855 /* Update send window */ 856 tipc_group_update_member(mb, blks); 857 858 /* A broadcast sent within next EXPIRE period must follow same path */ 859 method->rcast = true; 860 method->mandatory = true; 861 return dlen; 862 } 863 864 /** 865 * tipc_send_group_unicast - send message to a member in the group 866 * @sock: socket structure 867 * @m: message to send 868 * @dlen: total length of message data 869 * @timeout: timeout to wait for wakeup 870 * 871 * Called from function tipc_sendmsg(), which has done all sanity checks 872 * Returns the number of bytes sent on success, or errno 873 */ 874 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m, 875 int dlen, long timeout) 876 { 877 struct sock *sk = sock->sk; 878 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name); 879 int blks = tsk_blocks(GROUP_H_SIZE + dlen); 880 struct tipc_sock *tsk = tipc_sk(sk); 881 struct tipc_group *grp = tsk->group; 882 struct net *net = sock_net(sk); 883 struct tipc_member *mb = NULL; 884 u32 node, port; 885 int rc; 886 887 node = dest->addr.id.node; 888 port = dest->addr.id.ref; 889 if (!port && !node) 890 return -EHOSTUNREACH; 891 892 /* Block or return if destination link or member is congested */ 893 rc = tipc_wait_for_cond(sock, &timeout, 894 !tipc_dest_find(&tsk->cong_links, node, 0) && 895 !tipc_group_cong(grp, node, port, blks, &mb)); 896 if (unlikely(rc)) 897 return rc; 898 899 if (unlikely(!mb)) 900 return -EHOSTUNREACH; 901 902 rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen); 903 904 return rc ? rc : dlen; 905 } 906 907 /** 908 * tipc_send_group_anycast - send message to any member with given identity 909 * @sock: socket structure 910 * @m: message to send 911 * @dlen: total length of message data 912 * @timeout: timeout to wait for wakeup 913 * 914 * Called from function tipc_sendmsg(), which has done all sanity checks 915 * Returns the number of bytes sent on success, or errno 916 */ 917 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m, 918 int dlen, long timeout) 919 { 920 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name); 921 struct sock *sk = sock->sk; 922 struct tipc_sock *tsk = tipc_sk(sk); 923 struct list_head *cong_links = &tsk->cong_links; 924 int blks = tsk_blocks(GROUP_H_SIZE + dlen); 925 struct tipc_group *grp = tsk->group; 926 struct tipc_msg *hdr = &tsk->phdr; 927 struct tipc_member *first = NULL; 928 struct tipc_member *mbr = NULL; 929 struct net *net = sock_net(sk); 930 u32 node, port, exclude; 931 struct list_head dsts; 932 u32 type, inst, scope; 933 int lookups = 0; 934 int dstcnt, rc; 935 bool cong; 936 937 INIT_LIST_HEAD(&dsts); 938 939 type = msg_nametype(hdr); 940 inst = dest->addr.name.name.instance; 941 scope = msg_lookup_scope(hdr); 942 exclude = tipc_group_exclude(grp); 943 944 while (++lookups < 4) { 945 first = NULL; 946 947 /* Look for a non-congested destination member, if any */ 948 while (1) { 949 if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts, 950 &dstcnt, exclude, false)) 951 return -EHOSTUNREACH; 952 tipc_dest_pop(&dsts, &node, &port); 953 cong = tipc_group_cong(grp, node, port, blks, &mbr); 954 if (!cong) 955 break; 956 if (mbr == first) 957 break; 958 if (!first) 959 first = mbr; 960 } 961 962 /* Start over if destination was not in member list */ 963 if (unlikely(!mbr)) 964 continue; 965 966 if (likely(!cong && !tipc_dest_find(cong_links, node, 0))) 967 break; 968 969 /* Block or return if destination link or member is congested */ 970 rc = tipc_wait_for_cond(sock, &timeout, 971 !tipc_dest_find(cong_links, node, 0) && 972 !tipc_group_cong(grp, node, port, 973 blks, &mbr)); 974 if (unlikely(rc)) 975 return rc; 976 977 /* Send, unless destination disappeared while waiting */ 978 if (likely(mbr)) 979 break; 980 } 981 982 if (unlikely(lookups >= 4)) 983 return -EHOSTUNREACH; 984 985 rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen); 986 987 return rc ? rc : dlen; 988 } 989 990 /** 991 * tipc_send_group_bcast - send message to all members in communication group 992 * @sk: socket structure 993 * @m: message to send 994 * @dlen: total length of message data 995 * @timeout: timeout to wait for wakeup 996 * 997 * Called from function tipc_sendmsg(), which has done all sanity checks 998 * Returns the number of bytes sent on success, or errno 999 */ 1000 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m, 1001 int dlen, long timeout) 1002 { 1003 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name); 1004 struct sock *sk = sock->sk; 1005 struct net *net = sock_net(sk); 1006 struct tipc_sock *tsk = tipc_sk(sk); 1007 struct tipc_group *grp = tsk->group; 1008 struct tipc_nlist *dsts = tipc_group_dests(grp); 1009 struct tipc_mc_method *method = &tsk->mc_method; 1010 bool ack = method->mandatory && method->rcast; 1011 int blks = tsk_blocks(MCAST_H_SIZE + dlen); 1012 struct tipc_msg *hdr = &tsk->phdr; 1013 int mtu = tipc_bcast_get_mtu(net); 1014 struct sk_buff_head pkts; 1015 int rc = -EHOSTUNREACH; 1016 1017 if (!dsts->local && !dsts->remote) 1018 return -EHOSTUNREACH; 1019 1020 /* Block or return if any destination link or member is congested */ 1021 rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt && 1022 !tipc_group_bc_cong(grp, blks)); 1023 if (unlikely(rc)) 1024 return rc; 1025 1026 /* Complete message header */ 1027 if (dest) { 1028 msg_set_type(hdr, TIPC_GRP_MCAST_MSG); 1029 msg_set_nameinst(hdr, dest->addr.name.name.instance); 1030 } else { 1031 msg_set_type(hdr, TIPC_GRP_BCAST_MSG); 1032 msg_set_nameinst(hdr, 0); 1033 } 1034 msg_set_hdr_sz(hdr, GROUP_H_SIZE); 1035 msg_set_destport(hdr, 0); 1036 msg_set_destnode(hdr, 0); 1037 msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(grp)); 1038 1039 /* Avoid getting stuck with repeated forced replicasts */ 1040 msg_set_grp_bc_ack_req(hdr, ack); 1041 1042 /* Build message as chain of buffers */ 1043 skb_queue_head_init(&pkts); 1044 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts); 1045 if (unlikely(rc != dlen)) 1046 return rc; 1047 1048 /* Send message */ 1049 rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt); 1050 if (unlikely(rc)) 1051 return rc; 1052 1053 /* Update broadcast sequence number and send windows */ 1054 tipc_group_update_bc_members(tsk->group, blks, ack); 1055 1056 /* Broadcast link is now free to choose method for next broadcast */ 1057 method->mandatory = false; 1058 method->expires = jiffies; 1059 1060 return dlen; 1061 } 1062 1063 /** 1064 * tipc_send_group_mcast - send message to all members with given identity 1065 * @sock: socket structure 1066 * @m: message to send 1067 * @dlen: total length of message data 1068 * @timeout: timeout to wait for wakeup 1069 * 1070 * Called from function tipc_sendmsg(), which has done all sanity checks 1071 * Returns the number of bytes sent on success, or errno 1072 */ 1073 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m, 1074 int dlen, long timeout) 1075 { 1076 struct sock *sk = sock->sk; 1077 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name); 1078 struct tipc_sock *tsk = tipc_sk(sk); 1079 struct tipc_group *grp = tsk->group; 1080 struct tipc_msg *hdr = &tsk->phdr; 1081 struct net *net = sock_net(sk); 1082 u32 type, inst, scope, exclude; 1083 struct list_head dsts; 1084 u32 dstcnt; 1085 1086 INIT_LIST_HEAD(&dsts); 1087 1088 type = msg_nametype(hdr); 1089 inst = dest->addr.name.name.instance; 1090 scope = msg_lookup_scope(hdr); 1091 exclude = tipc_group_exclude(grp); 1092 1093 if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts, 1094 &dstcnt, exclude, true)) 1095 return -EHOSTUNREACH; 1096 1097 if (dstcnt == 1) { 1098 tipc_dest_pop(&dsts, &dest->addr.id.node, &dest->addr.id.ref); 1099 return tipc_send_group_unicast(sock, m, dlen, timeout); 1100 } 1101 1102 tipc_dest_list_purge(&dsts); 1103 return tipc_send_group_bcast(sock, m, dlen, timeout); 1104 } 1105 1106 /** 1107 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets 1108 * @arrvq: queue with arriving messages, to be cloned after destination lookup 1109 * @inputq: queue with cloned messages, delivered to socket after dest lookup 1110 * 1111 * Multi-threaded: parallel calls with reference to same queues may occur 1112 */ 1113 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq, 1114 struct sk_buff_head *inputq) 1115 { 1116 u32 self = tipc_own_addr(net); 1117 u32 type, lower, upper, scope; 1118 struct sk_buff *skb, *_skb; 1119 u32 portid, onode; 1120 struct sk_buff_head tmpq; 1121 struct list_head dports; 1122 struct tipc_msg *hdr; 1123 int user, mtyp, hlen; 1124 bool exact; 1125 1126 __skb_queue_head_init(&tmpq); 1127 INIT_LIST_HEAD(&dports); 1128 1129 skb = tipc_skb_peek(arrvq, &inputq->lock); 1130 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) { 1131 hdr = buf_msg(skb); 1132 user = msg_user(hdr); 1133 mtyp = msg_type(hdr); 1134 hlen = skb_headroom(skb) + msg_hdr_sz(hdr); 1135 onode = msg_orignode(hdr); 1136 type = msg_nametype(hdr); 1137 1138 if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) { 1139 spin_lock_bh(&inputq->lock); 1140 if (skb_peek(arrvq) == skb) { 1141 __skb_dequeue(arrvq); 1142 __skb_queue_tail(inputq, skb); 1143 } 1144 kfree_skb(skb); 1145 spin_unlock_bh(&inputq->lock); 1146 continue; 1147 } 1148 1149 /* Group messages require exact scope match */ 1150 if (msg_in_group(hdr)) { 1151 lower = 0; 1152 upper = ~0; 1153 scope = msg_lookup_scope(hdr); 1154 exact = true; 1155 } else { 1156 /* TIPC_NODE_SCOPE means "any scope" in this context */ 1157 if (onode == self) 1158 scope = TIPC_NODE_SCOPE; 1159 else 1160 scope = TIPC_CLUSTER_SCOPE; 1161 exact = false; 1162 lower = msg_namelower(hdr); 1163 upper = msg_nameupper(hdr); 1164 } 1165 1166 /* Create destination port list: */ 1167 tipc_nametbl_mc_lookup(net, type, lower, upper, 1168 scope, exact, &dports); 1169 1170 /* Clone message per destination */ 1171 while (tipc_dest_pop(&dports, NULL, &portid)) { 1172 _skb = __pskb_copy(skb, hlen, GFP_ATOMIC); 1173 if (_skb) { 1174 msg_set_destport(buf_msg(_skb), portid); 1175 __skb_queue_tail(&tmpq, _skb); 1176 continue; 1177 } 1178 pr_warn("Failed to clone mcast rcv buffer\n"); 1179 } 1180 /* Append to inputq if not already done by other thread */ 1181 spin_lock_bh(&inputq->lock); 1182 if (skb_peek(arrvq) == skb) { 1183 skb_queue_splice_tail_init(&tmpq, inputq); 1184 kfree_skb(__skb_dequeue(arrvq)); 1185 } 1186 spin_unlock_bh(&inputq->lock); 1187 __skb_queue_purge(&tmpq); 1188 kfree_skb(skb); 1189 } 1190 tipc_sk_rcv(net, inputq); 1191 } 1192 1193 /** 1194 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message 1195 * @tsk: receiving socket 1196 * @skb: pointer to message buffer. 1197 */ 1198 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb, 1199 struct sk_buff_head *inputq, 1200 struct sk_buff_head *xmitq) 1201 { 1202 struct tipc_msg *hdr = buf_msg(skb); 1203 u32 onode = tsk_own_node(tsk); 1204 struct sock *sk = &tsk->sk; 1205 int mtyp = msg_type(hdr); 1206 bool conn_cong; 1207 1208 /* Ignore if connection cannot be validated: */ 1209 if (!tsk_peer_msg(tsk, hdr)) 1210 goto exit; 1211 1212 if (unlikely(msg_errcode(hdr))) { 1213 tipc_set_sk_state(sk, TIPC_DISCONNECTING); 1214 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk), 1215 tsk_peer_port(tsk)); 1216 sk->sk_state_change(sk); 1217 1218 /* State change is ignored if socket already awake, 1219 * - convert msg to abort msg and add to inqueue 1220 */ 1221 msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE); 1222 msg_set_type(hdr, TIPC_CONN_MSG); 1223 msg_set_size(hdr, BASIC_H_SIZE); 1224 msg_set_hdr_sz(hdr, BASIC_H_SIZE); 1225 __skb_queue_tail(inputq, skb); 1226 return; 1227 } 1228 1229 tsk->probe_unacked = false; 1230 1231 if (mtyp == CONN_PROBE) { 1232 msg_set_type(hdr, CONN_PROBE_REPLY); 1233 if (tipc_msg_reverse(onode, &skb, TIPC_OK)) 1234 __skb_queue_tail(xmitq, skb); 1235 return; 1236 } else if (mtyp == CONN_ACK) { 1237 conn_cong = tsk_conn_cong(tsk); 1238 tsk->snt_unacked -= msg_conn_ack(hdr); 1239 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) 1240 tsk->snd_win = msg_adv_win(hdr); 1241 if (conn_cong) 1242 sk->sk_write_space(sk); 1243 } else if (mtyp != CONN_PROBE_REPLY) { 1244 pr_warn("Received unknown CONN_PROTO msg\n"); 1245 } 1246 exit: 1247 kfree_skb(skb); 1248 } 1249 1250 /** 1251 * tipc_sendmsg - send message in connectionless manner 1252 * @sock: socket structure 1253 * @m: message to send 1254 * @dsz: amount of user data to be sent 1255 * 1256 * Message must have an destination specified explicitly. 1257 * Used for SOCK_RDM and SOCK_DGRAM messages, 1258 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections. 1259 * (Note: 'SYN+' is prohibited on SOCK_STREAM.) 1260 * 1261 * Returns the number of bytes sent on success, or errno otherwise 1262 */ 1263 static int tipc_sendmsg(struct socket *sock, 1264 struct msghdr *m, size_t dsz) 1265 { 1266 struct sock *sk = sock->sk; 1267 int ret; 1268 1269 lock_sock(sk); 1270 ret = __tipc_sendmsg(sock, m, dsz); 1271 release_sock(sk); 1272 1273 return ret; 1274 } 1275 1276 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen) 1277 { 1278 struct sock *sk = sock->sk; 1279 struct net *net = sock_net(sk); 1280 struct tipc_sock *tsk = tipc_sk(sk); 1281 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name); 1282 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT); 1283 struct list_head *clinks = &tsk->cong_links; 1284 bool syn = !tipc_sk_type_connectionless(sk); 1285 struct tipc_group *grp = tsk->group; 1286 struct tipc_msg *hdr = &tsk->phdr; 1287 struct tipc_name_seq *seq; 1288 struct sk_buff_head pkts; 1289 u32 dport, dnode = 0; 1290 u32 type, inst; 1291 int mtu, rc; 1292 1293 if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE)) 1294 return -EMSGSIZE; 1295 1296 if (likely(dest)) { 1297 if (unlikely(m->msg_namelen < sizeof(*dest))) 1298 return -EINVAL; 1299 if (unlikely(dest->family != AF_TIPC)) 1300 return -EINVAL; 1301 } 1302 1303 if (grp) { 1304 if (!dest) 1305 return tipc_send_group_bcast(sock, m, dlen, timeout); 1306 if (dest->addrtype == TIPC_ADDR_NAME) 1307 return tipc_send_group_anycast(sock, m, dlen, timeout); 1308 if (dest->addrtype == TIPC_ADDR_ID) 1309 return tipc_send_group_unicast(sock, m, dlen, timeout); 1310 if (dest->addrtype == TIPC_ADDR_MCAST) 1311 return tipc_send_group_mcast(sock, m, dlen, timeout); 1312 return -EINVAL; 1313 } 1314 1315 if (unlikely(!dest)) { 1316 dest = &tsk->peer; 1317 if (!syn || dest->family != AF_TIPC) 1318 return -EDESTADDRREQ; 1319 } 1320 1321 if (unlikely(syn)) { 1322 if (sk->sk_state == TIPC_LISTEN) 1323 return -EPIPE; 1324 if (sk->sk_state != TIPC_OPEN) 1325 return -EISCONN; 1326 if (tsk->published) 1327 return -EOPNOTSUPP; 1328 if (dest->addrtype == TIPC_ADDR_NAME) { 1329 tsk->conn_type = dest->addr.name.name.type; 1330 tsk->conn_instance = dest->addr.name.name.instance; 1331 } 1332 } 1333 1334 seq = &dest->addr.nameseq; 1335 if (dest->addrtype == TIPC_ADDR_MCAST) 1336 return tipc_sendmcast(sock, seq, m, dlen, timeout); 1337 1338 if (dest->addrtype == TIPC_ADDR_NAME) { 1339 type = dest->addr.name.name.type; 1340 inst = dest->addr.name.name.instance; 1341 dnode = dest->addr.name.domain; 1342 msg_set_type(hdr, TIPC_NAMED_MSG); 1343 msg_set_hdr_sz(hdr, NAMED_H_SIZE); 1344 msg_set_nametype(hdr, type); 1345 msg_set_nameinst(hdr, inst); 1346 msg_set_lookup_scope(hdr, tipc_node2scope(dnode)); 1347 dport = tipc_nametbl_translate(net, type, inst, &dnode); 1348 msg_set_destnode(hdr, dnode); 1349 msg_set_destport(hdr, dport); 1350 if (unlikely(!dport && !dnode)) 1351 return -EHOSTUNREACH; 1352 } else if (dest->addrtype == TIPC_ADDR_ID) { 1353 dnode = dest->addr.id.node; 1354 msg_set_type(hdr, TIPC_DIRECT_MSG); 1355 msg_set_lookup_scope(hdr, 0); 1356 msg_set_destnode(hdr, dnode); 1357 msg_set_destport(hdr, dest->addr.id.ref); 1358 msg_set_hdr_sz(hdr, BASIC_H_SIZE); 1359 } else { 1360 return -EINVAL; 1361 } 1362 1363 /* Block or return if destination link is congested */ 1364 rc = tipc_wait_for_cond(sock, &timeout, 1365 !tipc_dest_find(clinks, dnode, 0)); 1366 if (unlikely(rc)) 1367 return rc; 1368 1369 skb_queue_head_init(&pkts); 1370 mtu = tipc_node_get_mtu(net, dnode, tsk->portid); 1371 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts); 1372 if (unlikely(rc != dlen)) 1373 return rc; 1374 1375 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid); 1376 if (unlikely(rc == -ELINKCONG)) { 1377 tipc_dest_push(clinks, dnode, 0); 1378 tsk->cong_link_cnt++; 1379 rc = 0; 1380 } 1381 1382 if (unlikely(syn && !rc)) 1383 tipc_set_sk_state(sk, TIPC_CONNECTING); 1384 1385 return rc ? rc : dlen; 1386 } 1387 1388 /** 1389 * tipc_sendstream - send stream-oriented data 1390 * @sock: socket structure 1391 * @m: data to send 1392 * @dsz: total length of data to be transmitted 1393 * 1394 * Used for SOCK_STREAM data. 1395 * 1396 * Returns the number of bytes sent on success (or partial success), 1397 * or errno if no data sent 1398 */ 1399 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz) 1400 { 1401 struct sock *sk = sock->sk; 1402 int ret; 1403 1404 lock_sock(sk); 1405 ret = __tipc_sendstream(sock, m, dsz); 1406 release_sock(sk); 1407 1408 return ret; 1409 } 1410 1411 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen) 1412 { 1413 struct sock *sk = sock->sk; 1414 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name); 1415 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT); 1416 struct tipc_sock *tsk = tipc_sk(sk); 1417 struct tipc_msg *hdr = &tsk->phdr; 1418 struct net *net = sock_net(sk); 1419 struct sk_buff_head pkts; 1420 u32 dnode = tsk_peer_node(tsk); 1421 int send, sent = 0; 1422 int rc = 0; 1423 1424 skb_queue_head_init(&pkts); 1425 1426 if (unlikely(dlen > INT_MAX)) 1427 return -EMSGSIZE; 1428 1429 /* Handle implicit connection setup */ 1430 if (unlikely(dest)) { 1431 rc = __tipc_sendmsg(sock, m, dlen); 1432 if (dlen && dlen == rc) { 1433 tsk->peer_caps = tipc_node_get_capabilities(net, dnode); 1434 tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr)); 1435 } 1436 return rc; 1437 } 1438 1439 do { 1440 rc = tipc_wait_for_cond(sock, &timeout, 1441 (!tsk->cong_link_cnt && 1442 !tsk_conn_cong(tsk) && 1443 tipc_sk_connected(sk))); 1444 if (unlikely(rc)) 1445 break; 1446 1447 send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE); 1448 rc = tipc_msg_build(hdr, m, sent, send, tsk->max_pkt, &pkts); 1449 if (unlikely(rc != send)) 1450 break; 1451 1452 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid); 1453 if (unlikely(rc == -ELINKCONG)) { 1454 tsk->cong_link_cnt = 1; 1455 rc = 0; 1456 } 1457 if (likely(!rc)) { 1458 tsk->snt_unacked += tsk_inc(tsk, send + MIN_H_SIZE); 1459 sent += send; 1460 } 1461 } while (sent < dlen && !rc); 1462 1463 return sent ? sent : rc; 1464 } 1465 1466 /** 1467 * tipc_send_packet - send a connection-oriented message 1468 * @sock: socket structure 1469 * @m: message to send 1470 * @dsz: length of data to be transmitted 1471 * 1472 * Used for SOCK_SEQPACKET messages. 1473 * 1474 * Returns the number of bytes sent on success, or errno otherwise 1475 */ 1476 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz) 1477 { 1478 if (dsz > TIPC_MAX_USER_MSG_SIZE) 1479 return -EMSGSIZE; 1480 1481 return tipc_sendstream(sock, m, dsz); 1482 } 1483 1484 /* tipc_sk_finish_conn - complete the setup of a connection 1485 */ 1486 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port, 1487 u32 peer_node) 1488 { 1489 struct sock *sk = &tsk->sk; 1490 struct net *net = sock_net(sk); 1491 struct tipc_msg *msg = &tsk->phdr; 1492 1493 msg_set_destnode(msg, peer_node); 1494 msg_set_destport(msg, peer_port); 1495 msg_set_type(msg, TIPC_CONN_MSG); 1496 msg_set_lookup_scope(msg, 0); 1497 msg_set_hdr_sz(msg, SHORT_H_SIZE); 1498 1499 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV); 1500 tipc_set_sk_state(sk, TIPC_ESTABLISHED); 1501 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port); 1502 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid); 1503 tsk->peer_caps = tipc_node_get_capabilities(net, peer_node); 1504 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) 1505 return; 1506 1507 /* Fall back to message based flow control */ 1508 tsk->rcv_win = FLOWCTL_MSG_WIN; 1509 tsk->snd_win = FLOWCTL_MSG_WIN; 1510 } 1511 1512 /** 1513 * tipc_sk_set_orig_addr - capture sender's address for received message 1514 * @m: descriptor for message info 1515 * @hdr: received message header 1516 * 1517 * Note: Address is not captured if not requested by receiver. 1518 */ 1519 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb) 1520 { 1521 DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name); 1522 struct tipc_msg *hdr = buf_msg(skb); 1523 1524 if (!srcaddr) 1525 return; 1526 1527 srcaddr->sock.family = AF_TIPC; 1528 srcaddr->sock.addrtype = TIPC_ADDR_ID; 1529 srcaddr->sock.scope = 0; 1530 srcaddr->sock.addr.id.ref = msg_origport(hdr); 1531 srcaddr->sock.addr.id.node = msg_orignode(hdr); 1532 srcaddr->sock.addr.name.domain = 0; 1533 m->msg_namelen = sizeof(struct sockaddr_tipc); 1534 1535 if (!msg_in_group(hdr)) 1536 return; 1537 1538 /* Group message users may also want to know sending member's id */ 1539 srcaddr->member.family = AF_TIPC; 1540 srcaddr->member.addrtype = TIPC_ADDR_NAME; 1541 srcaddr->member.scope = 0; 1542 srcaddr->member.addr.name.name.type = msg_nametype(hdr); 1543 srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member; 1544 srcaddr->member.addr.name.domain = 0; 1545 m->msg_namelen = sizeof(*srcaddr); 1546 } 1547 1548 /** 1549 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message 1550 * @m: descriptor for message info 1551 * @msg: received message header 1552 * @tsk: TIPC port associated with message 1553 * 1554 * Note: Ancillary data is not captured if not requested by receiver. 1555 * 1556 * Returns 0 if successful, otherwise errno 1557 */ 1558 static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg, 1559 struct tipc_sock *tsk) 1560 { 1561 u32 anc_data[3]; 1562 u32 err; 1563 u32 dest_type; 1564 int has_name; 1565 int res; 1566 1567 if (likely(m->msg_controllen == 0)) 1568 return 0; 1569 1570 /* Optionally capture errored message object(s) */ 1571 err = msg ? msg_errcode(msg) : 0; 1572 if (unlikely(err)) { 1573 anc_data[0] = err; 1574 anc_data[1] = msg_data_sz(msg); 1575 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data); 1576 if (res) 1577 return res; 1578 if (anc_data[1]) { 1579 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1], 1580 msg_data(msg)); 1581 if (res) 1582 return res; 1583 } 1584 } 1585 1586 /* Optionally capture message destination object */ 1587 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG; 1588 switch (dest_type) { 1589 case TIPC_NAMED_MSG: 1590 has_name = 1; 1591 anc_data[0] = msg_nametype(msg); 1592 anc_data[1] = msg_namelower(msg); 1593 anc_data[2] = msg_namelower(msg); 1594 break; 1595 case TIPC_MCAST_MSG: 1596 has_name = 1; 1597 anc_data[0] = msg_nametype(msg); 1598 anc_data[1] = msg_namelower(msg); 1599 anc_data[2] = msg_nameupper(msg); 1600 break; 1601 case TIPC_CONN_MSG: 1602 has_name = (tsk->conn_type != 0); 1603 anc_data[0] = tsk->conn_type; 1604 anc_data[1] = tsk->conn_instance; 1605 anc_data[2] = tsk->conn_instance; 1606 break; 1607 default: 1608 has_name = 0; 1609 } 1610 if (has_name) { 1611 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data); 1612 if (res) 1613 return res; 1614 } 1615 1616 return 0; 1617 } 1618 1619 static void tipc_sk_send_ack(struct tipc_sock *tsk) 1620 { 1621 struct sock *sk = &tsk->sk; 1622 struct net *net = sock_net(sk); 1623 struct sk_buff *skb = NULL; 1624 struct tipc_msg *msg; 1625 u32 peer_port = tsk_peer_port(tsk); 1626 u32 dnode = tsk_peer_node(tsk); 1627 1628 if (!tipc_sk_connected(sk)) 1629 return; 1630 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0, 1631 dnode, tsk_own_node(tsk), peer_port, 1632 tsk->portid, TIPC_OK); 1633 if (!skb) 1634 return; 1635 msg = buf_msg(skb); 1636 msg_set_conn_ack(msg, tsk->rcv_unacked); 1637 tsk->rcv_unacked = 0; 1638 1639 /* Adjust to and advertize the correct window limit */ 1640 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) { 1641 tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf); 1642 msg_set_adv_win(msg, tsk->rcv_win); 1643 } 1644 tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg)); 1645 } 1646 1647 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop) 1648 { 1649 struct sock *sk = sock->sk; 1650 DEFINE_WAIT(wait); 1651 long timeo = *timeop; 1652 int err = sock_error(sk); 1653 1654 if (err) 1655 return err; 1656 1657 for (;;) { 1658 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); 1659 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) { 1660 if (sk->sk_shutdown & RCV_SHUTDOWN) { 1661 err = -ENOTCONN; 1662 break; 1663 } 1664 release_sock(sk); 1665 timeo = schedule_timeout(timeo); 1666 lock_sock(sk); 1667 } 1668 err = 0; 1669 if (!skb_queue_empty(&sk->sk_receive_queue)) 1670 break; 1671 err = -EAGAIN; 1672 if (!timeo) 1673 break; 1674 err = sock_intr_errno(timeo); 1675 if (signal_pending(current)) 1676 break; 1677 1678 err = sock_error(sk); 1679 if (err) 1680 break; 1681 } 1682 finish_wait(sk_sleep(sk), &wait); 1683 *timeop = timeo; 1684 return err; 1685 } 1686 1687 /** 1688 * tipc_recvmsg - receive packet-oriented message 1689 * @m: descriptor for message info 1690 * @buflen: length of user buffer area 1691 * @flags: receive flags 1692 * 1693 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages. 1694 * If the complete message doesn't fit in user area, truncate it. 1695 * 1696 * Returns size of returned message data, errno otherwise 1697 */ 1698 static int tipc_recvmsg(struct socket *sock, struct msghdr *m, 1699 size_t buflen, int flags) 1700 { 1701 struct sock *sk = sock->sk; 1702 bool connected = !tipc_sk_type_connectionless(sk); 1703 struct tipc_sock *tsk = tipc_sk(sk); 1704 int rc, err, hlen, dlen, copy; 1705 struct sk_buff_head xmitq; 1706 struct tipc_msg *hdr; 1707 struct sk_buff *skb; 1708 bool grp_evt; 1709 long timeout; 1710 1711 /* Catch invalid receive requests */ 1712 if (unlikely(!buflen)) 1713 return -EINVAL; 1714 1715 lock_sock(sk); 1716 if (unlikely(connected && sk->sk_state == TIPC_OPEN)) { 1717 rc = -ENOTCONN; 1718 goto exit; 1719 } 1720 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT); 1721 1722 /* Step rcv queue to first msg with data or error; wait if necessary */ 1723 do { 1724 rc = tipc_wait_for_rcvmsg(sock, &timeout); 1725 if (unlikely(rc)) 1726 goto exit; 1727 skb = skb_peek(&sk->sk_receive_queue); 1728 hdr = buf_msg(skb); 1729 dlen = msg_data_sz(hdr); 1730 hlen = msg_hdr_sz(hdr); 1731 err = msg_errcode(hdr); 1732 grp_evt = msg_is_grp_evt(hdr); 1733 if (likely(dlen || err)) 1734 break; 1735 tsk_advance_rx_queue(sk); 1736 } while (1); 1737 1738 /* Collect msg meta data, including error code and rejected data */ 1739 tipc_sk_set_orig_addr(m, skb); 1740 rc = tipc_sk_anc_data_recv(m, hdr, tsk); 1741 if (unlikely(rc)) 1742 goto exit; 1743 1744 /* Capture data if non-error msg, otherwise just set return value */ 1745 if (likely(!err)) { 1746 copy = min_t(int, dlen, buflen); 1747 if (unlikely(copy != dlen)) 1748 m->msg_flags |= MSG_TRUNC; 1749 rc = skb_copy_datagram_msg(skb, hlen, m, copy); 1750 } else { 1751 copy = 0; 1752 rc = 0; 1753 if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control) 1754 rc = -ECONNRESET; 1755 } 1756 if (unlikely(rc)) 1757 goto exit; 1758 1759 /* Mark message as group event if applicable */ 1760 if (unlikely(grp_evt)) { 1761 if (msg_grp_evt(hdr) == TIPC_WITHDRAWN) 1762 m->msg_flags |= MSG_EOR; 1763 m->msg_flags |= MSG_OOB; 1764 copy = 0; 1765 } 1766 1767 /* Caption of data or error code/rejected data was successful */ 1768 if (unlikely(flags & MSG_PEEK)) 1769 goto exit; 1770 1771 /* Send group flow control advertisement when applicable */ 1772 if (tsk->group && msg_in_group(hdr) && !grp_evt) { 1773 skb_queue_head_init(&xmitq); 1774 tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen), 1775 msg_orignode(hdr), msg_origport(hdr), 1776 &xmitq); 1777 tipc_node_distr_xmit(sock_net(sk), &xmitq); 1778 } 1779 1780 tsk_advance_rx_queue(sk); 1781 1782 if (likely(!connected)) 1783 goto exit; 1784 1785 /* Send connection flow control advertisement when applicable */ 1786 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen); 1787 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE) 1788 tipc_sk_send_ack(tsk); 1789 exit: 1790 release_sock(sk); 1791 return rc ? rc : copy; 1792 } 1793 1794 /** 1795 * tipc_recvstream - receive stream-oriented data 1796 * @m: descriptor for message info 1797 * @buflen: total size of user buffer area 1798 * @flags: receive flags 1799 * 1800 * Used for SOCK_STREAM messages only. If not enough data is available 1801 * will optionally wait for more; never truncates data. 1802 * 1803 * Returns size of returned message data, errno otherwise 1804 */ 1805 static int tipc_recvstream(struct socket *sock, struct msghdr *m, 1806 size_t buflen, int flags) 1807 { 1808 struct sock *sk = sock->sk; 1809 struct tipc_sock *tsk = tipc_sk(sk); 1810 struct sk_buff *skb; 1811 struct tipc_msg *hdr; 1812 struct tipc_skb_cb *skb_cb; 1813 bool peek = flags & MSG_PEEK; 1814 int offset, required, copy, copied = 0; 1815 int hlen, dlen, err, rc; 1816 long timeout; 1817 1818 /* Catch invalid receive attempts */ 1819 if (unlikely(!buflen)) 1820 return -EINVAL; 1821 1822 lock_sock(sk); 1823 1824 if (unlikely(sk->sk_state == TIPC_OPEN)) { 1825 rc = -ENOTCONN; 1826 goto exit; 1827 } 1828 required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen); 1829 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT); 1830 1831 do { 1832 /* Look at first msg in receive queue; wait if necessary */ 1833 rc = tipc_wait_for_rcvmsg(sock, &timeout); 1834 if (unlikely(rc)) 1835 break; 1836 skb = skb_peek(&sk->sk_receive_queue); 1837 skb_cb = TIPC_SKB_CB(skb); 1838 hdr = buf_msg(skb); 1839 dlen = msg_data_sz(hdr); 1840 hlen = msg_hdr_sz(hdr); 1841 err = msg_errcode(hdr); 1842 1843 /* Discard any empty non-errored (SYN-) message */ 1844 if (unlikely(!dlen && !err)) { 1845 tsk_advance_rx_queue(sk); 1846 continue; 1847 } 1848 1849 /* Collect msg meta data, incl. error code and rejected data */ 1850 if (!copied) { 1851 tipc_sk_set_orig_addr(m, skb); 1852 rc = tipc_sk_anc_data_recv(m, hdr, tsk); 1853 if (rc) 1854 break; 1855 } 1856 1857 /* Copy data if msg ok, otherwise return error/partial data */ 1858 if (likely(!err)) { 1859 offset = skb_cb->bytes_read; 1860 copy = min_t(int, dlen - offset, buflen - copied); 1861 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy); 1862 if (unlikely(rc)) 1863 break; 1864 copied += copy; 1865 offset += copy; 1866 if (unlikely(offset < dlen)) { 1867 if (!peek) 1868 skb_cb->bytes_read = offset; 1869 break; 1870 } 1871 } else { 1872 rc = 0; 1873 if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control) 1874 rc = -ECONNRESET; 1875 if (copied || rc) 1876 break; 1877 } 1878 1879 if (unlikely(peek)) 1880 break; 1881 1882 tsk_advance_rx_queue(sk); 1883 1884 /* Send connection flow control advertisement when applicable */ 1885 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen); 1886 if (unlikely(tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)) 1887 tipc_sk_send_ack(tsk); 1888 1889 /* Exit if all requested data or FIN/error received */ 1890 if (copied == buflen || err) 1891 break; 1892 1893 } while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required); 1894 exit: 1895 release_sock(sk); 1896 return copied ? copied : rc; 1897 } 1898 1899 /** 1900 * tipc_write_space - wake up thread if port congestion is released 1901 * @sk: socket 1902 */ 1903 static void tipc_write_space(struct sock *sk) 1904 { 1905 struct socket_wq *wq; 1906 1907 rcu_read_lock(); 1908 wq = rcu_dereference(sk->sk_wq); 1909 if (skwq_has_sleeper(wq)) 1910 wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT | 1911 EPOLLWRNORM | EPOLLWRBAND); 1912 rcu_read_unlock(); 1913 } 1914 1915 /** 1916 * tipc_data_ready - wake up threads to indicate messages have been received 1917 * @sk: socket 1918 * @len: the length of messages 1919 */ 1920 static void tipc_data_ready(struct sock *sk) 1921 { 1922 struct socket_wq *wq; 1923 1924 rcu_read_lock(); 1925 wq = rcu_dereference(sk->sk_wq); 1926 if (skwq_has_sleeper(wq)) 1927 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN | 1928 EPOLLRDNORM | EPOLLRDBAND); 1929 rcu_read_unlock(); 1930 } 1931 1932 static void tipc_sock_destruct(struct sock *sk) 1933 { 1934 __skb_queue_purge(&sk->sk_receive_queue); 1935 } 1936 1937 static void tipc_sk_proto_rcv(struct sock *sk, 1938 struct sk_buff_head *inputq, 1939 struct sk_buff_head *xmitq) 1940 { 1941 struct sk_buff *skb = __skb_dequeue(inputq); 1942 struct tipc_sock *tsk = tipc_sk(sk); 1943 struct tipc_msg *hdr = buf_msg(skb); 1944 struct tipc_group *grp = tsk->group; 1945 bool wakeup = false; 1946 1947 switch (msg_user(hdr)) { 1948 case CONN_MANAGER: 1949 tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq); 1950 return; 1951 case SOCK_WAKEUP: 1952 tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0); 1953 tsk->cong_link_cnt--; 1954 wakeup = true; 1955 break; 1956 case GROUP_PROTOCOL: 1957 tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq); 1958 break; 1959 case TOP_SRV: 1960 tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf, 1961 hdr, inputq, xmitq); 1962 break; 1963 default: 1964 break; 1965 } 1966 1967 if (wakeup) 1968 sk->sk_write_space(sk); 1969 1970 kfree_skb(skb); 1971 } 1972 1973 /** 1974 * tipc_filter_connect - Handle incoming message for a connection-based socket 1975 * @tsk: TIPC socket 1976 * @skb: pointer to message buffer. Set to NULL if buffer is consumed 1977 * 1978 * Returns true if everything ok, false otherwise 1979 */ 1980 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb) 1981 { 1982 struct sock *sk = &tsk->sk; 1983 struct net *net = sock_net(sk); 1984 struct tipc_msg *hdr = buf_msg(skb); 1985 u32 pport = msg_origport(hdr); 1986 u32 pnode = msg_orignode(hdr); 1987 1988 if (unlikely(msg_mcast(hdr))) 1989 return false; 1990 1991 switch (sk->sk_state) { 1992 case TIPC_CONNECTING: 1993 /* Accept only ACK or NACK message */ 1994 if (unlikely(!msg_connected(hdr))) { 1995 if (pport != tsk_peer_port(tsk) || 1996 pnode != tsk_peer_node(tsk)) 1997 return false; 1998 1999 tipc_set_sk_state(sk, TIPC_DISCONNECTING); 2000 sk->sk_err = ECONNREFUSED; 2001 sk->sk_state_change(sk); 2002 return true; 2003 } 2004 2005 if (unlikely(msg_errcode(hdr))) { 2006 tipc_set_sk_state(sk, TIPC_DISCONNECTING); 2007 sk->sk_err = ECONNREFUSED; 2008 sk->sk_state_change(sk); 2009 return true; 2010 } 2011 2012 if (unlikely(!msg_isdata(hdr))) { 2013 tipc_set_sk_state(sk, TIPC_DISCONNECTING); 2014 sk->sk_err = EINVAL; 2015 sk->sk_state_change(sk); 2016 return true; 2017 } 2018 2019 tipc_sk_finish_conn(tsk, msg_origport(hdr), msg_orignode(hdr)); 2020 msg_set_importance(&tsk->phdr, msg_importance(hdr)); 2021 2022 /* If 'ACK+' message, add to socket receive queue */ 2023 if (msg_data_sz(hdr)) 2024 return true; 2025 2026 /* If empty 'ACK-' message, wake up sleeping connect() */ 2027 sk->sk_data_ready(sk); 2028 2029 /* 'ACK-' message is neither accepted nor rejected: */ 2030 msg_set_dest_droppable(hdr, 1); 2031 return false; 2032 2033 case TIPC_OPEN: 2034 case TIPC_DISCONNECTING: 2035 break; 2036 case TIPC_LISTEN: 2037 /* Accept only SYN message */ 2038 if (!msg_connected(hdr) && !(msg_errcode(hdr))) 2039 return true; 2040 break; 2041 case TIPC_ESTABLISHED: 2042 /* Accept only connection-based messages sent by peer */ 2043 if (unlikely(!tsk_peer_msg(tsk, hdr))) 2044 return false; 2045 2046 if (unlikely(msg_errcode(hdr))) { 2047 tipc_set_sk_state(sk, TIPC_DISCONNECTING); 2048 /* Let timer expire on it's own */ 2049 tipc_node_remove_conn(net, tsk_peer_node(tsk), 2050 tsk->portid); 2051 sk->sk_state_change(sk); 2052 } 2053 return true; 2054 default: 2055 pr_err("Unknown sk_state %u\n", sk->sk_state); 2056 } 2057 2058 return false; 2059 } 2060 2061 /** 2062 * rcvbuf_limit - get proper overload limit of socket receive queue 2063 * @sk: socket 2064 * @skb: message 2065 * 2066 * For connection oriented messages, irrespective of importance, 2067 * default queue limit is 2 MB. 2068 * 2069 * For connectionless messages, queue limits are based on message 2070 * importance as follows: 2071 * 2072 * TIPC_LOW_IMPORTANCE (2 MB) 2073 * TIPC_MEDIUM_IMPORTANCE (4 MB) 2074 * TIPC_HIGH_IMPORTANCE (8 MB) 2075 * TIPC_CRITICAL_IMPORTANCE (16 MB) 2076 * 2077 * Returns overload limit according to corresponding message importance 2078 */ 2079 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb) 2080 { 2081 struct tipc_sock *tsk = tipc_sk(sk); 2082 struct tipc_msg *hdr = buf_msg(skb); 2083 2084 if (unlikely(msg_in_group(hdr))) 2085 return sk->sk_rcvbuf; 2086 2087 if (unlikely(!msg_connected(hdr))) 2088 return sk->sk_rcvbuf << msg_importance(hdr); 2089 2090 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL)) 2091 return sk->sk_rcvbuf; 2092 2093 return FLOWCTL_MSG_LIM; 2094 } 2095 2096 /** 2097 * tipc_sk_filter_rcv - validate incoming message 2098 * @sk: socket 2099 * @skb: pointer to message. 2100 * 2101 * Enqueues message on receive queue if acceptable; optionally handles 2102 * disconnect indication for a connected socket. 2103 * 2104 * Called with socket lock already taken 2105 * 2106 */ 2107 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb, 2108 struct sk_buff_head *xmitq) 2109 { 2110 bool sk_conn = !tipc_sk_type_connectionless(sk); 2111 struct tipc_sock *tsk = tipc_sk(sk); 2112 struct tipc_group *grp = tsk->group; 2113 struct tipc_msg *hdr = buf_msg(skb); 2114 struct net *net = sock_net(sk); 2115 struct sk_buff_head inputq; 2116 int limit, err = TIPC_OK; 2117 2118 TIPC_SKB_CB(skb)->bytes_read = 0; 2119 __skb_queue_head_init(&inputq); 2120 __skb_queue_tail(&inputq, skb); 2121 2122 if (unlikely(!msg_isdata(hdr))) 2123 tipc_sk_proto_rcv(sk, &inputq, xmitq); 2124 2125 if (unlikely(grp)) 2126 tipc_group_filter_msg(grp, &inputq, xmitq); 2127 2128 /* Validate and add to receive buffer if there is space */ 2129 while ((skb = __skb_dequeue(&inputq))) { 2130 hdr = buf_msg(skb); 2131 limit = rcvbuf_limit(sk, skb); 2132 if ((sk_conn && !tipc_sk_filter_connect(tsk, skb)) || 2133 (!sk_conn && msg_connected(hdr)) || 2134 (!grp && msg_in_group(hdr))) 2135 err = TIPC_ERR_NO_PORT; 2136 else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) { 2137 atomic_inc(&sk->sk_drops); 2138 err = TIPC_ERR_OVERLOAD; 2139 } 2140 2141 if (unlikely(err)) { 2142 tipc_skb_reject(net, err, skb, xmitq); 2143 err = TIPC_OK; 2144 continue; 2145 } 2146 __skb_queue_tail(&sk->sk_receive_queue, skb); 2147 skb_set_owner_r(skb, sk); 2148 sk->sk_data_ready(sk); 2149 } 2150 } 2151 2152 /** 2153 * tipc_sk_backlog_rcv - handle incoming message from backlog queue 2154 * @sk: socket 2155 * @skb: message 2156 * 2157 * Caller must hold socket lock 2158 */ 2159 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb) 2160 { 2161 unsigned int before = sk_rmem_alloc_get(sk); 2162 struct sk_buff_head xmitq; 2163 unsigned int added; 2164 2165 __skb_queue_head_init(&xmitq); 2166 2167 tipc_sk_filter_rcv(sk, skb, &xmitq); 2168 added = sk_rmem_alloc_get(sk) - before; 2169 atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt); 2170 2171 /* Send pending response/rejected messages, if any */ 2172 tipc_node_distr_xmit(sock_net(sk), &xmitq); 2173 return 0; 2174 } 2175 2176 /** 2177 * tipc_sk_enqueue - extract all buffers with destination 'dport' from 2178 * inputq and try adding them to socket or backlog queue 2179 * @inputq: list of incoming buffers with potentially different destinations 2180 * @sk: socket where the buffers should be enqueued 2181 * @dport: port number for the socket 2182 * 2183 * Caller must hold socket lock 2184 */ 2185 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk, 2186 u32 dport, struct sk_buff_head *xmitq) 2187 { 2188 unsigned long time_limit = jiffies + 2; 2189 struct sk_buff *skb; 2190 unsigned int lim; 2191 atomic_t *dcnt; 2192 u32 onode; 2193 2194 while (skb_queue_len(inputq)) { 2195 if (unlikely(time_after_eq(jiffies, time_limit))) 2196 return; 2197 2198 skb = tipc_skb_dequeue(inputq, dport); 2199 if (unlikely(!skb)) 2200 return; 2201 2202 /* Add message directly to receive queue if possible */ 2203 if (!sock_owned_by_user(sk)) { 2204 tipc_sk_filter_rcv(sk, skb, xmitq); 2205 continue; 2206 } 2207 2208 /* Try backlog, compensating for double-counted bytes */ 2209 dcnt = &tipc_sk(sk)->dupl_rcvcnt; 2210 if (!sk->sk_backlog.len) 2211 atomic_set(dcnt, 0); 2212 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt); 2213 if (likely(!sk_add_backlog(sk, skb, lim))) 2214 continue; 2215 2216 /* Overload => reject message back to sender */ 2217 onode = tipc_own_addr(sock_net(sk)); 2218 atomic_inc(&sk->sk_drops); 2219 if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) 2220 __skb_queue_tail(xmitq, skb); 2221 break; 2222 } 2223 } 2224 2225 /** 2226 * tipc_sk_rcv - handle a chain of incoming buffers 2227 * @inputq: buffer list containing the buffers 2228 * Consumes all buffers in list until inputq is empty 2229 * Note: may be called in multiple threads referring to the same queue 2230 */ 2231 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq) 2232 { 2233 struct sk_buff_head xmitq; 2234 u32 dnode, dport = 0; 2235 int err; 2236 struct tipc_sock *tsk; 2237 struct sock *sk; 2238 struct sk_buff *skb; 2239 2240 __skb_queue_head_init(&xmitq); 2241 while (skb_queue_len(inputq)) { 2242 dport = tipc_skb_peek_port(inputq, dport); 2243 tsk = tipc_sk_lookup(net, dport); 2244 2245 if (likely(tsk)) { 2246 sk = &tsk->sk; 2247 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) { 2248 tipc_sk_enqueue(inputq, sk, dport, &xmitq); 2249 spin_unlock_bh(&sk->sk_lock.slock); 2250 } 2251 /* Send pending response/rejected messages, if any */ 2252 tipc_node_distr_xmit(sock_net(sk), &xmitq); 2253 sock_put(sk); 2254 continue; 2255 } 2256 /* No destination socket => dequeue skb if still there */ 2257 skb = tipc_skb_dequeue(inputq, dport); 2258 if (!skb) 2259 return; 2260 2261 /* Try secondary lookup if unresolved named message */ 2262 err = TIPC_ERR_NO_PORT; 2263 if (tipc_msg_lookup_dest(net, skb, &err)) 2264 goto xmit; 2265 2266 /* Prepare for message rejection */ 2267 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err)) 2268 continue; 2269 xmit: 2270 dnode = msg_destnode(buf_msg(skb)); 2271 tipc_node_xmit_skb(net, skb, dnode, dport); 2272 } 2273 } 2274 2275 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p) 2276 { 2277 DEFINE_WAIT_FUNC(wait, woken_wake_function); 2278 struct sock *sk = sock->sk; 2279 int done; 2280 2281 do { 2282 int err = sock_error(sk); 2283 if (err) 2284 return err; 2285 if (!*timeo_p) 2286 return -ETIMEDOUT; 2287 if (signal_pending(current)) 2288 return sock_intr_errno(*timeo_p); 2289 2290 add_wait_queue(sk_sleep(sk), &wait); 2291 done = sk_wait_event(sk, timeo_p, 2292 sk->sk_state != TIPC_CONNECTING, &wait); 2293 remove_wait_queue(sk_sleep(sk), &wait); 2294 } while (!done); 2295 return 0; 2296 } 2297 2298 /** 2299 * tipc_connect - establish a connection to another TIPC port 2300 * @sock: socket structure 2301 * @dest: socket address for destination port 2302 * @destlen: size of socket address data structure 2303 * @flags: file-related flags associated with socket 2304 * 2305 * Returns 0 on success, errno otherwise 2306 */ 2307 static int tipc_connect(struct socket *sock, struct sockaddr *dest, 2308 int destlen, int flags) 2309 { 2310 struct sock *sk = sock->sk; 2311 struct tipc_sock *tsk = tipc_sk(sk); 2312 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest; 2313 struct msghdr m = {NULL,}; 2314 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout; 2315 int previous; 2316 int res = 0; 2317 2318 if (destlen != sizeof(struct sockaddr_tipc)) 2319 return -EINVAL; 2320 2321 lock_sock(sk); 2322 2323 if (tsk->group) { 2324 res = -EINVAL; 2325 goto exit; 2326 } 2327 2328 if (dst->family == AF_UNSPEC) { 2329 memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc)); 2330 if (!tipc_sk_type_connectionless(sk)) 2331 res = -EINVAL; 2332 goto exit; 2333 } else if (dst->family != AF_TIPC) { 2334 res = -EINVAL; 2335 } 2336 if (dst->addrtype != TIPC_ADDR_ID && dst->addrtype != TIPC_ADDR_NAME) 2337 res = -EINVAL; 2338 if (res) 2339 goto exit; 2340 2341 /* DGRAM/RDM connect(), just save the destaddr */ 2342 if (tipc_sk_type_connectionless(sk)) { 2343 memcpy(&tsk->peer, dest, destlen); 2344 goto exit; 2345 } 2346 2347 previous = sk->sk_state; 2348 2349 switch (sk->sk_state) { 2350 case TIPC_OPEN: 2351 /* Send a 'SYN-' to destination */ 2352 m.msg_name = dest; 2353 m.msg_namelen = destlen; 2354 2355 /* If connect is in non-blocking case, set MSG_DONTWAIT to 2356 * indicate send_msg() is never blocked. 2357 */ 2358 if (!timeout) 2359 m.msg_flags = MSG_DONTWAIT; 2360 2361 res = __tipc_sendmsg(sock, &m, 0); 2362 if ((res < 0) && (res != -EWOULDBLOCK)) 2363 goto exit; 2364 2365 /* Just entered TIPC_CONNECTING state; the only 2366 * difference is that return value in non-blocking 2367 * case is EINPROGRESS, rather than EALREADY. 2368 */ 2369 res = -EINPROGRESS; 2370 /* fall thru' */ 2371 case TIPC_CONNECTING: 2372 if (!timeout) { 2373 if (previous == TIPC_CONNECTING) 2374 res = -EALREADY; 2375 goto exit; 2376 } 2377 timeout = msecs_to_jiffies(timeout); 2378 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */ 2379 res = tipc_wait_for_connect(sock, &timeout); 2380 break; 2381 case TIPC_ESTABLISHED: 2382 res = -EISCONN; 2383 break; 2384 default: 2385 res = -EINVAL; 2386 } 2387 2388 exit: 2389 release_sock(sk); 2390 return res; 2391 } 2392 2393 /** 2394 * tipc_listen - allow socket to listen for incoming connections 2395 * @sock: socket structure 2396 * @len: (unused) 2397 * 2398 * Returns 0 on success, errno otherwise 2399 */ 2400 static int tipc_listen(struct socket *sock, int len) 2401 { 2402 struct sock *sk = sock->sk; 2403 int res; 2404 2405 lock_sock(sk); 2406 res = tipc_set_sk_state(sk, TIPC_LISTEN); 2407 release_sock(sk); 2408 2409 return res; 2410 } 2411 2412 static int tipc_wait_for_accept(struct socket *sock, long timeo) 2413 { 2414 struct sock *sk = sock->sk; 2415 DEFINE_WAIT(wait); 2416 int err; 2417 2418 /* True wake-one mechanism for incoming connections: only 2419 * one process gets woken up, not the 'whole herd'. 2420 * Since we do not 'race & poll' for established sockets 2421 * anymore, the common case will execute the loop only once. 2422 */ 2423 for (;;) { 2424 prepare_to_wait_exclusive(sk_sleep(sk), &wait, 2425 TASK_INTERRUPTIBLE); 2426 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) { 2427 release_sock(sk); 2428 timeo = schedule_timeout(timeo); 2429 lock_sock(sk); 2430 } 2431 err = 0; 2432 if (!skb_queue_empty(&sk->sk_receive_queue)) 2433 break; 2434 err = -EAGAIN; 2435 if (!timeo) 2436 break; 2437 err = sock_intr_errno(timeo); 2438 if (signal_pending(current)) 2439 break; 2440 } 2441 finish_wait(sk_sleep(sk), &wait); 2442 return err; 2443 } 2444 2445 /** 2446 * tipc_accept - wait for connection request 2447 * @sock: listening socket 2448 * @newsock: new socket that is to be connected 2449 * @flags: file-related flags associated with socket 2450 * 2451 * Returns 0 on success, errno otherwise 2452 */ 2453 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags, 2454 bool kern) 2455 { 2456 struct sock *new_sk, *sk = sock->sk; 2457 struct sk_buff *buf; 2458 struct tipc_sock *new_tsock; 2459 struct tipc_msg *msg; 2460 long timeo; 2461 int res; 2462 2463 lock_sock(sk); 2464 2465 if (sk->sk_state != TIPC_LISTEN) { 2466 res = -EINVAL; 2467 goto exit; 2468 } 2469 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK); 2470 res = tipc_wait_for_accept(sock, timeo); 2471 if (res) 2472 goto exit; 2473 2474 buf = skb_peek(&sk->sk_receive_queue); 2475 2476 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern); 2477 if (res) 2478 goto exit; 2479 security_sk_clone(sock->sk, new_sock->sk); 2480 2481 new_sk = new_sock->sk; 2482 new_tsock = tipc_sk(new_sk); 2483 msg = buf_msg(buf); 2484 2485 /* we lock on new_sk; but lockdep sees the lock on sk */ 2486 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING); 2487 2488 /* 2489 * Reject any stray messages received by new socket 2490 * before the socket lock was taken (very, very unlikely) 2491 */ 2492 tsk_rej_rx_queue(new_sk); 2493 2494 /* Connect new socket to it's peer */ 2495 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg)); 2496 2497 tsk_set_importance(new_tsock, msg_importance(msg)); 2498 if (msg_named(msg)) { 2499 new_tsock->conn_type = msg_nametype(msg); 2500 new_tsock->conn_instance = msg_nameinst(msg); 2501 } 2502 2503 /* 2504 * Respond to 'SYN-' by discarding it & returning 'ACK'-. 2505 * Respond to 'SYN+' by queuing it on new socket. 2506 */ 2507 if (!msg_data_sz(msg)) { 2508 struct msghdr m = {NULL,}; 2509 2510 tsk_advance_rx_queue(sk); 2511 __tipc_sendstream(new_sock, &m, 0); 2512 } else { 2513 __skb_dequeue(&sk->sk_receive_queue); 2514 __skb_queue_head(&new_sk->sk_receive_queue, buf); 2515 skb_set_owner_r(buf, new_sk); 2516 } 2517 release_sock(new_sk); 2518 exit: 2519 release_sock(sk); 2520 return res; 2521 } 2522 2523 /** 2524 * tipc_shutdown - shutdown socket connection 2525 * @sock: socket structure 2526 * @how: direction to close (must be SHUT_RDWR) 2527 * 2528 * Terminates connection (if necessary), then purges socket's receive queue. 2529 * 2530 * Returns 0 on success, errno otherwise 2531 */ 2532 static int tipc_shutdown(struct socket *sock, int how) 2533 { 2534 struct sock *sk = sock->sk; 2535 int res; 2536 2537 if (how != SHUT_RDWR) 2538 return -EINVAL; 2539 2540 lock_sock(sk); 2541 2542 __tipc_shutdown(sock, TIPC_CONN_SHUTDOWN); 2543 sk->sk_shutdown = SEND_SHUTDOWN; 2544 2545 if (sk->sk_state == TIPC_DISCONNECTING) { 2546 /* Discard any unreceived messages */ 2547 __skb_queue_purge(&sk->sk_receive_queue); 2548 2549 /* Wake up anyone sleeping in poll */ 2550 sk->sk_state_change(sk); 2551 res = 0; 2552 } else { 2553 res = -ENOTCONN; 2554 } 2555 2556 release_sock(sk); 2557 return res; 2558 } 2559 2560 static void tipc_sk_timeout(struct timer_list *t) 2561 { 2562 struct sock *sk = from_timer(sk, t, sk_timer); 2563 struct tipc_sock *tsk = tipc_sk(sk); 2564 u32 peer_port = tsk_peer_port(tsk); 2565 u32 peer_node = tsk_peer_node(tsk); 2566 u32 own_node = tsk_own_node(tsk); 2567 u32 own_port = tsk->portid; 2568 struct net *net = sock_net(sk); 2569 struct sk_buff *skb = NULL; 2570 2571 bh_lock_sock(sk); 2572 if (!tipc_sk_connected(sk)) 2573 goto exit; 2574 2575 /* Try again later if socket is busy */ 2576 if (sock_owned_by_user(sk)) { 2577 sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20); 2578 goto exit; 2579 } 2580 2581 if (tsk->probe_unacked) { 2582 tipc_set_sk_state(sk, TIPC_DISCONNECTING); 2583 tipc_node_remove_conn(net, peer_node, peer_port); 2584 sk->sk_state_change(sk); 2585 goto exit; 2586 } 2587 /* Send new probe */ 2588 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0, 2589 peer_node, own_node, peer_port, own_port, 2590 TIPC_OK); 2591 tsk->probe_unacked = true; 2592 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV); 2593 exit: 2594 bh_unlock_sock(sk); 2595 if (skb) 2596 tipc_node_xmit_skb(net, skb, peer_node, own_port); 2597 sock_put(sk); 2598 } 2599 2600 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope, 2601 struct tipc_name_seq const *seq) 2602 { 2603 struct sock *sk = &tsk->sk; 2604 struct net *net = sock_net(sk); 2605 struct publication *publ; 2606 u32 key; 2607 2608 if (scope != TIPC_NODE_SCOPE) 2609 scope = TIPC_CLUSTER_SCOPE; 2610 2611 if (tipc_sk_connected(sk)) 2612 return -EINVAL; 2613 key = tsk->portid + tsk->pub_count + 1; 2614 if (key == tsk->portid) 2615 return -EADDRINUSE; 2616 2617 publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper, 2618 scope, tsk->portid, key); 2619 if (unlikely(!publ)) 2620 return -EINVAL; 2621 2622 list_add(&publ->binding_sock, &tsk->publications); 2623 tsk->pub_count++; 2624 tsk->published = 1; 2625 return 0; 2626 } 2627 2628 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope, 2629 struct tipc_name_seq const *seq) 2630 { 2631 struct net *net = sock_net(&tsk->sk); 2632 struct publication *publ; 2633 struct publication *safe; 2634 int rc = -EINVAL; 2635 2636 if (scope != TIPC_NODE_SCOPE) 2637 scope = TIPC_CLUSTER_SCOPE; 2638 2639 list_for_each_entry_safe(publ, safe, &tsk->publications, binding_sock) { 2640 if (seq) { 2641 if (publ->scope != scope) 2642 continue; 2643 if (publ->type != seq->type) 2644 continue; 2645 if (publ->lower != seq->lower) 2646 continue; 2647 if (publ->upper != seq->upper) 2648 break; 2649 tipc_nametbl_withdraw(net, publ->type, publ->lower, 2650 publ->upper, publ->key); 2651 rc = 0; 2652 break; 2653 } 2654 tipc_nametbl_withdraw(net, publ->type, publ->lower, 2655 publ->upper, publ->key); 2656 rc = 0; 2657 } 2658 if (list_empty(&tsk->publications)) 2659 tsk->published = 0; 2660 return rc; 2661 } 2662 2663 /* tipc_sk_reinit: set non-zero address in all existing sockets 2664 * when we go from standalone to network mode. 2665 */ 2666 void tipc_sk_reinit(struct net *net) 2667 { 2668 struct tipc_net *tn = net_generic(net, tipc_net_id); 2669 struct rhashtable_iter iter; 2670 struct tipc_sock *tsk; 2671 struct tipc_msg *msg; 2672 2673 rhashtable_walk_enter(&tn->sk_rht, &iter); 2674 2675 do { 2676 rhashtable_walk_start(&iter); 2677 2678 while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) { 2679 spin_lock_bh(&tsk->sk.sk_lock.slock); 2680 msg = &tsk->phdr; 2681 msg_set_prevnode(msg, tipc_own_addr(net)); 2682 msg_set_orignode(msg, tipc_own_addr(net)); 2683 spin_unlock_bh(&tsk->sk.sk_lock.slock); 2684 } 2685 2686 rhashtable_walk_stop(&iter); 2687 } while (tsk == ERR_PTR(-EAGAIN)); 2688 2689 rhashtable_walk_exit(&iter); 2690 } 2691 2692 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid) 2693 { 2694 struct tipc_net *tn = net_generic(net, tipc_net_id); 2695 struct tipc_sock *tsk; 2696 2697 rcu_read_lock(); 2698 tsk = rhashtable_lookup_fast(&tn->sk_rht, &portid, tsk_rht_params); 2699 if (tsk) 2700 sock_hold(&tsk->sk); 2701 rcu_read_unlock(); 2702 2703 return tsk; 2704 } 2705 2706 static int tipc_sk_insert(struct tipc_sock *tsk) 2707 { 2708 struct sock *sk = &tsk->sk; 2709 struct net *net = sock_net(sk); 2710 struct tipc_net *tn = net_generic(net, tipc_net_id); 2711 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1; 2712 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT; 2713 2714 while (remaining--) { 2715 portid++; 2716 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT)) 2717 portid = TIPC_MIN_PORT; 2718 tsk->portid = portid; 2719 sock_hold(&tsk->sk); 2720 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node, 2721 tsk_rht_params)) 2722 return 0; 2723 sock_put(&tsk->sk); 2724 } 2725 2726 return -1; 2727 } 2728 2729 static void tipc_sk_remove(struct tipc_sock *tsk) 2730 { 2731 struct sock *sk = &tsk->sk; 2732 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id); 2733 2734 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) { 2735 WARN_ON(refcount_read(&sk->sk_refcnt) == 1); 2736 __sock_put(sk); 2737 } 2738 } 2739 2740 static const struct rhashtable_params tsk_rht_params = { 2741 .nelem_hint = 192, 2742 .head_offset = offsetof(struct tipc_sock, node), 2743 .key_offset = offsetof(struct tipc_sock, portid), 2744 .key_len = sizeof(u32), /* portid */ 2745 .max_size = 1048576, 2746 .min_size = 256, 2747 .automatic_shrinking = true, 2748 }; 2749 2750 int tipc_sk_rht_init(struct net *net) 2751 { 2752 struct tipc_net *tn = net_generic(net, tipc_net_id); 2753 2754 return rhashtable_init(&tn->sk_rht, &tsk_rht_params); 2755 } 2756 2757 void tipc_sk_rht_destroy(struct net *net) 2758 { 2759 struct tipc_net *tn = net_generic(net, tipc_net_id); 2760 2761 /* Wait for socket readers to complete */ 2762 synchronize_net(); 2763 2764 rhashtable_destroy(&tn->sk_rht); 2765 } 2766 2767 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq) 2768 { 2769 struct net *net = sock_net(&tsk->sk); 2770 struct tipc_group *grp = tsk->group; 2771 struct tipc_msg *hdr = &tsk->phdr; 2772 struct tipc_name_seq seq; 2773 int rc; 2774 2775 if (mreq->type < TIPC_RESERVED_TYPES) 2776 return -EACCES; 2777 if (mreq->scope > TIPC_NODE_SCOPE) 2778 return -EINVAL; 2779 if (grp) 2780 return -EACCES; 2781 grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open); 2782 if (!grp) 2783 return -ENOMEM; 2784 tsk->group = grp; 2785 msg_set_lookup_scope(hdr, mreq->scope); 2786 msg_set_nametype(hdr, mreq->type); 2787 msg_set_dest_droppable(hdr, true); 2788 seq.type = mreq->type; 2789 seq.lower = mreq->instance; 2790 seq.upper = seq.lower; 2791 tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope); 2792 rc = tipc_sk_publish(tsk, mreq->scope, &seq); 2793 if (rc) { 2794 tipc_group_delete(net, grp); 2795 tsk->group = NULL; 2796 return rc; 2797 } 2798 /* Eliminate any risk that a broadcast overtakes sent JOINs */ 2799 tsk->mc_method.rcast = true; 2800 tsk->mc_method.mandatory = true; 2801 tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf); 2802 return rc; 2803 } 2804 2805 static int tipc_sk_leave(struct tipc_sock *tsk) 2806 { 2807 struct net *net = sock_net(&tsk->sk); 2808 struct tipc_group *grp = tsk->group; 2809 struct tipc_name_seq seq; 2810 int scope; 2811 2812 if (!grp) 2813 return -EINVAL; 2814 tipc_group_self(grp, &seq, &scope); 2815 tipc_group_delete(net, grp); 2816 tsk->group = NULL; 2817 tipc_sk_withdraw(tsk, scope, &seq); 2818 return 0; 2819 } 2820 2821 /** 2822 * tipc_setsockopt - set socket option 2823 * @sock: socket structure 2824 * @lvl: option level 2825 * @opt: option identifier 2826 * @ov: pointer to new option value 2827 * @ol: length of option value 2828 * 2829 * For stream sockets only, accepts and ignores all IPPROTO_TCP options 2830 * (to ease compatibility). 2831 * 2832 * Returns 0 on success, errno otherwise 2833 */ 2834 static int tipc_setsockopt(struct socket *sock, int lvl, int opt, 2835 char __user *ov, unsigned int ol) 2836 { 2837 struct sock *sk = sock->sk; 2838 struct tipc_sock *tsk = tipc_sk(sk); 2839 struct tipc_group_req mreq; 2840 u32 value = 0; 2841 int res = 0; 2842 2843 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM)) 2844 return 0; 2845 if (lvl != SOL_TIPC) 2846 return -ENOPROTOOPT; 2847 2848 switch (opt) { 2849 case TIPC_IMPORTANCE: 2850 case TIPC_SRC_DROPPABLE: 2851 case TIPC_DEST_DROPPABLE: 2852 case TIPC_CONN_TIMEOUT: 2853 if (ol < sizeof(value)) 2854 return -EINVAL; 2855 if (get_user(value, (u32 __user *)ov)) 2856 return -EFAULT; 2857 break; 2858 case TIPC_GROUP_JOIN: 2859 if (ol < sizeof(mreq)) 2860 return -EINVAL; 2861 if (copy_from_user(&mreq, ov, sizeof(mreq))) 2862 return -EFAULT; 2863 break; 2864 default: 2865 if (ov || ol) 2866 return -EINVAL; 2867 } 2868 2869 lock_sock(sk); 2870 2871 switch (opt) { 2872 case TIPC_IMPORTANCE: 2873 res = tsk_set_importance(tsk, value); 2874 break; 2875 case TIPC_SRC_DROPPABLE: 2876 if (sock->type != SOCK_STREAM) 2877 tsk_set_unreliable(tsk, value); 2878 else 2879 res = -ENOPROTOOPT; 2880 break; 2881 case TIPC_DEST_DROPPABLE: 2882 tsk_set_unreturnable(tsk, value); 2883 break; 2884 case TIPC_CONN_TIMEOUT: 2885 tipc_sk(sk)->conn_timeout = value; 2886 break; 2887 case TIPC_MCAST_BROADCAST: 2888 tsk->mc_method.rcast = false; 2889 tsk->mc_method.mandatory = true; 2890 break; 2891 case TIPC_MCAST_REPLICAST: 2892 tsk->mc_method.rcast = true; 2893 tsk->mc_method.mandatory = true; 2894 break; 2895 case TIPC_GROUP_JOIN: 2896 res = tipc_sk_join(tsk, &mreq); 2897 break; 2898 case TIPC_GROUP_LEAVE: 2899 res = tipc_sk_leave(tsk); 2900 break; 2901 default: 2902 res = -EINVAL; 2903 } 2904 2905 release_sock(sk); 2906 2907 return res; 2908 } 2909 2910 /** 2911 * tipc_getsockopt - get socket option 2912 * @sock: socket structure 2913 * @lvl: option level 2914 * @opt: option identifier 2915 * @ov: receptacle for option value 2916 * @ol: receptacle for length of option value 2917 * 2918 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options 2919 * (to ease compatibility). 2920 * 2921 * Returns 0 on success, errno otherwise 2922 */ 2923 static int tipc_getsockopt(struct socket *sock, int lvl, int opt, 2924 char __user *ov, int __user *ol) 2925 { 2926 struct sock *sk = sock->sk; 2927 struct tipc_sock *tsk = tipc_sk(sk); 2928 struct tipc_name_seq seq; 2929 int len, scope; 2930 u32 value; 2931 int res; 2932 2933 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM)) 2934 return put_user(0, ol); 2935 if (lvl != SOL_TIPC) 2936 return -ENOPROTOOPT; 2937 res = get_user(len, ol); 2938 if (res) 2939 return res; 2940 2941 lock_sock(sk); 2942 2943 switch (opt) { 2944 case TIPC_IMPORTANCE: 2945 value = tsk_importance(tsk); 2946 break; 2947 case TIPC_SRC_DROPPABLE: 2948 value = tsk_unreliable(tsk); 2949 break; 2950 case TIPC_DEST_DROPPABLE: 2951 value = tsk_unreturnable(tsk); 2952 break; 2953 case TIPC_CONN_TIMEOUT: 2954 value = tsk->conn_timeout; 2955 /* no need to set "res", since already 0 at this point */ 2956 break; 2957 case TIPC_NODE_RECVQ_DEPTH: 2958 value = 0; /* was tipc_queue_size, now obsolete */ 2959 break; 2960 case TIPC_SOCK_RECVQ_DEPTH: 2961 value = skb_queue_len(&sk->sk_receive_queue); 2962 break; 2963 case TIPC_GROUP_JOIN: 2964 seq.type = 0; 2965 if (tsk->group) 2966 tipc_group_self(tsk->group, &seq, &scope); 2967 value = seq.type; 2968 break; 2969 default: 2970 res = -EINVAL; 2971 } 2972 2973 release_sock(sk); 2974 2975 if (res) 2976 return res; /* "get" failed */ 2977 2978 if (len < sizeof(value)) 2979 return -EINVAL; 2980 2981 if (copy_to_user(ov, &value, sizeof(value))) 2982 return -EFAULT; 2983 2984 return put_user(sizeof(value), ol); 2985 } 2986 2987 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 2988 { 2989 struct net *net = sock_net(sock->sk); 2990 struct tipc_sioc_nodeid_req nr = {0}; 2991 struct tipc_sioc_ln_req lnr; 2992 void __user *argp = (void __user *)arg; 2993 2994 switch (cmd) { 2995 case SIOCGETLINKNAME: 2996 if (copy_from_user(&lnr, argp, sizeof(lnr))) 2997 return -EFAULT; 2998 if (!tipc_node_get_linkname(net, 2999 lnr.bearer_id & 0xffff, lnr.peer, 3000 lnr.linkname, TIPC_MAX_LINK_NAME)) { 3001 if (copy_to_user(argp, &lnr, sizeof(lnr))) 3002 return -EFAULT; 3003 return 0; 3004 } 3005 return -EADDRNOTAVAIL; 3006 case SIOCGETNODEID: 3007 if (copy_from_user(&nr, argp, sizeof(nr))) 3008 return -EFAULT; 3009 if (!tipc_node_get_id(net, nr.peer, nr.node_id)) 3010 return -EADDRNOTAVAIL; 3011 if (copy_to_user(argp, &nr, sizeof(nr))) 3012 return -EFAULT; 3013 return 0; 3014 default: 3015 return -ENOIOCTLCMD; 3016 } 3017 } 3018 3019 static int tipc_socketpair(struct socket *sock1, struct socket *sock2) 3020 { 3021 struct tipc_sock *tsk2 = tipc_sk(sock2->sk); 3022 struct tipc_sock *tsk1 = tipc_sk(sock1->sk); 3023 u32 onode = tipc_own_addr(sock_net(sock1->sk)); 3024 3025 tsk1->peer.family = AF_TIPC; 3026 tsk1->peer.addrtype = TIPC_ADDR_ID; 3027 tsk1->peer.scope = TIPC_NODE_SCOPE; 3028 tsk1->peer.addr.id.ref = tsk2->portid; 3029 tsk1->peer.addr.id.node = onode; 3030 tsk2->peer.family = AF_TIPC; 3031 tsk2->peer.addrtype = TIPC_ADDR_ID; 3032 tsk2->peer.scope = TIPC_NODE_SCOPE; 3033 tsk2->peer.addr.id.ref = tsk1->portid; 3034 tsk2->peer.addr.id.node = onode; 3035 3036 tipc_sk_finish_conn(tsk1, tsk2->portid, onode); 3037 tipc_sk_finish_conn(tsk2, tsk1->portid, onode); 3038 return 0; 3039 } 3040 3041 /* Protocol switches for the various types of TIPC sockets */ 3042 3043 static const struct proto_ops msg_ops = { 3044 .owner = THIS_MODULE, 3045 .family = AF_TIPC, 3046 .release = tipc_release, 3047 .bind = tipc_bind, 3048 .connect = tipc_connect, 3049 .socketpair = tipc_socketpair, 3050 .accept = sock_no_accept, 3051 .getname = tipc_getname, 3052 .poll = tipc_poll, 3053 .ioctl = tipc_ioctl, 3054 .listen = sock_no_listen, 3055 .shutdown = tipc_shutdown, 3056 .setsockopt = tipc_setsockopt, 3057 .getsockopt = tipc_getsockopt, 3058 .sendmsg = tipc_sendmsg, 3059 .recvmsg = tipc_recvmsg, 3060 .mmap = sock_no_mmap, 3061 .sendpage = sock_no_sendpage 3062 }; 3063 3064 static const struct proto_ops packet_ops = { 3065 .owner = THIS_MODULE, 3066 .family = AF_TIPC, 3067 .release = tipc_release, 3068 .bind = tipc_bind, 3069 .connect = tipc_connect, 3070 .socketpair = tipc_socketpair, 3071 .accept = tipc_accept, 3072 .getname = tipc_getname, 3073 .poll = tipc_poll, 3074 .ioctl = tipc_ioctl, 3075 .listen = tipc_listen, 3076 .shutdown = tipc_shutdown, 3077 .setsockopt = tipc_setsockopt, 3078 .getsockopt = tipc_getsockopt, 3079 .sendmsg = tipc_send_packet, 3080 .recvmsg = tipc_recvmsg, 3081 .mmap = sock_no_mmap, 3082 .sendpage = sock_no_sendpage 3083 }; 3084 3085 static const struct proto_ops stream_ops = { 3086 .owner = THIS_MODULE, 3087 .family = AF_TIPC, 3088 .release = tipc_release, 3089 .bind = tipc_bind, 3090 .connect = tipc_connect, 3091 .socketpair = tipc_socketpair, 3092 .accept = tipc_accept, 3093 .getname = tipc_getname, 3094 .poll = tipc_poll, 3095 .ioctl = tipc_ioctl, 3096 .listen = tipc_listen, 3097 .shutdown = tipc_shutdown, 3098 .setsockopt = tipc_setsockopt, 3099 .getsockopt = tipc_getsockopt, 3100 .sendmsg = tipc_sendstream, 3101 .recvmsg = tipc_recvstream, 3102 .mmap = sock_no_mmap, 3103 .sendpage = sock_no_sendpage 3104 }; 3105 3106 static const struct net_proto_family tipc_family_ops = { 3107 .owner = THIS_MODULE, 3108 .family = AF_TIPC, 3109 .create = tipc_sk_create 3110 }; 3111 3112 static struct proto tipc_proto = { 3113 .name = "TIPC", 3114 .owner = THIS_MODULE, 3115 .obj_size = sizeof(struct tipc_sock), 3116 .sysctl_rmem = sysctl_tipc_rmem 3117 }; 3118 3119 /** 3120 * tipc_socket_init - initialize TIPC socket interface 3121 * 3122 * Returns 0 on success, errno otherwise 3123 */ 3124 int tipc_socket_init(void) 3125 { 3126 int res; 3127 3128 res = proto_register(&tipc_proto, 1); 3129 if (res) { 3130 pr_err("Failed to register TIPC protocol type\n"); 3131 goto out; 3132 } 3133 3134 res = sock_register(&tipc_family_ops); 3135 if (res) { 3136 pr_err("Failed to register TIPC socket type\n"); 3137 proto_unregister(&tipc_proto); 3138 goto out; 3139 } 3140 out: 3141 return res; 3142 } 3143 3144 /** 3145 * tipc_socket_stop - stop TIPC socket interface 3146 */ 3147 void tipc_socket_stop(void) 3148 { 3149 sock_unregister(tipc_family_ops.family); 3150 proto_unregister(&tipc_proto); 3151 } 3152 3153 /* Caller should hold socket lock for the passed tipc socket. */ 3154 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk) 3155 { 3156 u32 peer_node; 3157 u32 peer_port; 3158 struct nlattr *nest; 3159 3160 peer_node = tsk_peer_node(tsk); 3161 peer_port = tsk_peer_port(tsk); 3162 3163 nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON); 3164 3165 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node)) 3166 goto msg_full; 3167 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port)) 3168 goto msg_full; 3169 3170 if (tsk->conn_type != 0) { 3171 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG)) 3172 goto msg_full; 3173 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type)) 3174 goto msg_full; 3175 if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance)) 3176 goto msg_full; 3177 } 3178 nla_nest_end(skb, nest); 3179 3180 return 0; 3181 3182 msg_full: 3183 nla_nest_cancel(skb, nest); 3184 3185 return -EMSGSIZE; 3186 } 3187 3188 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock 3189 *tsk) 3190 { 3191 struct net *net = sock_net(skb->sk); 3192 struct sock *sk = &tsk->sk; 3193 3194 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) || 3195 nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net))) 3196 return -EMSGSIZE; 3197 3198 if (tipc_sk_connected(sk)) { 3199 if (__tipc_nl_add_sk_con(skb, tsk)) 3200 return -EMSGSIZE; 3201 } else if (!list_empty(&tsk->publications)) { 3202 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL)) 3203 return -EMSGSIZE; 3204 } 3205 return 0; 3206 } 3207 3208 /* Caller should hold socket lock for the passed tipc socket. */ 3209 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb, 3210 struct tipc_sock *tsk) 3211 { 3212 struct nlattr *attrs; 3213 void *hdr; 3214 3215 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq, 3216 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET); 3217 if (!hdr) 3218 goto msg_cancel; 3219 3220 attrs = nla_nest_start(skb, TIPC_NLA_SOCK); 3221 if (!attrs) 3222 goto genlmsg_cancel; 3223 3224 if (__tipc_nl_add_sk_info(skb, tsk)) 3225 goto attr_msg_cancel; 3226 3227 nla_nest_end(skb, attrs); 3228 genlmsg_end(skb, hdr); 3229 3230 return 0; 3231 3232 attr_msg_cancel: 3233 nla_nest_cancel(skb, attrs); 3234 genlmsg_cancel: 3235 genlmsg_cancel(skb, hdr); 3236 msg_cancel: 3237 return -EMSGSIZE; 3238 } 3239 3240 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb, 3241 int (*skb_handler)(struct sk_buff *skb, 3242 struct netlink_callback *cb, 3243 struct tipc_sock *tsk)) 3244 { 3245 struct rhashtable_iter *iter = (void *)cb->args[4]; 3246 struct tipc_sock *tsk; 3247 int err; 3248 3249 rhashtable_walk_start(iter); 3250 while ((tsk = rhashtable_walk_next(iter)) != NULL) { 3251 if (IS_ERR(tsk)) { 3252 err = PTR_ERR(tsk); 3253 if (err == -EAGAIN) { 3254 err = 0; 3255 continue; 3256 } 3257 break; 3258 } 3259 3260 sock_hold(&tsk->sk); 3261 rhashtable_walk_stop(iter); 3262 lock_sock(&tsk->sk); 3263 err = skb_handler(skb, cb, tsk); 3264 if (err) { 3265 release_sock(&tsk->sk); 3266 sock_put(&tsk->sk); 3267 goto out; 3268 } 3269 release_sock(&tsk->sk); 3270 rhashtable_walk_start(iter); 3271 sock_put(&tsk->sk); 3272 } 3273 rhashtable_walk_stop(iter); 3274 out: 3275 return skb->len; 3276 } 3277 EXPORT_SYMBOL(tipc_nl_sk_walk); 3278 3279 int tipc_dump_start(struct netlink_callback *cb) 3280 { 3281 return __tipc_dump_start(cb, sock_net(cb->skb->sk)); 3282 } 3283 EXPORT_SYMBOL(tipc_dump_start); 3284 3285 int __tipc_dump_start(struct netlink_callback *cb, struct net *net) 3286 { 3287 /* tipc_nl_name_table_dump() uses cb->args[0...3]. */ 3288 struct rhashtable_iter *iter = (void *)cb->args[4]; 3289 struct tipc_net *tn = tipc_net(net); 3290 3291 if (!iter) { 3292 iter = kmalloc(sizeof(*iter), GFP_KERNEL); 3293 if (!iter) 3294 return -ENOMEM; 3295 3296 cb->args[4] = (long)iter; 3297 } 3298 3299 rhashtable_walk_enter(&tn->sk_rht, iter); 3300 return 0; 3301 } 3302 3303 int tipc_dump_done(struct netlink_callback *cb) 3304 { 3305 struct rhashtable_iter *hti = (void *)cb->args[4]; 3306 3307 rhashtable_walk_exit(hti); 3308 kfree(hti); 3309 return 0; 3310 } 3311 EXPORT_SYMBOL(tipc_dump_done); 3312 3313 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb, 3314 struct tipc_sock *tsk, u32 sk_filter_state, 3315 u64 (*tipc_diag_gen_cookie)(struct sock *sk)) 3316 { 3317 struct sock *sk = &tsk->sk; 3318 struct nlattr *attrs; 3319 struct nlattr *stat; 3320 3321 /*filter response w.r.t sk_state*/ 3322 if (!(sk_filter_state & (1 << sk->sk_state))) 3323 return 0; 3324 3325 attrs = nla_nest_start(skb, TIPC_NLA_SOCK); 3326 if (!attrs) 3327 goto msg_cancel; 3328 3329 if (__tipc_nl_add_sk_info(skb, tsk)) 3330 goto attr_msg_cancel; 3331 3332 if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) || 3333 nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) || 3334 nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) || 3335 nla_put_u32(skb, TIPC_NLA_SOCK_UID, 3336 from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk), 3337 sock_i_uid(sk))) || 3338 nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE, 3339 tipc_diag_gen_cookie(sk), 3340 TIPC_NLA_SOCK_PAD)) 3341 goto attr_msg_cancel; 3342 3343 stat = nla_nest_start(skb, TIPC_NLA_SOCK_STAT); 3344 if (!stat) 3345 goto attr_msg_cancel; 3346 3347 if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ, 3348 skb_queue_len(&sk->sk_receive_queue)) || 3349 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ, 3350 skb_queue_len(&sk->sk_write_queue)) || 3351 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP, 3352 atomic_read(&sk->sk_drops))) 3353 goto stat_msg_cancel; 3354 3355 if (tsk->cong_link_cnt && 3356 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG)) 3357 goto stat_msg_cancel; 3358 3359 if (tsk_conn_cong(tsk) && 3360 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG)) 3361 goto stat_msg_cancel; 3362 3363 nla_nest_end(skb, stat); 3364 3365 if (tsk->group) 3366 if (tipc_group_fill_sock_diag(tsk->group, skb)) 3367 goto stat_msg_cancel; 3368 3369 nla_nest_end(skb, attrs); 3370 3371 return 0; 3372 3373 stat_msg_cancel: 3374 nla_nest_cancel(skb, stat); 3375 attr_msg_cancel: 3376 nla_nest_cancel(skb, attrs); 3377 msg_cancel: 3378 return -EMSGSIZE; 3379 } 3380 EXPORT_SYMBOL(tipc_sk_fill_sock_diag); 3381 3382 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb) 3383 { 3384 return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk); 3385 } 3386 3387 /* Caller should hold socket lock for the passed tipc socket. */ 3388 static int __tipc_nl_add_sk_publ(struct sk_buff *skb, 3389 struct netlink_callback *cb, 3390 struct publication *publ) 3391 { 3392 void *hdr; 3393 struct nlattr *attrs; 3394 3395 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq, 3396 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET); 3397 if (!hdr) 3398 goto msg_cancel; 3399 3400 attrs = nla_nest_start(skb, TIPC_NLA_PUBL); 3401 if (!attrs) 3402 goto genlmsg_cancel; 3403 3404 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key)) 3405 goto attr_msg_cancel; 3406 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type)) 3407 goto attr_msg_cancel; 3408 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower)) 3409 goto attr_msg_cancel; 3410 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper)) 3411 goto attr_msg_cancel; 3412 3413 nla_nest_end(skb, attrs); 3414 genlmsg_end(skb, hdr); 3415 3416 return 0; 3417 3418 attr_msg_cancel: 3419 nla_nest_cancel(skb, attrs); 3420 genlmsg_cancel: 3421 genlmsg_cancel(skb, hdr); 3422 msg_cancel: 3423 return -EMSGSIZE; 3424 } 3425 3426 /* Caller should hold socket lock for the passed tipc socket. */ 3427 static int __tipc_nl_list_sk_publ(struct sk_buff *skb, 3428 struct netlink_callback *cb, 3429 struct tipc_sock *tsk, u32 *last_publ) 3430 { 3431 int err; 3432 struct publication *p; 3433 3434 if (*last_publ) { 3435 list_for_each_entry(p, &tsk->publications, binding_sock) { 3436 if (p->key == *last_publ) 3437 break; 3438 } 3439 if (p->key != *last_publ) { 3440 /* We never set seq or call nl_dump_check_consistent() 3441 * this means that setting prev_seq here will cause the 3442 * consistence check to fail in the netlink callback 3443 * handler. Resulting in the last NLMSG_DONE message 3444 * having the NLM_F_DUMP_INTR flag set. 3445 */ 3446 cb->prev_seq = 1; 3447 *last_publ = 0; 3448 return -EPIPE; 3449 } 3450 } else { 3451 p = list_first_entry(&tsk->publications, struct publication, 3452 binding_sock); 3453 } 3454 3455 list_for_each_entry_from(p, &tsk->publications, binding_sock) { 3456 err = __tipc_nl_add_sk_publ(skb, cb, p); 3457 if (err) { 3458 *last_publ = p->key; 3459 return err; 3460 } 3461 } 3462 *last_publ = 0; 3463 3464 return 0; 3465 } 3466 3467 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb) 3468 { 3469 int err; 3470 u32 tsk_portid = cb->args[0]; 3471 u32 last_publ = cb->args[1]; 3472 u32 done = cb->args[2]; 3473 struct net *net = sock_net(skb->sk); 3474 struct tipc_sock *tsk; 3475 3476 if (!tsk_portid) { 3477 struct nlattr **attrs; 3478 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1]; 3479 3480 err = tipc_nlmsg_parse(cb->nlh, &attrs); 3481 if (err) 3482 return err; 3483 3484 if (!attrs[TIPC_NLA_SOCK]) 3485 return -EINVAL; 3486 3487 err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX, 3488 attrs[TIPC_NLA_SOCK], 3489 tipc_nl_sock_policy, NULL); 3490 if (err) 3491 return err; 3492 3493 if (!sock[TIPC_NLA_SOCK_REF]) 3494 return -EINVAL; 3495 3496 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]); 3497 } 3498 3499 if (done) 3500 return 0; 3501 3502 tsk = tipc_sk_lookup(net, tsk_portid); 3503 if (!tsk) 3504 return -EINVAL; 3505 3506 lock_sock(&tsk->sk); 3507 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ); 3508 if (!err) 3509 done = 1; 3510 release_sock(&tsk->sk); 3511 sock_put(&tsk->sk); 3512 3513 cb->args[0] = tsk_portid; 3514 cb->args[1] = last_publ; 3515 cb->args[2] = done; 3516 3517 return skb->len; 3518 } 3519