1 /* 2 * net/tipc/socket.c: TIPC socket API 3 * 4 * Copyright (c) 2001-2007, 2012-2019, Ericsson AB 5 * Copyright (c) 2004-2008, 2010-2013, Wind River Systems 6 * Copyright (c) 2020-2021, Red Hat Inc 7 * All rights reserved. 8 * 9 * Redistribution and use in source and binary forms, with or without 10 * modification, are permitted provided that the following conditions are met: 11 * 12 * 1. Redistributions of source code must retain the above copyright 13 * notice, this list of conditions and the following disclaimer. 14 * 2. Redistributions in binary form must reproduce the above copyright 15 * notice, this list of conditions and the following disclaimer in the 16 * documentation and/or other materials provided with the distribution. 17 * 3. Neither the names of the copyright holders nor the names of its 18 * contributors may be used to endorse or promote products derived from 19 * this software without specific prior written permission. 20 * 21 * Alternatively, this software may be distributed under the terms of the 22 * GNU General Public License ("GPL") version 2 as published by the Free 23 * Software Foundation. 24 * 25 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 26 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 28 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE 29 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 30 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 31 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 32 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 33 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 34 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 35 * POSSIBILITY OF SUCH DAMAGE. 36 */ 37 38 #include <linux/rhashtable.h> 39 #include <linux/sched/signal.h> 40 41 #include "core.h" 42 #include "name_table.h" 43 #include "node.h" 44 #include "link.h" 45 #include "name_distr.h" 46 #include "socket.h" 47 #include "bcast.h" 48 #include "netlink.h" 49 #include "group.h" 50 #include "trace.h" 51 52 #define NAGLE_START_INIT 4 53 #define NAGLE_START_MAX 1024 54 #define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */ 55 #define CONN_PROBING_INTV msecs_to_jiffies(3600000) /* [ms] => 1 h */ 56 #define TIPC_MAX_PORT 0xffffffff 57 #define TIPC_MIN_PORT 1 58 #define TIPC_ACK_RATE 4 /* ACK at 1/4 of rcv window size */ 59 60 enum { 61 TIPC_LISTEN = TCP_LISTEN, 62 TIPC_ESTABLISHED = TCP_ESTABLISHED, 63 TIPC_OPEN = TCP_CLOSE, 64 TIPC_DISCONNECTING = TCP_CLOSE_WAIT, 65 TIPC_CONNECTING = TCP_SYN_SENT, 66 }; 67 68 struct sockaddr_pair { 69 struct sockaddr_tipc sock; 70 struct sockaddr_tipc member; 71 }; 72 73 /** 74 * struct tipc_sock - TIPC socket structure 75 * @sk: socket - interacts with 'port' and with user via the socket API 76 * @max_pkt: maximum packet size "hint" used when building messages sent by port 77 * @maxnagle: maximum size of msg which can be subject to nagle 78 * @portid: unique port identity in TIPC socket hash table 79 * @phdr: preformatted message header used when sending messages 80 * @cong_links: list of congested links 81 * @publications: list of publications for port 82 * @blocking_link: address of the congested link we are currently sleeping on 83 * @pub_count: total # of publications port has made during its lifetime 84 * @conn_timeout: the time we can wait for an unresponded setup request 85 * @probe_unacked: probe has not received ack yet 86 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue 87 * @cong_link_cnt: number of congested links 88 * @snt_unacked: # messages sent by socket, and not yet acked by peer 89 * @snd_win: send window size 90 * @peer_caps: peer capabilities mask 91 * @rcv_unacked: # messages read by user, but not yet acked back to peer 92 * @rcv_win: receive window size 93 * @peer: 'connected' peer for dgram/rdm 94 * @node: hash table node 95 * @mc_method: cookie for use between socket and broadcast layer 96 * @rcu: rcu struct for tipc_sock 97 * @group: TIPC communications group 98 * @oneway: message count in one direction (FIXME) 99 * @nagle_start: current nagle value 100 * @snd_backlog: send backlog count 101 * @msg_acc: messages accepted; used in managing backlog and nagle 102 * @pkt_cnt: TIPC socket packet count 103 * @expect_ack: whether this TIPC socket is expecting an ack 104 * @nodelay: setsockopt() TIPC_NODELAY setting 105 * @group_is_open: TIPC socket group is fully open (FIXME) 106 * @published: true if port has one or more associated names 107 * @conn_addrtype: address type used when establishing connection 108 */ 109 struct tipc_sock { 110 struct sock sk; 111 u32 max_pkt; 112 u32 maxnagle; 113 u32 portid; 114 struct tipc_msg phdr; 115 struct list_head cong_links; 116 struct list_head publications; 117 u32 pub_count; 118 atomic_t dupl_rcvcnt; 119 u16 conn_timeout; 120 bool probe_unacked; 121 u16 cong_link_cnt; 122 u16 snt_unacked; 123 u16 snd_win; 124 u16 peer_caps; 125 u16 rcv_unacked; 126 u16 rcv_win; 127 struct sockaddr_tipc peer; 128 struct rhash_head node; 129 struct tipc_mc_method mc_method; 130 struct rcu_head rcu; 131 struct tipc_group *group; 132 u32 oneway; 133 u32 nagle_start; 134 u16 snd_backlog; 135 u16 msg_acc; 136 u16 pkt_cnt; 137 bool expect_ack; 138 bool nodelay; 139 bool group_is_open; 140 bool published; 141 u8 conn_addrtype; 142 }; 143 144 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb); 145 static void tipc_data_ready(struct sock *sk); 146 static void tipc_write_space(struct sock *sk); 147 static void tipc_sock_destruct(struct sock *sk); 148 static int tipc_release(struct socket *sock); 149 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags, 150 bool kern); 151 static void tipc_sk_timeout(struct timer_list *t); 152 static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua); 153 static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua); 154 static int tipc_sk_leave(struct tipc_sock *tsk); 155 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid); 156 static int tipc_sk_insert(struct tipc_sock *tsk); 157 static void tipc_sk_remove(struct tipc_sock *tsk); 158 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz); 159 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz); 160 static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack); 161 162 static const struct proto_ops packet_ops; 163 static const struct proto_ops stream_ops; 164 static const struct proto_ops msg_ops; 165 static struct proto tipc_proto; 166 static const struct rhashtable_params tsk_rht_params; 167 168 static u32 tsk_own_node(struct tipc_sock *tsk) 169 { 170 return msg_prevnode(&tsk->phdr); 171 } 172 173 static u32 tsk_peer_node(struct tipc_sock *tsk) 174 { 175 return msg_destnode(&tsk->phdr); 176 } 177 178 static u32 tsk_peer_port(struct tipc_sock *tsk) 179 { 180 return msg_destport(&tsk->phdr); 181 } 182 183 static bool tsk_unreliable(struct tipc_sock *tsk) 184 { 185 return msg_src_droppable(&tsk->phdr) != 0; 186 } 187 188 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable) 189 { 190 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0); 191 } 192 193 static bool tsk_unreturnable(struct tipc_sock *tsk) 194 { 195 return msg_dest_droppable(&tsk->phdr) != 0; 196 } 197 198 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable) 199 { 200 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0); 201 } 202 203 static int tsk_importance(struct tipc_sock *tsk) 204 { 205 return msg_importance(&tsk->phdr); 206 } 207 208 static struct tipc_sock *tipc_sk(const struct sock *sk) 209 { 210 return container_of(sk, struct tipc_sock, sk); 211 } 212 213 int tsk_set_importance(struct sock *sk, int imp) 214 { 215 if (imp > TIPC_CRITICAL_IMPORTANCE) 216 return -EINVAL; 217 msg_set_importance(&tipc_sk(sk)->phdr, (u32)imp); 218 return 0; 219 } 220 221 static bool tsk_conn_cong(struct tipc_sock *tsk) 222 { 223 return tsk->snt_unacked > tsk->snd_win; 224 } 225 226 static u16 tsk_blocks(int len) 227 { 228 return ((len / FLOWCTL_BLK_SZ) + 1); 229 } 230 231 /* tsk_blocks(): translate a buffer size in bytes to number of 232 * advertisable blocks, taking into account the ratio truesize(len)/len 233 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ 234 */ 235 static u16 tsk_adv_blocks(int len) 236 { 237 return len / FLOWCTL_BLK_SZ / 4; 238 } 239 240 /* tsk_inc(): increment counter for sent or received data 241 * - If block based flow control is not supported by peer we 242 * fall back to message based ditto, incrementing the counter 243 */ 244 static u16 tsk_inc(struct tipc_sock *tsk, int msglen) 245 { 246 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL)) 247 return ((msglen / FLOWCTL_BLK_SZ) + 1); 248 return 1; 249 } 250 251 /* tsk_set_nagle - enable/disable nagle property by manipulating maxnagle 252 */ 253 static void tsk_set_nagle(struct tipc_sock *tsk) 254 { 255 struct sock *sk = &tsk->sk; 256 257 tsk->maxnagle = 0; 258 if (sk->sk_type != SOCK_STREAM) 259 return; 260 if (tsk->nodelay) 261 return; 262 if (!(tsk->peer_caps & TIPC_NAGLE)) 263 return; 264 /* Limit node local buffer size to avoid receive queue overflow */ 265 if (tsk->max_pkt == MAX_MSG_SIZE) 266 tsk->maxnagle = 1500; 267 else 268 tsk->maxnagle = tsk->max_pkt; 269 } 270 271 /** 272 * tsk_advance_rx_queue - discard first buffer in socket receive queue 273 * @sk: network socket 274 * 275 * Caller must hold socket lock 276 */ 277 static void tsk_advance_rx_queue(struct sock *sk) 278 { 279 trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " "); 280 kfree_skb(__skb_dequeue(&sk->sk_receive_queue)); 281 } 282 283 /* tipc_sk_respond() : send response message back to sender 284 */ 285 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err) 286 { 287 u32 selector; 288 u32 dnode; 289 u32 onode = tipc_own_addr(sock_net(sk)); 290 291 if (!tipc_msg_reverse(onode, &skb, err)) 292 return; 293 294 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!"); 295 dnode = msg_destnode(buf_msg(skb)); 296 selector = msg_origport(buf_msg(skb)); 297 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector); 298 } 299 300 /** 301 * tsk_rej_rx_queue - reject all buffers in socket receive queue 302 * @sk: network socket 303 * @error: response error code 304 * 305 * Caller must hold socket lock 306 */ 307 static void tsk_rej_rx_queue(struct sock *sk, int error) 308 { 309 struct sk_buff *skb; 310 311 while ((skb = __skb_dequeue(&sk->sk_receive_queue))) 312 tipc_sk_respond(sk, skb, error); 313 } 314 315 static bool tipc_sk_connected(struct sock *sk) 316 { 317 return sk->sk_state == TIPC_ESTABLISHED; 318 } 319 320 /* tipc_sk_type_connectionless - check if the socket is datagram socket 321 * @sk: socket 322 * 323 * Returns true if connection less, false otherwise 324 */ 325 static bool tipc_sk_type_connectionless(struct sock *sk) 326 { 327 return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM; 328 } 329 330 /* tsk_peer_msg - verify if message was sent by connected port's peer 331 * 332 * Handles cases where the node's network address has changed from 333 * the default of <0.0.0> to its configured setting. 334 */ 335 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg) 336 { 337 struct sock *sk = &tsk->sk; 338 u32 self = tipc_own_addr(sock_net(sk)); 339 u32 peer_port = tsk_peer_port(tsk); 340 u32 orig_node, peer_node; 341 342 if (unlikely(!tipc_sk_connected(sk))) 343 return false; 344 345 if (unlikely(msg_origport(msg) != peer_port)) 346 return false; 347 348 orig_node = msg_orignode(msg); 349 peer_node = tsk_peer_node(tsk); 350 351 if (likely(orig_node == peer_node)) 352 return true; 353 354 if (!orig_node && peer_node == self) 355 return true; 356 357 if (!peer_node && orig_node == self) 358 return true; 359 360 return false; 361 } 362 363 /* tipc_set_sk_state - set the sk_state of the socket 364 * @sk: socket 365 * 366 * Caller must hold socket lock 367 * 368 * Returns 0 on success, errno otherwise 369 */ 370 static int tipc_set_sk_state(struct sock *sk, int state) 371 { 372 int oldsk_state = sk->sk_state; 373 int res = -EINVAL; 374 375 switch (state) { 376 case TIPC_OPEN: 377 res = 0; 378 break; 379 case TIPC_LISTEN: 380 case TIPC_CONNECTING: 381 if (oldsk_state == TIPC_OPEN) 382 res = 0; 383 break; 384 case TIPC_ESTABLISHED: 385 if (oldsk_state == TIPC_CONNECTING || 386 oldsk_state == TIPC_OPEN) 387 res = 0; 388 break; 389 case TIPC_DISCONNECTING: 390 if (oldsk_state == TIPC_CONNECTING || 391 oldsk_state == TIPC_ESTABLISHED) 392 res = 0; 393 break; 394 } 395 396 if (!res) 397 sk->sk_state = state; 398 399 return res; 400 } 401 402 static int tipc_sk_sock_err(struct socket *sock, long *timeout) 403 { 404 struct sock *sk = sock->sk; 405 int err = sock_error(sk); 406 int typ = sock->type; 407 408 if (err) 409 return err; 410 if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) { 411 if (sk->sk_state == TIPC_DISCONNECTING) 412 return -EPIPE; 413 else if (!tipc_sk_connected(sk)) 414 return -ENOTCONN; 415 } 416 if (!*timeout) 417 return -EAGAIN; 418 if (signal_pending(current)) 419 return sock_intr_errno(*timeout); 420 421 return 0; 422 } 423 424 #define tipc_wait_for_cond(sock_, timeo_, condition_) \ 425 ({ \ 426 DEFINE_WAIT_FUNC(wait_, woken_wake_function); \ 427 struct sock *sk_; \ 428 int rc_; \ 429 \ 430 while ((rc_ = !(condition_))) { \ 431 /* coupled with smp_wmb() in tipc_sk_proto_rcv() */ \ 432 smp_rmb(); \ 433 sk_ = (sock_)->sk; \ 434 rc_ = tipc_sk_sock_err((sock_), timeo_); \ 435 if (rc_) \ 436 break; \ 437 add_wait_queue(sk_sleep(sk_), &wait_); \ 438 release_sock(sk_); \ 439 *(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \ 440 sched_annotate_sleep(); \ 441 lock_sock(sk_); \ 442 remove_wait_queue(sk_sleep(sk_), &wait_); \ 443 } \ 444 rc_; \ 445 }) 446 447 /** 448 * tipc_sk_create - create a TIPC socket 449 * @net: network namespace (must be default network) 450 * @sock: pre-allocated socket structure 451 * @protocol: protocol indicator (must be 0) 452 * @kern: caused by kernel or by userspace? 453 * 454 * This routine creates additional data structures used by the TIPC socket, 455 * initializes them, and links them together. 456 * 457 * Return: 0 on success, errno otherwise 458 */ 459 static int tipc_sk_create(struct net *net, struct socket *sock, 460 int protocol, int kern) 461 { 462 const struct proto_ops *ops; 463 struct sock *sk; 464 struct tipc_sock *tsk; 465 struct tipc_msg *msg; 466 467 /* Validate arguments */ 468 if (unlikely(protocol != 0)) 469 return -EPROTONOSUPPORT; 470 471 switch (sock->type) { 472 case SOCK_STREAM: 473 ops = &stream_ops; 474 break; 475 case SOCK_SEQPACKET: 476 ops = &packet_ops; 477 break; 478 case SOCK_DGRAM: 479 case SOCK_RDM: 480 ops = &msg_ops; 481 break; 482 default: 483 return -EPROTOTYPE; 484 } 485 486 /* Allocate socket's protocol area */ 487 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern); 488 if (sk == NULL) 489 return -ENOMEM; 490 491 tsk = tipc_sk(sk); 492 tsk->max_pkt = MAX_PKT_DEFAULT; 493 tsk->maxnagle = 0; 494 tsk->nagle_start = NAGLE_START_INIT; 495 INIT_LIST_HEAD(&tsk->publications); 496 INIT_LIST_HEAD(&tsk->cong_links); 497 msg = &tsk->phdr; 498 499 /* Finish initializing socket data structures */ 500 sock->ops = ops; 501 sock_init_data(sock, sk); 502 tipc_set_sk_state(sk, TIPC_OPEN); 503 if (tipc_sk_insert(tsk)) { 504 pr_warn("Socket create failed; port number exhausted\n"); 505 return -EINVAL; 506 } 507 508 /* Ensure tsk is visible before we read own_addr. */ 509 smp_mb(); 510 511 tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE, 512 TIPC_NAMED_MSG, NAMED_H_SIZE, 0); 513 514 msg_set_origport(msg, tsk->portid); 515 timer_setup(&sk->sk_timer, tipc_sk_timeout, 0); 516 sk->sk_shutdown = 0; 517 sk->sk_backlog_rcv = tipc_sk_backlog_rcv; 518 sk->sk_rcvbuf = sysctl_tipc_rmem[1]; 519 sk->sk_data_ready = tipc_data_ready; 520 sk->sk_write_space = tipc_write_space; 521 sk->sk_destruct = tipc_sock_destruct; 522 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT; 523 tsk->group_is_open = true; 524 atomic_set(&tsk->dupl_rcvcnt, 0); 525 526 /* Start out with safe limits until we receive an advertised window */ 527 tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN); 528 tsk->rcv_win = tsk->snd_win; 529 530 if (tipc_sk_type_connectionless(sk)) { 531 tsk_set_unreturnable(tsk, true); 532 if (sock->type == SOCK_DGRAM) 533 tsk_set_unreliable(tsk, true); 534 } 535 __skb_queue_head_init(&tsk->mc_method.deferredq); 536 trace_tipc_sk_create(sk, NULL, TIPC_DUMP_NONE, " "); 537 return 0; 538 } 539 540 static void tipc_sk_callback(struct rcu_head *head) 541 { 542 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu); 543 544 sock_put(&tsk->sk); 545 } 546 547 /* Caller should hold socket lock for the socket. */ 548 static void __tipc_shutdown(struct socket *sock, int error) 549 { 550 struct sock *sk = sock->sk; 551 struct tipc_sock *tsk = tipc_sk(sk); 552 struct net *net = sock_net(sk); 553 long timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT); 554 u32 dnode = tsk_peer_node(tsk); 555 struct sk_buff *skb; 556 557 /* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */ 558 tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt && 559 !tsk_conn_cong(tsk))); 560 561 /* Push out delayed messages if in Nagle mode */ 562 tipc_sk_push_backlog(tsk, false); 563 /* Remove pending SYN */ 564 __skb_queue_purge(&sk->sk_write_queue); 565 566 /* Remove partially received buffer if any */ 567 skb = skb_peek(&sk->sk_receive_queue); 568 if (skb && TIPC_SKB_CB(skb)->bytes_read) { 569 __skb_unlink(skb, &sk->sk_receive_queue); 570 kfree_skb(skb); 571 } 572 573 /* Reject all unreceived messages if connectionless */ 574 if (tipc_sk_type_connectionless(sk)) { 575 tsk_rej_rx_queue(sk, error); 576 return; 577 } 578 579 switch (sk->sk_state) { 580 case TIPC_CONNECTING: 581 case TIPC_ESTABLISHED: 582 tipc_set_sk_state(sk, TIPC_DISCONNECTING); 583 tipc_node_remove_conn(net, dnode, tsk->portid); 584 /* Send a FIN+/- to its peer */ 585 skb = __skb_dequeue(&sk->sk_receive_queue); 586 if (skb) { 587 __skb_queue_purge(&sk->sk_receive_queue); 588 tipc_sk_respond(sk, skb, error); 589 break; 590 } 591 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE, 592 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode, 593 tsk_own_node(tsk), tsk_peer_port(tsk), 594 tsk->portid, error); 595 if (skb) 596 tipc_node_xmit_skb(net, skb, dnode, tsk->portid); 597 break; 598 case TIPC_LISTEN: 599 /* Reject all SYN messages */ 600 tsk_rej_rx_queue(sk, error); 601 break; 602 default: 603 __skb_queue_purge(&sk->sk_receive_queue); 604 break; 605 } 606 } 607 608 /** 609 * tipc_release - destroy a TIPC socket 610 * @sock: socket to destroy 611 * 612 * This routine cleans up any messages that are still queued on the socket. 613 * For DGRAM and RDM socket types, all queued messages are rejected. 614 * For SEQPACKET and STREAM socket types, the first message is rejected 615 * and any others are discarded. (If the first message on a STREAM socket 616 * is partially-read, it is discarded and the next one is rejected instead.) 617 * 618 * NOTE: Rejected messages are not necessarily returned to the sender! They 619 * are returned or discarded according to the "destination droppable" setting 620 * specified for the message by the sender. 621 * 622 * Return: 0 on success, errno otherwise 623 */ 624 static int tipc_release(struct socket *sock) 625 { 626 struct sock *sk = sock->sk; 627 struct tipc_sock *tsk; 628 629 /* 630 * Exit if socket isn't fully initialized (occurs when a failed accept() 631 * releases a pre-allocated child socket that was never used) 632 */ 633 if (sk == NULL) 634 return 0; 635 636 tsk = tipc_sk(sk); 637 lock_sock(sk); 638 639 trace_tipc_sk_release(sk, NULL, TIPC_DUMP_ALL, " "); 640 __tipc_shutdown(sock, TIPC_ERR_NO_PORT); 641 sk->sk_shutdown = SHUTDOWN_MASK; 642 tipc_sk_leave(tsk); 643 tipc_sk_withdraw(tsk, NULL); 644 __skb_queue_purge(&tsk->mc_method.deferredq); 645 sk_stop_timer(sk, &sk->sk_timer); 646 tipc_sk_remove(tsk); 647 648 sock_orphan(sk); 649 /* Reject any messages that accumulated in backlog queue */ 650 release_sock(sk); 651 tipc_dest_list_purge(&tsk->cong_links); 652 tsk->cong_link_cnt = 0; 653 call_rcu(&tsk->rcu, tipc_sk_callback); 654 sock->sk = NULL; 655 656 return 0; 657 } 658 659 /** 660 * __tipc_bind - associate or disassocate TIPC name(s) with a socket 661 * @sock: socket structure 662 * @skaddr: socket address describing name(s) and desired operation 663 * @alen: size of socket address data structure 664 * 665 * Name and name sequence binding are indicated using a positive scope value; 666 * a negative scope value unbinds the specified name. Specifying no name 667 * (i.e. a socket address length of 0) unbinds all names from the socket. 668 * 669 * Return: 0 on success, errno otherwise 670 * 671 * NOTE: This routine doesn't need to take the socket lock since it doesn't 672 * access any non-constant socket information. 673 */ 674 static int __tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen) 675 { 676 struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr; 677 struct tipc_sock *tsk = tipc_sk(sock->sk); 678 bool unbind = false; 679 680 if (unlikely(!alen)) 681 return tipc_sk_withdraw(tsk, NULL); 682 683 if (ua->addrtype == TIPC_SERVICE_ADDR) { 684 ua->addrtype = TIPC_SERVICE_RANGE; 685 ua->sr.upper = ua->sr.lower; 686 } 687 if (ua->scope < 0) { 688 unbind = true; 689 ua->scope = -ua->scope; 690 } 691 /* Users may still use deprecated TIPC_ZONE_SCOPE */ 692 if (ua->scope != TIPC_NODE_SCOPE) 693 ua->scope = TIPC_CLUSTER_SCOPE; 694 695 if (tsk->group) 696 return -EACCES; 697 698 if (unbind) 699 return tipc_sk_withdraw(tsk, ua); 700 return tipc_sk_publish(tsk, ua); 701 } 702 703 int tipc_sk_bind(struct socket *sock, struct sockaddr *skaddr, int alen) 704 { 705 int res; 706 707 lock_sock(sock->sk); 708 res = __tipc_bind(sock, skaddr, alen); 709 release_sock(sock->sk); 710 return res; 711 } 712 713 static int tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen) 714 { 715 struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr; 716 u32 atype = ua->addrtype; 717 718 if (alen) { 719 if (!tipc_uaddr_valid(ua, alen)) 720 return -EINVAL; 721 if (atype == TIPC_SOCKET_ADDR) 722 return -EAFNOSUPPORT; 723 if (ua->sr.type < TIPC_RESERVED_TYPES) { 724 pr_warn_once("Can't bind to reserved service type %u\n", 725 ua->sr.type); 726 return -EACCES; 727 } 728 } 729 return tipc_sk_bind(sock, skaddr, alen); 730 } 731 732 /** 733 * tipc_getname - get port ID of socket or peer socket 734 * @sock: socket structure 735 * @uaddr: area for returned socket address 736 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID 737 * 738 * Return: 0 on success, errno otherwise 739 * 740 * NOTE: This routine doesn't need to take the socket lock since it only 741 * accesses socket information that is unchanging (or which changes in 742 * a completely predictable manner). 743 */ 744 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr, 745 int peer) 746 { 747 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr; 748 struct sock *sk = sock->sk; 749 struct tipc_sock *tsk = tipc_sk(sk); 750 751 memset(addr, 0, sizeof(*addr)); 752 if (peer) { 753 if ((!tipc_sk_connected(sk)) && 754 ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING))) 755 return -ENOTCONN; 756 addr->addr.id.ref = tsk_peer_port(tsk); 757 addr->addr.id.node = tsk_peer_node(tsk); 758 } else { 759 addr->addr.id.ref = tsk->portid; 760 addr->addr.id.node = tipc_own_addr(sock_net(sk)); 761 } 762 763 addr->addrtype = TIPC_SOCKET_ADDR; 764 addr->family = AF_TIPC; 765 addr->scope = 0; 766 addr->addr.name.domain = 0; 767 768 return sizeof(*addr); 769 } 770 771 /** 772 * tipc_poll - read and possibly block on pollmask 773 * @file: file structure associated with the socket 774 * @sock: socket for which to calculate the poll bits 775 * @wait: ??? 776 * 777 * Return: pollmask value 778 * 779 * COMMENTARY: 780 * It appears that the usual socket locking mechanisms are not useful here 781 * since the pollmask info is potentially out-of-date the moment this routine 782 * exits. TCP and other protocols seem to rely on higher level poll routines 783 * to handle any preventable race conditions, so TIPC will do the same ... 784 * 785 * IMPORTANT: The fact that a read or write operation is indicated does NOT 786 * imply that the operation will succeed, merely that it should be performed 787 * and will not block. 788 */ 789 static __poll_t tipc_poll(struct file *file, struct socket *sock, 790 poll_table *wait) 791 { 792 struct sock *sk = sock->sk; 793 struct tipc_sock *tsk = tipc_sk(sk); 794 __poll_t revents = 0; 795 796 sock_poll_wait(file, sock, wait); 797 trace_tipc_sk_poll(sk, NULL, TIPC_DUMP_ALL, " "); 798 799 if (sk->sk_shutdown & RCV_SHUTDOWN) 800 revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM; 801 if (sk->sk_shutdown == SHUTDOWN_MASK) 802 revents |= EPOLLHUP; 803 804 switch (sk->sk_state) { 805 case TIPC_ESTABLISHED: 806 if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk)) 807 revents |= EPOLLOUT; 808 fallthrough; 809 case TIPC_LISTEN: 810 case TIPC_CONNECTING: 811 if (!skb_queue_empty_lockless(&sk->sk_receive_queue)) 812 revents |= EPOLLIN | EPOLLRDNORM; 813 break; 814 case TIPC_OPEN: 815 if (tsk->group_is_open && !tsk->cong_link_cnt) 816 revents |= EPOLLOUT; 817 if (!tipc_sk_type_connectionless(sk)) 818 break; 819 if (skb_queue_empty_lockless(&sk->sk_receive_queue)) 820 break; 821 revents |= EPOLLIN | EPOLLRDNORM; 822 break; 823 case TIPC_DISCONNECTING: 824 revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP; 825 break; 826 } 827 return revents; 828 } 829 830 /** 831 * tipc_sendmcast - send multicast message 832 * @sock: socket structure 833 * @ua: destination address struct 834 * @msg: message to send 835 * @dlen: length of data to send 836 * @timeout: timeout to wait for wakeup 837 * 838 * Called from function tipc_sendmsg(), which has done all sanity checks 839 * Return: the number of bytes sent on success, or errno 840 */ 841 static int tipc_sendmcast(struct socket *sock, struct tipc_uaddr *ua, 842 struct msghdr *msg, size_t dlen, long timeout) 843 { 844 struct sock *sk = sock->sk; 845 struct tipc_sock *tsk = tipc_sk(sk); 846 struct tipc_msg *hdr = &tsk->phdr; 847 struct net *net = sock_net(sk); 848 int mtu = tipc_bcast_get_mtu(net); 849 struct sk_buff_head pkts; 850 struct tipc_nlist dsts; 851 int rc; 852 853 if (tsk->group) 854 return -EACCES; 855 856 /* Block or return if any destination link is congested */ 857 rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt); 858 if (unlikely(rc)) 859 return rc; 860 861 /* Lookup destination nodes */ 862 tipc_nlist_init(&dsts, tipc_own_addr(net)); 863 tipc_nametbl_lookup_mcast_nodes(net, ua, &dsts); 864 if (!dsts.local && !dsts.remote) 865 return -EHOSTUNREACH; 866 867 /* Build message header */ 868 msg_set_type(hdr, TIPC_MCAST_MSG); 869 msg_set_hdr_sz(hdr, MCAST_H_SIZE); 870 msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE); 871 msg_set_destport(hdr, 0); 872 msg_set_destnode(hdr, 0); 873 msg_set_nametype(hdr, ua->sr.type); 874 msg_set_namelower(hdr, ua->sr.lower); 875 msg_set_nameupper(hdr, ua->sr.upper); 876 877 /* Build message as chain of buffers */ 878 __skb_queue_head_init(&pkts); 879 rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts); 880 881 /* Send message if build was successful */ 882 if (unlikely(rc == dlen)) { 883 trace_tipc_sk_sendmcast(sk, skb_peek(&pkts), 884 TIPC_DUMP_SK_SNDQ, " "); 885 rc = tipc_mcast_xmit(net, &pkts, &tsk->mc_method, &dsts, 886 &tsk->cong_link_cnt); 887 } 888 889 tipc_nlist_purge(&dsts); 890 891 return rc ? rc : dlen; 892 } 893 894 /** 895 * tipc_send_group_msg - send a message to a member in the group 896 * @net: network namespace 897 * @tsk: tipc socket 898 * @m: message to send 899 * @mb: group member 900 * @dnode: destination node 901 * @dport: destination port 902 * @dlen: total length of message data 903 */ 904 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk, 905 struct msghdr *m, struct tipc_member *mb, 906 u32 dnode, u32 dport, int dlen) 907 { 908 u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group); 909 struct tipc_mc_method *method = &tsk->mc_method; 910 int blks = tsk_blocks(GROUP_H_SIZE + dlen); 911 struct tipc_msg *hdr = &tsk->phdr; 912 struct sk_buff_head pkts; 913 int mtu, rc; 914 915 /* Complete message header */ 916 msg_set_type(hdr, TIPC_GRP_UCAST_MSG); 917 msg_set_hdr_sz(hdr, GROUP_H_SIZE); 918 msg_set_destport(hdr, dport); 919 msg_set_destnode(hdr, dnode); 920 msg_set_grp_bc_seqno(hdr, bc_snd_nxt); 921 922 /* Build message as chain of buffers */ 923 __skb_queue_head_init(&pkts); 924 mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false); 925 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts); 926 if (unlikely(rc != dlen)) 927 return rc; 928 929 /* Send message */ 930 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid); 931 if (unlikely(rc == -ELINKCONG)) { 932 tipc_dest_push(&tsk->cong_links, dnode, 0); 933 tsk->cong_link_cnt++; 934 } 935 936 /* Update send window */ 937 tipc_group_update_member(mb, blks); 938 939 /* A broadcast sent within next EXPIRE period must follow same path */ 940 method->rcast = true; 941 method->mandatory = true; 942 return dlen; 943 } 944 945 /** 946 * tipc_send_group_unicast - send message to a member in the group 947 * @sock: socket structure 948 * @m: message to send 949 * @dlen: total length of message data 950 * @timeout: timeout to wait for wakeup 951 * 952 * Called from function tipc_sendmsg(), which has done all sanity checks 953 * Return: the number of bytes sent on success, or errno 954 */ 955 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m, 956 int dlen, long timeout) 957 { 958 struct sock *sk = sock->sk; 959 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name; 960 int blks = tsk_blocks(GROUP_H_SIZE + dlen); 961 struct tipc_sock *tsk = tipc_sk(sk); 962 struct net *net = sock_net(sk); 963 struct tipc_member *mb = NULL; 964 u32 node, port; 965 int rc; 966 967 node = ua->sk.node; 968 port = ua->sk.ref; 969 if (!port && !node) 970 return -EHOSTUNREACH; 971 972 /* Block or return if destination link or member is congested */ 973 rc = tipc_wait_for_cond(sock, &timeout, 974 !tipc_dest_find(&tsk->cong_links, node, 0) && 975 tsk->group && 976 !tipc_group_cong(tsk->group, node, port, blks, 977 &mb)); 978 if (unlikely(rc)) 979 return rc; 980 981 if (unlikely(!mb)) 982 return -EHOSTUNREACH; 983 984 rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen); 985 986 return rc ? rc : dlen; 987 } 988 989 /** 990 * tipc_send_group_anycast - send message to any member with given identity 991 * @sock: socket structure 992 * @m: message to send 993 * @dlen: total length of message data 994 * @timeout: timeout to wait for wakeup 995 * 996 * Called from function tipc_sendmsg(), which has done all sanity checks 997 * Return: the number of bytes sent on success, or errno 998 */ 999 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m, 1000 int dlen, long timeout) 1001 { 1002 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name; 1003 struct sock *sk = sock->sk; 1004 struct tipc_sock *tsk = tipc_sk(sk); 1005 struct list_head *cong_links = &tsk->cong_links; 1006 int blks = tsk_blocks(GROUP_H_SIZE + dlen); 1007 struct tipc_msg *hdr = &tsk->phdr; 1008 struct tipc_member *first = NULL; 1009 struct tipc_member *mbr = NULL; 1010 struct net *net = sock_net(sk); 1011 u32 node, port, exclude; 1012 struct list_head dsts; 1013 int lookups = 0; 1014 int dstcnt, rc; 1015 bool cong; 1016 1017 INIT_LIST_HEAD(&dsts); 1018 ua->sa.type = msg_nametype(hdr); 1019 ua->scope = msg_lookup_scope(hdr); 1020 1021 while (++lookups < 4) { 1022 exclude = tipc_group_exclude(tsk->group); 1023 1024 first = NULL; 1025 1026 /* Look for a non-congested destination member, if any */ 1027 while (1) { 1028 if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt, 1029 exclude, false)) 1030 return -EHOSTUNREACH; 1031 tipc_dest_pop(&dsts, &node, &port); 1032 cong = tipc_group_cong(tsk->group, node, port, blks, 1033 &mbr); 1034 if (!cong) 1035 break; 1036 if (mbr == first) 1037 break; 1038 if (!first) 1039 first = mbr; 1040 } 1041 1042 /* Start over if destination was not in member list */ 1043 if (unlikely(!mbr)) 1044 continue; 1045 1046 if (likely(!cong && !tipc_dest_find(cong_links, node, 0))) 1047 break; 1048 1049 /* Block or return if destination link or member is congested */ 1050 rc = tipc_wait_for_cond(sock, &timeout, 1051 !tipc_dest_find(cong_links, node, 0) && 1052 tsk->group && 1053 !tipc_group_cong(tsk->group, node, port, 1054 blks, &mbr)); 1055 if (unlikely(rc)) 1056 return rc; 1057 1058 /* Send, unless destination disappeared while waiting */ 1059 if (likely(mbr)) 1060 break; 1061 } 1062 1063 if (unlikely(lookups >= 4)) 1064 return -EHOSTUNREACH; 1065 1066 rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen); 1067 1068 return rc ? rc : dlen; 1069 } 1070 1071 /** 1072 * tipc_send_group_bcast - send message to all members in communication group 1073 * @sock: socket structure 1074 * @m: message to send 1075 * @dlen: total length of message data 1076 * @timeout: timeout to wait for wakeup 1077 * 1078 * Called from function tipc_sendmsg(), which has done all sanity checks 1079 * Return: the number of bytes sent on success, or errno 1080 */ 1081 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m, 1082 int dlen, long timeout) 1083 { 1084 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name; 1085 struct sock *sk = sock->sk; 1086 struct net *net = sock_net(sk); 1087 struct tipc_sock *tsk = tipc_sk(sk); 1088 struct tipc_nlist *dsts; 1089 struct tipc_mc_method *method = &tsk->mc_method; 1090 bool ack = method->mandatory && method->rcast; 1091 int blks = tsk_blocks(MCAST_H_SIZE + dlen); 1092 struct tipc_msg *hdr = &tsk->phdr; 1093 int mtu = tipc_bcast_get_mtu(net); 1094 struct sk_buff_head pkts; 1095 int rc = -EHOSTUNREACH; 1096 1097 /* Block or return if any destination link or member is congested */ 1098 rc = tipc_wait_for_cond(sock, &timeout, 1099 !tsk->cong_link_cnt && tsk->group && 1100 !tipc_group_bc_cong(tsk->group, blks)); 1101 if (unlikely(rc)) 1102 return rc; 1103 1104 dsts = tipc_group_dests(tsk->group); 1105 if (!dsts->local && !dsts->remote) 1106 return -EHOSTUNREACH; 1107 1108 /* Complete message header */ 1109 if (ua) { 1110 msg_set_type(hdr, TIPC_GRP_MCAST_MSG); 1111 msg_set_nameinst(hdr, ua->sa.instance); 1112 } else { 1113 msg_set_type(hdr, TIPC_GRP_BCAST_MSG); 1114 msg_set_nameinst(hdr, 0); 1115 } 1116 msg_set_hdr_sz(hdr, GROUP_H_SIZE); 1117 msg_set_destport(hdr, 0); 1118 msg_set_destnode(hdr, 0); 1119 msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group)); 1120 1121 /* Avoid getting stuck with repeated forced replicasts */ 1122 msg_set_grp_bc_ack_req(hdr, ack); 1123 1124 /* Build message as chain of buffers */ 1125 __skb_queue_head_init(&pkts); 1126 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts); 1127 if (unlikely(rc != dlen)) 1128 return rc; 1129 1130 /* Send message */ 1131 rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt); 1132 if (unlikely(rc)) 1133 return rc; 1134 1135 /* Update broadcast sequence number and send windows */ 1136 tipc_group_update_bc_members(tsk->group, blks, ack); 1137 1138 /* Broadcast link is now free to choose method for next broadcast */ 1139 method->mandatory = false; 1140 method->expires = jiffies; 1141 1142 return dlen; 1143 } 1144 1145 /** 1146 * tipc_send_group_mcast - send message to all members with given identity 1147 * @sock: socket structure 1148 * @m: message to send 1149 * @dlen: total length of message data 1150 * @timeout: timeout to wait for wakeup 1151 * 1152 * Called from function tipc_sendmsg(), which has done all sanity checks 1153 * Return: the number of bytes sent on success, or errno 1154 */ 1155 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m, 1156 int dlen, long timeout) 1157 { 1158 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name; 1159 struct sock *sk = sock->sk; 1160 struct tipc_sock *tsk = tipc_sk(sk); 1161 struct tipc_group *grp = tsk->group; 1162 struct tipc_msg *hdr = &tsk->phdr; 1163 struct net *net = sock_net(sk); 1164 struct list_head dsts; 1165 u32 dstcnt, exclude; 1166 1167 INIT_LIST_HEAD(&dsts); 1168 ua->sa.type = msg_nametype(hdr); 1169 ua->scope = msg_lookup_scope(hdr); 1170 exclude = tipc_group_exclude(grp); 1171 1172 if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt, exclude, true)) 1173 return -EHOSTUNREACH; 1174 1175 if (dstcnt == 1) { 1176 tipc_dest_pop(&dsts, &ua->sk.node, &ua->sk.ref); 1177 return tipc_send_group_unicast(sock, m, dlen, timeout); 1178 } 1179 1180 tipc_dest_list_purge(&dsts); 1181 return tipc_send_group_bcast(sock, m, dlen, timeout); 1182 } 1183 1184 /** 1185 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets 1186 * @net: the associated network namespace 1187 * @arrvq: queue with arriving messages, to be cloned after destination lookup 1188 * @inputq: queue with cloned messages, delivered to socket after dest lookup 1189 * 1190 * Multi-threaded: parallel calls with reference to same queues may occur 1191 */ 1192 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq, 1193 struct sk_buff_head *inputq) 1194 { 1195 u32 self = tipc_own_addr(net); 1196 struct sk_buff *skb, *_skb; 1197 u32 portid, onode; 1198 struct sk_buff_head tmpq; 1199 struct list_head dports; 1200 struct tipc_msg *hdr; 1201 struct tipc_uaddr ua; 1202 int user, mtyp, hlen; 1203 1204 __skb_queue_head_init(&tmpq); 1205 INIT_LIST_HEAD(&dports); 1206 ua.addrtype = TIPC_SERVICE_RANGE; 1207 1208 /* tipc_skb_peek() increments the head skb's reference counter */ 1209 skb = tipc_skb_peek(arrvq, &inputq->lock); 1210 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) { 1211 hdr = buf_msg(skb); 1212 user = msg_user(hdr); 1213 mtyp = msg_type(hdr); 1214 hlen = skb_headroom(skb) + msg_hdr_sz(hdr); 1215 onode = msg_orignode(hdr); 1216 ua.sr.type = msg_nametype(hdr); 1217 ua.sr.lower = msg_namelower(hdr); 1218 ua.sr.upper = msg_nameupper(hdr); 1219 if (onode == self) 1220 ua.scope = TIPC_ANY_SCOPE; 1221 else 1222 ua.scope = TIPC_CLUSTER_SCOPE; 1223 1224 if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) { 1225 spin_lock_bh(&inputq->lock); 1226 if (skb_peek(arrvq) == skb) { 1227 __skb_dequeue(arrvq); 1228 __skb_queue_tail(inputq, skb); 1229 } 1230 kfree_skb(skb); 1231 spin_unlock_bh(&inputq->lock); 1232 continue; 1233 } 1234 1235 /* Group messages require exact scope match */ 1236 if (msg_in_group(hdr)) { 1237 ua.sr.lower = 0; 1238 ua.sr.upper = ~0; 1239 ua.scope = msg_lookup_scope(hdr); 1240 } 1241 1242 /* Create destination port list: */ 1243 tipc_nametbl_lookup_mcast_sockets(net, &ua, &dports); 1244 1245 /* Clone message per destination */ 1246 while (tipc_dest_pop(&dports, NULL, &portid)) { 1247 _skb = __pskb_copy(skb, hlen, GFP_ATOMIC); 1248 if (_skb) { 1249 msg_set_destport(buf_msg(_skb), portid); 1250 __skb_queue_tail(&tmpq, _skb); 1251 continue; 1252 } 1253 pr_warn("Failed to clone mcast rcv buffer\n"); 1254 } 1255 /* Append clones to inputq only if skb is still head of arrvq */ 1256 spin_lock_bh(&inputq->lock); 1257 if (skb_peek(arrvq) == skb) { 1258 skb_queue_splice_tail_init(&tmpq, inputq); 1259 /* Decrement the skb's refcnt */ 1260 kfree_skb(__skb_dequeue(arrvq)); 1261 } 1262 spin_unlock_bh(&inputq->lock); 1263 __skb_queue_purge(&tmpq); 1264 kfree_skb(skb); 1265 } 1266 tipc_sk_rcv(net, inputq); 1267 } 1268 1269 /* tipc_sk_push_backlog(): send accumulated buffers in socket write queue 1270 * when socket is in Nagle mode 1271 */ 1272 static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack) 1273 { 1274 struct sk_buff_head *txq = &tsk->sk.sk_write_queue; 1275 struct sk_buff *skb = skb_peek_tail(txq); 1276 struct net *net = sock_net(&tsk->sk); 1277 u32 dnode = tsk_peer_node(tsk); 1278 int rc; 1279 1280 if (nagle_ack) { 1281 tsk->pkt_cnt += skb_queue_len(txq); 1282 if (!tsk->pkt_cnt || tsk->msg_acc / tsk->pkt_cnt < 2) { 1283 tsk->oneway = 0; 1284 if (tsk->nagle_start < NAGLE_START_MAX) 1285 tsk->nagle_start *= 2; 1286 tsk->expect_ack = false; 1287 pr_debug("tsk %10u: bad nagle %u -> %u, next start %u!\n", 1288 tsk->portid, tsk->msg_acc, tsk->pkt_cnt, 1289 tsk->nagle_start); 1290 } else { 1291 tsk->nagle_start = NAGLE_START_INIT; 1292 if (skb) { 1293 msg_set_ack_required(buf_msg(skb)); 1294 tsk->expect_ack = true; 1295 } else { 1296 tsk->expect_ack = false; 1297 } 1298 } 1299 tsk->msg_acc = 0; 1300 tsk->pkt_cnt = 0; 1301 } 1302 1303 if (!skb || tsk->cong_link_cnt) 1304 return; 1305 1306 /* Do not send SYN again after congestion */ 1307 if (msg_is_syn(buf_msg(skb))) 1308 return; 1309 1310 if (tsk->msg_acc) 1311 tsk->pkt_cnt += skb_queue_len(txq); 1312 tsk->snt_unacked += tsk->snd_backlog; 1313 tsk->snd_backlog = 0; 1314 rc = tipc_node_xmit(net, txq, dnode, tsk->portid); 1315 if (rc == -ELINKCONG) 1316 tsk->cong_link_cnt = 1; 1317 } 1318 1319 /** 1320 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message 1321 * @tsk: receiving socket 1322 * @skb: pointer to message buffer. 1323 * @inputq: buffer list containing the buffers 1324 * @xmitq: output message area 1325 */ 1326 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb, 1327 struct sk_buff_head *inputq, 1328 struct sk_buff_head *xmitq) 1329 { 1330 struct tipc_msg *hdr = buf_msg(skb); 1331 u32 onode = tsk_own_node(tsk); 1332 struct sock *sk = &tsk->sk; 1333 int mtyp = msg_type(hdr); 1334 bool was_cong; 1335 1336 /* Ignore if connection cannot be validated: */ 1337 if (!tsk_peer_msg(tsk, hdr)) { 1338 trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!"); 1339 goto exit; 1340 } 1341 1342 if (unlikely(msg_errcode(hdr))) { 1343 tipc_set_sk_state(sk, TIPC_DISCONNECTING); 1344 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk), 1345 tsk_peer_port(tsk)); 1346 sk->sk_state_change(sk); 1347 1348 /* State change is ignored if socket already awake, 1349 * - convert msg to abort msg and add to inqueue 1350 */ 1351 msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE); 1352 msg_set_type(hdr, TIPC_CONN_MSG); 1353 msg_set_size(hdr, BASIC_H_SIZE); 1354 msg_set_hdr_sz(hdr, BASIC_H_SIZE); 1355 __skb_queue_tail(inputq, skb); 1356 return; 1357 } 1358 1359 tsk->probe_unacked = false; 1360 1361 if (mtyp == CONN_PROBE) { 1362 msg_set_type(hdr, CONN_PROBE_REPLY); 1363 if (tipc_msg_reverse(onode, &skb, TIPC_OK)) 1364 __skb_queue_tail(xmitq, skb); 1365 return; 1366 } else if (mtyp == CONN_ACK) { 1367 was_cong = tsk_conn_cong(tsk); 1368 tipc_sk_push_backlog(tsk, msg_nagle_ack(hdr)); 1369 tsk->snt_unacked -= msg_conn_ack(hdr); 1370 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) 1371 tsk->snd_win = msg_adv_win(hdr); 1372 if (was_cong && !tsk_conn_cong(tsk)) 1373 sk->sk_write_space(sk); 1374 } else if (mtyp != CONN_PROBE_REPLY) { 1375 pr_warn("Received unknown CONN_PROTO msg\n"); 1376 } 1377 exit: 1378 kfree_skb(skb); 1379 } 1380 1381 /** 1382 * tipc_sendmsg - send message in connectionless manner 1383 * @sock: socket structure 1384 * @m: message to send 1385 * @dsz: amount of user data to be sent 1386 * 1387 * Message must have an destination specified explicitly. 1388 * Used for SOCK_RDM and SOCK_DGRAM messages, 1389 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections. 1390 * (Note: 'SYN+' is prohibited on SOCK_STREAM.) 1391 * 1392 * Return: the number of bytes sent on success, or errno otherwise 1393 */ 1394 static int tipc_sendmsg(struct socket *sock, 1395 struct msghdr *m, size_t dsz) 1396 { 1397 struct sock *sk = sock->sk; 1398 int ret; 1399 1400 lock_sock(sk); 1401 ret = __tipc_sendmsg(sock, m, dsz); 1402 release_sock(sk); 1403 1404 return ret; 1405 } 1406 1407 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen) 1408 { 1409 struct sock *sk = sock->sk; 1410 struct net *net = sock_net(sk); 1411 struct tipc_sock *tsk = tipc_sk(sk); 1412 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name; 1413 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT); 1414 struct list_head *clinks = &tsk->cong_links; 1415 bool syn = !tipc_sk_type_connectionless(sk); 1416 struct tipc_group *grp = tsk->group; 1417 struct tipc_msg *hdr = &tsk->phdr; 1418 struct tipc_socket_addr skaddr; 1419 struct sk_buff_head pkts; 1420 int atype, mtu, rc; 1421 1422 if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE)) 1423 return -EMSGSIZE; 1424 1425 if (ua) { 1426 if (!tipc_uaddr_valid(ua, m->msg_namelen)) 1427 return -EINVAL; 1428 atype = ua->addrtype; 1429 } 1430 1431 /* If socket belongs to a communication group follow other paths */ 1432 if (grp) { 1433 if (!ua) 1434 return tipc_send_group_bcast(sock, m, dlen, timeout); 1435 if (atype == TIPC_SERVICE_ADDR) 1436 return tipc_send_group_anycast(sock, m, dlen, timeout); 1437 if (atype == TIPC_SOCKET_ADDR) 1438 return tipc_send_group_unicast(sock, m, dlen, timeout); 1439 if (atype == TIPC_SERVICE_RANGE) 1440 return tipc_send_group_mcast(sock, m, dlen, timeout); 1441 return -EINVAL; 1442 } 1443 1444 if (!ua) { 1445 ua = (struct tipc_uaddr *)&tsk->peer; 1446 if (!syn && ua->family != AF_TIPC) 1447 return -EDESTADDRREQ; 1448 atype = ua->addrtype; 1449 } 1450 1451 if (unlikely(syn)) { 1452 if (sk->sk_state == TIPC_LISTEN) 1453 return -EPIPE; 1454 if (sk->sk_state != TIPC_OPEN) 1455 return -EISCONN; 1456 if (tsk->published) 1457 return -EOPNOTSUPP; 1458 if (atype == TIPC_SERVICE_ADDR) 1459 tsk->conn_addrtype = atype; 1460 msg_set_syn(hdr, 1); 1461 } 1462 1463 /* Determine destination */ 1464 if (atype == TIPC_SERVICE_RANGE) { 1465 return tipc_sendmcast(sock, ua, m, dlen, timeout); 1466 } else if (atype == TIPC_SERVICE_ADDR) { 1467 skaddr.node = ua->lookup_node; 1468 ua->scope = tipc_node2scope(skaddr.node); 1469 if (!tipc_nametbl_lookup_anycast(net, ua, &skaddr)) 1470 return -EHOSTUNREACH; 1471 } else if (atype == TIPC_SOCKET_ADDR) { 1472 skaddr = ua->sk; 1473 } else { 1474 return -EINVAL; 1475 } 1476 1477 /* Block or return if destination link is congested */ 1478 rc = tipc_wait_for_cond(sock, &timeout, 1479 !tipc_dest_find(clinks, skaddr.node, 0)); 1480 if (unlikely(rc)) 1481 return rc; 1482 1483 /* Finally build message header */ 1484 msg_set_destnode(hdr, skaddr.node); 1485 msg_set_destport(hdr, skaddr.ref); 1486 if (atype == TIPC_SERVICE_ADDR) { 1487 msg_set_type(hdr, TIPC_NAMED_MSG); 1488 msg_set_hdr_sz(hdr, NAMED_H_SIZE); 1489 msg_set_nametype(hdr, ua->sa.type); 1490 msg_set_nameinst(hdr, ua->sa.instance); 1491 msg_set_lookup_scope(hdr, ua->scope); 1492 } else { /* TIPC_SOCKET_ADDR */ 1493 msg_set_type(hdr, TIPC_DIRECT_MSG); 1494 msg_set_lookup_scope(hdr, 0); 1495 msg_set_hdr_sz(hdr, BASIC_H_SIZE); 1496 } 1497 1498 /* Add message body */ 1499 __skb_queue_head_init(&pkts); 1500 mtu = tipc_node_get_mtu(net, skaddr.node, tsk->portid, true); 1501 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts); 1502 if (unlikely(rc != dlen)) 1503 return rc; 1504 if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue))) { 1505 __skb_queue_purge(&pkts); 1506 return -ENOMEM; 1507 } 1508 1509 /* Send message */ 1510 trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " "); 1511 rc = tipc_node_xmit(net, &pkts, skaddr.node, tsk->portid); 1512 if (unlikely(rc == -ELINKCONG)) { 1513 tipc_dest_push(clinks, skaddr.node, 0); 1514 tsk->cong_link_cnt++; 1515 rc = 0; 1516 } 1517 1518 if (unlikely(syn && !rc)) 1519 tipc_set_sk_state(sk, TIPC_CONNECTING); 1520 1521 return rc ? rc : dlen; 1522 } 1523 1524 /** 1525 * tipc_sendstream - send stream-oriented data 1526 * @sock: socket structure 1527 * @m: data to send 1528 * @dsz: total length of data to be transmitted 1529 * 1530 * Used for SOCK_STREAM data. 1531 * 1532 * Return: the number of bytes sent on success (or partial success), 1533 * or errno if no data sent 1534 */ 1535 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz) 1536 { 1537 struct sock *sk = sock->sk; 1538 int ret; 1539 1540 lock_sock(sk); 1541 ret = __tipc_sendstream(sock, m, dsz); 1542 release_sock(sk); 1543 1544 return ret; 1545 } 1546 1547 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen) 1548 { 1549 struct sock *sk = sock->sk; 1550 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name); 1551 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT); 1552 struct sk_buff_head *txq = &sk->sk_write_queue; 1553 struct tipc_sock *tsk = tipc_sk(sk); 1554 struct tipc_msg *hdr = &tsk->phdr; 1555 struct net *net = sock_net(sk); 1556 struct sk_buff *skb; 1557 u32 dnode = tsk_peer_node(tsk); 1558 int maxnagle = tsk->maxnagle; 1559 int maxpkt = tsk->max_pkt; 1560 int send, sent = 0; 1561 int blocks, rc = 0; 1562 1563 if (unlikely(dlen > INT_MAX)) 1564 return -EMSGSIZE; 1565 1566 /* Handle implicit connection setup */ 1567 if (unlikely(dest)) { 1568 rc = __tipc_sendmsg(sock, m, dlen); 1569 if (dlen && dlen == rc) { 1570 tsk->peer_caps = tipc_node_get_capabilities(net, dnode); 1571 tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr)); 1572 } 1573 return rc; 1574 } 1575 1576 do { 1577 rc = tipc_wait_for_cond(sock, &timeout, 1578 (!tsk->cong_link_cnt && 1579 !tsk_conn_cong(tsk) && 1580 tipc_sk_connected(sk))); 1581 if (unlikely(rc)) 1582 break; 1583 send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE); 1584 blocks = tsk->snd_backlog; 1585 if (tsk->oneway++ >= tsk->nagle_start && maxnagle && 1586 send <= maxnagle) { 1587 rc = tipc_msg_append(hdr, m, send, maxnagle, txq); 1588 if (unlikely(rc < 0)) 1589 break; 1590 blocks += rc; 1591 tsk->msg_acc++; 1592 if (blocks <= 64 && tsk->expect_ack) { 1593 tsk->snd_backlog = blocks; 1594 sent += send; 1595 break; 1596 } else if (blocks > 64) { 1597 tsk->pkt_cnt += skb_queue_len(txq); 1598 } else { 1599 skb = skb_peek_tail(txq); 1600 if (skb) { 1601 msg_set_ack_required(buf_msg(skb)); 1602 tsk->expect_ack = true; 1603 } else { 1604 tsk->expect_ack = false; 1605 } 1606 tsk->msg_acc = 0; 1607 tsk->pkt_cnt = 0; 1608 } 1609 } else { 1610 rc = tipc_msg_build(hdr, m, sent, send, maxpkt, txq); 1611 if (unlikely(rc != send)) 1612 break; 1613 blocks += tsk_inc(tsk, send + MIN_H_SIZE); 1614 } 1615 trace_tipc_sk_sendstream(sk, skb_peek(txq), 1616 TIPC_DUMP_SK_SNDQ, " "); 1617 rc = tipc_node_xmit(net, txq, dnode, tsk->portid); 1618 if (unlikely(rc == -ELINKCONG)) { 1619 tsk->cong_link_cnt = 1; 1620 rc = 0; 1621 } 1622 if (likely(!rc)) { 1623 tsk->snt_unacked += blocks; 1624 tsk->snd_backlog = 0; 1625 sent += send; 1626 } 1627 } while (sent < dlen && !rc); 1628 1629 return sent ? sent : rc; 1630 } 1631 1632 /** 1633 * tipc_send_packet - send a connection-oriented message 1634 * @sock: socket structure 1635 * @m: message to send 1636 * @dsz: length of data to be transmitted 1637 * 1638 * Used for SOCK_SEQPACKET messages. 1639 * 1640 * Return: the number of bytes sent on success, or errno otherwise 1641 */ 1642 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz) 1643 { 1644 if (dsz > TIPC_MAX_USER_MSG_SIZE) 1645 return -EMSGSIZE; 1646 1647 return tipc_sendstream(sock, m, dsz); 1648 } 1649 1650 /* tipc_sk_finish_conn - complete the setup of a connection 1651 */ 1652 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port, 1653 u32 peer_node) 1654 { 1655 struct sock *sk = &tsk->sk; 1656 struct net *net = sock_net(sk); 1657 struct tipc_msg *msg = &tsk->phdr; 1658 1659 msg_set_syn(msg, 0); 1660 msg_set_destnode(msg, peer_node); 1661 msg_set_destport(msg, peer_port); 1662 msg_set_type(msg, TIPC_CONN_MSG); 1663 msg_set_lookup_scope(msg, 0); 1664 msg_set_hdr_sz(msg, SHORT_H_SIZE); 1665 1666 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV); 1667 tipc_set_sk_state(sk, TIPC_ESTABLISHED); 1668 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port); 1669 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid, true); 1670 tsk->peer_caps = tipc_node_get_capabilities(net, peer_node); 1671 tsk_set_nagle(tsk); 1672 __skb_queue_purge(&sk->sk_write_queue); 1673 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) 1674 return; 1675 1676 /* Fall back to message based flow control */ 1677 tsk->rcv_win = FLOWCTL_MSG_WIN; 1678 tsk->snd_win = FLOWCTL_MSG_WIN; 1679 } 1680 1681 /** 1682 * tipc_sk_set_orig_addr - capture sender's address for received message 1683 * @m: descriptor for message info 1684 * @skb: received message 1685 * 1686 * Note: Address is not captured if not requested by receiver. 1687 */ 1688 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb) 1689 { 1690 DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name); 1691 struct tipc_msg *hdr = buf_msg(skb); 1692 1693 if (!srcaddr) 1694 return; 1695 1696 srcaddr->sock.family = AF_TIPC; 1697 srcaddr->sock.addrtype = TIPC_SOCKET_ADDR; 1698 srcaddr->sock.scope = 0; 1699 srcaddr->sock.addr.id.ref = msg_origport(hdr); 1700 srcaddr->sock.addr.id.node = msg_orignode(hdr); 1701 srcaddr->sock.addr.name.domain = 0; 1702 m->msg_namelen = sizeof(struct sockaddr_tipc); 1703 1704 if (!msg_in_group(hdr)) 1705 return; 1706 1707 /* Group message users may also want to know sending member's id */ 1708 srcaddr->member.family = AF_TIPC; 1709 srcaddr->member.addrtype = TIPC_SERVICE_ADDR; 1710 srcaddr->member.scope = 0; 1711 srcaddr->member.addr.name.name.type = msg_nametype(hdr); 1712 srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member; 1713 srcaddr->member.addr.name.domain = 0; 1714 m->msg_namelen = sizeof(*srcaddr); 1715 } 1716 1717 /** 1718 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message 1719 * @m: descriptor for message info 1720 * @skb: received message buffer 1721 * @tsk: TIPC port associated with message 1722 * 1723 * Note: Ancillary data is not captured if not requested by receiver. 1724 * 1725 * Return: 0 if successful, otherwise errno 1726 */ 1727 static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb, 1728 struct tipc_sock *tsk) 1729 { 1730 struct tipc_msg *hdr; 1731 u32 data[3] = {0,}; 1732 bool has_addr; 1733 int dlen, rc; 1734 1735 if (likely(m->msg_controllen == 0)) 1736 return 0; 1737 1738 hdr = buf_msg(skb); 1739 dlen = msg_data_sz(hdr); 1740 1741 /* Capture errored message object, if any */ 1742 if (msg_errcode(hdr)) { 1743 if (skb_linearize(skb)) 1744 return -ENOMEM; 1745 hdr = buf_msg(skb); 1746 data[0] = msg_errcode(hdr); 1747 data[1] = dlen; 1748 rc = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, data); 1749 if (rc || !dlen) 1750 return rc; 1751 rc = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, dlen, msg_data(hdr)); 1752 if (rc) 1753 return rc; 1754 } 1755 1756 /* Capture TIPC_SERVICE_ADDR/RANGE destination address, if any */ 1757 switch (msg_type(hdr)) { 1758 case TIPC_NAMED_MSG: 1759 has_addr = true; 1760 data[0] = msg_nametype(hdr); 1761 data[1] = msg_namelower(hdr); 1762 data[2] = data[1]; 1763 break; 1764 case TIPC_MCAST_MSG: 1765 has_addr = true; 1766 data[0] = msg_nametype(hdr); 1767 data[1] = msg_namelower(hdr); 1768 data[2] = msg_nameupper(hdr); 1769 break; 1770 case TIPC_CONN_MSG: 1771 has_addr = !!tsk->conn_addrtype; 1772 data[0] = msg_nametype(&tsk->phdr); 1773 data[1] = msg_nameinst(&tsk->phdr); 1774 data[2] = data[1]; 1775 break; 1776 default: 1777 has_addr = false; 1778 } 1779 if (!has_addr) 1780 return 0; 1781 return put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, data); 1782 } 1783 1784 static struct sk_buff *tipc_sk_build_ack(struct tipc_sock *tsk) 1785 { 1786 struct sock *sk = &tsk->sk; 1787 struct sk_buff *skb = NULL; 1788 struct tipc_msg *msg; 1789 u32 peer_port = tsk_peer_port(tsk); 1790 u32 dnode = tsk_peer_node(tsk); 1791 1792 if (!tipc_sk_connected(sk)) 1793 return NULL; 1794 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0, 1795 dnode, tsk_own_node(tsk), peer_port, 1796 tsk->portid, TIPC_OK); 1797 if (!skb) 1798 return NULL; 1799 msg = buf_msg(skb); 1800 msg_set_conn_ack(msg, tsk->rcv_unacked); 1801 tsk->rcv_unacked = 0; 1802 1803 /* Adjust to and advertize the correct window limit */ 1804 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) { 1805 tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf); 1806 msg_set_adv_win(msg, tsk->rcv_win); 1807 } 1808 return skb; 1809 } 1810 1811 static void tipc_sk_send_ack(struct tipc_sock *tsk) 1812 { 1813 struct sk_buff *skb; 1814 1815 skb = tipc_sk_build_ack(tsk); 1816 if (!skb) 1817 return; 1818 1819 tipc_node_xmit_skb(sock_net(&tsk->sk), skb, tsk_peer_node(tsk), 1820 msg_link_selector(buf_msg(skb))); 1821 } 1822 1823 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop) 1824 { 1825 struct sock *sk = sock->sk; 1826 DEFINE_WAIT_FUNC(wait, woken_wake_function); 1827 long timeo = *timeop; 1828 int err = sock_error(sk); 1829 1830 if (err) 1831 return err; 1832 1833 for (;;) { 1834 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) { 1835 if (sk->sk_shutdown & RCV_SHUTDOWN) { 1836 err = -ENOTCONN; 1837 break; 1838 } 1839 add_wait_queue(sk_sleep(sk), &wait); 1840 release_sock(sk); 1841 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo); 1842 sched_annotate_sleep(); 1843 lock_sock(sk); 1844 remove_wait_queue(sk_sleep(sk), &wait); 1845 } 1846 err = 0; 1847 if (!skb_queue_empty(&sk->sk_receive_queue)) 1848 break; 1849 err = -EAGAIN; 1850 if (!timeo) 1851 break; 1852 err = sock_intr_errno(timeo); 1853 if (signal_pending(current)) 1854 break; 1855 1856 err = sock_error(sk); 1857 if (err) 1858 break; 1859 } 1860 *timeop = timeo; 1861 return err; 1862 } 1863 1864 /** 1865 * tipc_recvmsg - receive packet-oriented message 1866 * @sock: network socket 1867 * @m: descriptor for message info 1868 * @buflen: length of user buffer area 1869 * @flags: receive flags 1870 * 1871 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages. 1872 * If the complete message doesn't fit in user area, truncate it. 1873 * 1874 * Return: size of returned message data, errno otherwise 1875 */ 1876 static int tipc_recvmsg(struct socket *sock, struct msghdr *m, 1877 size_t buflen, int flags) 1878 { 1879 struct sock *sk = sock->sk; 1880 bool connected = !tipc_sk_type_connectionless(sk); 1881 struct tipc_sock *tsk = tipc_sk(sk); 1882 int rc, err, hlen, dlen, copy; 1883 struct tipc_skb_cb *skb_cb; 1884 struct sk_buff_head xmitq; 1885 struct tipc_msg *hdr; 1886 struct sk_buff *skb; 1887 bool grp_evt; 1888 long timeout; 1889 1890 /* Catch invalid receive requests */ 1891 if (unlikely(!buflen)) 1892 return -EINVAL; 1893 1894 lock_sock(sk); 1895 if (unlikely(connected && sk->sk_state == TIPC_OPEN)) { 1896 rc = -ENOTCONN; 1897 goto exit; 1898 } 1899 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT); 1900 1901 /* Step rcv queue to first msg with data or error; wait if necessary */ 1902 do { 1903 rc = tipc_wait_for_rcvmsg(sock, &timeout); 1904 if (unlikely(rc)) 1905 goto exit; 1906 skb = skb_peek(&sk->sk_receive_queue); 1907 skb_cb = TIPC_SKB_CB(skb); 1908 hdr = buf_msg(skb); 1909 dlen = msg_data_sz(hdr); 1910 hlen = msg_hdr_sz(hdr); 1911 err = msg_errcode(hdr); 1912 grp_evt = msg_is_grp_evt(hdr); 1913 if (likely(dlen || err)) 1914 break; 1915 tsk_advance_rx_queue(sk); 1916 } while (1); 1917 1918 /* Collect msg meta data, including error code and rejected data */ 1919 tipc_sk_set_orig_addr(m, skb); 1920 rc = tipc_sk_anc_data_recv(m, skb, tsk); 1921 if (unlikely(rc)) 1922 goto exit; 1923 hdr = buf_msg(skb); 1924 1925 /* Capture data if non-error msg, otherwise just set return value */ 1926 if (likely(!err)) { 1927 int offset = skb_cb->bytes_read; 1928 1929 copy = min_t(int, dlen - offset, buflen); 1930 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy); 1931 if (unlikely(rc)) 1932 goto exit; 1933 if (unlikely(offset + copy < dlen)) { 1934 if (flags & MSG_EOR) { 1935 if (!(flags & MSG_PEEK)) 1936 skb_cb->bytes_read = offset + copy; 1937 } else { 1938 m->msg_flags |= MSG_TRUNC; 1939 skb_cb->bytes_read = 0; 1940 } 1941 } else { 1942 if (flags & MSG_EOR) 1943 m->msg_flags |= MSG_EOR; 1944 skb_cb->bytes_read = 0; 1945 } 1946 } else { 1947 copy = 0; 1948 rc = 0; 1949 if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control) { 1950 rc = -ECONNRESET; 1951 goto exit; 1952 } 1953 } 1954 1955 /* Mark message as group event if applicable */ 1956 if (unlikely(grp_evt)) { 1957 if (msg_grp_evt(hdr) == TIPC_WITHDRAWN) 1958 m->msg_flags |= MSG_EOR; 1959 m->msg_flags |= MSG_OOB; 1960 copy = 0; 1961 } 1962 1963 /* Caption of data or error code/rejected data was successful */ 1964 if (unlikely(flags & MSG_PEEK)) 1965 goto exit; 1966 1967 /* Send group flow control advertisement when applicable */ 1968 if (tsk->group && msg_in_group(hdr) && !grp_evt) { 1969 __skb_queue_head_init(&xmitq); 1970 tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen), 1971 msg_orignode(hdr), msg_origport(hdr), 1972 &xmitq); 1973 tipc_node_distr_xmit(sock_net(sk), &xmitq); 1974 } 1975 1976 if (skb_cb->bytes_read) 1977 goto exit; 1978 1979 tsk_advance_rx_queue(sk); 1980 1981 if (likely(!connected)) 1982 goto exit; 1983 1984 /* Send connection flow control advertisement when applicable */ 1985 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen); 1986 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE) 1987 tipc_sk_send_ack(tsk); 1988 exit: 1989 release_sock(sk); 1990 return rc ? rc : copy; 1991 } 1992 1993 /** 1994 * tipc_recvstream - receive stream-oriented data 1995 * @sock: network socket 1996 * @m: descriptor for message info 1997 * @buflen: total size of user buffer area 1998 * @flags: receive flags 1999 * 2000 * Used for SOCK_STREAM messages only. If not enough data is available 2001 * will optionally wait for more; never truncates data. 2002 * 2003 * Return: size of returned message data, errno otherwise 2004 */ 2005 static int tipc_recvstream(struct socket *sock, struct msghdr *m, 2006 size_t buflen, int flags) 2007 { 2008 struct sock *sk = sock->sk; 2009 struct tipc_sock *tsk = tipc_sk(sk); 2010 struct sk_buff *skb; 2011 struct tipc_msg *hdr; 2012 struct tipc_skb_cb *skb_cb; 2013 bool peek = flags & MSG_PEEK; 2014 int offset, required, copy, copied = 0; 2015 int hlen, dlen, err, rc; 2016 long timeout; 2017 2018 /* Catch invalid receive attempts */ 2019 if (unlikely(!buflen)) 2020 return -EINVAL; 2021 2022 lock_sock(sk); 2023 2024 if (unlikely(sk->sk_state == TIPC_OPEN)) { 2025 rc = -ENOTCONN; 2026 goto exit; 2027 } 2028 required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen); 2029 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT); 2030 2031 do { 2032 /* Look at first msg in receive queue; wait if necessary */ 2033 rc = tipc_wait_for_rcvmsg(sock, &timeout); 2034 if (unlikely(rc)) 2035 break; 2036 skb = skb_peek(&sk->sk_receive_queue); 2037 skb_cb = TIPC_SKB_CB(skb); 2038 hdr = buf_msg(skb); 2039 dlen = msg_data_sz(hdr); 2040 hlen = msg_hdr_sz(hdr); 2041 err = msg_errcode(hdr); 2042 2043 /* Discard any empty non-errored (SYN-) message */ 2044 if (unlikely(!dlen && !err)) { 2045 tsk_advance_rx_queue(sk); 2046 continue; 2047 } 2048 2049 /* Collect msg meta data, incl. error code and rejected data */ 2050 if (!copied) { 2051 tipc_sk_set_orig_addr(m, skb); 2052 rc = tipc_sk_anc_data_recv(m, skb, tsk); 2053 if (rc) 2054 break; 2055 hdr = buf_msg(skb); 2056 } 2057 2058 /* Copy data if msg ok, otherwise return error/partial data */ 2059 if (likely(!err)) { 2060 offset = skb_cb->bytes_read; 2061 copy = min_t(int, dlen - offset, buflen - copied); 2062 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy); 2063 if (unlikely(rc)) 2064 break; 2065 copied += copy; 2066 offset += copy; 2067 if (unlikely(offset < dlen)) { 2068 if (!peek) 2069 skb_cb->bytes_read = offset; 2070 break; 2071 } 2072 } else { 2073 rc = 0; 2074 if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control) 2075 rc = -ECONNRESET; 2076 if (copied || rc) 2077 break; 2078 } 2079 2080 if (unlikely(peek)) 2081 break; 2082 2083 tsk_advance_rx_queue(sk); 2084 2085 /* Send connection flow control advertisement when applicable */ 2086 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen); 2087 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE) 2088 tipc_sk_send_ack(tsk); 2089 2090 /* Exit if all requested data or FIN/error received */ 2091 if (copied == buflen || err) 2092 break; 2093 2094 } while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required); 2095 exit: 2096 release_sock(sk); 2097 return copied ? copied : rc; 2098 } 2099 2100 /** 2101 * tipc_write_space - wake up thread if port congestion is released 2102 * @sk: socket 2103 */ 2104 static void tipc_write_space(struct sock *sk) 2105 { 2106 struct socket_wq *wq; 2107 2108 rcu_read_lock(); 2109 wq = rcu_dereference(sk->sk_wq); 2110 if (skwq_has_sleeper(wq)) 2111 wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT | 2112 EPOLLWRNORM | EPOLLWRBAND); 2113 rcu_read_unlock(); 2114 } 2115 2116 /** 2117 * tipc_data_ready - wake up threads to indicate messages have been received 2118 * @sk: socket 2119 */ 2120 static void tipc_data_ready(struct sock *sk) 2121 { 2122 struct socket_wq *wq; 2123 2124 rcu_read_lock(); 2125 wq = rcu_dereference(sk->sk_wq); 2126 if (skwq_has_sleeper(wq)) 2127 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN | 2128 EPOLLRDNORM | EPOLLRDBAND); 2129 rcu_read_unlock(); 2130 } 2131 2132 static void tipc_sock_destruct(struct sock *sk) 2133 { 2134 __skb_queue_purge(&sk->sk_receive_queue); 2135 } 2136 2137 static void tipc_sk_proto_rcv(struct sock *sk, 2138 struct sk_buff_head *inputq, 2139 struct sk_buff_head *xmitq) 2140 { 2141 struct sk_buff *skb = __skb_dequeue(inputq); 2142 struct tipc_sock *tsk = tipc_sk(sk); 2143 struct tipc_msg *hdr = buf_msg(skb); 2144 struct tipc_group *grp = tsk->group; 2145 bool wakeup = false; 2146 2147 switch (msg_user(hdr)) { 2148 case CONN_MANAGER: 2149 tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq); 2150 return; 2151 case SOCK_WAKEUP: 2152 tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0); 2153 /* coupled with smp_rmb() in tipc_wait_for_cond() */ 2154 smp_wmb(); 2155 tsk->cong_link_cnt--; 2156 wakeup = true; 2157 tipc_sk_push_backlog(tsk, false); 2158 break; 2159 case GROUP_PROTOCOL: 2160 tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq); 2161 break; 2162 case TOP_SRV: 2163 tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf, 2164 hdr, inputq, xmitq); 2165 break; 2166 default: 2167 break; 2168 } 2169 2170 if (wakeup) 2171 sk->sk_write_space(sk); 2172 2173 kfree_skb(skb); 2174 } 2175 2176 /** 2177 * tipc_sk_filter_connect - check incoming message for a connection-based socket 2178 * @tsk: TIPC socket 2179 * @skb: pointer to message buffer. 2180 * @xmitq: for Nagle ACK if any 2181 * Return: true if message should be added to receive queue, false otherwise 2182 */ 2183 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb, 2184 struct sk_buff_head *xmitq) 2185 { 2186 struct sock *sk = &tsk->sk; 2187 struct net *net = sock_net(sk); 2188 struct tipc_msg *hdr = buf_msg(skb); 2189 bool con_msg = msg_connected(hdr); 2190 u32 pport = tsk_peer_port(tsk); 2191 u32 pnode = tsk_peer_node(tsk); 2192 u32 oport = msg_origport(hdr); 2193 u32 onode = msg_orignode(hdr); 2194 int err = msg_errcode(hdr); 2195 unsigned long delay; 2196 2197 if (unlikely(msg_mcast(hdr))) 2198 return false; 2199 tsk->oneway = 0; 2200 2201 switch (sk->sk_state) { 2202 case TIPC_CONNECTING: 2203 /* Setup ACK */ 2204 if (likely(con_msg)) { 2205 if (err) 2206 break; 2207 tipc_sk_finish_conn(tsk, oport, onode); 2208 msg_set_importance(&tsk->phdr, msg_importance(hdr)); 2209 /* ACK+ message with data is added to receive queue */ 2210 if (msg_data_sz(hdr)) 2211 return true; 2212 /* Empty ACK-, - wake up sleeping connect() and drop */ 2213 sk->sk_state_change(sk); 2214 msg_set_dest_droppable(hdr, 1); 2215 return false; 2216 } 2217 /* Ignore connectionless message if not from listening socket */ 2218 if (oport != pport || onode != pnode) 2219 return false; 2220 2221 /* Rejected SYN */ 2222 if (err != TIPC_ERR_OVERLOAD) 2223 break; 2224 2225 /* Prepare for new setup attempt if we have a SYN clone */ 2226 if (skb_queue_empty(&sk->sk_write_queue)) 2227 break; 2228 get_random_bytes(&delay, 2); 2229 delay %= (tsk->conn_timeout / 4); 2230 delay = msecs_to_jiffies(delay + 100); 2231 sk_reset_timer(sk, &sk->sk_timer, jiffies + delay); 2232 return false; 2233 case TIPC_OPEN: 2234 case TIPC_DISCONNECTING: 2235 return false; 2236 case TIPC_LISTEN: 2237 /* Accept only SYN message */ 2238 if (!msg_is_syn(hdr) && 2239 tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT) 2240 return false; 2241 if (!con_msg && !err) 2242 return true; 2243 return false; 2244 case TIPC_ESTABLISHED: 2245 if (!skb_queue_empty(&sk->sk_write_queue)) 2246 tipc_sk_push_backlog(tsk, false); 2247 /* Accept only connection-based messages sent by peer */ 2248 if (likely(con_msg && !err && pport == oport && 2249 pnode == onode)) { 2250 if (msg_ack_required(hdr)) { 2251 struct sk_buff *skb; 2252 2253 skb = tipc_sk_build_ack(tsk); 2254 if (skb) { 2255 msg_set_nagle_ack(buf_msg(skb)); 2256 __skb_queue_tail(xmitq, skb); 2257 } 2258 } 2259 return true; 2260 } 2261 if (!tsk_peer_msg(tsk, hdr)) 2262 return false; 2263 if (!err) 2264 return true; 2265 tipc_set_sk_state(sk, TIPC_DISCONNECTING); 2266 tipc_node_remove_conn(net, pnode, tsk->portid); 2267 sk->sk_state_change(sk); 2268 return true; 2269 default: 2270 pr_err("Unknown sk_state %u\n", sk->sk_state); 2271 } 2272 /* Abort connection setup attempt */ 2273 tipc_set_sk_state(sk, TIPC_DISCONNECTING); 2274 sk->sk_err = ECONNREFUSED; 2275 sk->sk_state_change(sk); 2276 return true; 2277 } 2278 2279 /** 2280 * rcvbuf_limit - get proper overload limit of socket receive queue 2281 * @sk: socket 2282 * @skb: message 2283 * 2284 * For connection oriented messages, irrespective of importance, 2285 * default queue limit is 2 MB. 2286 * 2287 * For connectionless messages, queue limits are based on message 2288 * importance as follows: 2289 * 2290 * TIPC_LOW_IMPORTANCE (2 MB) 2291 * TIPC_MEDIUM_IMPORTANCE (4 MB) 2292 * TIPC_HIGH_IMPORTANCE (8 MB) 2293 * TIPC_CRITICAL_IMPORTANCE (16 MB) 2294 * 2295 * Return: overload limit according to corresponding message importance 2296 */ 2297 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb) 2298 { 2299 struct tipc_sock *tsk = tipc_sk(sk); 2300 struct tipc_msg *hdr = buf_msg(skb); 2301 2302 if (unlikely(msg_in_group(hdr))) 2303 return READ_ONCE(sk->sk_rcvbuf); 2304 2305 if (unlikely(!msg_connected(hdr))) 2306 return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr); 2307 2308 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL)) 2309 return READ_ONCE(sk->sk_rcvbuf); 2310 2311 return FLOWCTL_MSG_LIM; 2312 } 2313 2314 /** 2315 * tipc_sk_filter_rcv - validate incoming message 2316 * @sk: socket 2317 * @skb: pointer to message. 2318 * @xmitq: output message area (FIXME) 2319 * 2320 * Enqueues message on receive queue if acceptable; optionally handles 2321 * disconnect indication for a connected socket. 2322 * 2323 * Called with socket lock already taken 2324 */ 2325 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb, 2326 struct sk_buff_head *xmitq) 2327 { 2328 bool sk_conn = !tipc_sk_type_connectionless(sk); 2329 struct tipc_sock *tsk = tipc_sk(sk); 2330 struct tipc_group *grp = tsk->group; 2331 struct tipc_msg *hdr = buf_msg(skb); 2332 struct net *net = sock_net(sk); 2333 struct sk_buff_head inputq; 2334 int mtyp = msg_type(hdr); 2335 int limit, err = TIPC_OK; 2336 2337 trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " "); 2338 TIPC_SKB_CB(skb)->bytes_read = 0; 2339 __skb_queue_head_init(&inputq); 2340 __skb_queue_tail(&inputq, skb); 2341 2342 if (unlikely(!msg_isdata(hdr))) 2343 tipc_sk_proto_rcv(sk, &inputq, xmitq); 2344 2345 if (unlikely(grp)) 2346 tipc_group_filter_msg(grp, &inputq, xmitq); 2347 2348 if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG) 2349 tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq); 2350 2351 /* Validate and add to receive buffer if there is space */ 2352 while ((skb = __skb_dequeue(&inputq))) { 2353 hdr = buf_msg(skb); 2354 limit = rcvbuf_limit(sk, skb); 2355 if ((sk_conn && !tipc_sk_filter_connect(tsk, skb, xmitq)) || 2356 (!sk_conn && msg_connected(hdr)) || 2357 (!grp && msg_in_group(hdr))) 2358 err = TIPC_ERR_NO_PORT; 2359 else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) { 2360 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, 2361 "err_overload2!"); 2362 atomic_inc(&sk->sk_drops); 2363 err = TIPC_ERR_OVERLOAD; 2364 } 2365 2366 if (unlikely(err)) { 2367 if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) { 2368 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, 2369 "@filter_rcv!"); 2370 __skb_queue_tail(xmitq, skb); 2371 } 2372 err = TIPC_OK; 2373 continue; 2374 } 2375 __skb_queue_tail(&sk->sk_receive_queue, skb); 2376 skb_set_owner_r(skb, sk); 2377 trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL, 2378 "rcvq >90% allocated!"); 2379 sk->sk_data_ready(sk); 2380 } 2381 } 2382 2383 /** 2384 * tipc_sk_backlog_rcv - handle incoming message from backlog queue 2385 * @sk: socket 2386 * @skb: message 2387 * 2388 * Caller must hold socket lock 2389 */ 2390 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb) 2391 { 2392 unsigned int before = sk_rmem_alloc_get(sk); 2393 struct sk_buff_head xmitq; 2394 unsigned int added; 2395 2396 __skb_queue_head_init(&xmitq); 2397 2398 tipc_sk_filter_rcv(sk, skb, &xmitq); 2399 added = sk_rmem_alloc_get(sk) - before; 2400 atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt); 2401 2402 /* Send pending response/rejected messages, if any */ 2403 tipc_node_distr_xmit(sock_net(sk), &xmitq); 2404 return 0; 2405 } 2406 2407 /** 2408 * tipc_sk_enqueue - extract all buffers with destination 'dport' from 2409 * inputq and try adding them to socket or backlog queue 2410 * @inputq: list of incoming buffers with potentially different destinations 2411 * @sk: socket where the buffers should be enqueued 2412 * @dport: port number for the socket 2413 * @xmitq: output queue 2414 * 2415 * Caller must hold socket lock 2416 */ 2417 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk, 2418 u32 dport, struct sk_buff_head *xmitq) 2419 { 2420 unsigned long time_limit = jiffies + 2; 2421 struct sk_buff *skb; 2422 unsigned int lim; 2423 atomic_t *dcnt; 2424 u32 onode; 2425 2426 while (skb_queue_len(inputq)) { 2427 if (unlikely(time_after_eq(jiffies, time_limit))) 2428 return; 2429 2430 skb = tipc_skb_dequeue(inputq, dport); 2431 if (unlikely(!skb)) 2432 return; 2433 2434 /* Add message directly to receive queue if possible */ 2435 if (!sock_owned_by_user(sk)) { 2436 tipc_sk_filter_rcv(sk, skb, xmitq); 2437 continue; 2438 } 2439 2440 /* Try backlog, compensating for double-counted bytes */ 2441 dcnt = &tipc_sk(sk)->dupl_rcvcnt; 2442 if (!sk->sk_backlog.len) 2443 atomic_set(dcnt, 0); 2444 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt); 2445 if (likely(!sk_add_backlog(sk, skb, lim))) { 2446 trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL, 2447 "bklg & rcvq >90% allocated!"); 2448 continue; 2449 } 2450 2451 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!"); 2452 /* Overload => reject message back to sender */ 2453 onode = tipc_own_addr(sock_net(sk)); 2454 atomic_inc(&sk->sk_drops); 2455 if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) { 2456 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL, 2457 "@sk_enqueue!"); 2458 __skb_queue_tail(xmitq, skb); 2459 } 2460 break; 2461 } 2462 } 2463 2464 /** 2465 * tipc_sk_rcv - handle a chain of incoming buffers 2466 * @net: the associated network namespace 2467 * @inputq: buffer list containing the buffers 2468 * Consumes all buffers in list until inputq is empty 2469 * Note: may be called in multiple threads referring to the same queue 2470 */ 2471 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq) 2472 { 2473 struct sk_buff_head xmitq; 2474 u32 dnode, dport = 0; 2475 int err; 2476 struct tipc_sock *tsk; 2477 struct sock *sk; 2478 struct sk_buff *skb; 2479 2480 __skb_queue_head_init(&xmitq); 2481 while (skb_queue_len(inputq)) { 2482 dport = tipc_skb_peek_port(inputq, dport); 2483 tsk = tipc_sk_lookup(net, dport); 2484 2485 if (likely(tsk)) { 2486 sk = &tsk->sk; 2487 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) { 2488 tipc_sk_enqueue(inputq, sk, dport, &xmitq); 2489 spin_unlock_bh(&sk->sk_lock.slock); 2490 } 2491 /* Send pending response/rejected messages, if any */ 2492 tipc_node_distr_xmit(sock_net(sk), &xmitq); 2493 sock_put(sk); 2494 continue; 2495 } 2496 /* No destination socket => dequeue skb if still there */ 2497 skb = tipc_skb_dequeue(inputq, dport); 2498 if (!skb) 2499 return; 2500 2501 /* Try secondary lookup if unresolved named message */ 2502 err = TIPC_ERR_NO_PORT; 2503 if (tipc_msg_lookup_dest(net, skb, &err)) 2504 goto xmit; 2505 2506 /* Prepare for message rejection */ 2507 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err)) 2508 continue; 2509 2510 trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!"); 2511 xmit: 2512 dnode = msg_destnode(buf_msg(skb)); 2513 tipc_node_xmit_skb(net, skb, dnode, dport); 2514 } 2515 } 2516 2517 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p) 2518 { 2519 DEFINE_WAIT_FUNC(wait, woken_wake_function); 2520 struct sock *sk = sock->sk; 2521 int done; 2522 2523 do { 2524 int err = sock_error(sk); 2525 if (err) 2526 return err; 2527 if (!*timeo_p) 2528 return -ETIMEDOUT; 2529 if (signal_pending(current)) 2530 return sock_intr_errno(*timeo_p); 2531 if (sk->sk_state == TIPC_DISCONNECTING) 2532 break; 2533 2534 add_wait_queue(sk_sleep(sk), &wait); 2535 done = sk_wait_event(sk, timeo_p, tipc_sk_connected(sk), 2536 &wait); 2537 remove_wait_queue(sk_sleep(sk), &wait); 2538 } while (!done); 2539 return 0; 2540 } 2541 2542 static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr) 2543 { 2544 if (addr->family != AF_TIPC) 2545 return false; 2546 if (addr->addrtype == TIPC_SERVICE_RANGE) 2547 return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper); 2548 return (addr->addrtype == TIPC_SERVICE_ADDR || 2549 addr->addrtype == TIPC_SOCKET_ADDR); 2550 } 2551 2552 /** 2553 * tipc_connect - establish a connection to another TIPC port 2554 * @sock: socket structure 2555 * @dest: socket address for destination port 2556 * @destlen: size of socket address data structure 2557 * @flags: file-related flags associated with socket 2558 * 2559 * Return: 0 on success, errno otherwise 2560 */ 2561 static int tipc_connect(struct socket *sock, struct sockaddr *dest, 2562 int destlen, int flags) 2563 { 2564 struct sock *sk = sock->sk; 2565 struct tipc_sock *tsk = tipc_sk(sk); 2566 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest; 2567 struct msghdr m = {NULL,}; 2568 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout; 2569 int previous; 2570 int res = 0; 2571 2572 if (destlen != sizeof(struct sockaddr_tipc)) 2573 return -EINVAL; 2574 2575 lock_sock(sk); 2576 2577 if (tsk->group) { 2578 res = -EINVAL; 2579 goto exit; 2580 } 2581 2582 if (dst->family == AF_UNSPEC) { 2583 memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc)); 2584 if (!tipc_sk_type_connectionless(sk)) 2585 res = -EINVAL; 2586 goto exit; 2587 } 2588 if (!tipc_sockaddr_is_sane(dst)) { 2589 res = -EINVAL; 2590 goto exit; 2591 } 2592 /* DGRAM/RDM connect(), just save the destaddr */ 2593 if (tipc_sk_type_connectionless(sk)) { 2594 memcpy(&tsk->peer, dest, destlen); 2595 goto exit; 2596 } else if (dst->addrtype == TIPC_SERVICE_RANGE) { 2597 res = -EINVAL; 2598 goto exit; 2599 } 2600 2601 previous = sk->sk_state; 2602 2603 switch (sk->sk_state) { 2604 case TIPC_OPEN: 2605 /* Send a 'SYN-' to destination */ 2606 m.msg_name = dest; 2607 m.msg_namelen = destlen; 2608 2609 /* If connect is in non-blocking case, set MSG_DONTWAIT to 2610 * indicate send_msg() is never blocked. 2611 */ 2612 if (!timeout) 2613 m.msg_flags = MSG_DONTWAIT; 2614 2615 res = __tipc_sendmsg(sock, &m, 0); 2616 if ((res < 0) && (res != -EWOULDBLOCK)) 2617 goto exit; 2618 2619 /* Just entered TIPC_CONNECTING state; the only 2620 * difference is that return value in non-blocking 2621 * case is EINPROGRESS, rather than EALREADY. 2622 */ 2623 res = -EINPROGRESS; 2624 fallthrough; 2625 case TIPC_CONNECTING: 2626 if (!timeout) { 2627 if (previous == TIPC_CONNECTING) 2628 res = -EALREADY; 2629 goto exit; 2630 } 2631 timeout = msecs_to_jiffies(timeout); 2632 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */ 2633 res = tipc_wait_for_connect(sock, &timeout); 2634 break; 2635 case TIPC_ESTABLISHED: 2636 res = -EISCONN; 2637 break; 2638 default: 2639 res = -EINVAL; 2640 } 2641 2642 exit: 2643 release_sock(sk); 2644 return res; 2645 } 2646 2647 /** 2648 * tipc_listen - allow socket to listen for incoming connections 2649 * @sock: socket structure 2650 * @len: (unused) 2651 * 2652 * Return: 0 on success, errno otherwise 2653 */ 2654 static int tipc_listen(struct socket *sock, int len) 2655 { 2656 struct sock *sk = sock->sk; 2657 int res; 2658 2659 lock_sock(sk); 2660 res = tipc_set_sk_state(sk, TIPC_LISTEN); 2661 release_sock(sk); 2662 2663 return res; 2664 } 2665 2666 static int tipc_wait_for_accept(struct socket *sock, long timeo) 2667 { 2668 struct sock *sk = sock->sk; 2669 DEFINE_WAIT(wait); 2670 int err; 2671 2672 /* True wake-one mechanism for incoming connections: only 2673 * one process gets woken up, not the 'whole herd'. 2674 * Since we do not 'race & poll' for established sockets 2675 * anymore, the common case will execute the loop only once. 2676 */ 2677 for (;;) { 2678 prepare_to_wait_exclusive(sk_sleep(sk), &wait, 2679 TASK_INTERRUPTIBLE); 2680 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) { 2681 release_sock(sk); 2682 timeo = schedule_timeout(timeo); 2683 lock_sock(sk); 2684 } 2685 err = 0; 2686 if (!skb_queue_empty(&sk->sk_receive_queue)) 2687 break; 2688 err = -EAGAIN; 2689 if (!timeo) 2690 break; 2691 err = sock_intr_errno(timeo); 2692 if (signal_pending(current)) 2693 break; 2694 } 2695 finish_wait(sk_sleep(sk), &wait); 2696 return err; 2697 } 2698 2699 /** 2700 * tipc_accept - wait for connection request 2701 * @sock: listening socket 2702 * @new_sock: new socket that is to be connected 2703 * @flags: file-related flags associated with socket 2704 * @kern: caused by kernel or by userspace? 2705 * 2706 * Return: 0 on success, errno otherwise 2707 */ 2708 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags, 2709 bool kern) 2710 { 2711 struct sock *new_sk, *sk = sock->sk; 2712 struct sk_buff *buf; 2713 struct tipc_sock *new_tsock; 2714 struct tipc_msg *msg; 2715 long timeo; 2716 int res; 2717 2718 lock_sock(sk); 2719 2720 if (sk->sk_state != TIPC_LISTEN) { 2721 res = -EINVAL; 2722 goto exit; 2723 } 2724 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK); 2725 res = tipc_wait_for_accept(sock, timeo); 2726 if (res) 2727 goto exit; 2728 2729 buf = skb_peek(&sk->sk_receive_queue); 2730 2731 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern); 2732 if (res) 2733 goto exit; 2734 security_sk_clone(sock->sk, new_sock->sk); 2735 2736 new_sk = new_sock->sk; 2737 new_tsock = tipc_sk(new_sk); 2738 msg = buf_msg(buf); 2739 2740 /* we lock on new_sk; but lockdep sees the lock on sk */ 2741 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING); 2742 2743 /* 2744 * Reject any stray messages received by new socket 2745 * before the socket lock was taken (very, very unlikely) 2746 */ 2747 tsk_rej_rx_queue(new_sk, TIPC_ERR_NO_PORT); 2748 2749 /* Connect new socket to it's peer */ 2750 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg)); 2751 2752 tsk_set_importance(new_sk, msg_importance(msg)); 2753 if (msg_named(msg)) { 2754 new_tsock->conn_addrtype = TIPC_SERVICE_ADDR; 2755 msg_set_nametype(&new_tsock->phdr, msg_nametype(msg)); 2756 msg_set_nameinst(&new_tsock->phdr, msg_nameinst(msg)); 2757 } 2758 2759 /* 2760 * Respond to 'SYN-' by discarding it & returning 'ACK'-. 2761 * Respond to 'SYN+' by queuing it on new socket. 2762 */ 2763 if (!msg_data_sz(msg)) { 2764 struct msghdr m = {NULL,}; 2765 2766 tsk_advance_rx_queue(sk); 2767 __tipc_sendstream(new_sock, &m, 0); 2768 } else { 2769 __skb_dequeue(&sk->sk_receive_queue); 2770 __skb_queue_head(&new_sk->sk_receive_queue, buf); 2771 skb_set_owner_r(buf, new_sk); 2772 } 2773 release_sock(new_sk); 2774 exit: 2775 release_sock(sk); 2776 return res; 2777 } 2778 2779 /** 2780 * tipc_shutdown - shutdown socket connection 2781 * @sock: socket structure 2782 * @how: direction to close (must be SHUT_RDWR) 2783 * 2784 * Terminates connection (if necessary), then purges socket's receive queue. 2785 * 2786 * Return: 0 on success, errno otherwise 2787 */ 2788 static int tipc_shutdown(struct socket *sock, int how) 2789 { 2790 struct sock *sk = sock->sk; 2791 int res; 2792 2793 if (how != SHUT_RDWR) 2794 return -EINVAL; 2795 2796 lock_sock(sk); 2797 2798 trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " "); 2799 __tipc_shutdown(sock, TIPC_CONN_SHUTDOWN); 2800 sk->sk_shutdown = SHUTDOWN_MASK; 2801 2802 if (sk->sk_state == TIPC_DISCONNECTING) { 2803 /* Discard any unreceived messages */ 2804 __skb_queue_purge(&sk->sk_receive_queue); 2805 2806 res = 0; 2807 } else { 2808 res = -ENOTCONN; 2809 } 2810 /* Wake up anyone sleeping in poll. */ 2811 sk->sk_state_change(sk); 2812 2813 release_sock(sk); 2814 return res; 2815 } 2816 2817 static void tipc_sk_check_probing_state(struct sock *sk, 2818 struct sk_buff_head *list) 2819 { 2820 struct tipc_sock *tsk = tipc_sk(sk); 2821 u32 pnode = tsk_peer_node(tsk); 2822 u32 pport = tsk_peer_port(tsk); 2823 u32 self = tsk_own_node(tsk); 2824 u32 oport = tsk->portid; 2825 struct sk_buff *skb; 2826 2827 if (tsk->probe_unacked) { 2828 tipc_set_sk_state(sk, TIPC_DISCONNECTING); 2829 sk->sk_err = ECONNABORTED; 2830 tipc_node_remove_conn(sock_net(sk), pnode, pport); 2831 sk->sk_state_change(sk); 2832 return; 2833 } 2834 /* Prepare new probe */ 2835 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0, 2836 pnode, self, pport, oport, TIPC_OK); 2837 if (skb) 2838 __skb_queue_tail(list, skb); 2839 tsk->probe_unacked = true; 2840 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV); 2841 } 2842 2843 static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list) 2844 { 2845 struct tipc_sock *tsk = tipc_sk(sk); 2846 2847 /* Try again later if dest link is congested */ 2848 if (tsk->cong_link_cnt) { 2849 sk_reset_timer(sk, &sk->sk_timer, msecs_to_jiffies(100)); 2850 return; 2851 } 2852 /* Prepare SYN for retransmit */ 2853 tipc_msg_skb_clone(&sk->sk_write_queue, list); 2854 } 2855 2856 static void tipc_sk_timeout(struct timer_list *t) 2857 { 2858 struct sock *sk = from_timer(sk, t, sk_timer); 2859 struct tipc_sock *tsk = tipc_sk(sk); 2860 u32 pnode = tsk_peer_node(tsk); 2861 struct sk_buff_head list; 2862 int rc = 0; 2863 2864 __skb_queue_head_init(&list); 2865 bh_lock_sock(sk); 2866 2867 /* Try again later if socket is busy */ 2868 if (sock_owned_by_user(sk)) { 2869 sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20); 2870 bh_unlock_sock(sk); 2871 sock_put(sk); 2872 return; 2873 } 2874 2875 if (sk->sk_state == TIPC_ESTABLISHED) 2876 tipc_sk_check_probing_state(sk, &list); 2877 else if (sk->sk_state == TIPC_CONNECTING) 2878 tipc_sk_retry_connect(sk, &list); 2879 2880 bh_unlock_sock(sk); 2881 2882 if (!skb_queue_empty(&list)) 2883 rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid); 2884 2885 /* SYN messages may cause link congestion */ 2886 if (rc == -ELINKCONG) { 2887 tipc_dest_push(&tsk->cong_links, pnode, 0); 2888 tsk->cong_link_cnt = 1; 2889 } 2890 sock_put(sk); 2891 } 2892 2893 static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua) 2894 { 2895 struct sock *sk = &tsk->sk; 2896 struct net *net = sock_net(sk); 2897 struct tipc_socket_addr skaddr; 2898 struct publication *p; 2899 u32 key; 2900 2901 if (tipc_sk_connected(sk)) 2902 return -EINVAL; 2903 key = tsk->portid + tsk->pub_count + 1; 2904 if (key == tsk->portid) 2905 return -EADDRINUSE; 2906 skaddr.ref = tsk->portid; 2907 skaddr.node = tipc_own_addr(net); 2908 p = tipc_nametbl_publish(net, ua, &skaddr, key); 2909 if (unlikely(!p)) 2910 return -EINVAL; 2911 2912 list_add(&p->binding_sock, &tsk->publications); 2913 tsk->pub_count++; 2914 tsk->published = true; 2915 return 0; 2916 } 2917 2918 static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua) 2919 { 2920 struct net *net = sock_net(&tsk->sk); 2921 struct publication *safe, *p; 2922 struct tipc_uaddr _ua; 2923 int rc = -EINVAL; 2924 2925 list_for_each_entry_safe(p, safe, &tsk->publications, binding_sock) { 2926 if (!ua) { 2927 tipc_uaddr(&_ua, TIPC_SERVICE_RANGE, p->scope, 2928 p->sr.type, p->sr.lower, p->sr.upper); 2929 tipc_nametbl_withdraw(net, &_ua, &p->sk, p->key); 2930 continue; 2931 } 2932 /* Unbind specific publication */ 2933 if (p->scope != ua->scope) 2934 continue; 2935 if (p->sr.type != ua->sr.type) 2936 continue; 2937 if (p->sr.lower != ua->sr.lower) 2938 continue; 2939 if (p->sr.upper != ua->sr.upper) 2940 break; 2941 tipc_nametbl_withdraw(net, ua, &p->sk, p->key); 2942 rc = 0; 2943 break; 2944 } 2945 if (list_empty(&tsk->publications)) { 2946 tsk->published = 0; 2947 rc = 0; 2948 } 2949 return rc; 2950 } 2951 2952 /* tipc_sk_reinit: set non-zero address in all existing sockets 2953 * when we go from standalone to network mode. 2954 */ 2955 void tipc_sk_reinit(struct net *net) 2956 { 2957 struct tipc_net *tn = net_generic(net, tipc_net_id); 2958 struct rhashtable_iter iter; 2959 struct tipc_sock *tsk; 2960 struct tipc_msg *msg; 2961 2962 rhashtable_walk_enter(&tn->sk_rht, &iter); 2963 2964 do { 2965 rhashtable_walk_start(&iter); 2966 2967 while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) { 2968 sock_hold(&tsk->sk); 2969 rhashtable_walk_stop(&iter); 2970 lock_sock(&tsk->sk); 2971 msg = &tsk->phdr; 2972 msg_set_prevnode(msg, tipc_own_addr(net)); 2973 msg_set_orignode(msg, tipc_own_addr(net)); 2974 release_sock(&tsk->sk); 2975 rhashtable_walk_start(&iter); 2976 sock_put(&tsk->sk); 2977 } 2978 2979 rhashtable_walk_stop(&iter); 2980 } while (tsk == ERR_PTR(-EAGAIN)); 2981 2982 rhashtable_walk_exit(&iter); 2983 } 2984 2985 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid) 2986 { 2987 struct tipc_net *tn = net_generic(net, tipc_net_id); 2988 struct tipc_sock *tsk; 2989 2990 rcu_read_lock(); 2991 tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params); 2992 if (tsk) 2993 sock_hold(&tsk->sk); 2994 rcu_read_unlock(); 2995 2996 return tsk; 2997 } 2998 2999 static int tipc_sk_insert(struct tipc_sock *tsk) 3000 { 3001 struct sock *sk = &tsk->sk; 3002 struct net *net = sock_net(sk); 3003 struct tipc_net *tn = net_generic(net, tipc_net_id); 3004 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1; 3005 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT; 3006 3007 while (remaining--) { 3008 portid++; 3009 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT)) 3010 portid = TIPC_MIN_PORT; 3011 tsk->portid = portid; 3012 sock_hold(&tsk->sk); 3013 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node, 3014 tsk_rht_params)) 3015 return 0; 3016 sock_put(&tsk->sk); 3017 } 3018 3019 return -1; 3020 } 3021 3022 static void tipc_sk_remove(struct tipc_sock *tsk) 3023 { 3024 struct sock *sk = &tsk->sk; 3025 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id); 3026 3027 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) { 3028 WARN_ON(refcount_read(&sk->sk_refcnt) == 1); 3029 __sock_put(sk); 3030 } 3031 } 3032 3033 static const struct rhashtable_params tsk_rht_params = { 3034 .nelem_hint = 192, 3035 .head_offset = offsetof(struct tipc_sock, node), 3036 .key_offset = offsetof(struct tipc_sock, portid), 3037 .key_len = sizeof(u32), /* portid */ 3038 .max_size = 1048576, 3039 .min_size = 256, 3040 .automatic_shrinking = true, 3041 }; 3042 3043 int tipc_sk_rht_init(struct net *net) 3044 { 3045 struct tipc_net *tn = net_generic(net, tipc_net_id); 3046 3047 return rhashtable_init(&tn->sk_rht, &tsk_rht_params); 3048 } 3049 3050 void tipc_sk_rht_destroy(struct net *net) 3051 { 3052 struct tipc_net *tn = net_generic(net, tipc_net_id); 3053 3054 /* Wait for socket readers to complete */ 3055 synchronize_net(); 3056 3057 rhashtable_destroy(&tn->sk_rht); 3058 } 3059 3060 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq) 3061 { 3062 struct net *net = sock_net(&tsk->sk); 3063 struct tipc_group *grp = tsk->group; 3064 struct tipc_msg *hdr = &tsk->phdr; 3065 struct tipc_uaddr ua; 3066 int rc; 3067 3068 if (mreq->type < TIPC_RESERVED_TYPES) 3069 return -EACCES; 3070 if (mreq->scope > TIPC_NODE_SCOPE) 3071 return -EINVAL; 3072 if (mreq->scope != TIPC_NODE_SCOPE) 3073 mreq->scope = TIPC_CLUSTER_SCOPE; 3074 if (grp) 3075 return -EACCES; 3076 grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open); 3077 if (!grp) 3078 return -ENOMEM; 3079 tsk->group = grp; 3080 msg_set_lookup_scope(hdr, mreq->scope); 3081 msg_set_nametype(hdr, mreq->type); 3082 msg_set_dest_droppable(hdr, true); 3083 tipc_uaddr(&ua, TIPC_SERVICE_RANGE, mreq->scope, 3084 mreq->type, mreq->instance, mreq->instance); 3085 tipc_nametbl_build_group(net, grp, &ua); 3086 rc = tipc_sk_publish(tsk, &ua); 3087 if (rc) { 3088 tipc_group_delete(net, grp); 3089 tsk->group = NULL; 3090 return rc; 3091 } 3092 /* Eliminate any risk that a broadcast overtakes sent JOINs */ 3093 tsk->mc_method.rcast = true; 3094 tsk->mc_method.mandatory = true; 3095 tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf); 3096 return rc; 3097 } 3098 3099 static int tipc_sk_leave(struct tipc_sock *tsk) 3100 { 3101 struct net *net = sock_net(&tsk->sk); 3102 struct tipc_group *grp = tsk->group; 3103 struct tipc_uaddr ua; 3104 int scope; 3105 3106 if (!grp) 3107 return -EINVAL; 3108 ua.addrtype = TIPC_SERVICE_RANGE; 3109 tipc_group_self(grp, &ua.sr, &scope); 3110 ua.scope = scope; 3111 tipc_group_delete(net, grp); 3112 tsk->group = NULL; 3113 tipc_sk_withdraw(tsk, &ua); 3114 return 0; 3115 } 3116 3117 /** 3118 * tipc_setsockopt - set socket option 3119 * @sock: socket structure 3120 * @lvl: option level 3121 * @opt: option identifier 3122 * @ov: pointer to new option value 3123 * @ol: length of option value 3124 * 3125 * For stream sockets only, accepts and ignores all IPPROTO_TCP options 3126 * (to ease compatibility). 3127 * 3128 * Return: 0 on success, errno otherwise 3129 */ 3130 static int tipc_setsockopt(struct socket *sock, int lvl, int opt, 3131 sockptr_t ov, unsigned int ol) 3132 { 3133 struct sock *sk = sock->sk; 3134 struct tipc_sock *tsk = tipc_sk(sk); 3135 struct tipc_group_req mreq; 3136 u32 value = 0; 3137 int res = 0; 3138 3139 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM)) 3140 return 0; 3141 if (lvl != SOL_TIPC) 3142 return -ENOPROTOOPT; 3143 3144 switch (opt) { 3145 case TIPC_IMPORTANCE: 3146 case TIPC_SRC_DROPPABLE: 3147 case TIPC_DEST_DROPPABLE: 3148 case TIPC_CONN_TIMEOUT: 3149 case TIPC_NODELAY: 3150 if (ol < sizeof(value)) 3151 return -EINVAL; 3152 if (copy_from_sockptr(&value, ov, sizeof(u32))) 3153 return -EFAULT; 3154 break; 3155 case TIPC_GROUP_JOIN: 3156 if (ol < sizeof(mreq)) 3157 return -EINVAL; 3158 if (copy_from_sockptr(&mreq, ov, sizeof(mreq))) 3159 return -EFAULT; 3160 break; 3161 default: 3162 if (!sockptr_is_null(ov) || ol) 3163 return -EINVAL; 3164 } 3165 3166 lock_sock(sk); 3167 3168 switch (opt) { 3169 case TIPC_IMPORTANCE: 3170 res = tsk_set_importance(sk, value); 3171 break; 3172 case TIPC_SRC_DROPPABLE: 3173 if (sock->type != SOCK_STREAM) 3174 tsk_set_unreliable(tsk, value); 3175 else 3176 res = -ENOPROTOOPT; 3177 break; 3178 case TIPC_DEST_DROPPABLE: 3179 tsk_set_unreturnable(tsk, value); 3180 break; 3181 case TIPC_CONN_TIMEOUT: 3182 tipc_sk(sk)->conn_timeout = value; 3183 break; 3184 case TIPC_MCAST_BROADCAST: 3185 tsk->mc_method.rcast = false; 3186 tsk->mc_method.mandatory = true; 3187 break; 3188 case TIPC_MCAST_REPLICAST: 3189 tsk->mc_method.rcast = true; 3190 tsk->mc_method.mandatory = true; 3191 break; 3192 case TIPC_GROUP_JOIN: 3193 res = tipc_sk_join(tsk, &mreq); 3194 break; 3195 case TIPC_GROUP_LEAVE: 3196 res = tipc_sk_leave(tsk); 3197 break; 3198 case TIPC_NODELAY: 3199 tsk->nodelay = !!value; 3200 tsk_set_nagle(tsk); 3201 break; 3202 default: 3203 res = -EINVAL; 3204 } 3205 3206 release_sock(sk); 3207 3208 return res; 3209 } 3210 3211 /** 3212 * tipc_getsockopt - get socket option 3213 * @sock: socket structure 3214 * @lvl: option level 3215 * @opt: option identifier 3216 * @ov: receptacle for option value 3217 * @ol: receptacle for length of option value 3218 * 3219 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options 3220 * (to ease compatibility). 3221 * 3222 * Return: 0 on success, errno otherwise 3223 */ 3224 static int tipc_getsockopt(struct socket *sock, int lvl, int opt, 3225 char __user *ov, int __user *ol) 3226 { 3227 struct sock *sk = sock->sk; 3228 struct tipc_sock *tsk = tipc_sk(sk); 3229 struct tipc_service_range seq; 3230 int len, scope; 3231 u32 value; 3232 int res; 3233 3234 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM)) 3235 return put_user(0, ol); 3236 if (lvl != SOL_TIPC) 3237 return -ENOPROTOOPT; 3238 res = get_user(len, ol); 3239 if (res) 3240 return res; 3241 3242 lock_sock(sk); 3243 3244 switch (opt) { 3245 case TIPC_IMPORTANCE: 3246 value = tsk_importance(tsk); 3247 break; 3248 case TIPC_SRC_DROPPABLE: 3249 value = tsk_unreliable(tsk); 3250 break; 3251 case TIPC_DEST_DROPPABLE: 3252 value = tsk_unreturnable(tsk); 3253 break; 3254 case TIPC_CONN_TIMEOUT: 3255 value = tsk->conn_timeout; 3256 /* no need to set "res", since already 0 at this point */ 3257 break; 3258 case TIPC_NODE_RECVQ_DEPTH: 3259 value = 0; /* was tipc_queue_size, now obsolete */ 3260 break; 3261 case TIPC_SOCK_RECVQ_DEPTH: 3262 value = skb_queue_len(&sk->sk_receive_queue); 3263 break; 3264 case TIPC_SOCK_RECVQ_USED: 3265 value = sk_rmem_alloc_get(sk); 3266 break; 3267 case TIPC_GROUP_JOIN: 3268 seq.type = 0; 3269 if (tsk->group) 3270 tipc_group_self(tsk->group, &seq, &scope); 3271 value = seq.type; 3272 break; 3273 default: 3274 res = -EINVAL; 3275 } 3276 3277 release_sock(sk); 3278 3279 if (res) 3280 return res; /* "get" failed */ 3281 3282 if (len < sizeof(value)) 3283 return -EINVAL; 3284 3285 if (copy_to_user(ov, &value, sizeof(value))) 3286 return -EFAULT; 3287 3288 return put_user(sizeof(value), ol); 3289 } 3290 3291 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 3292 { 3293 struct net *net = sock_net(sock->sk); 3294 struct tipc_sioc_nodeid_req nr = {0}; 3295 struct tipc_sioc_ln_req lnr; 3296 void __user *argp = (void __user *)arg; 3297 3298 switch (cmd) { 3299 case SIOCGETLINKNAME: 3300 if (copy_from_user(&lnr, argp, sizeof(lnr))) 3301 return -EFAULT; 3302 if (!tipc_node_get_linkname(net, 3303 lnr.bearer_id & 0xffff, lnr.peer, 3304 lnr.linkname, TIPC_MAX_LINK_NAME)) { 3305 if (copy_to_user(argp, &lnr, sizeof(lnr))) 3306 return -EFAULT; 3307 return 0; 3308 } 3309 return -EADDRNOTAVAIL; 3310 case SIOCGETNODEID: 3311 if (copy_from_user(&nr, argp, sizeof(nr))) 3312 return -EFAULT; 3313 if (!tipc_node_get_id(net, nr.peer, nr.node_id)) 3314 return -EADDRNOTAVAIL; 3315 if (copy_to_user(argp, &nr, sizeof(nr))) 3316 return -EFAULT; 3317 return 0; 3318 default: 3319 return -ENOIOCTLCMD; 3320 } 3321 } 3322 3323 static int tipc_socketpair(struct socket *sock1, struct socket *sock2) 3324 { 3325 struct tipc_sock *tsk2 = tipc_sk(sock2->sk); 3326 struct tipc_sock *tsk1 = tipc_sk(sock1->sk); 3327 u32 onode = tipc_own_addr(sock_net(sock1->sk)); 3328 3329 tsk1->peer.family = AF_TIPC; 3330 tsk1->peer.addrtype = TIPC_SOCKET_ADDR; 3331 tsk1->peer.scope = TIPC_NODE_SCOPE; 3332 tsk1->peer.addr.id.ref = tsk2->portid; 3333 tsk1->peer.addr.id.node = onode; 3334 tsk2->peer.family = AF_TIPC; 3335 tsk2->peer.addrtype = TIPC_SOCKET_ADDR; 3336 tsk2->peer.scope = TIPC_NODE_SCOPE; 3337 tsk2->peer.addr.id.ref = tsk1->portid; 3338 tsk2->peer.addr.id.node = onode; 3339 3340 tipc_sk_finish_conn(tsk1, tsk2->portid, onode); 3341 tipc_sk_finish_conn(tsk2, tsk1->portid, onode); 3342 return 0; 3343 } 3344 3345 /* Protocol switches for the various types of TIPC sockets */ 3346 3347 static const struct proto_ops msg_ops = { 3348 .owner = THIS_MODULE, 3349 .family = AF_TIPC, 3350 .release = tipc_release, 3351 .bind = tipc_bind, 3352 .connect = tipc_connect, 3353 .socketpair = tipc_socketpair, 3354 .accept = sock_no_accept, 3355 .getname = tipc_getname, 3356 .poll = tipc_poll, 3357 .ioctl = tipc_ioctl, 3358 .listen = sock_no_listen, 3359 .shutdown = tipc_shutdown, 3360 .setsockopt = tipc_setsockopt, 3361 .getsockopt = tipc_getsockopt, 3362 .sendmsg = tipc_sendmsg, 3363 .recvmsg = tipc_recvmsg, 3364 .mmap = sock_no_mmap, 3365 .sendpage = sock_no_sendpage 3366 }; 3367 3368 static const struct proto_ops packet_ops = { 3369 .owner = THIS_MODULE, 3370 .family = AF_TIPC, 3371 .release = tipc_release, 3372 .bind = tipc_bind, 3373 .connect = tipc_connect, 3374 .socketpair = tipc_socketpair, 3375 .accept = tipc_accept, 3376 .getname = tipc_getname, 3377 .poll = tipc_poll, 3378 .ioctl = tipc_ioctl, 3379 .listen = tipc_listen, 3380 .shutdown = tipc_shutdown, 3381 .setsockopt = tipc_setsockopt, 3382 .getsockopt = tipc_getsockopt, 3383 .sendmsg = tipc_send_packet, 3384 .recvmsg = tipc_recvmsg, 3385 .mmap = sock_no_mmap, 3386 .sendpage = sock_no_sendpage 3387 }; 3388 3389 static const struct proto_ops stream_ops = { 3390 .owner = THIS_MODULE, 3391 .family = AF_TIPC, 3392 .release = tipc_release, 3393 .bind = tipc_bind, 3394 .connect = tipc_connect, 3395 .socketpair = tipc_socketpair, 3396 .accept = tipc_accept, 3397 .getname = tipc_getname, 3398 .poll = tipc_poll, 3399 .ioctl = tipc_ioctl, 3400 .listen = tipc_listen, 3401 .shutdown = tipc_shutdown, 3402 .setsockopt = tipc_setsockopt, 3403 .getsockopt = tipc_getsockopt, 3404 .sendmsg = tipc_sendstream, 3405 .recvmsg = tipc_recvstream, 3406 .mmap = sock_no_mmap, 3407 .sendpage = sock_no_sendpage 3408 }; 3409 3410 static const struct net_proto_family tipc_family_ops = { 3411 .owner = THIS_MODULE, 3412 .family = AF_TIPC, 3413 .create = tipc_sk_create 3414 }; 3415 3416 static struct proto tipc_proto = { 3417 .name = "TIPC", 3418 .owner = THIS_MODULE, 3419 .obj_size = sizeof(struct tipc_sock), 3420 .sysctl_rmem = sysctl_tipc_rmem 3421 }; 3422 3423 /** 3424 * tipc_socket_init - initialize TIPC socket interface 3425 * 3426 * Return: 0 on success, errno otherwise 3427 */ 3428 int tipc_socket_init(void) 3429 { 3430 int res; 3431 3432 res = proto_register(&tipc_proto, 1); 3433 if (res) { 3434 pr_err("Failed to register TIPC protocol type\n"); 3435 goto out; 3436 } 3437 3438 res = sock_register(&tipc_family_ops); 3439 if (res) { 3440 pr_err("Failed to register TIPC socket type\n"); 3441 proto_unregister(&tipc_proto); 3442 goto out; 3443 } 3444 out: 3445 return res; 3446 } 3447 3448 /** 3449 * tipc_socket_stop - stop TIPC socket interface 3450 */ 3451 void tipc_socket_stop(void) 3452 { 3453 sock_unregister(tipc_family_ops.family); 3454 proto_unregister(&tipc_proto); 3455 } 3456 3457 /* Caller should hold socket lock for the passed tipc socket. */ 3458 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk) 3459 { 3460 u32 peer_node, peer_port; 3461 u32 conn_type, conn_instance; 3462 struct nlattr *nest; 3463 3464 peer_node = tsk_peer_node(tsk); 3465 peer_port = tsk_peer_port(tsk); 3466 conn_type = msg_nametype(&tsk->phdr); 3467 conn_instance = msg_nameinst(&tsk->phdr); 3468 nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON); 3469 if (!nest) 3470 return -EMSGSIZE; 3471 3472 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node)) 3473 goto msg_full; 3474 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port)) 3475 goto msg_full; 3476 3477 if (tsk->conn_addrtype != 0) { 3478 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG)) 3479 goto msg_full; 3480 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, conn_type)) 3481 goto msg_full; 3482 if (nla_put_u32(skb, TIPC_NLA_CON_INST, conn_instance)) 3483 goto msg_full; 3484 } 3485 nla_nest_end(skb, nest); 3486 3487 return 0; 3488 3489 msg_full: 3490 nla_nest_cancel(skb, nest); 3491 3492 return -EMSGSIZE; 3493 } 3494 3495 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock 3496 *tsk) 3497 { 3498 struct net *net = sock_net(skb->sk); 3499 struct sock *sk = &tsk->sk; 3500 3501 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) || 3502 nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net))) 3503 return -EMSGSIZE; 3504 3505 if (tipc_sk_connected(sk)) { 3506 if (__tipc_nl_add_sk_con(skb, tsk)) 3507 return -EMSGSIZE; 3508 } else if (!list_empty(&tsk->publications)) { 3509 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL)) 3510 return -EMSGSIZE; 3511 } 3512 return 0; 3513 } 3514 3515 /* Caller should hold socket lock for the passed tipc socket. */ 3516 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb, 3517 struct tipc_sock *tsk) 3518 { 3519 struct nlattr *attrs; 3520 void *hdr; 3521 3522 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq, 3523 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET); 3524 if (!hdr) 3525 goto msg_cancel; 3526 3527 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK); 3528 if (!attrs) 3529 goto genlmsg_cancel; 3530 3531 if (__tipc_nl_add_sk_info(skb, tsk)) 3532 goto attr_msg_cancel; 3533 3534 nla_nest_end(skb, attrs); 3535 genlmsg_end(skb, hdr); 3536 3537 return 0; 3538 3539 attr_msg_cancel: 3540 nla_nest_cancel(skb, attrs); 3541 genlmsg_cancel: 3542 genlmsg_cancel(skb, hdr); 3543 msg_cancel: 3544 return -EMSGSIZE; 3545 } 3546 3547 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb, 3548 int (*skb_handler)(struct sk_buff *skb, 3549 struct netlink_callback *cb, 3550 struct tipc_sock *tsk)) 3551 { 3552 struct rhashtable_iter *iter = (void *)cb->args[4]; 3553 struct tipc_sock *tsk; 3554 int err; 3555 3556 rhashtable_walk_start(iter); 3557 while ((tsk = rhashtable_walk_next(iter)) != NULL) { 3558 if (IS_ERR(tsk)) { 3559 err = PTR_ERR(tsk); 3560 if (err == -EAGAIN) { 3561 err = 0; 3562 continue; 3563 } 3564 break; 3565 } 3566 3567 sock_hold(&tsk->sk); 3568 rhashtable_walk_stop(iter); 3569 lock_sock(&tsk->sk); 3570 err = skb_handler(skb, cb, tsk); 3571 if (err) { 3572 release_sock(&tsk->sk); 3573 sock_put(&tsk->sk); 3574 goto out; 3575 } 3576 release_sock(&tsk->sk); 3577 rhashtable_walk_start(iter); 3578 sock_put(&tsk->sk); 3579 } 3580 rhashtable_walk_stop(iter); 3581 out: 3582 return skb->len; 3583 } 3584 EXPORT_SYMBOL(tipc_nl_sk_walk); 3585 3586 int tipc_dump_start(struct netlink_callback *cb) 3587 { 3588 return __tipc_dump_start(cb, sock_net(cb->skb->sk)); 3589 } 3590 EXPORT_SYMBOL(tipc_dump_start); 3591 3592 int __tipc_dump_start(struct netlink_callback *cb, struct net *net) 3593 { 3594 /* tipc_nl_name_table_dump() uses cb->args[0...3]. */ 3595 struct rhashtable_iter *iter = (void *)cb->args[4]; 3596 struct tipc_net *tn = tipc_net(net); 3597 3598 if (!iter) { 3599 iter = kmalloc(sizeof(*iter), GFP_KERNEL); 3600 if (!iter) 3601 return -ENOMEM; 3602 3603 cb->args[4] = (long)iter; 3604 } 3605 3606 rhashtable_walk_enter(&tn->sk_rht, iter); 3607 return 0; 3608 } 3609 3610 int tipc_dump_done(struct netlink_callback *cb) 3611 { 3612 struct rhashtable_iter *hti = (void *)cb->args[4]; 3613 3614 rhashtable_walk_exit(hti); 3615 kfree(hti); 3616 return 0; 3617 } 3618 EXPORT_SYMBOL(tipc_dump_done); 3619 3620 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb, 3621 struct tipc_sock *tsk, u32 sk_filter_state, 3622 u64 (*tipc_diag_gen_cookie)(struct sock *sk)) 3623 { 3624 struct sock *sk = &tsk->sk; 3625 struct nlattr *attrs; 3626 struct nlattr *stat; 3627 3628 /*filter response w.r.t sk_state*/ 3629 if (!(sk_filter_state & (1 << sk->sk_state))) 3630 return 0; 3631 3632 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK); 3633 if (!attrs) 3634 goto msg_cancel; 3635 3636 if (__tipc_nl_add_sk_info(skb, tsk)) 3637 goto attr_msg_cancel; 3638 3639 if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) || 3640 nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) || 3641 nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) || 3642 nla_put_u32(skb, TIPC_NLA_SOCK_UID, 3643 from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk), 3644 sock_i_uid(sk))) || 3645 nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE, 3646 tipc_diag_gen_cookie(sk), 3647 TIPC_NLA_SOCK_PAD)) 3648 goto attr_msg_cancel; 3649 3650 stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT); 3651 if (!stat) 3652 goto attr_msg_cancel; 3653 3654 if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ, 3655 skb_queue_len(&sk->sk_receive_queue)) || 3656 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ, 3657 skb_queue_len(&sk->sk_write_queue)) || 3658 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP, 3659 atomic_read(&sk->sk_drops))) 3660 goto stat_msg_cancel; 3661 3662 if (tsk->cong_link_cnt && 3663 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG)) 3664 goto stat_msg_cancel; 3665 3666 if (tsk_conn_cong(tsk) && 3667 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG)) 3668 goto stat_msg_cancel; 3669 3670 nla_nest_end(skb, stat); 3671 3672 if (tsk->group) 3673 if (tipc_group_fill_sock_diag(tsk->group, skb)) 3674 goto stat_msg_cancel; 3675 3676 nla_nest_end(skb, attrs); 3677 3678 return 0; 3679 3680 stat_msg_cancel: 3681 nla_nest_cancel(skb, stat); 3682 attr_msg_cancel: 3683 nla_nest_cancel(skb, attrs); 3684 msg_cancel: 3685 return -EMSGSIZE; 3686 } 3687 EXPORT_SYMBOL(tipc_sk_fill_sock_diag); 3688 3689 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb) 3690 { 3691 return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk); 3692 } 3693 3694 /* Caller should hold socket lock for the passed tipc socket. */ 3695 static int __tipc_nl_add_sk_publ(struct sk_buff *skb, 3696 struct netlink_callback *cb, 3697 struct publication *publ) 3698 { 3699 void *hdr; 3700 struct nlattr *attrs; 3701 3702 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq, 3703 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET); 3704 if (!hdr) 3705 goto msg_cancel; 3706 3707 attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL); 3708 if (!attrs) 3709 goto genlmsg_cancel; 3710 3711 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key)) 3712 goto attr_msg_cancel; 3713 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->sr.type)) 3714 goto attr_msg_cancel; 3715 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->sr.lower)) 3716 goto attr_msg_cancel; 3717 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->sr.upper)) 3718 goto attr_msg_cancel; 3719 3720 nla_nest_end(skb, attrs); 3721 genlmsg_end(skb, hdr); 3722 3723 return 0; 3724 3725 attr_msg_cancel: 3726 nla_nest_cancel(skb, attrs); 3727 genlmsg_cancel: 3728 genlmsg_cancel(skb, hdr); 3729 msg_cancel: 3730 return -EMSGSIZE; 3731 } 3732 3733 /* Caller should hold socket lock for the passed tipc socket. */ 3734 static int __tipc_nl_list_sk_publ(struct sk_buff *skb, 3735 struct netlink_callback *cb, 3736 struct tipc_sock *tsk, u32 *last_publ) 3737 { 3738 int err; 3739 struct publication *p; 3740 3741 if (*last_publ) { 3742 list_for_each_entry(p, &tsk->publications, binding_sock) { 3743 if (p->key == *last_publ) 3744 break; 3745 } 3746 if (p->key != *last_publ) { 3747 /* We never set seq or call nl_dump_check_consistent() 3748 * this means that setting prev_seq here will cause the 3749 * consistence check to fail in the netlink callback 3750 * handler. Resulting in the last NLMSG_DONE message 3751 * having the NLM_F_DUMP_INTR flag set. 3752 */ 3753 cb->prev_seq = 1; 3754 *last_publ = 0; 3755 return -EPIPE; 3756 } 3757 } else { 3758 p = list_first_entry(&tsk->publications, struct publication, 3759 binding_sock); 3760 } 3761 3762 list_for_each_entry_from(p, &tsk->publications, binding_sock) { 3763 err = __tipc_nl_add_sk_publ(skb, cb, p); 3764 if (err) { 3765 *last_publ = p->key; 3766 return err; 3767 } 3768 } 3769 *last_publ = 0; 3770 3771 return 0; 3772 } 3773 3774 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb) 3775 { 3776 int err; 3777 u32 tsk_portid = cb->args[0]; 3778 u32 last_publ = cb->args[1]; 3779 u32 done = cb->args[2]; 3780 struct net *net = sock_net(skb->sk); 3781 struct tipc_sock *tsk; 3782 3783 if (!tsk_portid) { 3784 struct nlattr **attrs = genl_dumpit_info(cb)->attrs; 3785 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1]; 3786 3787 if (!attrs[TIPC_NLA_SOCK]) 3788 return -EINVAL; 3789 3790 err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX, 3791 attrs[TIPC_NLA_SOCK], 3792 tipc_nl_sock_policy, NULL); 3793 if (err) 3794 return err; 3795 3796 if (!sock[TIPC_NLA_SOCK_REF]) 3797 return -EINVAL; 3798 3799 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]); 3800 } 3801 3802 if (done) 3803 return 0; 3804 3805 tsk = tipc_sk_lookup(net, tsk_portid); 3806 if (!tsk) 3807 return -EINVAL; 3808 3809 lock_sock(&tsk->sk); 3810 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ); 3811 if (!err) 3812 done = 1; 3813 release_sock(&tsk->sk); 3814 sock_put(&tsk->sk); 3815 3816 cb->args[0] = tsk_portid; 3817 cb->args[1] = last_publ; 3818 cb->args[2] = done; 3819 3820 return skb->len; 3821 } 3822 3823 /** 3824 * tipc_sk_filtering - check if a socket should be traced 3825 * @sk: the socket to be examined 3826 * 3827 * @sysctl_tipc_sk_filter is used as the socket tuple for filtering: 3828 * (portid, sock type, name type, name lower, name upper) 3829 * 3830 * Return: true if the socket meets the socket tuple data 3831 * (value 0 = 'any') or when there is no tuple set (all = 0), 3832 * otherwise false 3833 */ 3834 bool tipc_sk_filtering(struct sock *sk) 3835 { 3836 struct tipc_sock *tsk; 3837 struct publication *p; 3838 u32 _port, _sktype, _type, _lower, _upper; 3839 u32 type = 0, lower = 0, upper = 0; 3840 3841 if (!sk) 3842 return true; 3843 3844 tsk = tipc_sk(sk); 3845 3846 _port = sysctl_tipc_sk_filter[0]; 3847 _sktype = sysctl_tipc_sk_filter[1]; 3848 _type = sysctl_tipc_sk_filter[2]; 3849 _lower = sysctl_tipc_sk_filter[3]; 3850 _upper = sysctl_tipc_sk_filter[4]; 3851 3852 if (!_port && !_sktype && !_type && !_lower && !_upper) 3853 return true; 3854 3855 if (_port) 3856 return (_port == tsk->portid); 3857 3858 if (_sktype && _sktype != sk->sk_type) 3859 return false; 3860 3861 if (tsk->published) { 3862 p = list_first_entry_or_null(&tsk->publications, 3863 struct publication, binding_sock); 3864 if (p) { 3865 type = p->sr.type; 3866 lower = p->sr.lower; 3867 upper = p->sr.upper; 3868 } 3869 } 3870 3871 if (!tipc_sk_type_connectionless(sk)) { 3872 type = msg_nametype(&tsk->phdr); 3873 lower = msg_nameinst(&tsk->phdr); 3874 upper = lower; 3875 } 3876 3877 if ((_type && _type != type) || (_lower && _lower != lower) || 3878 (_upper && _upper != upper)) 3879 return false; 3880 3881 return true; 3882 } 3883 3884 u32 tipc_sock_get_portid(struct sock *sk) 3885 { 3886 return (sk) ? (tipc_sk(sk))->portid : 0; 3887 } 3888 3889 /** 3890 * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded, 3891 * both the rcv and backlog queues are considered 3892 * @sk: tipc sk to be checked 3893 * @skb: tipc msg to be checked 3894 * 3895 * Return: true if the socket rx queue allocation is > 90%, otherwise false 3896 */ 3897 3898 bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb) 3899 { 3900 atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt; 3901 unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt); 3902 unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk); 3903 3904 return (qsize > lim * 90 / 100); 3905 } 3906 3907 /** 3908 * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded, 3909 * only the rcv queue is considered 3910 * @sk: tipc sk to be checked 3911 * @skb: tipc msg to be checked 3912 * 3913 * Return: true if the socket rx queue allocation is > 90%, otherwise false 3914 */ 3915 3916 bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb) 3917 { 3918 unsigned int lim = rcvbuf_limit(sk, skb); 3919 unsigned int qsize = sk_rmem_alloc_get(sk); 3920 3921 return (qsize > lim * 90 / 100); 3922 } 3923 3924 /** 3925 * tipc_sk_dump - dump TIPC socket 3926 * @sk: tipc sk to be dumped 3927 * @dqueues: bitmask to decide if any socket queue to be dumped? 3928 * - TIPC_DUMP_NONE: don't dump socket queues 3929 * - TIPC_DUMP_SK_SNDQ: dump socket send queue 3930 * - TIPC_DUMP_SK_RCVQ: dump socket rcv queue 3931 * - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue 3932 * - TIPC_DUMP_ALL: dump all the socket queues above 3933 * @buf: returned buffer of dump data in format 3934 */ 3935 int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf) 3936 { 3937 int i = 0; 3938 size_t sz = (dqueues) ? SK_LMAX : SK_LMIN; 3939 u32 conn_type, conn_instance; 3940 struct tipc_sock *tsk; 3941 struct publication *p; 3942 bool tsk_connected; 3943 3944 if (!sk) { 3945 i += scnprintf(buf, sz, "sk data: (null)\n"); 3946 return i; 3947 } 3948 3949 tsk = tipc_sk(sk); 3950 tsk_connected = !tipc_sk_type_connectionless(sk); 3951 3952 i += scnprintf(buf, sz, "sk data: %u", sk->sk_type); 3953 i += scnprintf(buf + i, sz - i, " %d", sk->sk_state); 3954 i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk)); 3955 i += scnprintf(buf + i, sz - i, " %u", tsk->portid); 3956 i += scnprintf(buf + i, sz - i, " | %u", tsk_connected); 3957 if (tsk_connected) { 3958 i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk)); 3959 i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk)); 3960 conn_type = msg_nametype(&tsk->phdr); 3961 conn_instance = msg_nameinst(&tsk->phdr); 3962 i += scnprintf(buf + i, sz - i, " %u", conn_type); 3963 i += scnprintf(buf + i, sz - i, " %u", conn_instance); 3964 } 3965 i += scnprintf(buf + i, sz - i, " | %u", tsk->published); 3966 if (tsk->published) { 3967 p = list_first_entry_or_null(&tsk->publications, 3968 struct publication, binding_sock); 3969 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.type : 0); 3970 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.lower : 0); 3971 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.upper : 0); 3972 } 3973 i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win); 3974 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win); 3975 i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt); 3976 i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps); 3977 i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt); 3978 i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked); 3979 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked); 3980 i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt)); 3981 i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown); 3982 i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk)); 3983 i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf); 3984 i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk)); 3985 i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf); 3986 i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len)); 3987 3988 if (dqueues & TIPC_DUMP_SK_SNDQ) { 3989 i += scnprintf(buf + i, sz - i, "sk_write_queue: "); 3990 i += tipc_list_dump(&sk->sk_write_queue, false, buf + i); 3991 } 3992 3993 if (dqueues & TIPC_DUMP_SK_RCVQ) { 3994 i += scnprintf(buf + i, sz - i, "sk_receive_queue: "); 3995 i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i); 3996 } 3997 3998 if (dqueues & TIPC_DUMP_SK_BKLGQ) { 3999 i += scnprintf(buf + i, sz - i, "sk_backlog:\n head "); 4000 i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i); 4001 if (sk->sk_backlog.tail != sk->sk_backlog.head) { 4002 i += scnprintf(buf + i, sz - i, " tail "); 4003 i += tipc_skb_dump(sk->sk_backlog.tail, false, 4004 buf + i); 4005 } 4006 } 4007 4008 return i; 4009 } 4010