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