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 #include <linux/uio.h> 41 #include <trace/events/sock.h> 42 43 #include "core.h" 44 #include "name_table.h" 45 #include "node.h" 46 #include "link.h" 47 #include "name_distr.h" 48 #include "socket.h" 49 #include "bcast.h" 50 #include "netlink.h" 51 #include "group.h" 52 #include "trace.h" 53 54 #define NAGLE_START_INIT 4 55 #define NAGLE_START_MAX 1024 56 #define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */ 57 #define CONN_PROBING_INTV msecs_to_jiffies(3600000) /* [ms] => 1 h */ 58 #define TIPC_MAX_PORT 0xffffffff 59 #define TIPC_MIN_PORT 1 60 #define TIPC_ACK_RATE 4 /* ACK at 1/4 of rcv window size */ 61 62 enum { 63 TIPC_LISTEN = TCP_LISTEN, 64 TIPC_ESTABLISHED = TCP_ESTABLISHED, 65 TIPC_OPEN = TCP_CLOSE, 66 TIPC_DISCONNECTING = TCP_CLOSE_WAIT, 67 TIPC_CONNECTING = TCP_SYN_SENT, 68 }; 69 70 struct sockaddr_pair { 71 struct sockaddr_tipc sock; 72 struct sockaddr_tipc member; 73 }; 74 75 /** 76 * struct tipc_sock - TIPC socket structure 77 * @sk: socket - interacts with 'port' and with user via the socket API 78 * @max_pkt: maximum packet size "hint" used when building messages sent by port 79 * @maxnagle: maximum size of msg which can be subject to nagle 80 * @portid: unique port identity in TIPC socket hash table 81 * @phdr: preformatted message header used when sending messages 82 * @cong_links: list of congested links 83 * @publications: list of publications for port 84 * @pub_count: total # of publications port has made during its lifetime 85 * @conn_timeout: the time we can wait for an unresponded setup request 86 * @probe_unacked: probe has not received ack yet 87 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue 88 * @cong_link_cnt: number of congested links 89 * @snt_unacked: # messages sent by socket, and not yet acked by peer 90 * @snd_win: send window size 91 * @peer_caps: peer capabilities mask 92 * @rcv_unacked: # messages read by user, but not yet acked back to peer 93 * @rcv_win: receive window size 94 * @peer: 'connected' peer for dgram/rdm 95 * @node: hash table node 96 * @mc_method: cookie for use between socket and broadcast layer 97 * @rcu: rcu struct for tipc_sock 98 * @group: TIPC communications group 99 * @oneway: message count in one direction (FIXME) 100 * @nagle_start: current nagle value 101 * @snd_backlog: send backlog count 102 * @msg_acc: messages accepted; used in managing backlog and nagle 103 * @pkt_cnt: TIPC socket packet count 104 * @expect_ack: whether this TIPC socket is expecting an ack 105 * @nodelay: setsockopt() TIPC_NODELAY setting 106 * @group_is_open: TIPC socket group is fully open (FIXME) 107 * @published: true if port has one or more associated names 108 * @conn_addrtype: address type used when establishing connection 109 */ 110 struct tipc_sock { 111 struct sock sk; 112 u32 max_pkt; 113 u32 maxnagle; 114 u32 portid; 115 struct tipc_msg phdr; 116 struct list_head cong_links; 117 struct list_head publications; 118 u32 pub_count; 119 atomic_t dupl_rcvcnt; 120 u16 conn_timeout; 121 bool probe_unacked; 122 u16 cong_link_cnt; 123 u16 snt_unacked; 124 u16 snd_win; 125 u16 peer_caps; 126 u16 rcv_unacked; 127 u16 rcv_win; 128 struct sockaddr_tipc peer; 129 struct rhash_head node; 130 struct tipc_mc_method mc_method; 131 struct rcu_head rcu; 132 struct tipc_group *group; 133 u32 oneway; 134 u32 nagle_start; 135 u16 snd_backlog; 136 u16 msg_acc; 137 u16 pkt_cnt; 138 bool expect_ack; 139 bool nodelay; 140 bool group_is_open; 141 bool published; 142 u8 conn_addrtype; 143 }; 144 145 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb); 146 static void tipc_data_ready(struct sock *sk); 147 static void tipc_write_space(struct sock *sk); 148 static void tipc_sock_destruct(struct sock *sk); 149 static int tipc_release(struct socket *sock); 150 static void tipc_sk_timeout(struct timer_list *t); 151 static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua); 152 static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua); 153 static int tipc_sk_leave(struct tipc_sock *tsk); 154 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid); 155 static int tipc_sk_insert(struct tipc_sock *tsk); 156 static void tipc_sk_remove(struct tipc_sock *tsk); 157 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz); 158 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz); 159 static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack); 160 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p); 161 162 static const struct proto_ops packet_ops; 163 static const struct proto_ops stream_ops; 164 static const struct proto_ops msg_ops; 165 static struct proto tipc_proto; 166 static const struct rhashtable_params tsk_rht_params; 167 168 static u32 tsk_own_node(struct tipc_sock *tsk) 169 { 170 return msg_prevnode(&tsk->phdr); 171 } 172 173 static u32 tsk_peer_node(struct tipc_sock *tsk) 174 { 175 return msg_destnode(&tsk->phdr); 176 } 177 178 static u32 tsk_peer_port(struct tipc_sock *tsk) 179 { 180 return msg_destport(&tsk->phdr); 181 } 182 183 static bool tsk_unreliable(struct tipc_sock *tsk) 184 { 185 return msg_src_droppable(&tsk->phdr) != 0; 186 } 187 188 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable) 189 { 190 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0); 191 } 192 193 static bool tsk_unreturnable(struct tipc_sock *tsk) 194 { 195 return msg_dest_droppable(&tsk->phdr) != 0; 196 } 197 198 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable) 199 { 200 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0); 201 } 202 203 static int tsk_importance(struct tipc_sock *tsk) 204 { 205 return msg_importance(&tsk->phdr); 206 } 207 208 static struct tipc_sock *tipc_sk(const struct sock *sk) 209 { 210 return container_of(sk, struct tipc_sock, sk); 211 } 212 213 int tsk_set_importance(struct sock *sk, int imp) 214 { 215 if (imp > TIPC_CRITICAL_IMPORTANCE) 216 return -EINVAL; 217 msg_set_importance(&tipc_sk(sk)->phdr, (u32)imp); 218 return 0; 219 } 220 221 static bool tsk_conn_cong(struct tipc_sock *tsk) 222 { 223 return tsk->snt_unacked > tsk->snd_win; 224 } 225 226 static u16 tsk_blocks(int len) 227 { 228 return ((len / FLOWCTL_BLK_SZ) + 1); 229 } 230 231 /* tsk_blocks(): translate a buffer size in bytes to number of 232 * advertisable blocks, taking into account the ratio truesize(len)/len 233 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ 234 */ 235 static u16 tsk_adv_blocks(int len) 236 { 237 return len / FLOWCTL_BLK_SZ / 4; 238 } 239 240 /* tsk_inc(): increment counter for sent or received data 241 * - If block based flow control is not supported by peer we 242 * fall back to message based ditto, incrementing the counter 243 */ 244 static u16 tsk_inc(struct tipc_sock *tsk, int msglen) 245 { 246 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL)) 247 return ((msglen / FLOWCTL_BLK_SZ) + 1); 248 return 1; 249 } 250 251 /* tsk_set_nagle - enable/disable nagle property by manipulating maxnagle 252 */ 253 static void tsk_set_nagle(struct tipc_sock *tsk) 254 { 255 struct sock *sk = &tsk->sk; 256 257 tsk->maxnagle = 0; 258 if (sk->sk_type != SOCK_STREAM) 259 return; 260 if (tsk->nodelay) 261 return; 262 if (!(tsk->peer_caps & TIPC_NAGLE)) 263 return; 264 /* Limit node local buffer size to avoid receive queue overflow */ 265 if (tsk->max_pkt == MAX_MSG_SIZE) 266 tsk->maxnagle = 1500; 267 else 268 tsk->maxnagle = tsk->max_pkt; 269 } 270 271 /** 272 * tsk_advance_rx_queue - discard first buffer in socket receive queue 273 * @sk: network socket 274 * 275 * Caller must hold socket lock 276 */ 277 static void tsk_advance_rx_queue(struct sock *sk) 278 { 279 trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " "); 280 kfree_skb(__skb_dequeue(&sk->sk_receive_queue)); 281 } 282 283 /* tipc_sk_respond() : send response message back to sender 284 */ 285 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err) 286 { 287 u32 selector; 288 u32 dnode; 289 u32 onode = tipc_own_addr(sock_net(sk)); 290 291 if (!tipc_msg_reverse(onode, &skb, err)) 292 return; 293 294 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!"); 295 dnode = msg_destnode(buf_msg(skb)); 296 selector = msg_origport(buf_msg(skb)); 297 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector); 298 } 299 300 /** 301 * tsk_rej_rx_queue - reject all buffers in socket receive queue 302 * @sk: network socket 303 * @error: response error code 304 * 305 * Caller must hold socket lock 306 */ 307 static void tsk_rej_rx_queue(struct sock *sk, int error) 308 { 309 struct sk_buff *skb; 310 311 while ((skb = __skb_dequeue(&sk->sk_receive_queue))) 312 tipc_sk_respond(sk, skb, error); 313 } 314 315 static bool tipc_sk_connected(const struct sock *sk) 316 { 317 return READ_ONCE(sk->sk_state) == TIPC_ESTABLISHED; 318 } 319 320 /* tipc_sk_type_connectionless - check if the socket is datagram socket 321 * @sk: socket 322 * 323 * Returns true if connection less, false otherwise 324 */ 325 static bool tipc_sk_type_connectionless(struct sock *sk) 326 { 327 return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM; 328 } 329 330 /* tsk_peer_msg - verify if message was sent by connected port's peer 331 * 332 * Handles cases where the node's network address has changed from 333 * the default of <0.0.0> to its configured setting. 334 */ 335 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg) 336 { 337 struct sock *sk = &tsk->sk; 338 u32 self = tipc_own_addr(sock_net(sk)); 339 u32 peer_port = tsk_peer_port(tsk); 340 u32 orig_node, peer_node; 341 342 if (unlikely(!tipc_sk_connected(sk))) 343 return false; 344 345 if (unlikely(msg_origport(msg) != peer_port)) 346 return false; 347 348 orig_node = msg_orignode(msg); 349 peer_node = tsk_peer_node(tsk); 350 351 if (likely(orig_node == peer_node)) 352 return true; 353 354 if (!orig_node && peer_node == self) 355 return true; 356 357 if (!peer_node && orig_node == self) 358 return true; 359 360 return false; 361 } 362 363 /* tipc_set_sk_state - set the sk_state of the socket 364 * @sk: socket 365 * 366 * Caller must hold socket lock 367 * 368 * Returns 0 on success, errno otherwise 369 */ 370 static int tipc_set_sk_state(struct sock *sk, int state) 371 { 372 int oldsk_state = sk->sk_state; 373 int res = -EINVAL; 374 375 switch (state) { 376 case TIPC_OPEN: 377 res = 0; 378 break; 379 case TIPC_LISTEN: 380 case TIPC_CONNECTING: 381 if (oldsk_state == TIPC_OPEN) 382 res = 0; 383 break; 384 case TIPC_ESTABLISHED: 385 if (oldsk_state == TIPC_CONNECTING || 386 oldsk_state == TIPC_OPEN) 387 res = 0; 388 break; 389 case TIPC_DISCONNECTING: 390 if (oldsk_state == TIPC_CONNECTING || 391 oldsk_state == TIPC_ESTABLISHED) 392 res = 0; 393 break; 394 } 395 396 if (!res) 397 sk->sk_state = state; 398 399 return res; 400 } 401 402 static int tipc_sk_sock_err(struct socket *sock, long *timeout) 403 { 404 struct sock *sk = sock->sk; 405 int err = sock_error(sk); 406 int typ = sock->type; 407 408 if (err) 409 return err; 410 if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) { 411 if (sk->sk_state == TIPC_DISCONNECTING) 412 return -EPIPE; 413 else if (!tipc_sk_connected(sk)) 414 return -ENOTCONN; 415 } 416 if (!*timeout) 417 return -EAGAIN; 418 if (signal_pending(current)) 419 return sock_intr_errno(*timeout); 420 421 return 0; 422 } 423 424 #define tipc_wait_for_cond(sock_, timeo_, condition_) \ 425 ({ \ 426 DEFINE_WAIT_FUNC(wait_, woken_wake_function); \ 427 struct sock *sk_; \ 428 int rc_; \ 429 \ 430 while ((rc_ = !(condition_))) { \ 431 /* coupled with smp_wmb() in tipc_sk_proto_rcv() */ \ 432 smp_rmb(); \ 433 sk_ = (sock_)->sk; \ 434 rc_ = tipc_sk_sock_err((sock_), timeo_); \ 435 if (rc_) \ 436 break; \ 437 add_wait_queue(sk_sleep(sk_), &wait_); \ 438 release_sock(sk_); \ 439 *(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \ 440 sched_annotate_sleep(); \ 441 lock_sock(sk_); \ 442 remove_wait_queue(sk_sleep(sk_), &wait_); \ 443 } \ 444 rc_; \ 445 }) 446 447 /** 448 * tipc_sk_create - create a TIPC socket 449 * @net: network namespace (must be default network) 450 * @sock: pre-allocated socket structure 451 * @protocol: protocol indicator (must be 0) 452 * @kern: caused by kernel or by userspace? 453 * 454 * This routine creates additional data structures used by the TIPC socket, 455 * initializes them, and links them together. 456 * 457 * Return: 0 on success, errno otherwise 458 */ 459 static int tipc_sk_create(struct net *net, struct socket *sock, 460 int protocol, int kern) 461 { 462 const struct proto_ops *ops; 463 struct sock *sk; 464 struct tipc_sock *tsk; 465 struct tipc_msg *msg; 466 467 /* Validate arguments */ 468 if (unlikely(protocol != 0)) 469 return -EPROTONOSUPPORT; 470 471 switch (sock->type) { 472 case SOCK_STREAM: 473 ops = &stream_ops; 474 break; 475 case SOCK_SEQPACKET: 476 ops = &packet_ops; 477 break; 478 case SOCK_DGRAM: 479 case SOCK_RDM: 480 ops = &msg_ops; 481 break; 482 default: 483 return -EPROTOTYPE; 484 } 485 486 /* Allocate socket's protocol area */ 487 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern); 488 if (sk == NULL) 489 return -ENOMEM; 490 491 tsk = tipc_sk(sk); 492 tsk->max_pkt = MAX_PKT_DEFAULT; 493 tsk->maxnagle = 0; 494 tsk->nagle_start = NAGLE_START_INIT; 495 INIT_LIST_HEAD(&tsk->publications); 496 INIT_LIST_HEAD(&tsk->cong_links); 497 msg = &tsk->phdr; 498 499 /* Finish initializing socket data structures */ 500 sock->ops = ops; 501 sock_init_data(sock, sk); 502 tipc_set_sk_state(sk, TIPC_OPEN); 503 if (tipc_sk_insert(tsk)) { 504 sk_free(sk); 505 pr_warn("Socket create failed; port number exhausted\n"); 506 return -EINVAL; 507 } 508 509 /* Ensure tsk is visible before we read own_addr. */ 510 smp_mb(); 511 512 tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE, 513 TIPC_NAMED_MSG, NAMED_H_SIZE, 0); 514 515 msg_set_origport(msg, tsk->portid); 516 timer_setup(&sk->sk_timer, tipc_sk_timeout, 0); 517 sk->sk_shutdown = 0; 518 sk->sk_backlog_rcv = tipc_sk_backlog_rcv; 519 sk->sk_rcvbuf = READ_ONCE(sysctl_tipc_rmem[1]); 520 sk->sk_data_ready = tipc_data_ready; 521 sk->sk_write_space = tipc_write_space; 522 sk->sk_destruct = tipc_sock_destruct; 523 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT; 524 tsk->group_is_open = true; 525 atomic_set(&tsk->dupl_rcvcnt, 0); 526 527 /* Start out with safe limits until we receive an advertised window */ 528 tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN); 529 tsk->rcv_win = tsk->snd_win; 530 531 if (tipc_sk_type_connectionless(sk)) { 532 tsk_set_unreturnable(tsk, true); 533 if (sock->type == SOCK_DGRAM) 534 tsk_set_unreliable(tsk, true); 535 } 536 __skb_queue_head_init(&tsk->mc_method.deferredq); 537 trace_tipc_sk_create(sk, NULL, TIPC_DUMP_NONE, " "); 538 return 0; 539 } 540 541 static void tipc_sk_callback(struct rcu_head *head) 542 { 543 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu); 544 545 sock_put(&tsk->sk); 546 } 547 548 /* Caller should hold socket lock for the socket. */ 549 static void __tipc_shutdown(struct socket *sock, int error) 550 { 551 struct sock *sk = sock->sk; 552 struct tipc_sock *tsk = tipc_sk(sk); 553 struct net *net = sock_net(sk); 554 long timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT); 555 u32 dnode = tsk_peer_node(tsk); 556 struct sk_buff *skb; 557 558 /* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */ 559 tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt && 560 !tsk_conn_cong(tsk))); 561 562 /* Push out delayed messages if in Nagle mode */ 563 tipc_sk_push_backlog(tsk, false); 564 /* Remove pending SYN */ 565 __skb_queue_purge(&sk->sk_write_queue); 566 567 /* Remove partially received buffer if any */ 568 skb = skb_peek(&sk->sk_receive_queue); 569 if (skb && TIPC_SKB_CB(skb)->bytes_read) { 570 __skb_unlink(skb, &sk->sk_receive_queue); 571 kfree_skb(skb); 572 } 573 574 /* Reject all unreceived messages if connectionless */ 575 if (tipc_sk_type_connectionless(sk)) { 576 tsk_rej_rx_queue(sk, error); 577 return; 578 } 579 580 switch (sk->sk_state) { 581 case TIPC_CONNECTING: 582 case TIPC_ESTABLISHED: 583 tipc_set_sk_state(sk, TIPC_DISCONNECTING); 584 tipc_node_remove_conn(net, dnode, tsk->portid); 585 /* Send a FIN+/- to its peer */ 586 skb = __skb_dequeue(&sk->sk_receive_queue); 587 if (skb) { 588 __skb_queue_purge(&sk->sk_receive_queue); 589 tipc_sk_respond(sk, skb, error); 590 break; 591 } 592 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE, 593 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode, 594 tsk_own_node(tsk), tsk_peer_port(tsk), 595 tsk->portid, error); 596 if (skb) 597 tipc_node_xmit_skb(net, skb, dnode, tsk->portid); 598 break; 599 case TIPC_LISTEN: 600 /* Reject all SYN messages */ 601 tsk_rej_rx_queue(sk, error); 602 break; 603 default: 604 __skb_queue_purge(&sk->sk_receive_queue); 605 break; 606 } 607 } 608 609 /** 610 * tipc_release - destroy a TIPC socket 611 * @sock: socket to destroy 612 * 613 * This routine cleans up any messages that are still queued on the socket. 614 * For DGRAM and RDM socket types, all queued messages are rejected. 615 * For SEQPACKET and STREAM socket types, the first message is rejected 616 * and any others are discarded. (If the first message on a STREAM socket 617 * is partially-read, it is discarded and the next one is rejected instead.) 618 * 619 * NOTE: Rejected messages are not necessarily returned to the sender! They 620 * are returned or discarded according to the "destination droppable" setting 621 * specified for the message by the sender. 622 * 623 * Return: 0 on success, errno otherwise 624 */ 625 static int tipc_release(struct socket *sock) 626 { 627 struct sock *sk = sock->sk; 628 struct tipc_sock *tsk; 629 630 /* 631 * Exit if socket isn't fully initialized (occurs when a failed accept() 632 * releases a pre-allocated child socket that was never used) 633 */ 634 if (sk == NULL) 635 return 0; 636 637 tsk = tipc_sk(sk); 638 lock_sock(sk); 639 640 trace_tipc_sk_release(sk, NULL, TIPC_DUMP_ALL, " "); 641 __tipc_shutdown(sock, TIPC_ERR_NO_PORT); 642 sk->sk_shutdown = SHUTDOWN_MASK; 643 tipc_sk_leave(tsk); 644 tipc_sk_withdraw(tsk, NULL); 645 __skb_queue_purge(&tsk->mc_method.deferredq); 646 sk_stop_timer(sk, &sk->sk_timer); 647 tipc_sk_remove(tsk); 648 649 sock_orphan(sk); 650 /* Reject any messages that accumulated in backlog queue */ 651 release_sock(sk); 652 tipc_dest_list_purge(&tsk->cong_links); 653 tsk->cong_link_cnt = 0; 654 call_rcu(&tsk->rcu, tipc_sk_callback); 655 sock->sk = NULL; 656 657 return 0; 658 } 659 660 /** 661 * __tipc_bind - associate or disassociate TIPC name(s) with a socket 662 * @sock: socket structure 663 * @skaddr: socket address describing name(s) and desired operation 664 * @alen: size of socket address data structure 665 * 666 * Name and name sequence binding are indicated using a positive scope value; 667 * a negative scope value unbinds the specified name. Specifying no name 668 * (i.e. a socket address length of 0) unbinds all names from the socket. 669 * 670 * Return: 0 on success, errno otherwise 671 * 672 * NOTE: This routine doesn't need to take the socket lock since it doesn't 673 * access any non-constant socket information. 674 */ 675 static int __tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen) 676 { 677 struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr; 678 struct tipc_sock *tsk = tipc_sk(sock->sk); 679 bool unbind = false; 680 681 if (unlikely(!alen)) 682 return tipc_sk_withdraw(tsk, NULL); 683 684 if (ua->addrtype == TIPC_SERVICE_ADDR) { 685 ua->addrtype = TIPC_SERVICE_RANGE; 686 ua->sr.upper = ua->sr.lower; 687 } 688 if (ua->scope < 0) { 689 unbind = true; 690 ua->scope = -ua->scope; 691 } 692 /* Users may still use deprecated TIPC_ZONE_SCOPE */ 693 if (ua->scope != TIPC_NODE_SCOPE) 694 ua->scope = TIPC_CLUSTER_SCOPE; 695 696 if (tsk->group) 697 return -EACCES; 698 699 if (unbind) 700 return tipc_sk_withdraw(tsk, ua); 701 return tipc_sk_publish(tsk, ua); 702 } 703 704 int tipc_sk_bind(struct socket *sock, struct sockaddr *skaddr, int alen) 705 { 706 int res; 707 708 lock_sock(sock->sk); 709 res = __tipc_bind(sock, skaddr, alen); 710 release_sock(sock->sk); 711 return res; 712 } 713 714 static int tipc_bind(struct socket *sock, struct sockaddr_unsized *skaddr, int alen) 715 { 716 struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr; 717 u32 atype = ua->addrtype; 718 719 if (alen) { 720 if (!tipc_uaddr_valid(ua, alen)) 721 return -EINVAL; 722 if (atype == TIPC_SOCKET_ADDR) 723 return -EAFNOSUPPORT; 724 if (ua->sr.type < TIPC_RESERVED_TYPES) { 725 pr_warn_once("Can't bind to reserved service type %u\n", 726 ua->sr.type); 727 return -EACCES; 728 } 729 } 730 return tipc_sk_bind(sock, (struct sockaddr *)skaddr, alen); 731 } 732 733 /** 734 * tipc_getname - get port ID of socket or peer socket 735 * @sock: socket structure 736 * @uaddr: area for returned socket address 737 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID 738 * 739 * Return: 0 on success, errno otherwise 740 * 741 * NOTE: This routine doesn't need to take the socket lock since it only 742 * accesses socket information that is unchanging (or which changes in 743 * a completely predictable manner). 744 */ 745 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr, 746 int peer) 747 { 748 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr; 749 struct sock *sk = sock->sk; 750 struct tipc_sock *tsk = tipc_sk(sk); 751 752 memset(addr, 0, sizeof(*addr)); 753 if (peer) { 754 if ((!tipc_sk_connected(sk)) && 755 ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING))) 756 return -ENOTCONN; 757 addr->addr.id.ref = tsk_peer_port(tsk); 758 addr->addr.id.node = tsk_peer_node(tsk); 759 } else { 760 addr->addr.id.ref = tsk->portid; 761 addr->addr.id.node = tipc_own_addr(sock_net(sk)); 762 } 763 764 addr->addrtype = TIPC_SOCKET_ADDR; 765 addr->family = AF_TIPC; 766 addr->scope = 0; 767 addr->addr.name.domain = 0; 768 769 return sizeof(*addr); 770 } 771 772 /** 773 * tipc_poll - read and possibly block on pollmask 774 * @file: file structure associated with the socket 775 * @sock: socket for which to calculate the poll bits 776 * @wait: ??? 777 * 778 * Return: pollmask value 779 * 780 * COMMENTARY: 781 * It appears that the usual socket locking mechanisms are not useful here 782 * since the pollmask info is potentially out-of-date the moment this routine 783 * exits. TCP and other protocols seem to rely on higher level poll routines 784 * to handle any preventable race conditions, so TIPC will do the same ... 785 * 786 * IMPORTANT: The fact that a read or write operation is indicated does NOT 787 * imply that the operation will succeed, merely that it should be performed 788 * and will not block. 789 */ 790 static __poll_t tipc_poll(struct file *file, struct socket *sock, 791 poll_table *wait) 792 { 793 struct sock *sk = sock->sk; 794 struct tipc_sock *tsk = tipc_sk(sk); 795 __poll_t revents = 0; 796 797 sock_poll_wait(file, sock, wait); 798 trace_tipc_sk_poll(sk, NULL, TIPC_DUMP_ALL, " "); 799 800 if (sk->sk_shutdown & RCV_SHUTDOWN) 801 revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM; 802 if (sk->sk_shutdown == SHUTDOWN_MASK) 803 revents |= EPOLLHUP; 804 805 switch (sk->sk_state) { 806 case TIPC_ESTABLISHED: 807 if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk)) 808 revents |= EPOLLOUT; 809 fallthrough; 810 case TIPC_LISTEN: 811 case TIPC_CONNECTING: 812 if (!skb_queue_empty_lockless(&sk->sk_receive_queue)) 813 revents |= EPOLLIN | EPOLLRDNORM; 814 break; 815 case TIPC_OPEN: 816 if (tsk->group_is_open && !tsk->cong_link_cnt) 817 revents |= EPOLLOUT; 818 if (!tipc_sk_type_connectionless(sk)) 819 break; 820 if (skb_queue_empty_lockless(&sk->sk_receive_queue)) 821 break; 822 revents |= EPOLLIN | EPOLLRDNORM; 823 break; 824 case TIPC_DISCONNECTING: 825 revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP; 826 break; 827 } 828 return revents; 829 } 830 831 /** 832 * tipc_sendmcast - send multicast message 833 * @sock: socket structure 834 * @ua: destination address struct 835 * @msg: message to send 836 * @dlen: length of data to send 837 * @timeout: timeout to wait for wakeup 838 * 839 * Called from function tipc_sendmsg(), which has done all sanity checks 840 * Return: the number of bytes sent on success, or errno 841 */ 842 static int tipc_sendmcast(struct socket *sock, struct tipc_uaddr *ua, 843 struct msghdr *msg, size_t dlen, long timeout) 844 { 845 struct sock *sk = sock->sk; 846 struct tipc_sock *tsk = tipc_sk(sk); 847 struct tipc_msg *hdr = &tsk->phdr; 848 struct net *net = sock_net(sk); 849 int mtu = tipc_bcast_get_mtu(net); 850 struct sk_buff_head pkts; 851 struct tipc_nlist dsts; 852 int rc; 853 854 if (tsk->group) 855 return -EACCES; 856 857 /* Block or return if any destination link is congested */ 858 rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt); 859 if (unlikely(rc)) 860 return rc; 861 862 /* Lookup destination nodes */ 863 tipc_nlist_init(&dsts, tipc_own_addr(net)); 864 tipc_nametbl_lookup_mcast_nodes(net, ua, &dsts); 865 if (!dsts.local && !dsts.remote) 866 return -EHOSTUNREACH; 867 868 /* Build message header */ 869 msg_set_type(hdr, TIPC_MCAST_MSG); 870 msg_set_hdr_sz(hdr, MCAST_H_SIZE); 871 msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE); 872 msg_set_destport(hdr, 0); 873 msg_set_destnode(hdr, 0); 874 msg_set_nametype(hdr, ua->sr.type); 875 msg_set_namelower(hdr, ua->sr.lower); 876 msg_set_nameupper(hdr, ua->sr.upper); 877 878 /* Build message as chain of buffers */ 879 __skb_queue_head_init(&pkts); 880 rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts); 881 882 /* Send message if build was successful */ 883 if (unlikely(rc == dlen)) { 884 trace_tipc_sk_sendmcast(sk, skb_peek(&pkts), 885 TIPC_DUMP_SK_SNDQ, " "); 886 rc = tipc_mcast_xmit(net, &pkts, &tsk->mc_method, &dsts, 887 &tsk->cong_link_cnt); 888 } 889 890 tipc_nlist_purge(&dsts); 891 892 return rc ? rc : dlen; 893 } 894 895 /** 896 * tipc_send_group_msg - send a message to a member in the group 897 * @net: network namespace 898 * @tsk: tipc socket 899 * @m: message to send 900 * @mb: group member 901 * @dnode: destination node 902 * @dport: destination port 903 * @dlen: total length of message data 904 */ 905 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk, 906 struct msghdr *m, struct tipc_member *mb, 907 u32 dnode, u32 dport, int dlen) 908 { 909 u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group); 910 struct tipc_mc_method *method = &tsk->mc_method; 911 int blks = tsk_blocks(GROUP_H_SIZE + dlen); 912 struct tipc_msg *hdr = &tsk->phdr; 913 struct sk_buff_head pkts; 914 int mtu, rc; 915 916 /* Complete message header */ 917 msg_set_type(hdr, TIPC_GRP_UCAST_MSG); 918 msg_set_hdr_sz(hdr, GROUP_H_SIZE); 919 msg_set_destport(hdr, dport); 920 msg_set_destnode(hdr, dnode); 921 msg_set_grp_bc_seqno(hdr, bc_snd_nxt); 922 923 /* Build message as chain of buffers */ 924 __skb_queue_head_init(&pkts); 925 mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false); 926 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts); 927 if (unlikely(rc != dlen)) 928 return rc; 929 930 /* Send message */ 931 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid); 932 if (unlikely(rc == -ELINKCONG)) { 933 tipc_dest_push(&tsk->cong_links, dnode, 0); 934 tsk->cong_link_cnt++; 935 } 936 937 /* Update send window */ 938 tipc_group_update_member(mb, blks); 939 940 /* A broadcast sent within next EXPIRE period must follow same path */ 941 method->rcast = true; 942 method->mandatory = true; 943 return dlen; 944 } 945 946 /** 947 * tipc_send_group_unicast - send message to a member in the group 948 * @sock: socket structure 949 * @m: message to send 950 * @dlen: total length of message data 951 * @timeout: timeout to wait for wakeup 952 * 953 * Called from function tipc_sendmsg(), which has done all sanity checks 954 * Return: the number of bytes sent on success, or errno 955 */ 956 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m, 957 int dlen, long timeout) 958 { 959 struct sock *sk = sock->sk; 960 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name; 961 int blks = tsk_blocks(GROUP_H_SIZE + dlen); 962 struct tipc_sock *tsk = tipc_sk(sk); 963 struct net *net = sock_net(sk); 964 struct tipc_member *mb = NULL; 965 u32 node, port; 966 int rc; 967 968 node = ua->sk.node; 969 port = ua->sk.ref; 970 if (!port && !node) 971 return -EHOSTUNREACH; 972 973 /* Block or return if destination link or member is congested */ 974 rc = tipc_wait_for_cond(sock, &timeout, 975 !tipc_dest_find(&tsk->cong_links, node, 0) && 976 tsk->group && 977 !tipc_group_cong(tsk->group, node, port, blks, 978 &mb)); 979 if (unlikely(rc)) 980 return rc; 981 982 if (unlikely(!mb)) 983 return -EHOSTUNREACH; 984 985 rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen); 986 987 return rc ? rc : dlen; 988 } 989 990 /** 991 * tipc_send_group_anycast - send message to any member with given identity 992 * @sock: socket structure 993 * @m: message to send 994 * @dlen: total length of message data 995 * @timeout: timeout to wait for wakeup 996 * 997 * Called from function tipc_sendmsg(), which has done all sanity checks 998 * Return: the number of bytes sent on success, or errno 999 */ 1000 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m, 1001 int dlen, long timeout) 1002 { 1003 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name; 1004 struct sock *sk = sock->sk; 1005 struct tipc_sock *tsk = tipc_sk(sk); 1006 struct list_head *cong_links = &tsk->cong_links; 1007 int blks = tsk_blocks(GROUP_H_SIZE + dlen); 1008 struct tipc_msg *hdr = &tsk->phdr; 1009 struct tipc_member *first = NULL; 1010 struct tipc_member *mbr = NULL; 1011 struct net *net = sock_net(sk); 1012 u32 node, port, exclude; 1013 LIST_HEAD(dsts); 1014 int lookups = 0; 1015 int dstcnt, rc; 1016 bool cong; 1017 1018 ua->sa.type = msg_nametype(hdr); 1019 ua->scope = msg_lookup_scope(hdr); 1020 1021 while (++lookups < 4) { 1022 exclude = tipc_group_exclude(tsk->group); 1023 1024 first = NULL; 1025 1026 /* Look for a non-congested destination member, if any */ 1027 while (1) { 1028 if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt, 1029 exclude, false)) 1030 return -EHOSTUNREACH; 1031 tipc_dest_pop(&dsts, &node, &port); 1032 cong = tipc_group_cong(tsk->group, node, port, blks, 1033 &mbr); 1034 if (!cong) 1035 break; 1036 if (mbr == first) 1037 break; 1038 if (!first) 1039 first = mbr; 1040 } 1041 1042 /* Start over if destination was not in member list */ 1043 if (unlikely(!mbr)) 1044 continue; 1045 1046 if (likely(!cong && !tipc_dest_find(cong_links, node, 0))) 1047 break; 1048 1049 /* Block or return if destination link or member is congested */ 1050 rc = tipc_wait_for_cond(sock, &timeout, 1051 !tipc_dest_find(cong_links, node, 0) && 1052 tsk->group && 1053 !tipc_group_cong(tsk->group, node, port, 1054 blks, &mbr)); 1055 if (unlikely(rc)) 1056 return rc; 1057 1058 /* Send, unless destination disappeared while waiting */ 1059 if (likely(mbr)) 1060 break; 1061 } 1062 1063 if (unlikely(lookups >= 4)) 1064 return -EHOSTUNREACH; 1065 1066 rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen); 1067 1068 return rc ? rc : dlen; 1069 } 1070 1071 /** 1072 * tipc_send_group_bcast - send message to all members in communication group 1073 * @sock: socket structure 1074 * @m: message to send 1075 * @dlen: total length of message data 1076 * @timeout: timeout to wait for wakeup 1077 * 1078 * Called from function tipc_sendmsg(), which has done all sanity checks 1079 * Return: the number of bytes sent on success, or errno 1080 */ 1081 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m, 1082 int dlen, long timeout) 1083 { 1084 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name; 1085 struct sock *sk = sock->sk; 1086 struct net *net = sock_net(sk); 1087 struct tipc_sock *tsk = tipc_sk(sk); 1088 struct tipc_nlist *dsts; 1089 struct tipc_mc_method *method = &tsk->mc_method; 1090 bool ack = method->mandatory && method->rcast; 1091 int blks = tsk_blocks(MCAST_H_SIZE + dlen); 1092 struct tipc_msg *hdr = &tsk->phdr; 1093 int mtu = tipc_bcast_get_mtu(net); 1094 struct sk_buff_head pkts; 1095 int rc = -EHOSTUNREACH; 1096 1097 /* Block or return if any destination link or member is congested */ 1098 rc = tipc_wait_for_cond(sock, &timeout, 1099 !tsk->cong_link_cnt && tsk->group && 1100 !tipc_group_bc_cong(tsk->group, blks)); 1101 if (unlikely(rc)) 1102 return rc; 1103 1104 dsts = tipc_group_dests(tsk->group); 1105 if (!dsts->local && !dsts->remote) 1106 return -EHOSTUNREACH; 1107 1108 /* Complete message header */ 1109 if (ua) { 1110 msg_set_type(hdr, TIPC_GRP_MCAST_MSG); 1111 msg_set_nameinst(hdr, ua->sa.instance); 1112 } else { 1113 msg_set_type(hdr, TIPC_GRP_BCAST_MSG); 1114 msg_set_nameinst(hdr, 0); 1115 } 1116 msg_set_hdr_sz(hdr, GROUP_H_SIZE); 1117 msg_set_destport(hdr, 0); 1118 msg_set_destnode(hdr, 0); 1119 msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group)); 1120 1121 /* Avoid getting stuck with repeated forced replicasts */ 1122 msg_set_grp_bc_ack_req(hdr, ack); 1123 1124 /* Build message as chain of buffers */ 1125 __skb_queue_head_init(&pkts); 1126 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts); 1127 if (unlikely(rc != dlen)) 1128 return rc; 1129 1130 /* Send message */ 1131 rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt); 1132 if (unlikely(rc)) 1133 return rc; 1134 1135 /* Update broadcast sequence number and send windows */ 1136 tipc_group_update_bc_members(tsk->group, blks, ack); 1137 1138 /* Broadcast link is now free to choose method for next broadcast */ 1139 method->mandatory = false; 1140 method->expires = jiffies; 1141 1142 return dlen; 1143 } 1144 1145 /** 1146 * tipc_send_group_mcast - send message to all members with given identity 1147 * @sock: socket structure 1148 * @m: message to send 1149 * @dlen: total length of message data 1150 * @timeout: timeout to wait for wakeup 1151 * 1152 * Called from function tipc_sendmsg(), which has done all sanity checks 1153 * Return: the number of bytes sent on success, or errno 1154 */ 1155 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m, 1156 int dlen, long timeout) 1157 { 1158 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name; 1159 struct sock *sk = sock->sk; 1160 struct tipc_sock *tsk = tipc_sk(sk); 1161 struct tipc_group *grp = tsk->group; 1162 struct tipc_msg *hdr = &tsk->phdr; 1163 struct net *net = sock_net(sk); 1164 u32 dstcnt, exclude; 1165 LIST_HEAD(dsts); 1166 1167 ua->sa.type = msg_nametype(hdr); 1168 ua->scope = msg_lookup_scope(hdr); 1169 exclude = tipc_group_exclude(grp); 1170 1171 if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt, exclude, true)) 1172 return -EHOSTUNREACH; 1173 1174 if (dstcnt == 1) { 1175 tipc_dest_pop(&dsts, &ua->sk.node, &ua->sk.ref); 1176 return tipc_send_group_unicast(sock, m, dlen, timeout); 1177 } 1178 1179 tipc_dest_list_purge(&dsts); 1180 return tipc_send_group_bcast(sock, m, dlen, timeout); 1181 } 1182 1183 /** 1184 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets 1185 * @net: the associated network namespace 1186 * @arrvq: queue with arriving messages, to be cloned after destination lookup 1187 * @inputq: queue with cloned messages, delivered to socket after dest lookup 1188 * 1189 * Multi-threaded: parallel calls with reference to same queues may occur 1190 */ 1191 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq, 1192 struct sk_buff_head *inputq) 1193 { 1194 u32 self = tipc_own_addr(net); 1195 struct sk_buff *skb, *_skb; 1196 u32 portid, onode; 1197 struct sk_buff_head tmpq; 1198 struct list_head dports; 1199 struct tipc_msg *hdr; 1200 struct tipc_uaddr ua; 1201 int user, mtyp, hlen; 1202 1203 __skb_queue_head_init(&tmpq); 1204 INIT_LIST_HEAD(&dports); 1205 ua.addrtype = TIPC_SERVICE_RANGE; 1206 1207 /* tipc_skb_peek() increments the head skb's reference counter */ 1208 skb = tipc_skb_peek(arrvq, &inputq->lock); 1209 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) { 1210 hdr = buf_msg(skb); 1211 user = msg_user(hdr); 1212 mtyp = msg_type(hdr); 1213 hlen = skb_headroom(skb) + msg_hdr_sz(hdr); 1214 onode = msg_orignode(hdr); 1215 ua.sr.type = msg_nametype(hdr); 1216 ua.sr.lower = msg_namelower(hdr); 1217 ua.sr.upper = msg_nameupper(hdr); 1218 if (onode == self) 1219 ua.scope = TIPC_ANY_SCOPE; 1220 else 1221 ua.scope = TIPC_CLUSTER_SCOPE; 1222 1223 if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) { 1224 spin_lock_bh(&inputq->lock); 1225 if (skb_peek(arrvq) == skb) { 1226 __skb_dequeue(arrvq); 1227 __skb_queue_tail(inputq, skb); 1228 } 1229 kfree_skb(skb); 1230 spin_unlock_bh(&inputq->lock); 1231 continue; 1232 } 1233 1234 /* Group messages require exact scope match */ 1235 if (msg_in_group(hdr)) { 1236 ua.sr.lower = 0; 1237 ua.sr.upper = ~0; 1238 ua.scope = msg_lookup_scope(hdr); 1239 } 1240 1241 /* Create destination port list: */ 1242 tipc_nametbl_lookup_mcast_sockets(net, &ua, &dports); 1243 1244 /* Clone message per destination */ 1245 while (tipc_dest_pop(&dports, NULL, &portid)) { 1246 _skb = __pskb_copy(skb, hlen, GFP_ATOMIC); 1247 if (_skb) { 1248 msg_set_destport(buf_msg(_skb), portid); 1249 __skb_queue_tail(&tmpq, _skb); 1250 continue; 1251 } 1252 pr_warn("Failed to clone mcast rcv buffer\n"); 1253 } 1254 /* Append clones to inputq only if skb is still head of arrvq */ 1255 spin_lock_bh(&inputq->lock); 1256 if (skb_peek(arrvq) == skb) { 1257 skb_queue_splice_tail_init(&tmpq, inputq); 1258 /* Decrement the skb's refcnt */ 1259 kfree_skb(__skb_dequeue(arrvq)); 1260 } 1261 spin_unlock_bh(&inputq->lock); 1262 __skb_queue_purge(&tmpq); 1263 kfree_skb(skb); 1264 } 1265 tipc_sk_rcv(net, inputq); 1266 } 1267 1268 /* tipc_sk_push_backlog(): send accumulated buffers in socket write queue 1269 * when socket is in Nagle mode 1270 */ 1271 static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack) 1272 { 1273 struct sk_buff_head *txq = &tsk->sk.sk_write_queue; 1274 struct sk_buff *skb = skb_peek_tail(txq); 1275 struct net *net = sock_net(&tsk->sk); 1276 u32 dnode = tsk_peer_node(tsk); 1277 int rc; 1278 1279 if (nagle_ack) { 1280 tsk->pkt_cnt += skb_queue_len(txq); 1281 if (!tsk->pkt_cnt || tsk->msg_acc / tsk->pkt_cnt < 2) { 1282 tsk->oneway = 0; 1283 if (tsk->nagle_start < NAGLE_START_MAX) 1284 tsk->nagle_start *= 2; 1285 tsk->expect_ack = false; 1286 pr_debug("tsk %10u: bad nagle %u -> %u, next start %u!\n", 1287 tsk->portid, tsk->msg_acc, tsk->pkt_cnt, 1288 tsk->nagle_start); 1289 } else { 1290 tsk->nagle_start = NAGLE_START_INIT; 1291 if (skb) { 1292 msg_set_ack_required(buf_msg(skb)); 1293 tsk->expect_ack = true; 1294 } else { 1295 tsk->expect_ack = false; 1296 } 1297 } 1298 tsk->msg_acc = 0; 1299 tsk->pkt_cnt = 0; 1300 } 1301 1302 if (!skb || tsk->cong_link_cnt) 1303 return; 1304 1305 /* Do not send SYN again after congestion */ 1306 if (msg_is_syn(buf_msg(skb))) 1307 return; 1308 1309 if (tsk->msg_acc) 1310 tsk->pkt_cnt += skb_queue_len(txq); 1311 tsk->snt_unacked += tsk->snd_backlog; 1312 tsk->snd_backlog = 0; 1313 rc = tipc_node_xmit(net, txq, dnode, tsk->portid); 1314 if (rc == -ELINKCONG) 1315 tsk->cong_link_cnt = 1; 1316 } 1317 1318 /** 1319 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message 1320 * @tsk: receiving socket 1321 * @skb: pointer to message buffer. 1322 * @inputq: buffer list containing the buffers 1323 * @xmitq: output message area 1324 */ 1325 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb, 1326 struct sk_buff_head *inputq, 1327 struct sk_buff_head *xmitq) 1328 { 1329 struct tipc_msg *hdr = buf_msg(skb); 1330 u32 onode = tsk_own_node(tsk); 1331 struct sock *sk = &tsk->sk; 1332 int mtyp = msg_type(hdr); 1333 bool was_cong; 1334 1335 /* Ignore if connection cannot be validated: */ 1336 if (!tsk_peer_msg(tsk, hdr)) { 1337 trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!"); 1338 goto exit; 1339 } 1340 1341 if (unlikely(msg_errcode(hdr))) { 1342 tipc_set_sk_state(sk, TIPC_DISCONNECTING); 1343 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk), 1344 tsk_peer_port(tsk)); 1345 sk->sk_state_change(sk); 1346 1347 /* State change is ignored if socket already awake, 1348 * - convert msg to abort msg and add to inqueue 1349 */ 1350 msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE); 1351 msg_set_type(hdr, TIPC_CONN_MSG); 1352 msg_set_size(hdr, BASIC_H_SIZE); 1353 msg_set_hdr_sz(hdr, BASIC_H_SIZE); 1354 __skb_queue_tail(inputq, skb); 1355 return; 1356 } 1357 1358 tsk->probe_unacked = false; 1359 1360 if (mtyp == CONN_PROBE) { 1361 msg_set_type(hdr, CONN_PROBE_REPLY); 1362 if (tipc_msg_reverse(onode, &skb, TIPC_OK)) 1363 __skb_queue_tail(xmitq, skb); 1364 return; 1365 } else if (mtyp == CONN_ACK) { 1366 was_cong = tsk_conn_cong(tsk); 1367 tipc_sk_push_backlog(tsk, msg_nagle_ack(hdr)); 1368 tsk->snt_unacked -= msg_conn_ack(hdr); 1369 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) 1370 tsk->snd_win = msg_adv_win(hdr); 1371 if (was_cong && !tsk_conn_cong(tsk)) 1372 sk->sk_write_space(sk); 1373 } else if (mtyp != CONN_PROBE_REPLY) { 1374 pr_warn("Received unknown CONN_PROTO msg\n"); 1375 } 1376 exit: 1377 kfree_skb(skb); 1378 } 1379 1380 /** 1381 * tipc_sendmsg - send message in connectionless manner 1382 * @sock: socket structure 1383 * @m: message to send 1384 * @dsz: amount of user data to be sent 1385 * 1386 * Message must have an destination specified explicitly. 1387 * Used for SOCK_RDM and SOCK_DGRAM messages, 1388 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections. 1389 * (Note: 'SYN+' is prohibited on SOCK_STREAM.) 1390 * 1391 * Return: the number of bytes sent on success, or errno otherwise 1392 */ 1393 static int tipc_sendmsg(struct socket *sock, 1394 struct msghdr *m, size_t dsz) 1395 { 1396 struct sock *sk = sock->sk; 1397 int ret; 1398 1399 lock_sock(sk); 1400 ret = __tipc_sendmsg(sock, m, dsz); 1401 release_sock(sk); 1402 1403 return ret; 1404 } 1405 1406 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen) 1407 { 1408 struct sock *sk = sock->sk; 1409 struct net *net = sock_net(sk); 1410 struct tipc_sock *tsk = tipc_sk(sk); 1411 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name; 1412 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT); 1413 struct list_head *clinks = &tsk->cong_links; 1414 bool syn = !tipc_sk_type_connectionless(sk); 1415 struct tipc_group *grp = tsk->group; 1416 struct tipc_msg *hdr = &tsk->phdr; 1417 struct tipc_socket_addr skaddr; 1418 struct sk_buff_head pkts; 1419 int atype, mtu, rc; 1420 1421 if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE)) 1422 return -EMSGSIZE; 1423 1424 if (ua) { 1425 if (!tipc_uaddr_valid(ua, m->msg_namelen)) 1426 return -EINVAL; 1427 atype = ua->addrtype; 1428 } 1429 1430 /* If socket belongs to a communication group follow other paths */ 1431 if (grp) { 1432 if (!ua) 1433 return tipc_send_group_bcast(sock, m, dlen, timeout); 1434 if (atype == TIPC_SERVICE_ADDR) 1435 return tipc_send_group_anycast(sock, m, dlen, timeout); 1436 if (atype == TIPC_SOCKET_ADDR) 1437 return tipc_send_group_unicast(sock, m, dlen, timeout); 1438 if (atype == TIPC_SERVICE_RANGE) 1439 return tipc_send_group_mcast(sock, m, dlen, timeout); 1440 return -EINVAL; 1441 } 1442 1443 if (!ua) { 1444 ua = (struct tipc_uaddr *)&tsk->peer; 1445 if (!syn && ua->family != AF_TIPC) 1446 return -EDESTADDRREQ; 1447 atype = ua->addrtype; 1448 } 1449 1450 if (unlikely(syn)) { 1451 if (sk->sk_state == TIPC_LISTEN) 1452 return -EPIPE; 1453 if (sk->sk_state != TIPC_OPEN) 1454 return -EISCONN; 1455 if (tsk->published) 1456 return -EOPNOTSUPP; 1457 if (atype == TIPC_SERVICE_ADDR) 1458 tsk->conn_addrtype = atype; 1459 msg_set_syn(hdr, 1); 1460 } 1461 1462 memset(&skaddr, 0, sizeof(skaddr)); 1463 1464 /* Determine destination */ 1465 if (atype == TIPC_SERVICE_RANGE) { 1466 return tipc_sendmcast(sock, ua, m, dlen, timeout); 1467 } else if (atype == TIPC_SERVICE_ADDR) { 1468 skaddr.node = ua->lookup_node; 1469 ua->scope = tipc_node2scope(skaddr.node); 1470 if (!tipc_nametbl_lookup_anycast(net, ua, &skaddr)) 1471 return -EHOSTUNREACH; 1472 } else if (atype == TIPC_SOCKET_ADDR) { 1473 skaddr = ua->sk; 1474 } else { 1475 return -EINVAL; 1476 } 1477 1478 /* Block or return if destination link is congested */ 1479 rc = tipc_wait_for_cond(sock, &timeout, 1480 !tipc_dest_find(clinks, skaddr.node, 0)); 1481 if (unlikely(rc)) 1482 return rc; 1483 1484 /* Finally build message header */ 1485 msg_set_destnode(hdr, skaddr.node); 1486 msg_set_destport(hdr, skaddr.ref); 1487 if (atype == TIPC_SERVICE_ADDR) { 1488 msg_set_type(hdr, TIPC_NAMED_MSG); 1489 msg_set_hdr_sz(hdr, NAMED_H_SIZE); 1490 msg_set_nametype(hdr, ua->sa.type); 1491 msg_set_nameinst(hdr, ua->sa.instance); 1492 msg_set_lookup_scope(hdr, ua->scope); 1493 } else { /* TIPC_SOCKET_ADDR */ 1494 msg_set_type(hdr, TIPC_DIRECT_MSG); 1495 msg_set_lookup_scope(hdr, 0); 1496 msg_set_hdr_sz(hdr, BASIC_H_SIZE); 1497 } 1498 1499 /* Add message body */ 1500 __skb_queue_head_init(&pkts); 1501 mtu = tipc_node_get_mtu(net, skaddr.node, tsk->portid, true); 1502 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts); 1503 if (unlikely(rc != dlen)) 1504 return rc; 1505 if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue))) { 1506 __skb_queue_purge(&pkts); 1507 return -ENOMEM; 1508 } 1509 1510 /* Send message */ 1511 trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " "); 1512 rc = tipc_node_xmit(net, &pkts, skaddr.node, tsk->portid); 1513 if (unlikely(rc == -ELINKCONG)) { 1514 tipc_dest_push(clinks, skaddr.node, 0); 1515 tsk->cong_link_cnt++; 1516 rc = 0; 1517 } 1518 1519 if (unlikely(syn && !rc)) { 1520 tipc_set_sk_state(sk, TIPC_CONNECTING); 1521 if (dlen && timeout) { 1522 timeout = msecs_to_jiffies(timeout); 1523 tipc_wait_for_connect(sock, &timeout); 1524 } 1525 } 1526 1527 return rc ? rc : dlen; 1528 } 1529 1530 /** 1531 * tipc_sendstream - send stream-oriented data 1532 * @sock: socket structure 1533 * @m: data to send 1534 * @dsz: total length of data to be transmitted 1535 * 1536 * Used for SOCK_STREAM data. 1537 * 1538 * Return: the number of bytes sent on success (or partial success), 1539 * or errno if no data sent 1540 */ 1541 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz) 1542 { 1543 struct sock *sk = sock->sk; 1544 int ret; 1545 1546 lock_sock(sk); 1547 ret = __tipc_sendstream(sock, m, dsz); 1548 release_sock(sk); 1549 1550 return ret; 1551 } 1552 1553 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen) 1554 { 1555 struct sock *sk = sock->sk; 1556 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name); 1557 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT); 1558 struct sk_buff_head *txq = &sk->sk_write_queue; 1559 struct tipc_sock *tsk = tipc_sk(sk); 1560 struct tipc_msg *hdr = &tsk->phdr; 1561 struct net *net = sock_net(sk); 1562 struct sk_buff *skb; 1563 u32 dnode = tsk_peer_node(tsk); 1564 int maxnagle = tsk->maxnagle; 1565 int maxpkt = tsk->max_pkt; 1566 int send, sent = 0; 1567 int blocks, rc = 0; 1568 1569 if (unlikely(dlen > INT_MAX)) 1570 return -EMSGSIZE; 1571 1572 /* Handle implicit connection setup */ 1573 if (unlikely(dest && sk->sk_state == TIPC_OPEN)) { 1574 rc = __tipc_sendmsg(sock, m, dlen); 1575 if (dlen && dlen == rc) { 1576 tsk->peer_caps = tipc_node_get_capabilities(net, dnode); 1577 tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr)); 1578 } 1579 return rc; 1580 } 1581 1582 do { 1583 rc = tipc_wait_for_cond(sock, &timeout, 1584 (!tsk->cong_link_cnt && 1585 !tsk_conn_cong(tsk) && 1586 tipc_sk_connected(sk))); 1587 if (unlikely(rc)) 1588 break; 1589 send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE); 1590 blocks = tsk->snd_backlog; 1591 if (tsk->oneway++ >= tsk->nagle_start && maxnagle && 1592 send <= maxnagle) { 1593 rc = tipc_msg_append(hdr, m, send, maxnagle, txq); 1594 if (unlikely(rc < 0)) 1595 break; 1596 blocks += rc; 1597 tsk->msg_acc++; 1598 if (blocks <= 64 && tsk->expect_ack) { 1599 tsk->snd_backlog = blocks; 1600 sent += send; 1601 break; 1602 } else if (blocks > 64) { 1603 tsk->pkt_cnt += skb_queue_len(txq); 1604 } else { 1605 skb = skb_peek_tail(txq); 1606 if (skb) { 1607 msg_set_ack_required(buf_msg(skb)); 1608 tsk->expect_ack = true; 1609 } else { 1610 tsk->expect_ack = false; 1611 } 1612 tsk->msg_acc = 0; 1613 tsk->pkt_cnt = 0; 1614 } 1615 } else { 1616 rc = tipc_msg_build(hdr, m, sent, send, maxpkt, txq); 1617 if (unlikely(rc != send)) 1618 break; 1619 blocks += tsk_inc(tsk, send + MIN_H_SIZE); 1620 } 1621 trace_tipc_sk_sendstream(sk, skb_peek(txq), 1622 TIPC_DUMP_SK_SNDQ, " "); 1623 rc = tipc_node_xmit(net, txq, dnode, tsk->portid); 1624 if (unlikely(rc == -ELINKCONG)) { 1625 tsk->cong_link_cnt = 1; 1626 rc = 0; 1627 } 1628 if (likely(!rc)) { 1629 tsk->snt_unacked += blocks; 1630 tsk->snd_backlog = 0; 1631 sent += send; 1632 } 1633 } while (sent < dlen && !rc); 1634 1635 return sent ? sent : rc; 1636 } 1637 1638 /** 1639 * tipc_send_packet - send a connection-oriented message 1640 * @sock: socket structure 1641 * @m: message to send 1642 * @dsz: length of data to be transmitted 1643 * 1644 * Used for SOCK_SEQPACKET messages. 1645 * 1646 * Return: the number of bytes sent on success, or errno otherwise 1647 */ 1648 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz) 1649 { 1650 if (dsz > TIPC_MAX_USER_MSG_SIZE) 1651 return -EMSGSIZE; 1652 1653 return tipc_sendstream(sock, m, dsz); 1654 } 1655 1656 /* tipc_sk_finish_conn - complete the setup of a connection 1657 */ 1658 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port, 1659 u32 peer_node) 1660 { 1661 struct sock *sk = &tsk->sk; 1662 struct net *net = sock_net(sk); 1663 struct tipc_msg *msg = &tsk->phdr; 1664 1665 msg_set_syn(msg, 0); 1666 msg_set_destnode(msg, peer_node); 1667 msg_set_destport(msg, peer_port); 1668 msg_set_type(msg, TIPC_CONN_MSG); 1669 msg_set_lookup_scope(msg, 0); 1670 msg_set_hdr_sz(msg, SHORT_H_SIZE); 1671 1672 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV); 1673 tipc_set_sk_state(sk, TIPC_ESTABLISHED); 1674 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port); 1675 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid, true); 1676 tsk->peer_caps = tipc_node_get_capabilities(net, peer_node); 1677 tsk_set_nagle(tsk); 1678 __skb_queue_purge(&sk->sk_write_queue); 1679 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) 1680 return; 1681 1682 /* Fall back to message based flow control */ 1683 tsk->rcv_win = FLOWCTL_MSG_WIN; 1684 tsk->snd_win = FLOWCTL_MSG_WIN; 1685 } 1686 1687 /** 1688 * tipc_sk_set_orig_addr - capture sender's address for received message 1689 * @m: descriptor for message info 1690 * @skb: received message 1691 * 1692 * Note: Address is not captured if not requested by receiver. 1693 */ 1694 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb) 1695 { 1696 DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name); 1697 struct tipc_msg *hdr = buf_msg(skb); 1698 1699 if (!srcaddr) 1700 return; 1701 1702 srcaddr->sock.family = AF_TIPC; 1703 srcaddr->sock.addrtype = TIPC_SOCKET_ADDR; 1704 srcaddr->sock.scope = 0; 1705 srcaddr->sock.addr.id.ref = msg_origport(hdr); 1706 srcaddr->sock.addr.id.node = msg_orignode(hdr); 1707 srcaddr->sock.addr.name.domain = 0; 1708 m->msg_namelen = sizeof(struct sockaddr_tipc); 1709 1710 if (!msg_in_group(hdr)) 1711 return; 1712 1713 /* Group message users may also want to know sending member's id */ 1714 srcaddr->member.family = AF_TIPC; 1715 srcaddr->member.addrtype = TIPC_SERVICE_ADDR; 1716 srcaddr->member.scope = 0; 1717 srcaddr->member.addr.name.name.type = msg_nametype(hdr); 1718 srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member; 1719 srcaddr->member.addr.name.domain = 0; 1720 m->msg_namelen = sizeof(*srcaddr); 1721 } 1722 1723 /** 1724 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message 1725 * @m: descriptor for message info 1726 * @skb: received message buffer 1727 * @tsk: TIPC port associated with message 1728 * 1729 * Note: Ancillary data is not captured if not requested by receiver. 1730 * 1731 * Return: 0 if successful, otherwise errno 1732 */ 1733 static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb, 1734 struct tipc_sock *tsk) 1735 { 1736 struct tipc_msg *hdr; 1737 u32 data[3] = {0,}; 1738 bool has_addr; 1739 int dlen, rc; 1740 1741 if (likely(m->msg_controllen == 0)) 1742 return 0; 1743 1744 hdr = buf_msg(skb); 1745 dlen = msg_data_sz(hdr); 1746 1747 /* Capture errored message object, if any */ 1748 if (msg_errcode(hdr)) { 1749 if (skb_linearize(skb)) 1750 return -ENOMEM; 1751 hdr = buf_msg(skb); 1752 data[0] = msg_errcode(hdr); 1753 data[1] = dlen; 1754 rc = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, data); 1755 if (rc || !dlen) 1756 return rc; 1757 rc = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, dlen, msg_data(hdr)); 1758 if (rc) 1759 return rc; 1760 } 1761 1762 /* Capture TIPC_SERVICE_ADDR/RANGE destination address, if any */ 1763 switch (msg_type(hdr)) { 1764 case TIPC_NAMED_MSG: 1765 has_addr = true; 1766 data[0] = msg_nametype(hdr); 1767 data[1] = msg_namelower(hdr); 1768 data[2] = data[1]; 1769 break; 1770 case TIPC_MCAST_MSG: 1771 has_addr = true; 1772 data[0] = msg_nametype(hdr); 1773 data[1] = msg_namelower(hdr); 1774 data[2] = msg_nameupper(hdr); 1775 break; 1776 case TIPC_CONN_MSG: 1777 has_addr = !!tsk->conn_addrtype; 1778 data[0] = msg_nametype(&tsk->phdr); 1779 data[1] = msg_nameinst(&tsk->phdr); 1780 data[2] = data[1]; 1781 break; 1782 default: 1783 has_addr = false; 1784 } 1785 if (!has_addr) 1786 return 0; 1787 return put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, data); 1788 } 1789 1790 static struct sk_buff *tipc_sk_build_ack(struct tipc_sock *tsk) 1791 { 1792 struct sock *sk = &tsk->sk; 1793 struct sk_buff *skb = NULL; 1794 struct tipc_msg *msg; 1795 u32 peer_port = tsk_peer_port(tsk); 1796 u32 dnode = tsk_peer_node(tsk); 1797 1798 if (!tipc_sk_connected(sk)) 1799 return NULL; 1800 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0, 1801 dnode, tsk_own_node(tsk), peer_port, 1802 tsk->portid, TIPC_OK); 1803 if (!skb) 1804 return NULL; 1805 msg = buf_msg(skb); 1806 msg_set_conn_ack(msg, tsk->rcv_unacked); 1807 tsk->rcv_unacked = 0; 1808 1809 /* Adjust to and advertize the correct window limit */ 1810 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) { 1811 tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf); 1812 msg_set_adv_win(msg, tsk->rcv_win); 1813 } 1814 return skb; 1815 } 1816 1817 static void tipc_sk_send_ack(struct tipc_sock *tsk) 1818 { 1819 struct sk_buff *skb; 1820 1821 skb = tipc_sk_build_ack(tsk); 1822 if (!skb) 1823 return; 1824 1825 tipc_node_xmit_skb(sock_net(&tsk->sk), skb, tsk_peer_node(tsk), 1826 msg_link_selector(buf_msg(skb))); 1827 } 1828 1829 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop) 1830 { 1831 struct sock *sk = sock->sk; 1832 DEFINE_WAIT_FUNC(wait, woken_wake_function); 1833 long timeo = *timeop; 1834 int err = sock_error(sk); 1835 1836 if (err) 1837 return err; 1838 1839 for (;;) { 1840 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) { 1841 if (sk->sk_shutdown & RCV_SHUTDOWN) { 1842 err = -ENOTCONN; 1843 break; 1844 } 1845 add_wait_queue(sk_sleep(sk), &wait); 1846 release_sock(sk); 1847 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo); 1848 sched_annotate_sleep(); 1849 lock_sock(sk); 1850 remove_wait_queue(sk_sleep(sk), &wait); 1851 } 1852 err = 0; 1853 if (!skb_queue_empty(&sk->sk_receive_queue)) 1854 break; 1855 err = -EAGAIN; 1856 if (!timeo) 1857 break; 1858 err = sock_intr_errno(timeo); 1859 if (signal_pending(current)) 1860 break; 1861 1862 err = sock_error(sk); 1863 if (err) 1864 break; 1865 } 1866 *timeop = timeo; 1867 return err; 1868 } 1869 1870 /** 1871 * tipc_recvmsg - receive packet-oriented message 1872 * @sock: network socket 1873 * @m: descriptor for message info 1874 * @buflen: length of user buffer area 1875 * @flags: receive flags 1876 * 1877 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages. 1878 * If the complete message doesn't fit in user area, truncate it. 1879 * 1880 * Return: size of returned message data, errno otherwise 1881 */ 1882 static int tipc_recvmsg(struct socket *sock, struct msghdr *m, 1883 size_t buflen, int flags) 1884 { 1885 struct sock *sk = sock->sk; 1886 bool connected = !tipc_sk_type_connectionless(sk); 1887 struct tipc_sock *tsk = tipc_sk(sk); 1888 int rc, err, hlen, dlen, copy; 1889 struct tipc_skb_cb *skb_cb; 1890 struct sk_buff_head xmitq; 1891 struct tipc_msg *hdr; 1892 struct sk_buff *skb; 1893 bool grp_evt; 1894 long timeout; 1895 1896 /* Catch invalid receive requests */ 1897 if (unlikely(!buflen)) 1898 return -EINVAL; 1899 1900 lock_sock(sk); 1901 if (unlikely(connected && sk->sk_state == TIPC_OPEN)) { 1902 rc = -ENOTCONN; 1903 goto exit; 1904 } 1905 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT); 1906 1907 /* Step rcv queue to first msg with data or error; wait if necessary */ 1908 do { 1909 rc = tipc_wait_for_rcvmsg(sock, &timeout); 1910 if (unlikely(rc)) 1911 goto exit; 1912 skb = skb_peek(&sk->sk_receive_queue); 1913 skb_cb = TIPC_SKB_CB(skb); 1914 hdr = buf_msg(skb); 1915 dlen = msg_data_sz(hdr); 1916 hlen = msg_hdr_sz(hdr); 1917 err = msg_errcode(hdr); 1918 grp_evt = msg_is_grp_evt(hdr); 1919 if (likely(dlen || err)) 1920 break; 1921 tsk_advance_rx_queue(sk); 1922 } while (1); 1923 1924 /* Collect msg meta data, including error code and rejected data */ 1925 tipc_sk_set_orig_addr(m, skb); 1926 rc = tipc_sk_anc_data_recv(m, skb, tsk); 1927 if (unlikely(rc)) 1928 goto exit; 1929 hdr = buf_msg(skb); 1930 1931 /* Capture data if non-error msg, otherwise just set return value */ 1932 if (likely(!err)) { 1933 int offset = skb_cb->bytes_read; 1934 1935 copy = min_t(int, dlen - offset, buflen); 1936 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy); 1937 if (unlikely(rc)) 1938 goto exit; 1939 if (unlikely(offset + copy < dlen)) { 1940 if (flags & MSG_EOR) { 1941 if (!(flags & MSG_PEEK)) 1942 skb_cb->bytes_read = offset + copy; 1943 } else { 1944 m->msg_flags |= MSG_TRUNC; 1945 skb_cb->bytes_read = 0; 1946 } 1947 } else { 1948 if (flags & MSG_EOR) 1949 m->msg_flags |= MSG_EOR; 1950 skb_cb->bytes_read = 0; 1951 } 1952 } else { 1953 copy = 0; 1954 rc = 0; 1955 if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control) { 1956 rc = -ECONNRESET; 1957 goto exit; 1958 } 1959 } 1960 1961 /* Mark message as group event if applicable */ 1962 if (unlikely(grp_evt)) { 1963 if (msg_grp_evt(hdr) == TIPC_WITHDRAWN) 1964 m->msg_flags |= MSG_EOR; 1965 m->msg_flags |= MSG_OOB; 1966 copy = 0; 1967 } 1968 1969 /* Caption of data or error code/rejected data was successful */ 1970 if (unlikely(flags & MSG_PEEK)) 1971 goto exit; 1972 1973 /* Send group flow control advertisement when applicable */ 1974 if (tsk->group && msg_in_group(hdr) && !grp_evt) { 1975 __skb_queue_head_init(&xmitq); 1976 tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen), 1977 msg_orignode(hdr), msg_origport(hdr), 1978 &xmitq); 1979 tipc_node_distr_xmit(sock_net(sk), &xmitq); 1980 } 1981 1982 if (skb_cb->bytes_read) 1983 goto exit; 1984 1985 tsk_advance_rx_queue(sk); 1986 1987 if (likely(!connected)) 1988 goto exit; 1989 1990 /* Send connection flow control advertisement when applicable */ 1991 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen); 1992 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE) 1993 tipc_sk_send_ack(tsk); 1994 exit: 1995 release_sock(sk); 1996 return rc ? rc : copy; 1997 } 1998 1999 /** 2000 * tipc_recvstream - receive stream-oriented data 2001 * @sock: network socket 2002 * @m: descriptor for message info 2003 * @buflen: total size of user buffer area 2004 * @flags: receive flags 2005 * 2006 * Used for SOCK_STREAM messages only. If not enough data is available 2007 * will optionally wait for more; never truncates data. 2008 * 2009 * Return: size of returned message data, errno otherwise 2010 */ 2011 static int tipc_recvstream(struct socket *sock, struct msghdr *m, 2012 size_t buflen, int flags) 2013 { 2014 struct sock *sk = sock->sk; 2015 struct tipc_sock *tsk = tipc_sk(sk); 2016 struct sk_buff *skb; 2017 struct tipc_msg *hdr; 2018 struct tipc_skb_cb *skb_cb; 2019 bool peek = flags & MSG_PEEK; 2020 int offset, required, copy, copied = 0; 2021 int hlen, dlen, err, rc; 2022 long timeout; 2023 2024 /* Catch invalid receive attempts */ 2025 if (unlikely(!buflen)) 2026 return -EINVAL; 2027 2028 lock_sock(sk); 2029 2030 if (unlikely(sk->sk_state == TIPC_OPEN)) { 2031 rc = -ENOTCONN; 2032 goto exit; 2033 } 2034 required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen); 2035 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT); 2036 2037 do { 2038 /* Look at first msg in receive queue; wait if necessary */ 2039 rc = tipc_wait_for_rcvmsg(sock, &timeout); 2040 if (unlikely(rc)) 2041 break; 2042 skb = skb_peek(&sk->sk_receive_queue); 2043 skb_cb = TIPC_SKB_CB(skb); 2044 hdr = buf_msg(skb); 2045 dlen = msg_data_sz(hdr); 2046 hlen = msg_hdr_sz(hdr); 2047 err = msg_errcode(hdr); 2048 2049 /* Discard any empty non-errored (SYN-) message */ 2050 if (unlikely(!dlen && !err)) { 2051 tsk_advance_rx_queue(sk); 2052 continue; 2053 } 2054 2055 /* Collect msg meta data, incl. error code and rejected data */ 2056 if (!copied) { 2057 tipc_sk_set_orig_addr(m, skb); 2058 rc = tipc_sk_anc_data_recv(m, skb, tsk); 2059 if (rc) 2060 break; 2061 hdr = buf_msg(skb); 2062 } 2063 2064 /* Copy data if msg ok, otherwise return error/partial data */ 2065 if (likely(!err)) { 2066 offset = skb_cb->bytes_read; 2067 copy = min_t(int, dlen - offset, buflen - copied); 2068 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy); 2069 if (unlikely(rc)) 2070 break; 2071 copied += copy; 2072 offset += copy; 2073 if (unlikely(offset < dlen)) { 2074 if (!peek) 2075 skb_cb->bytes_read = offset; 2076 break; 2077 } 2078 } else { 2079 rc = 0; 2080 if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control) 2081 rc = -ECONNRESET; 2082 if (copied || rc) 2083 break; 2084 } 2085 2086 if (unlikely(peek)) 2087 break; 2088 2089 tsk_advance_rx_queue(sk); 2090 2091 /* Send connection flow control advertisement when applicable */ 2092 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen); 2093 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE) 2094 tipc_sk_send_ack(tsk); 2095 2096 /* Exit if all requested data or FIN/error received */ 2097 if (copied == buflen || err) 2098 break; 2099 2100 } while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required); 2101 exit: 2102 release_sock(sk); 2103 return copied ? copied : rc; 2104 } 2105 2106 /** 2107 * tipc_write_space - wake up thread if port congestion is released 2108 * @sk: socket 2109 */ 2110 static void tipc_write_space(struct sock *sk) 2111 { 2112 struct socket_wq *wq; 2113 2114 rcu_read_lock(); 2115 wq = rcu_dereference(sk->sk_wq); 2116 if (skwq_has_sleeper(wq)) 2117 wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT | 2118 EPOLLWRNORM | EPOLLWRBAND); 2119 rcu_read_unlock(); 2120 } 2121 2122 /** 2123 * tipc_data_ready - wake up threads to indicate messages have been received 2124 * @sk: socket 2125 */ 2126 static void tipc_data_ready(struct sock *sk) 2127 { 2128 struct socket_wq *wq; 2129 2130 trace_sk_data_ready(sk); 2131 2132 rcu_read_lock(); 2133 wq = rcu_dereference(sk->sk_wq); 2134 if (skwq_has_sleeper(wq)) 2135 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN | 2136 EPOLLRDNORM | EPOLLRDBAND); 2137 rcu_read_unlock(); 2138 } 2139 2140 static void tipc_sock_destruct(struct sock *sk) 2141 { 2142 __skb_queue_purge(&sk->sk_receive_queue); 2143 } 2144 2145 static void tipc_sk_proto_rcv(struct sock *sk, 2146 struct sk_buff_head *inputq, 2147 struct sk_buff_head *xmitq) 2148 { 2149 struct sk_buff *skb = __skb_dequeue(inputq); 2150 struct tipc_sock *tsk = tipc_sk(sk); 2151 struct tipc_msg *hdr = buf_msg(skb); 2152 struct tipc_group *grp = tsk->group; 2153 bool wakeup = false; 2154 2155 switch (msg_user(hdr)) { 2156 case CONN_MANAGER: 2157 tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq); 2158 return; 2159 case SOCK_WAKEUP: 2160 tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0); 2161 /* coupled with smp_rmb() in tipc_wait_for_cond() */ 2162 smp_wmb(); 2163 tsk->cong_link_cnt--; 2164 wakeup = true; 2165 tipc_sk_push_backlog(tsk, false); 2166 break; 2167 case GROUP_PROTOCOL: 2168 tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq); 2169 break; 2170 case TOP_SRV: 2171 tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf, 2172 hdr, inputq, xmitq); 2173 break; 2174 default: 2175 break; 2176 } 2177 2178 if (wakeup) 2179 sk->sk_write_space(sk); 2180 2181 kfree_skb(skb); 2182 } 2183 2184 /** 2185 * tipc_sk_filter_connect - check incoming message for a connection-based socket 2186 * @tsk: TIPC socket 2187 * @skb: pointer to message buffer. 2188 * @xmitq: for Nagle ACK if any 2189 * Return: true if message should be added to receive queue, false otherwise 2190 */ 2191 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb, 2192 struct sk_buff_head *xmitq) 2193 { 2194 struct sock *sk = &tsk->sk; 2195 struct net *net = sock_net(sk); 2196 struct tipc_msg *hdr = buf_msg(skb); 2197 bool con_msg = msg_connected(hdr); 2198 u32 pport = tsk_peer_port(tsk); 2199 u32 pnode = tsk_peer_node(tsk); 2200 u32 oport = msg_origport(hdr); 2201 u32 onode = msg_orignode(hdr); 2202 int err = msg_errcode(hdr); 2203 unsigned long delay; 2204 2205 if (unlikely(msg_mcast(hdr))) 2206 return false; 2207 tsk->oneway = 0; 2208 2209 switch (sk->sk_state) { 2210 case TIPC_CONNECTING: 2211 /* Setup ACK */ 2212 if (likely(con_msg)) { 2213 if (err) 2214 break; 2215 tipc_sk_finish_conn(tsk, oport, onode); 2216 msg_set_importance(&tsk->phdr, msg_importance(hdr)); 2217 /* ACK+ message with data is added to receive queue */ 2218 if (msg_data_sz(hdr)) 2219 return true; 2220 /* Empty ACK-, - wake up sleeping connect() and drop */ 2221 sk->sk_state_change(sk); 2222 msg_set_dest_droppable(hdr, 1); 2223 return false; 2224 } 2225 /* Ignore connectionless message if not from listening socket */ 2226 if (oport != pport || onode != pnode) 2227 return false; 2228 2229 /* Rejected SYN */ 2230 if (err != TIPC_ERR_OVERLOAD) 2231 break; 2232 2233 /* Prepare for new setup attempt if we have a SYN clone */ 2234 if (skb_queue_empty(&sk->sk_write_queue)) 2235 break; 2236 get_random_bytes(&delay, 2); 2237 if (tsk->conn_timeout < 4) 2238 tsk->conn_timeout = 4; 2239 delay %= (tsk->conn_timeout / 4); 2240 delay = msecs_to_jiffies(delay + 100); 2241 sk_reset_timer(sk, &sk->sk_timer, jiffies + delay); 2242 return false; 2243 case TIPC_OPEN: 2244 case TIPC_DISCONNECTING: 2245 return false; 2246 case TIPC_LISTEN: 2247 /* Accept only SYN message */ 2248 if (!msg_is_syn(hdr) && 2249 tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT) 2250 return false; 2251 if (!con_msg && !err) 2252 return true; 2253 return false; 2254 case TIPC_ESTABLISHED: 2255 if (!skb_queue_empty(&sk->sk_write_queue)) 2256 tipc_sk_push_backlog(tsk, false); 2257 /* Accept only connection-based messages sent by peer */ 2258 if (likely(con_msg && !err && pport == oport && 2259 pnode == onode)) { 2260 if (msg_ack_required(hdr)) { 2261 struct sk_buff *skb; 2262 2263 skb = tipc_sk_build_ack(tsk); 2264 if (skb) { 2265 msg_set_nagle_ack(buf_msg(skb)); 2266 __skb_queue_tail(xmitq, skb); 2267 } 2268 } 2269 return true; 2270 } 2271 if (!tsk_peer_msg(tsk, hdr)) 2272 return false; 2273 if (!err) 2274 return true; 2275 tipc_set_sk_state(sk, TIPC_DISCONNECTING); 2276 tipc_node_remove_conn(net, pnode, tsk->portid); 2277 sk->sk_state_change(sk); 2278 return true; 2279 default: 2280 pr_err("Unknown sk_state %u\n", sk->sk_state); 2281 } 2282 /* Abort connection setup attempt */ 2283 tipc_set_sk_state(sk, TIPC_DISCONNECTING); 2284 sk->sk_err = ECONNREFUSED; 2285 sk->sk_state_change(sk); 2286 return true; 2287 } 2288 2289 /** 2290 * rcvbuf_limit - get proper overload limit of socket receive queue 2291 * @sk: socket 2292 * @skb: message 2293 * 2294 * For connection oriented messages, irrespective of importance, 2295 * default queue limit is 2 MB. 2296 * 2297 * For connectionless messages, queue limits are based on message 2298 * importance as follows: 2299 * 2300 * TIPC_LOW_IMPORTANCE (2 MB) 2301 * TIPC_MEDIUM_IMPORTANCE (4 MB) 2302 * TIPC_HIGH_IMPORTANCE (8 MB) 2303 * TIPC_CRITICAL_IMPORTANCE (16 MB) 2304 * 2305 * Return: overload limit according to corresponding message importance 2306 */ 2307 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb) 2308 { 2309 struct tipc_sock *tsk = tipc_sk(sk); 2310 struct tipc_msg *hdr = buf_msg(skb); 2311 2312 if (unlikely(msg_in_group(hdr))) 2313 return READ_ONCE(sk->sk_rcvbuf); 2314 2315 if (unlikely(!msg_connected(hdr))) 2316 return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr); 2317 2318 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL)) 2319 return READ_ONCE(sk->sk_rcvbuf); 2320 2321 return FLOWCTL_MSG_LIM; 2322 } 2323 2324 /** 2325 * tipc_sk_filter_rcv - validate incoming message 2326 * @sk: socket 2327 * @skb: pointer to message. 2328 * @xmitq: output message area (FIXME) 2329 * 2330 * Enqueues message on receive queue if acceptable; optionally handles 2331 * disconnect indication for a connected socket. 2332 * 2333 * Called with socket lock already taken 2334 */ 2335 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb, 2336 struct sk_buff_head *xmitq) 2337 { 2338 bool sk_conn = !tipc_sk_type_connectionless(sk); 2339 struct tipc_sock *tsk = tipc_sk(sk); 2340 struct tipc_group *grp = tsk->group; 2341 struct tipc_msg *hdr = buf_msg(skb); 2342 struct net *net = sock_net(sk); 2343 struct sk_buff_head inputq; 2344 int mtyp = msg_type(hdr); 2345 int limit, err = TIPC_OK; 2346 2347 trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " "); 2348 TIPC_SKB_CB(skb)->bytes_read = 0; 2349 __skb_queue_head_init(&inputq); 2350 __skb_queue_tail(&inputq, skb); 2351 2352 if (unlikely(!msg_isdata(hdr))) 2353 tipc_sk_proto_rcv(sk, &inputq, xmitq); 2354 2355 if (unlikely(grp)) 2356 tipc_group_filter_msg(grp, &inputq, xmitq); 2357 2358 if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG) 2359 tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq); 2360 2361 /* Validate and add to receive buffer if there is space */ 2362 while ((skb = __skb_dequeue(&inputq))) { 2363 hdr = buf_msg(skb); 2364 limit = rcvbuf_limit(sk, skb); 2365 if ((sk_conn && !tipc_sk_filter_connect(tsk, skb, xmitq)) || 2366 (!sk_conn && msg_connected(hdr)) || 2367 (!grp && msg_in_group(hdr))) 2368 err = TIPC_ERR_NO_PORT; 2369 else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) { 2370 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, 2371 "err_overload2!"); 2372 sk_drops_inc(sk); 2373 err = TIPC_ERR_OVERLOAD; 2374 } 2375 2376 if (unlikely(err)) { 2377 if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) { 2378 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, 2379 "@filter_rcv!"); 2380 __skb_queue_tail(xmitq, skb); 2381 } 2382 err = TIPC_OK; 2383 continue; 2384 } 2385 __skb_queue_tail(&sk->sk_receive_queue, skb); 2386 skb_set_owner_r(skb, sk); 2387 trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL, 2388 "rcvq >90% allocated!"); 2389 sk->sk_data_ready(sk); 2390 } 2391 } 2392 2393 /** 2394 * tipc_sk_backlog_rcv - handle incoming message from backlog queue 2395 * @sk: socket 2396 * @skb: message 2397 * 2398 * Caller must hold socket lock 2399 */ 2400 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb) 2401 { 2402 unsigned int before = sk_rmem_alloc_get(sk); 2403 struct sk_buff_head xmitq; 2404 unsigned int added; 2405 2406 __skb_queue_head_init(&xmitq); 2407 2408 tipc_sk_filter_rcv(sk, skb, &xmitq); 2409 added = sk_rmem_alloc_get(sk) - before; 2410 atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt); 2411 2412 /* Send pending response/rejected messages, if any */ 2413 tipc_node_distr_xmit(sock_net(sk), &xmitq); 2414 return 0; 2415 } 2416 2417 /** 2418 * tipc_sk_enqueue - extract all buffers with destination 'dport' from 2419 * inputq and try adding them to socket or backlog queue 2420 * @inputq: list of incoming buffers with potentially different destinations 2421 * @sk: socket where the buffers should be enqueued 2422 * @dport: port number for the socket 2423 * @xmitq: output queue 2424 * 2425 * Caller must hold socket lock 2426 */ 2427 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk, 2428 u32 dport, struct sk_buff_head *xmitq) 2429 { 2430 unsigned long time_limit = jiffies + usecs_to_jiffies(20000); 2431 struct sk_buff *skb; 2432 unsigned int lim; 2433 atomic_t *dcnt; 2434 u32 onode; 2435 2436 while (skb_queue_len(inputq)) { 2437 if (unlikely(time_after_eq(jiffies, time_limit))) 2438 return; 2439 2440 skb = tipc_skb_dequeue(inputq, dport); 2441 if (unlikely(!skb)) 2442 return; 2443 2444 /* Add message directly to receive queue if possible */ 2445 if (!sock_owned_by_user(sk)) { 2446 tipc_sk_filter_rcv(sk, skb, xmitq); 2447 continue; 2448 } 2449 2450 /* Try backlog, compensating for double-counted bytes */ 2451 dcnt = &tipc_sk(sk)->dupl_rcvcnt; 2452 if (!sk->sk_backlog.len) 2453 atomic_set(dcnt, 0); 2454 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt); 2455 if (likely(!sk_add_backlog(sk, skb, lim))) { 2456 trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL, 2457 "bklg & rcvq >90% allocated!"); 2458 continue; 2459 } 2460 2461 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!"); 2462 /* Overload => reject message back to sender */ 2463 onode = tipc_own_addr(sock_net(sk)); 2464 sk_drops_inc(sk); 2465 if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) { 2466 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL, 2467 "@sk_enqueue!"); 2468 __skb_queue_tail(xmitq, skb); 2469 } 2470 break; 2471 } 2472 } 2473 2474 /** 2475 * tipc_sk_rcv - handle a chain of incoming buffers 2476 * @net: the associated network namespace 2477 * @inputq: buffer list containing the buffers 2478 * Consumes all buffers in list until inputq is empty 2479 * Note: may be called in multiple threads referring to the same queue 2480 */ 2481 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq) 2482 { 2483 struct sk_buff_head xmitq; 2484 u32 dnode, dport = 0; 2485 int err; 2486 struct tipc_sock *tsk; 2487 struct sock *sk; 2488 struct sk_buff *skb; 2489 2490 __skb_queue_head_init(&xmitq); 2491 while (skb_queue_len(inputq)) { 2492 dport = tipc_skb_peek_port(inputq, dport); 2493 tsk = tipc_sk_lookup(net, dport); 2494 2495 if (likely(tsk)) { 2496 sk = &tsk->sk; 2497 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) { 2498 tipc_sk_enqueue(inputq, sk, dport, &xmitq); 2499 spin_unlock_bh(&sk->sk_lock.slock); 2500 } 2501 /* Send pending response/rejected messages, if any */ 2502 tipc_node_distr_xmit(sock_net(sk), &xmitq); 2503 sock_put(sk); 2504 continue; 2505 } 2506 /* No destination socket => dequeue skb if still there */ 2507 skb = tipc_skb_dequeue(inputq, dport); 2508 if (!skb) 2509 return; 2510 2511 /* Try secondary lookup if unresolved named message */ 2512 err = TIPC_ERR_NO_PORT; 2513 if (tipc_msg_lookup_dest(net, skb, &err)) 2514 goto xmit; 2515 2516 /* Prepare for message rejection */ 2517 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err)) 2518 continue; 2519 2520 trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!"); 2521 xmit: 2522 dnode = msg_destnode(buf_msg(skb)); 2523 tipc_node_xmit_skb(net, skb, dnode, dport); 2524 } 2525 } 2526 2527 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p) 2528 { 2529 DEFINE_WAIT_FUNC(wait, woken_wake_function); 2530 struct sock *sk = sock->sk; 2531 int done; 2532 2533 do { 2534 int err = sock_error(sk); 2535 if (err) 2536 return err; 2537 if (!*timeo_p) 2538 return -ETIMEDOUT; 2539 if (signal_pending(current)) 2540 return sock_intr_errno(*timeo_p); 2541 if (sk->sk_state == TIPC_DISCONNECTING) 2542 break; 2543 2544 add_wait_queue(sk_sleep(sk), &wait); 2545 done = sk_wait_event(sk, timeo_p, tipc_sk_connected(sk), 2546 &wait); 2547 remove_wait_queue(sk_sleep(sk), &wait); 2548 } while (!done); 2549 return 0; 2550 } 2551 2552 static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr) 2553 { 2554 if (addr->family != AF_TIPC) 2555 return false; 2556 if (addr->addrtype == TIPC_SERVICE_RANGE) 2557 return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper); 2558 return (addr->addrtype == TIPC_SERVICE_ADDR || 2559 addr->addrtype == TIPC_SOCKET_ADDR); 2560 } 2561 2562 /** 2563 * tipc_connect - establish a connection to another TIPC port 2564 * @sock: socket structure 2565 * @dest: socket address for destination port 2566 * @destlen: size of socket address data structure 2567 * @flags: file-related flags associated with socket 2568 * 2569 * Return: 0 on success, errno otherwise 2570 */ 2571 static int tipc_connect(struct socket *sock, struct sockaddr_unsized *dest, 2572 int destlen, int flags) 2573 { 2574 struct sock *sk = sock->sk; 2575 struct tipc_sock *tsk = tipc_sk(sk); 2576 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest; 2577 struct msghdr m = {NULL,}; 2578 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout; 2579 int previous; 2580 int res = 0; 2581 2582 if (destlen != sizeof(struct sockaddr_tipc)) 2583 return -EINVAL; 2584 2585 lock_sock(sk); 2586 2587 if (tsk->group) { 2588 res = -EINVAL; 2589 goto exit; 2590 } 2591 2592 if (dst->family == AF_UNSPEC) { 2593 memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc)); 2594 if (!tipc_sk_type_connectionless(sk)) 2595 res = -EINVAL; 2596 goto exit; 2597 } 2598 if (!tipc_sockaddr_is_sane(dst)) { 2599 res = -EINVAL; 2600 goto exit; 2601 } 2602 /* DGRAM/RDM connect(), just save the destaddr */ 2603 if (tipc_sk_type_connectionless(sk)) { 2604 memcpy(&tsk->peer, dest, destlen); 2605 goto exit; 2606 } else if (dst->addrtype == TIPC_SERVICE_RANGE) { 2607 res = -EINVAL; 2608 goto exit; 2609 } 2610 2611 previous = sk->sk_state; 2612 2613 switch (sk->sk_state) { 2614 case TIPC_OPEN: 2615 /* Send a 'SYN-' to destination */ 2616 m.msg_name = dest; 2617 m.msg_namelen = destlen; 2618 iov_iter_kvec(&m.msg_iter, ITER_SOURCE, NULL, 0, 0); 2619 2620 /* If connect is in non-blocking case, set MSG_DONTWAIT to 2621 * indicate send_msg() is never blocked. 2622 */ 2623 if (!timeout) 2624 m.msg_flags = MSG_DONTWAIT; 2625 2626 res = __tipc_sendmsg(sock, &m, 0); 2627 if ((res < 0) && (res != -EWOULDBLOCK)) 2628 goto exit; 2629 2630 /* Just entered TIPC_CONNECTING state; the only 2631 * difference is that return value in non-blocking 2632 * case is EINPROGRESS, rather than EALREADY. 2633 */ 2634 res = -EINPROGRESS; 2635 fallthrough; 2636 case TIPC_CONNECTING: 2637 if (!timeout) { 2638 if (previous == TIPC_CONNECTING) 2639 res = -EALREADY; 2640 goto exit; 2641 } 2642 timeout = msecs_to_jiffies(timeout); 2643 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */ 2644 res = tipc_wait_for_connect(sock, &timeout); 2645 break; 2646 case TIPC_ESTABLISHED: 2647 res = -EISCONN; 2648 break; 2649 default: 2650 res = -EINVAL; 2651 } 2652 2653 exit: 2654 release_sock(sk); 2655 return res; 2656 } 2657 2658 /** 2659 * tipc_listen - allow socket to listen for incoming connections 2660 * @sock: socket structure 2661 * @len: (unused) 2662 * 2663 * Return: 0 on success, errno otherwise 2664 */ 2665 static int tipc_listen(struct socket *sock, int len) 2666 { 2667 struct sock *sk = sock->sk; 2668 int res; 2669 2670 lock_sock(sk); 2671 res = tipc_set_sk_state(sk, TIPC_LISTEN); 2672 release_sock(sk); 2673 2674 return res; 2675 } 2676 2677 static int tipc_wait_for_accept(struct socket *sock, long timeo) 2678 { 2679 struct sock *sk = sock->sk; 2680 DEFINE_WAIT_FUNC(wait, woken_wake_function); 2681 int err; 2682 2683 /* True wake-one mechanism for incoming connections: only 2684 * one process gets woken up, not the 'whole herd'. 2685 * Since we do not 'race & poll' for established sockets 2686 * anymore, the common case will execute the loop only once. 2687 */ 2688 for (;;) { 2689 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) { 2690 add_wait_queue(sk_sleep(sk), &wait); 2691 release_sock(sk); 2692 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo); 2693 lock_sock(sk); 2694 remove_wait_queue(sk_sleep(sk), &wait); 2695 } 2696 err = 0; 2697 if (!skb_queue_empty(&sk->sk_receive_queue)) 2698 break; 2699 err = -EAGAIN; 2700 if (!timeo) 2701 break; 2702 err = sock_intr_errno(timeo); 2703 if (signal_pending(current)) 2704 break; 2705 } 2706 return err; 2707 } 2708 2709 /** 2710 * tipc_accept - wait for connection request 2711 * @sock: listening socket 2712 * @new_sock: new socket that is to be connected 2713 * @arg: arguments for accept 2714 * 2715 * Return: 0 on success, errno otherwise 2716 */ 2717 static int tipc_accept(struct socket *sock, struct socket *new_sock, 2718 struct proto_accept_arg *arg) 2719 { 2720 struct sock *new_sk, *sk = sock->sk; 2721 struct tipc_sock *new_tsock; 2722 struct msghdr m = {NULL,}; 2723 struct tipc_msg *msg; 2724 struct sk_buff *buf; 2725 long timeo; 2726 int res; 2727 2728 lock_sock(sk); 2729 2730 if (sk->sk_state != TIPC_LISTEN) { 2731 res = -EINVAL; 2732 goto exit; 2733 } 2734 timeo = sock_rcvtimeo(sk, arg->flags & O_NONBLOCK); 2735 res = tipc_wait_for_accept(sock, timeo); 2736 if (res) 2737 goto exit; 2738 2739 buf = skb_peek(&sk->sk_receive_queue); 2740 2741 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, arg->kern); 2742 if (res) 2743 goto exit; 2744 security_sk_clone(sock->sk, new_sock->sk); 2745 2746 new_sk = new_sock->sk; 2747 new_tsock = tipc_sk(new_sk); 2748 msg = buf_msg(buf); 2749 2750 /* we lock on new_sk; but lockdep sees the lock on sk */ 2751 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING); 2752 2753 /* 2754 * Reject any stray messages received by new socket 2755 * before the socket lock was taken (very, very unlikely) 2756 */ 2757 tsk_rej_rx_queue(new_sk, TIPC_ERR_NO_PORT); 2758 2759 /* Connect new socket to it's peer */ 2760 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg)); 2761 2762 tsk_set_importance(new_sk, msg_importance(msg)); 2763 if (msg_named(msg)) { 2764 new_tsock->conn_addrtype = TIPC_SERVICE_ADDR; 2765 msg_set_nametype(&new_tsock->phdr, msg_nametype(msg)); 2766 msg_set_nameinst(&new_tsock->phdr, msg_nameinst(msg)); 2767 } 2768 2769 /* 2770 * Respond to 'SYN-' by discarding it & returning 'ACK'. 2771 * Respond to 'SYN+' by queuing it on new socket & returning 'ACK'. 2772 */ 2773 if (!msg_data_sz(msg)) { 2774 tsk_advance_rx_queue(sk); 2775 } else { 2776 __skb_dequeue(&sk->sk_receive_queue); 2777 __skb_queue_head(&new_sk->sk_receive_queue, buf); 2778 skb_set_owner_r(buf, new_sk); 2779 } 2780 iov_iter_kvec(&m.msg_iter, ITER_SOURCE, NULL, 0, 0); 2781 __tipc_sendstream(new_sock, &m, 0); 2782 release_sock(new_sk); 2783 exit: 2784 release_sock(sk); 2785 return res; 2786 } 2787 2788 /** 2789 * tipc_shutdown - shutdown socket connection 2790 * @sock: socket structure 2791 * @how: direction to close (must be SHUT_RDWR) 2792 * 2793 * Terminates connection (if necessary), then purges socket's receive queue. 2794 * 2795 * Return: 0 on success, errno otherwise 2796 */ 2797 static int tipc_shutdown(struct socket *sock, int how) 2798 { 2799 struct sock *sk = sock->sk; 2800 int res; 2801 2802 if (how != SHUT_RDWR) 2803 return -EINVAL; 2804 2805 lock_sock(sk); 2806 2807 trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " "); 2808 __tipc_shutdown(sock, TIPC_CONN_SHUTDOWN); 2809 sk->sk_shutdown = SHUTDOWN_MASK; 2810 2811 if (sk->sk_state == TIPC_DISCONNECTING) { 2812 /* Discard any unreceived messages */ 2813 __skb_queue_purge(&sk->sk_receive_queue); 2814 2815 res = 0; 2816 } else { 2817 res = -ENOTCONN; 2818 } 2819 /* Wake up anyone sleeping in poll. */ 2820 sk->sk_state_change(sk); 2821 2822 release_sock(sk); 2823 return res; 2824 } 2825 2826 static void tipc_sk_check_probing_state(struct sock *sk, 2827 struct sk_buff_head *list) 2828 { 2829 struct tipc_sock *tsk = tipc_sk(sk); 2830 u32 pnode = tsk_peer_node(tsk); 2831 u32 pport = tsk_peer_port(tsk); 2832 u32 self = tsk_own_node(tsk); 2833 u32 oport = tsk->portid; 2834 struct sk_buff *skb; 2835 2836 if (tsk->probe_unacked) { 2837 tipc_set_sk_state(sk, TIPC_DISCONNECTING); 2838 sk->sk_err = ECONNABORTED; 2839 tipc_node_remove_conn(sock_net(sk), pnode, pport); 2840 sk->sk_state_change(sk); 2841 return; 2842 } 2843 /* Prepare new probe */ 2844 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0, 2845 pnode, self, pport, oport, TIPC_OK); 2846 if (skb) 2847 __skb_queue_tail(list, skb); 2848 tsk->probe_unacked = true; 2849 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV); 2850 } 2851 2852 static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list) 2853 { 2854 struct tipc_sock *tsk = tipc_sk(sk); 2855 2856 /* Try again later if dest link is congested */ 2857 if (tsk->cong_link_cnt) { 2858 sk_reset_timer(sk, &sk->sk_timer, 2859 jiffies + msecs_to_jiffies(100)); 2860 return; 2861 } 2862 /* Prepare SYN for retransmit */ 2863 tipc_msg_skb_clone(&sk->sk_write_queue, list); 2864 } 2865 2866 static void tipc_sk_timeout(struct timer_list *t) 2867 { 2868 struct sock *sk = timer_container_of(sk, t, sk_timer); 2869 struct tipc_sock *tsk = tipc_sk(sk); 2870 u32 pnode = tsk_peer_node(tsk); 2871 struct sk_buff_head list; 2872 int rc = 0; 2873 2874 __skb_queue_head_init(&list); 2875 bh_lock_sock(sk); 2876 2877 /* Try again later if socket is busy */ 2878 if (sock_owned_by_user(sk)) { 2879 sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20); 2880 bh_unlock_sock(sk); 2881 sock_put(sk); 2882 return; 2883 } 2884 2885 if (sk->sk_state == TIPC_ESTABLISHED) 2886 tipc_sk_check_probing_state(sk, &list); 2887 else if (sk->sk_state == TIPC_CONNECTING) 2888 tipc_sk_retry_connect(sk, &list); 2889 2890 bh_unlock_sock(sk); 2891 2892 if (!skb_queue_empty(&list)) 2893 rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid); 2894 2895 /* SYN messages may cause link congestion */ 2896 if (rc == -ELINKCONG) { 2897 tipc_dest_push(&tsk->cong_links, pnode, 0); 2898 tsk->cong_link_cnt = 1; 2899 } 2900 sock_put(sk); 2901 } 2902 2903 static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua) 2904 { 2905 struct sock *sk = &tsk->sk; 2906 struct net *net = sock_net(sk); 2907 struct tipc_socket_addr skaddr; 2908 struct publication *p; 2909 u32 key; 2910 2911 if (tipc_sk_connected(sk)) 2912 return -EINVAL; 2913 key = tsk->portid + tsk->pub_count + 1; 2914 if (key == tsk->portid) 2915 return -EADDRINUSE; 2916 skaddr.ref = tsk->portid; 2917 skaddr.node = tipc_own_addr(net); 2918 p = tipc_nametbl_publish(net, ua, &skaddr, key); 2919 if (unlikely(!p)) 2920 return -EINVAL; 2921 2922 list_add(&p->binding_sock, &tsk->publications); 2923 tsk->pub_count++; 2924 tsk->published = true; 2925 return 0; 2926 } 2927 2928 static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua) 2929 { 2930 struct net *net = sock_net(&tsk->sk); 2931 struct publication *safe, *p; 2932 struct tipc_uaddr _ua; 2933 int rc = -EINVAL; 2934 2935 list_for_each_entry_safe(p, safe, &tsk->publications, binding_sock) { 2936 if (!ua) { 2937 tipc_uaddr(&_ua, TIPC_SERVICE_RANGE, p->scope, 2938 p->sr.type, p->sr.lower, p->sr.upper); 2939 tipc_nametbl_withdraw(net, &_ua, &p->sk, p->key); 2940 continue; 2941 } 2942 /* Unbind specific publication */ 2943 if (p->scope != ua->scope) 2944 continue; 2945 if (p->sr.type != ua->sr.type) 2946 continue; 2947 if (p->sr.lower != ua->sr.lower) 2948 continue; 2949 if (p->sr.upper != ua->sr.upper) 2950 break; 2951 tipc_nametbl_withdraw(net, ua, &p->sk, p->key); 2952 rc = 0; 2953 break; 2954 } 2955 if (list_empty(&tsk->publications)) { 2956 tsk->published = 0; 2957 rc = 0; 2958 } 2959 return rc; 2960 } 2961 2962 /* tipc_sk_reinit: set non-zero address in all existing sockets 2963 * when we go from standalone to network mode. 2964 */ 2965 void tipc_sk_reinit(struct net *net) 2966 { 2967 struct tipc_net *tn = net_generic(net, tipc_net_id); 2968 struct rhashtable_iter iter; 2969 struct tipc_sock *tsk; 2970 struct tipc_msg *msg; 2971 2972 rhashtable_walk_enter(&tn->sk_rht, &iter); 2973 2974 do { 2975 rhashtable_walk_start(&iter); 2976 2977 while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) { 2978 sock_hold(&tsk->sk); 2979 rhashtable_walk_stop(&iter); 2980 lock_sock(&tsk->sk); 2981 msg = &tsk->phdr; 2982 msg_set_prevnode(msg, tipc_own_addr(net)); 2983 msg_set_orignode(msg, tipc_own_addr(net)); 2984 release_sock(&tsk->sk); 2985 rhashtable_walk_start(&iter); 2986 sock_put(&tsk->sk); 2987 } 2988 2989 rhashtable_walk_stop(&iter); 2990 } while (tsk == ERR_PTR(-EAGAIN)); 2991 2992 rhashtable_walk_exit(&iter); 2993 } 2994 2995 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid) 2996 { 2997 struct tipc_net *tn = net_generic(net, tipc_net_id); 2998 struct tipc_sock *tsk; 2999 3000 rcu_read_lock(); 3001 tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params); 3002 if (tsk) 3003 sock_hold(&tsk->sk); 3004 rcu_read_unlock(); 3005 3006 return tsk; 3007 } 3008 3009 static int tipc_sk_insert(struct tipc_sock *tsk) 3010 { 3011 struct sock *sk = &tsk->sk; 3012 struct net *net = sock_net(sk); 3013 struct tipc_net *tn = net_generic(net, tipc_net_id); 3014 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1; 3015 u32 portid = get_random_u32_below(remaining) + TIPC_MIN_PORT; 3016 3017 while (remaining--) { 3018 portid++; 3019 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT)) 3020 portid = TIPC_MIN_PORT; 3021 tsk->portid = portid; 3022 sock_hold(&tsk->sk); 3023 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node, 3024 tsk_rht_params)) 3025 return 0; 3026 sock_put(&tsk->sk); 3027 } 3028 3029 return -1; 3030 } 3031 3032 static void tipc_sk_remove(struct tipc_sock *tsk) 3033 { 3034 struct sock *sk = &tsk->sk; 3035 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id); 3036 3037 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) 3038 __sock_put(sk); 3039 } 3040 3041 static const struct rhashtable_params tsk_rht_params = { 3042 .nelem_hint = 192, 3043 .head_offset = offsetof(struct tipc_sock, node), 3044 .key_offset = offsetof(struct tipc_sock, portid), 3045 .key_len = sizeof(u32), /* portid */ 3046 .max_size = 1048576, 3047 .min_size = 256, 3048 .automatic_shrinking = true, 3049 }; 3050 3051 int tipc_sk_rht_init(struct net *net) 3052 { 3053 struct tipc_net *tn = net_generic(net, tipc_net_id); 3054 3055 return rhashtable_init(&tn->sk_rht, &tsk_rht_params); 3056 } 3057 3058 void tipc_sk_rht_destroy(struct net *net) 3059 { 3060 struct tipc_net *tn = net_generic(net, tipc_net_id); 3061 3062 /* Wait for socket readers to complete */ 3063 synchronize_net(); 3064 3065 rhashtable_destroy(&tn->sk_rht); 3066 } 3067 3068 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq) 3069 { 3070 struct net *net = sock_net(&tsk->sk); 3071 struct tipc_group *grp = tsk->group; 3072 struct tipc_msg *hdr = &tsk->phdr; 3073 struct tipc_uaddr ua; 3074 int rc; 3075 3076 if (mreq->type < TIPC_RESERVED_TYPES) 3077 return -EACCES; 3078 if (mreq->scope > TIPC_NODE_SCOPE) 3079 return -EINVAL; 3080 if (mreq->scope != TIPC_NODE_SCOPE) 3081 mreq->scope = TIPC_CLUSTER_SCOPE; 3082 if (grp) 3083 return -EACCES; 3084 grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open); 3085 if (!grp) 3086 return -ENOMEM; 3087 tsk->group = grp; 3088 msg_set_lookup_scope(hdr, mreq->scope); 3089 msg_set_nametype(hdr, mreq->type); 3090 msg_set_dest_droppable(hdr, true); 3091 tipc_uaddr(&ua, TIPC_SERVICE_RANGE, mreq->scope, 3092 mreq->type, mreq->instance, mreq->instance); 3093 tipc_nametbl_build_group(net, grp, &ua); 3094 rc = tipc_sk_publish(tsk, &ua); 3095 if (rc) { 3096 tipc_group_delete(net, grp); 3097 tsk->group = NULL; 3098 return rc; 3099 } 3100 /* Eliminate any risk that a broadcast overtakes sent JOINs */ 3101 tsk->mc_method.rcast = true; 3102 tsk->mc_method.mandatory = true; 3103 tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf); 3104 return rc; 3105 } 3106 3107 static int tipc_sk_leave(struct tipc_sock *tsk) 3108 { 3109 struct net *net = sock_net(&tsk->sk); 3110 struct tipc_group *grp = tsk->group; 3111 struct tipc_uaddr ua; 3112 int scope; 3113 3114 if (!grp) 3115 return -EINVAL; 3116 ua.addrtype = TIPC_SERVICE_RANGE; 3117 tipc_group_self(grp, &ua.sr, &scope); 3118 ua.scope = scope; 3119 tipc_group_delete(net, grp); 3120 tsk->group = NULL; 3121 tipc_sk_withdraw(tsk, &ua); 3122 return 0; 3123 } 3124 3125 /** 3126 * tipc_setsockopt - set socket option 3127 * @sock: socket structure 3128 * @lvl: option level 3129 * @opt: option identifier 3130 * @ov: pointer to new option value 3131 * @ol: length of option value 3132 * 3133 * For stream sockets only, accepts and ignores all IPPROTO_TCP options 3134 * (to ease compatibility). 3135 * 3136 * Return: 0 on success, errno otherwise 3137 */ 3138 static int tipc_setsockopt(struct socket *sock, int lvl, int opt, 3139 sockptr_t ov, unsigned int ol) 3140 { 3141 struct sock *sk = sock->sk; 3142 struct tipc_sock *tsk = tipc_sk(sk); 3143 struct tipc_group_req mreq; 3144 u32 value = 0; 3145 int res = 0; 3146 3147 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM)) 3148 return 0; 3149 if (lvl != SOL_TIPC) 3150 return -ENOPROTOOPT; 3151 3152 switch (opt) { 3153 case TIPC_IMPORTANCE: 3154 case TIPC_SRC_DROPPABLE: 3155 case TIPC_DEST_DROPPABLE: 3156 case TIPC_CONN_TIMEOUT: 3157 case TIPC_NODELAY: 3158 if (ol < sizeof(value)) 3159 return -EINVAL; 3160 if (copy_from_sockptr(&value, ov, sizeof(u32))) 3161 return -EFAULT; 3162 break; 3163 case TIPC_GROUP_JOIN: 3164 if (ol < sizeof(mreq)) 3165 return -EINVAL; 3166 if (copy_from_sockptr(&mreq, ov, sizeof(mreq))) 3167 return -EFAULT; 3168 break; 3169 default: 3170 if (!sockptr_is_null(ov) || ol) 3171 return -EINVAL; 3172 } 3173 3174 lock_sock(sk); 3175 3176 switch (opt) { 3177 case TIPC_IMPORTANCE: 3178 res = tsk_set_importance(sk, value); 3179 break; 3180 case TIPC_SRC_DROPPABLE: 3181 if (sock->type != SOCK_STREAM) 3182 tsk_set_unreliable(tsk, value); 3183 else 3184 res = -ENOPROTOOPT; 3185 break; 3186 case TIPC_DEST_DROPPABLE: 3187 tsk_set_unreturnable(tsk, value); 3188 break; 3189 case TIPC_CONN_TIMEOUT: 3190 tipc_sk(sk)->conn_timeout = value; 3191 break; 3192 case TIPC_MCAST_BROADCAST: 3193 tsk->mc_method.rcast = false; 3194 tsk->mc_method.mandatory = true; 3195 break; 3196 case TIPC_MCAST_REPLICAST: 3197 tsk->mc_method.rcast = true; 3198 tsk->mc_method.mandatory = true; 3199 break; 3200 case TIPC_GROUP_JOIN: 3201 res = tipc_sk_join(tsk, &mreq); 3202 break; 3203 case TIPC_GROUP_LEAVE: 3204 res = tipc_sk_leave(tsk); 3205 break; 3206 case TIPC_NODELAY: 3207 tsk->nodelay = !!value; 3208 tsk_set_nagle(tsk); 3209 break; 3210 default: 3211 res = -EINVAL; 3212 } 3213 3214 release_sock(sk); 3215 3216 return res; 3217 } 3218 3219 /** 3220 * tipc_getsockopt - get socket option 3221 * @sock: socket structure 3222 * @lvl: option level 3223 * @opt: option identifier 3224 * @sopt: socket option container (input buffer length, output value/length) 3225 * 3226 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options 3227 * (to ease compatibility). 3228 * 3229 * Return: 0 on success, errno otherwise 3230 */ 3231 static int tipc_getsockopt(struct socket *sock, int lvl, int opt, 3232 sockopt_t *sopt) 3233 { 3234 struct sock *sk = sock->sk; 3235 struct tipc_sock *tsk = tipc_sk(sk); 3236 struct tipc_service_range seq; 3237 int len, scope; 3238 int res = 0; 3239 u32 value; 3240 3241 if (lvl == IPPROTO_TCP && sock->type == SOCK_STREAM) { 3242 sopt->optlen = 0; 3243 return 0; 3244 } 3245 if (lvl != SOL_TIPC) 3246 return -ENOPROTOOPT; 3247 len = sopt->optlen; 3248 3249 lock_sock(sk); 3250 3251 switch (opt) { 3252 case TIPC_IMPORTANCE: 3253 value = tsk_importance(tsk); 3254 break; 3255 case TIPC_SRC_DROPPABLE: 3256 value = tsk_unreliable(tsk); 3257 break; 3258 case TIPC_DEST_DROPPABLE: 3259 value = tsk_unreturnable(tsk); 3260 break; 3261 case TIPC_CONN_TIMEOUT: 3262 value = tsk->conn_timeout; 3263 break; 3264 case TIPC_NODE_RECVQ_DEPTH: 3265 value = 0; /* was tipc_queue_size, now obsolete */ 3266 break; 3267 case TIPC_SOCK_RECVQ_DEPTH: 3268 value = skb_queue_len(&sk->sk_receive_queue); 3269 break; 3270 case TIPC_SOCK_RECVQ_USED: 3271 value = sk_rmem_alloc_get(sk); 3272 break; 3273 case TIPC_GROUP_JOIN: 3274 seq.type = 0; 3275 if (tsk->group) 3276 tipc_group_self(tsk->group, &seq, &scope); 3277 value = seq.type; 3278 break; 3279 default: 3280 res = -EINVAL; 3281 } 3282 3283 release_sock(sk); 3284 3285 if (res) 3286 return res; 3287 3288 if (len < sizeof(value)) 3289 return -EINVAL; 3290 3291 if (copy_to_iter(&value, sizeof(value), &sopt->iter_out) != 3292 sizeof(value)) 3293 return -EFAULT; 3294 sopt->optlen = sizeof(value); 3295 3296 return 0; 3297 } 3298 3299 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 3300 { 3301 struct net *net = sock_net(sock->sk); 3302 struct tipc_sioc_nodeid_req nr = {0}; 3303 struct tipc_sioc_ln_req lnr; 3304 void __user *argp = (void __user *)arg; 3305 3306 switch (cmd) { 3307 case SIOCGETLINKNAME: 3308 if (copy_from_user(&lnr, argp, sizeof(lnr))) 3309 return -EFAULT; 3310 if (!tipc_node_get_linkname(net, 3311 lnr.bearer_id & 0xffff, lnr.peer, 3312 lnr.linkname, TIPC_MAX_LINK_NAME)) { 3313 if (copy_to_user(argp, &lnr, sizeof(lnr))) 3314 return -EFAULT; 3315 return 0; 3316 } 3317 return -EADDRNOTAVAIL; 3318 case SIOCGETNODEID: 3319 if (copy_from_user(&nr, argp, sizeof(nr))) 3320 return -EFAULT; 3321 if (!tipc_node_get_id(net, nr.peer, nr.node_id)) 3322 return -EADDRNOTAVAIL; 3323 if (copy_to_user(argp, &nr, sizeof(nr))) 3324 return -EFAULT; 3325 return 0; 3326 default: 3327 return -ENOIOCTLCMD; 3328 } 3329 } 3330 3331 static int tipc_socketpair(struct socket *sock1, struct socket *sock2) 3332 { 3333 struct tipc_sock *tsk2 = tipc_sk(sock2->sk); 3334 struct tipc_sock *tsk1 = tipc_sk(sock1->sk); 3335 u32 onode = tipc_own_addr(sock_net(sock1->sk)); 3336 3337 tsk1->peer.family = AF_TIPC; 3338 tsk1->peer.addrtype = TIPC_SOCKET_ADDR; 3339 tsk1->peer.scope = TIPC_NODE_SCOPE; 3340 tsk1->peer.addr.id.ref = tsk2->portid; 3341 tsk1->peer.addr.id.node = onode; 3342 tsk2->peer.family = AF_TIPC; 3343 tsk2->peer.addrtype = TIPC_SOCKET_ADDR; 3344 tsk2->peer.scope = TIPC_NODE_SCOPE; 3345 tsk2->peer.addr.id.ref = tsk1->portid; 3346 tsk2->peer.addr.id.node = onode; 3347 3348 tipc_sk_finish_conn(tsk1, tsk2->portid, onode); 3349 tipc_sk_finish_conn(tsk2, tsk1->portid, onode); 3350 return 0; 3351 } 3352 3353 /* Protocol switches for the various types of TIPC sockets */ 3354 3355 static const struct proto_ops msg_ops = { 3356 .owner = THIS_MODULE, 3357 .family = AF_TIPC, 3358 .release = tipc_release, 3359 .bind = tipc_bind, 3360 .connect = tipc_connect, 3361 .socketpair = tipc_socketpair, 3362 .accept = sock_no_accept, 3363 .getname = tipc_getname, 3364 .poll = tipc_poll, 3365 .ioctl = tipc_ioctl, 3366 .listen = sock_no_listen, 3367 .shutdown = tipc_shutdown, 3368 .setsockopt = tipc_setsockopt, 3369 .getsockopt_iter = tipc_getsockopt, 3370 .sendmsg = tipc_sendmsg, 3371 .recvmsg = tipc_recvmsg, 3372 .mmap = sock_no_mmap, 3373 }; 3374 3375 static const struct proto_ops packet_ops = { 3376 .owner = THIS_MODULE, 3377 .family = AF_TIPC, 3378 .release = tipc_release, 3379 .bind = tipc_bind, 3380 .connect = tipc_connect, 3381 .socketpair = tipc_socketpair, 3382 .accept = tipc_accept, 3383 .getname = tipc_getname, 3384 .poll = tipc_poll, 3385 .ioctl = tipc_ioctl, 3386 .listen = tipc_listen, 3387 .shutdown = tipc_shutdown, 3388 .setsockopt = tipc_setsockopt, 3389 .getsockopt_iter = tipc_getsockopt, 3390 .sendmsg = tipc_send_packet, 3391 .recvmsg = tipc_recvmsg, 3392 .mmap = sock_no_mmap, 3393 }; 3394 3395 static const struct proto_ops stream_ops = { 3396 .owner = THIS_MODULE, 3397 .family = AF_TIPC, 3398 .release = tipc_release, 3399 .bind = tipc_bind, 3400 .connect = tipc_connect, 3401 .socketpair = tipc_socketpair, 3402 .accept = tipc_accept, 3403 .getname = tipc_getname, 3404 .poll = tipc_poll, 3405 .ioctl = tipc_ioctl, 3406 .listen = tipc_listen, 3407 .shutdown = tipc_shutdown, 3408 .setsockopt = tipc_setsockopt, 3409 .getsockopt_iter = tipc_getsockopt, 3410 .sendmsg = tipc_sendstream, 3411 .recvmsg = tipc_recvstream, 3412 .mmap = sock_no_mmap, 3413 }; 3414 3415 static const struct net_proto_family tipc_family_ops = { 3416 .owner = THIS_MODULE, 3417 .family = AF_TIPC, 3418 .create = tipc_sk_create 3419 }; 3420 3421 static struct proto tipc_proto = { 3422 .name = "TIPC", 3423 .owner = THIS_MODULE, 3424 .obj_size = sizeof(struct tipc_sock), 3425 .sysctl_rmem = sysctl_tipc_rmem 3426 }; 3427 3428 /** 3429 * tipc_socket_init - initialize TIPC socket interface 3430 * 3431 * Return: 0 on success, errno otherwise 3432 */ 3433 int tipc_socket_init(void) 3434 { 3435 int res; 3436 3437 res = proto_register(&tipc_proto, 1); 3438 if (res) { 3439 pr_err("Failed to register TIPC protocol type\n"); 3440 goto out; 3441 } 3442 3443 res = sock_register(&tipc_family_ops); 3444 if (res) { 3445 pr_err("Failed to register TIPC socket type\n"); 3446 proto_unregister(&tipc_proto); 3447 goto out; 3448 } 3449 out: 3450 return res; 3451 } 3452 3453 /** 3454 * tipc_socket_stop - stop TIPC socket interface 3455 */ 3456 void tipc_socket_stop(void) 3457 { 3458 sock_unregister(tipc_family_ops.family); 3459 proto_unregister(&tipc_proto); 3460 } 3461 3462 /* Caller should hold socket lock for the passed tipc socket. */ 3463 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk) 3464 { 3465 u32 peer_node, peer_port; 3466 u32 conn_type, conn_instance; 3467 struct nlattr *nest; 3468 3469 peer_node = tsk_peer_node(tsk); 3470 peer_port = tsk_peer_port(tsk); 3471 conn_type = msg_nametype(&tsk->phdr); 3472 conn_instance = msg_nameinst(&tsk->phdr); 3473 nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON); 3474 if (!nest) 3475 return -EMSGSIZE; 3476 3477 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node)) 3478 goto msg_full; 3479 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port)) 3480 goto msg_full; 3481 3482 if (tsk->conn_addrtype != 0) { 3483 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG)) 3484 goto msg_full; 3485 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, conn_type)) 3486 goto msg_full; 3487 if (nla_put_u32(skb, TIPC_NLA_CON_INST, conn_instance)) 3488 goto msg_full; 3489 } 3490 nla_nest_end(skb, nest); 3491 3492 return 0; 3493 3494 msg_full: 3495 nla_nest_cancel(skb, nest); 3496 3497 return -EMSGSIZE; 3498 } 3499 3500 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock 3501 *tsk) 3502 { 3503 struct net *net = sock_net(skb->sk); 3504 struct sock *sk = &tsk->sk; 3505 3506 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) || 3507 nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net))) 3508 return -EMSGSIZE; 3509 3510 if (tipc_sk_connected(sk)) { 3511 if (__tipc_nl_add_sk_con(skb, tsk)) 3512 return -EMSGSIZE; 3513 } else if (!list_empty(&tsk->publications)) { 3514 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL)) 3515 return -EMSGSIZE; 3516 } 3517 return 0; 3518 } 3519 3520 /* Caller should hold socket lock for the passed tipc socket. */ 3521 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb, 3522 struct tipc_sock *tsk) 3523 { 3524 struct nlattr *attrs; 3525 void *hdr; 3526 3527 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq, 3528 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET); 3529 if (!hdr) 3530 goto msg_cancel; 3531 3532 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK); 3533 if (!attrs) 3534 goto genlmsg_cancel; 3535 3536 if (__tipc_nl_add_sk_info(skb, tsk)) 3537 goto attr_msg_cancel; 3538 3539 nla_nest_end(skb, attrs); 3540 genlmsg_end(skb, hdr); 3541 3542 return 0; 3543 3544 attr_msg_cancel: 3545 nla_nest_cancel(skb, attrs); 3546 genlmsg_cancel: 3547 genlmsg_cancel(skb, hdr); 3548 msg_cancel: 3549 return -EMSGSIZE; 3550 } 3551 3552 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb, 3553 int (*skb_handler)(struct sk_buff *skb, 3554 struct netlink_callback *cb, 3555 struct tipc_sock *tsk)) 3556 { 3557 struct rhashtable_iter *iter = (void *)cb->args[4]; 3558 struct tipc_sock *tsk; 3559 int err; 3560 3561 rhashtable_walk_start(iter); 3562 while ((tsk = rhashtable_walk_next(iter)) != NULL) { 3563 if (IS_ERR(tsk)) { 3564 if (PTR_ERR(tsk) == -EAGAIN) 3565 continue; 3566 break; 3567 } 3568 3569 sock_hold(&tsk->sk); 3570 rhashtable_walk_stop(iter); 3571 lock_sock(&tsk->sk); 3572 err = skb_handler(skb, cb, tsk); 3573 if (err) { 3574 release_sock(&tsk->sk); 3575 sock_put(&tsk->sk); 3576 goto out; 3577 } 3578 release_sock(&tsk->sk); 3579 rhashtable_walk_start(iter); 3580 sock_put(&tsk->sk); 3581 } 3582 rhashtable_walk_stop(iter); 3583 out: 3584 return skb->len; 3585 } 3586 EXPORT_SYMBOL(tipc_nl_sk_walk); 3587 3588 int tipc_dump_start(struct netlink_callback *cb) 3589 { 3590 return __tipc_dump_start(cb, sock_net(cb->skb->sk)); 3591 } 3592 EXPORT_SYMBOL(tipc_dump_start); 3593 3594 int __tipc_dump_start(struct netlink_callback *cb, struct net *net) 3595 { 3596 /* tipc_nl_name_table_dump() uses cb->args[0...3]. */ 3597 struct rhashtable_iter *iter = (void *)cb->args[4]; 3598 struct tipc_net *tn = tipc_net(net); 3599 3600 if (!iter) { 3601 iter = kmalloc_obj(*iter); 3602 if (!iter) 3603 return -ENOMEM; 3604 3605 cb->args[4] = (long)iter; 3606 } 3607 3608 rhashtable_walk_enter(&tn->sk_rht, iter); 3609 return 0; 3610 } 3611 3612 int tipc_dump_done(struct netlink_callback *cb) 3613 { 3614 struct rhashtable_iter *hti = (void *)cb->args[4]; 3615 3616 rhashtable_walk_exit(hti); 3617 kfree(hti); 3618 return 0; 3619 } 3620 EXPORT_SYMBOL(tipc_dump_done); 3621 3622 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb, 3623 struct tipc_sock *tsk, u32 sk_filter_state, 3624 u64 (*tipc_diag_gen_cookie)(struct sock *sk)) 3625 { 3626 struct sock *sk = &tsk->sk; 3627 struct nlattr *attrs; 3628 struct nlattr *stat; 3629 3630 /*filter response w.r.t sk_state*/ 3631 if (!(sk_filter_state & (1 << sk->sk_state))) 3632 return 0; 3633 3634 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK); 3635 if (!attrs) 3636 goto msg_cancel; 3637 3638 if (__tipc_nl_add_sk_info(skb, tsk)) 3639 goto attr_msg_cancel; 3640 3641 if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) || 3642 nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) || 3643 nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) || 3644 nla_put_u32(skb, TIPC_NLA_SOCK_UID, 3645 from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk), 3646 sk_uid(sk))) || 3647 nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE, 3648 tipc_diag_gen_cookie(sk), 3649 TIPC_NLA_SOCK_PAD)) 3650 goto attr_msg_cancel; 3651 3652 stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT); 3653 if (!stat) 3654 goto attr_msg_cancel; 3655 3656 if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ, 3657 skb_queue_len(&sk->sk_receive_queue)) || 3658 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ, 3659 skb_queue_len(&sk->sk_write_queue)) || 3660 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP, 3661 sk_drops_read(sk))) 3662 goto stat_msg_cancel; 3663 3664 if (tsk->cong_link_cnt && 3665 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG)) 3666 goto stat_msg_cancel; 3667 3668 if (tsk_conn_cong(tsk) && 3669 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG)) 3670 goto stat_msg_cancel; 3671 3672 nla_nest_end(skb, stat); 3673 3674 if (tsk->group) 3675 if (tipc_group_fill_sock_diag(tsk->group, skb)) 3676 goto stat_msg_cancel; 3677 3678 nla_nest_end(skb, attrs); 3679 3680 return 0; 3681 3682 stat_msg_cancel: 3683 nla_nest_cancel(skb, stat); 3684 attr_msg_cancel: 3685 nla_nest_cancel(skb, attrs); 3686 msg_cancel: 3687 return -EMSGSIZE; 3688 } 3689 EXPORT_SYMBOL(tipc_sk_fill_sock_diag); 3690 3691 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb) 3692 { 3693 return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk); 3694 } 3695 3696 /* Caller should hold socket lock for the passed tipc socket. */ 3697 static int __tipc_nl_add_sk_publ(struct sk_buff *skb, 3698 struct netlink_callback *cb, 3699 struct publication *publ) 3700 { 3701 void *hdr; 3702 struct nlattr *attrs; 3703 3704 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq, 3705 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET); 3706 if (!hdr) 3707 goto msg_cancel; 3708 3709 attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL); 3710 if (!attrs) 3711 goto genlmsg_cancel; 3712 3713 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key)) 3714 goto attr_msg_cancel; 3715 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->sr.type)) 3716 goto attr_msg_cancel; 3717 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->sr.lower)) 3718 goto attr_msg_cancel; 3719 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->sr.upper)) 3720 goto attr_msg_cancel; 3721 3722 nla_nest_end(skb, attrs); 3723 genlmsg_end(skb, hdr); 3724 3725 return 0; 3726 3727 attr_msg_cancel: 3728 nla_nest_cancel(skb, attrs); 3729 genlmsg_cancel: 3730 genlmsg_cancel(skb, hdr); 3731 msg_cancel: 3732 return -EMSGSIZE; 3733 } 3734 3735 /* Caller should hold socket lock for the passed tipc socket. */ 3736 static int __tipc_nl_list_sk_publ(struct sk_buff *skb, 3737 struct netlink_callback *cb, 3738 struct tipc_sock *tsk, u32 *last_publ) 3739 { 3740 int err; 3741 struct publication *p; 3742 3743 if (*last_publ) { 3744 list_for_each_entry(p, &tsk->publications, binding_sock) { 3745 if (p->key == *last_publ) 3746 break; 3747 } 3748 if (list_entry_is_head(p, &tsk->publications, binding_sock)) { 3749 /* We never set seq or call nl_dump_check_consistent() 3750 * this means that setting prev_seq here will cause the 3751 * consistence check to fail in the netlink callback 3752 * handler. Resulting in the last NLMSG_DONE message 3753 * having the NLM_F_DUMP_INTR flag set. 3754 */ 3755 cb->prev_seq = 1; 3756 *last_publ = 0; 3757 return -EPIPE; 3758 } 3759 } else { 3760 p = list_first_entry(&tsk->publications, struct publication, 3761 binding_sock); 3762 } 3763 3764 list_for_each_entry_from(p, &tsk->publications, binding_sock) { 3765 err = __tipc_nl_add_sk_publ(skb, cb, p); 3766 if (err) { 3767 *last_publ = p->key; 3768 return err; 3769 } 3770 } 3771 *last_publ = 0; 3772 3773 return 0; 3774 } 3775 3776 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb) 3777 { 3778 int err; 3779 u32 tsk_portid = cb->args[0]; 3780 u32 last_publ = cb->args[1]; 3781 u32 done = cb->args[2]; 3782 struct net *net = sock_net(skb->sk); 3783 struct tipc_sock *tsk; 3784 3785 if (!tsk_portid) { 3786 struct nlattr **attrs = genl_dumpit_info(cb)->info.attrs; 3787 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1]; 3788 3789 if (!attrs[TIPC_NLA_SOCK]) 3790 return -EINVAL; 3791 3792 err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX, 3793 attrs[TIPC_NLA_SOCK], 3794 tipc_nl_sock_policy, NULL); 3795 if (err) 3796 return err; 3797 3798 if (!sock[TIPC_NLA_SOCK_REF]) 3799 return -EINVAL; 3800 3801 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]); 3802 } 3803 3804 if (done) 3805 return 0; 3806 3807 tsk = tipc_sk_lookup(net, tsk_portid); 3808 if (!tsk) 3809 return -EINVAL; 3810 3811 lock_sock(&tsk->sk); 3812 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ); 3813 if (!err) 3814 done = 1; 3815 release_sock(&tsk->sk); 3816 sock_put(&tsk->sk); 3817 3818 cb->args[0] = tsk_portid; 3819 cb->args[1] = last_publ; 3820 cb->args[2] = done; 3821 3822 return skb->len; 3823 } 3824 3825 /** 3826 * tipc_sk_filtering - check if a socket should be traced 3827 * @sk: the socket to be examined 3828 * 3829 * @sysctl_tipc_sk_filter is used as the socket tuple for filtering: 3830 * (portid, sock type, name type, name lower, name upper) 3831 * 3832 * Return: true if the socket meets the socket tuple data 3833 * (value 0 = 'any') or when there is no tuple set (all = 0), 3834 * otherwise false 3835 */ 3836 bool tipc_sk_filtering(struct sock *sk) 3837 { 3838 struct tipc_sock *tsk; 3839 struct publication *p; 3840 u32 _port, _sktype, _type, _lower, _upper; 3841 u32 type = 0, lower = 0, upper = 0; 3842 3843 if (!sk) 3844 return true; 3845 3846 tsk = tipc_sk(sk); 3847 3848 _port = sysctl_tipc_sk_filter[0]; 3849 _sktype = sysctl_tipc_sk_filter[1]; 3850 _type = sysctl_tipc_sk_filter[2]; 3851 _lower = sysctl_tipc_sk_filter[3]; 3852 _upper = sysctl_tipc_sk_filter[4]; 3853 3854 if (!_port && !_sktype && !_type && !_lower && !_upper) 3855 return true; 3856 3857 if (_port) 3858 return (_port == tsk->portid); 3859 3860 if (_sktype && _sktype != sk->sk_type) 3861 return false; 3862 3863 if (tsk->published) { 3864 p = list_first_entry_or_null(&tsk->publications, 3865 struct publication, binding_sock); 3866 if (p) { 3867 type = p->sr.type; 3868 lower = p->sr.lower; 3869 upper = p->sr.upper; 3870 } 3871 } 3872 3873 if (!tipc_sk_type_connectionless(sk)) { 3874 type = msg_nametype(&tsk->phdr); 3875 lower = msg_nameinst(&tsk->phdr); 3876 upper = lower; 3877 } 3878 3879 if ((_type && _type != type) || (_lower && _lower != lower) || 3880 (_upper && _upper != upper)) 3881 return false; 3882 3883 return true; 3884 } 3885 3886 u32 tipc_sock_get_portid(struct sock *sk) 3887 { 3888 return (sk) ? (tipc_sk(sk))->portid : 0; 3889 } 3890 3891 /** 3892 * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded, 3893 * both the rcv and backlog queues are considered 3894 * @sk: tipc sk to be checked 3895 * @skb: tipc msg to be checked 3896 * 3897 * Return: true if the socket rx queue allocation is > 90%, otherwise false 3898 */ 3899 3900 bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb) 3901 { 3902 atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt; 3903 unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt); 3904 unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk); 3905 3906 return (qsize > lim * 90 / 100); 3907 } 3908 3909 /** 3910 * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded, 3911 * only the rcv queue is considered 3912 * @sk: tipc sk to be checked 3913 * @skb: tipc msg to be checked 3914 * 3915 * Return: true if the socket rx queue allocation is > 90%, otherwise false 3916 */ 3917 3918 bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb) 3919 { 3920 unsigned int lim = rcvbuf_limit(sk, skb); 3921 unsigned int qsize = sk_rmem_alloc_get(sk); 3922 3923 return (qsize > lim * 90 / 100); 3924 } 3925 3926 /** 3927 * tipc_sk_dump - dump TIPC socket 3928 * @sk: tipc sk to be dumped 3929 * @dqueues: bitmask to decide if any socket queue to be dumped? 3930 * - TIPC_DUMP_NONE: don't dump socket queues 3931 * - TIPC_DUMP_SK_SNDQ: dump socket send queue 3932 * - TIPC_DUMP_SK_RCVQ: dump socket rcv queue 3933 * - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue 3934 * - TIPC_DUMP_ALL: dump all the socket queues above 3935 * @buf: returned buffer of dump data in format 3936 */ 3937 int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf) 3938 { 3939 int i = 0; 3940 size_t sz = (dqueues) ? SK_LMAX : SK_LMIN; 3941 u32 conn_type, conn_instance; 3942 struct tipc_sock *tsk; 3943 struct publication *p; 3944 bool tsk_connected; 3945 3946 if (!sk) { 3947 i += scnprintf(buf, sz, "sk data: (null)\n"); 3948 return i; 3949 } 3950 3951 tsk = tipc_sk(sk); 3952 tsk_connected = !tipc_sk_type_connectionless(sk); 3953 3954 i += scnprintf(buf, sz, "sk data: %u", sk->sk_type); 3955 i += scnprintf(buf + i, sz - i, " %d", sk->sk_state); 3956 i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk)); 3957 i += scnprintf(buf + i, sz - i, " %u", tsk->portid); 3958 i += scnprintf(buf + i, sz - i, " | %u", tsk_connected); 3959 if (tsk_connected) { 3960 i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk)); 3961 i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk)); 3962 conn_type = msg_nametype(&tsk->phdr); 3963 conn_instance = msg_nameinst(&tsk->phdr); 3964 i += scnprintf(buf + i, sz - i, " %u", conn_type); 3965 i += scnprintf(buf + i, sz - i, " %u", conn_instance); 3966 } 3967 i += scnprintf(buf + i, sz - i, " | %u", tsk->published); 3968 if (tsk->published) { 3969 p = list_first_entry_or_null(&tsk->publications, 3970 struct publication, binding_sock); 3971 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.type : 0); 3972 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.lower : 0); 3973 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.upper : 0); 3974 } 3975 i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win); 3976 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win); 3977 i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt); 3978 i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps); 3979 i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt); 3980 i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked); 3981 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked); 3982 i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt)); 3983 i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown); 3984 i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk)); 3985 i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf); 3986 i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk)); 3987 i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf); 3988 i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len)); 3989 3990 if (dqueues & TIPC_DUMP_SK_SNDQ) { 3991 i += scnprintf(buf + i, sz - i, "sk_write_queue: "); 3992 i += tipc_list_dump(&sk->sk_write_queue, false, buf + i); 3993 } 3994 3995 if (dqueues & TIPC_DUMP_SK_RCVQ) { 3996 i += scnprintf(buf + i, sz - i, "sk_receive_queue: "); 3997 i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i); 3998 } 3999 4000 if (dqueues & TIPC_DUMP_SK_BKLGQ) { 4001 i += scnprintf(buf + i, sz - i, "sk_backlog:\n head "); 4002 i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i); 4003 if (sk->sk_backlog.tail != sk->sk_backlog.head) { 4004 i += scnprintf(buf + i, sz - i, " tail "); 4005 i += tipc_skb_dump(sk->sk_backlog.tail, false, 4006 buf + i); 4007 } 4008 } 4009 4010 return i; 4011 } 4012