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 <trace/events/sock.h> 41 42 #include "core.h" 43 #include "name_table.h" 44 #include "node.h" 45 #include "link.h" 46 #include "name_distr.h" 47 #include "socket.h" 48 #include "bcast.h" 49 #include "netlink.h" 50 #include "group.h" 51 #include "trace.h" 52 53 #define NAGLE_START_INIT 4 54 #define NAGLE_START_MAX 1024 55 #define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */ 56 #define CONN_PROBING_INTV msecs_to_jiffies(3600000) /* [ms] => 1 h */ 57 #define TIPC_MAX_PORT 0xffffffff 58 #define TIPC_MIN_PORT 1 59 #define TIPC_ACK_RATE 4 /* ACK at 1/4 of rcv window size */ 60 61 enum { 62 TIPC_LISTEN = TCP_LISTEN, 63 TIPC_ESTABLISHED = TCP_ESTABLISHED, 64 TIPC_OPEN = TCP_CLOSE, 65 TIPC_DISCONNECTING = TCP_CLOSE_WAIT, 66 TIPC_CONNECTING = TCP_SYN_SENT, 67 }; 68 69 struct sockaddr_pair { 70 struct sockaddr_tipc sock; 71 struct sockaddr_tipc member; 72 }; 73 74 /** 75 * struct tipc_sock - TIPC socket structure 76 * @sk: socket - interacts with 'port' and with user via the socket API 77 * @max_pkt: maximum packet size "hint" used when building messages sent by port 78 * @maxnagle: maximum size of msg which can be subject to nagle 79 * @portid: unique port identity in TIPC socket hash table 80 * @phdr: preformatted message header used when sending messages 81 * @cong_links: list of congested links 82 * @publications: list of publications for port 83 * @pub_count: total # of publications port has made during its lifetime 84 * @conn_timeout: the time we can wait for an unresponded setup request 85 * @probe_unacked: probe has not received ack yet 86 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue 87 * @cong_link_cnt: number of congested links 88 * @snt_unacked: # messages sent by socket, and not yet acked by peer 89 * @snd_win: send window size 90 * @peer_caps: peer capabilities mask 91 * @rcv_unacked: # messages read by user, but not yet acked back to peer 92 * @rcv_win: receive window size 93 * @peer: 'connected' peer for dgram/rdm 94 * @node: hash table node 95 * @mc_method: cookie for use between socket and broadcast layer 96 * @rcu: rcu struct for tipc_sock 97 * @group: TIPC communications group 98 * @oneway: message count in one direction (FIXME) 99 * @nagle_start: current nagle value 100 * @snd_backlog: send backlog count 101 * @msg_acc: messages accepted; used in managing backlog and nagle 102 * @pkt_cnt: TIPC socket packet count 103 * @expect_ack: whether this TIPC socket is expecting an ack 104 * @nodelay: setsockopt() TIPC_NODELAY setting 105 * @group_is_open: TIPC socket group is fully open (FIXME) 106 * @published: true if port has one or more associated names 107 * @conn_addrtype: address type used when establishing connection 108 */ 109 struct tipc_sock { 110 struct sock sk; 111 u32 max_pkt; 112 u32 maxnagle; 113 u32 portid; 114 struct tipc_msg phdr; 115 struct list_head cong_links; 116 struct list_head publications; 117 u32 pub_count; 118 atomic_t dupl_rcvcnt; 119 u16 conn_timeout; 120 bool probe_unacked; 121 u16 cong_link_cnt; 122 u16 snt_unacked; 123 u16 snd_win; 124 u16 peer_caps; 125 u16 rcv_unacked; 126 u16 rcv_win; 127 struct sockaddr_tipc peer; 128 struct rhash_head node; 129 struct tipc_mc_method mc_method; 130 struct rcu_head rcu; 131 struct tipc_group *group; 132 u32 oneway; 133 u32 nagle_start; 134 u16 snd_backlog; 135 u16 msg_acc; 136 u16 pkt_cnt; 137 bool expect_ack; 138 bool nodelay; 139 bool group_is_open; 140 bool published; 141 u8 conn_addrtype; 142 }; 143 144 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb); 145 static void tipc_data_ready(struct sock *sk); 146 static void tipc_write_space(struct sock *sk); 147 static void tipc_sock_destruct(struct sock *sk); 148 static int tipc_release(struct socket *sock); 149 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags, 150 bool kern); 151 static void tipc_sk_timeout(struct timer_list *t); 152 static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua); 153 static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua); 154 static int tipc_sk_leave(struct tipc_sock *tsk); 155 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid); 156 static int tipc_sk_insert(struct tipc_sock *tsk); 157 static void tipc_sk_remove(struct tipc_sock *tsk); 158 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz); 159 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz); 160 static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack); 161 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p); 162 163 static const struct proto_ops packet_ops; 164 static const struct proto_ops stream_ops; 165 static const struct proto_ops msg_ops; 166 static struct proto tipc_proto; 167 static const struct rhashtable_params tsk_rht_params; 168 169 static u32 tsk_own_node(struct tipc_sock *tsk) 170 { 171 return msg_prevnode(&tsk->phdr); 172 } 173 174 static u32 tsk_peer_node(struct tipc_sock *tsk) 175 { 176 return msg_destnode(&tsk->phdr); 177 } 178 179 static u32 tsk_peer_port(struct tipc_sock *tsk) 180 { 181 return msg_destport(&tsk->phdr); 182 } 183 184 static bool tsk_unreliable(struct tipc_sock *tsk) 185 { 186 return msg_src_droppable(&tsk->phdr) != 0; 187 } 188 189 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable) 190 { 191 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0); 192 } 193 194 static bool tsk_unreturnable(struct tipc_sock *tsk) 195 { 196 return msg_dest_droppable(&tsk->phdr) != 0; 197 } 198 199 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable) 200 { 201 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0); 202 } 203 204 static int tsk_importance(struct tipc_sock *tsk) 205 { 206 return msg_importance(&tsk->phdr); 207 } 208 209 static struct tipc_sock *tipc_sk(const struct sock *sk) 210 { 211 return container_of(sk, struct tipc_sock, sk); 212 } 213 214 int tsk_set_importance(struct sock *sk, int imp) 215 { 216 if (imp > TIPC_CRITICAL_IMPORTANCE) 217 return -EINVAL; 218 msg_set_importance(&tipc_sk(sk)->phdr, (u32)imp); 219 return 0; 220 } 221 222 static bool tsk_conn_cong(struct tipc_sock *tsk) 223 { 224 return tsk->snt_unacked > tsk->snd_win; 225 } 226 227 static u16 tsk_blocks(int len) 228 { 229 return ((len / FLOWCTL_BLK_SZ) + 1); 230 } 231 232 /* tsk_blocks(): translate a buffer size in bytes to number of 233 * advertisable blocks, taking into account the ratio truesize(len)/len 234 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ 235 */ 236 static u16 tsk_adv_blocks(int len) 237 { 238 return len / FLOWCTL_BLK_SZ / 4; 239 } 240 241 /* tsk_inc(): increment counter for sent or received data 242 * - If block based flow control is not supported by peer we 243 * fall back to message based ditto, incrementing the counter 244 */ 245 static u16 tsk_inc(struct tipc_sock *tsk, int msglen) 246 { 247 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL)) 248 return ((msglen / FLOWCTL_BLK_SZ) + 1); 249 return 1; 250 } 251 252 /* tsk_set_nagle - enable/disable nagle property by manipulating maxnagle 253 */ 254 static void tsk_set_nagle(struct tipc_sock *tsk) 255 { 256 struct sock *sk = &tsk->sk; 257 258 tsk->maxnagle = 0; 259 if (sk->sk_type != SOCK_STREAM) 260 return; 261 if (tsk->nodelay) 262 return; 263 if (!(tsk->peer_caps & TIPC_NAGLE)) 264 return; 265 /* Limit node local buffer size to avoid receive queue overflow */ 266 if (tsk->max_pkt == MAX_MSG_SIZE) 267 tsk->maxnagle = 1500; 268 else 269 tsk->maxnagle = tsk->max_pkt; 270 } 271 272 /** 273 * tsk_advance_rx_queue - discard first buffer in socket receive queue 274 * @sk: network socket 275 * 276 * Caller must hold socket lock 277 */ 278 static void tsk_advance_rx_queue(struct sock *sk) 279 { 280 trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " "); 281 kfree_skb(__skb_dequeue(&sk->sk_receive_queue)); 282 } 283 284 /* tipc_sk_respond() : send response message back to sender 285 */ 286 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err) 287 { 288 u32 selector; 289 u32 dnode; 290 u32 onode = tipc_own_addr(sock_net(sk)); 291 292 if (!tipc_msg_reverse(onode, &skb, err)) 293 return; 294 295 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!"); 296 dnode = msg_destnode(buf_msg(skb)); 297 selector = msg_origport(buf_msg(skb)); 298 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector); 299 } 300 301 /** 302 * tsk_rej_rx_queue - reject all buffers in socket receive queue 303 * @sk: network socket 304 * @error: response error code 305 * 306 * Caller must hold socket lock 307 */ 308 static void tsk_rej_rx_queue(struct sock *sk, int error) 309 { 310 struct sk_buff *skb; 311 312 while ((skb = __skb_dequeue(&sk->sk_receive_queue))) 313 tipc_sk_respond(sk, skb, error); 314 } 315 316 static bool tipc_sk_connected(const struct sock *sk) 317 { 318 return READ_ONCE(sk->sk_state) == TIPC_ESTABLISHED; 319 } 320 321 /* tipc_sk_type_connectionless - check if the socket is datagram socket 322 * @sk: socket 323 * 324 * Returns true if connection less, false otherwise 325 */ 326 static bool tipc_sk_type_connectionless(struct sock *sk) 327 { 328 return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM; 329 } 330 331 /* tsk_peer_msg - verify if message was sent by connected port's peer 332 * 333 * Handles cases where the node's network address has changed from 334 * the default of <0.0.0> to its configured setting. 335 */ 336 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg) 337 { 338 struct sock *sk = &tsk->sk; 339 u32 self = tipc_own_addr(sock_net(sk)); 340 u32 peer_port = tsk_peer_port(tsk); 341 u32 orig_node, peer_node; 342 343 if (unlikely(!tipc_sk_connected(sk))) 344 return false; 345 346 if (unlikely(msg_origport(msg) != peer_port)) 347 return false; 348 349 orig_node = msg_orignode(msg); 350 peer_node = tsk_peer_node(tsk); 351 352 if (likely(orig_node == peer_node)) 353 return true; 354 355 if (!orig_node && peer_node == self) 356 return true; 357 358 if (!peer_node && orig_node == self) 359 return true; 360 361 return false; 362 } 363 364 /* tipc_set_sk_state - set the sk_state of the socket 365 * @sk: socket 366 * 367 * Caller must hold socket lock 368 * 369 * Returns 0 on success, errno otherwise 370 */ 371 static int tipc_set_sk_state(struct sock *sk, int state) 372 { 373 int oldsk_state = sk->sk_state; 374 int res = -EINVAL; 375 376 switch (state) { 377 case TIPC_OPEN: 378 res = 0; 379 break; 380 case TIPC_LISTEN: 381 case TIPC_CONNECTING: 382 if (oldsk_state == TIPC_OPEN) 383 res = 0; 384 break; 385 case TIPC_ESTABLISHED: 386 if (oldsk_state == TIPC_CONNECTING || 387 oldsk_state == TIPC_OPEN) 388 res = 0; 389 break; 390 case TIPC_DISCONNECTING: 391 if (oldsk_state == TIPC_CONNECTING || 392 oldsk_state == TIPC_ESTABLISHED) 393 res = 0; 394 break; 395 } 396 397 if (!res) 398 sk->sk_state = state; 399 400 return res; 401 } 402 403 static int tipc_sk_sock_err(struct socket *sock, long *timeout) 404 { 405 struct sock *sk = sock->sk; 406 int err = sock_error(sk); 407 int typ = sock->type; 408 409 if (err) 410 return err; 411 if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) { 412 if (sk->sk_state == TIPC_DISCONNECTING) 413 return -EPIPE; 414 else if (!tipc_sk_connected(sk)) 415 return -ENOTCONN; 416 } 417 if (!*timeout) 418 return -EAGAIN; 419 if (signal_pending(current)) 420 return sock_intr_errno(*timeout); 421 422 return 0; 423 } 424 425 #define tipc_wait_for_cond(sock_, timeo_, condition_) \ 426 ({ \ 427 DEFINE_WAIT_FUNC(wait_, woken_wake_function); \ 428 struct sock *sk_; \ 429 int rc_; \ 430 \ 431 while ((rc_ = !(condition_))) { \ 432 /* coupled with smp_wmb() in tipc_sk_proto_rcv() */ \ 433 smp_rmb(); \ 434 sk_ = (sock_)->sk; \ 435 rc_ = tipc_sk_sock_err((sock_), timeo_); \ 436 if (rc_) \ 437 break; \ 438 add_wait_queue(sk_sleep(sk_), &wait_); \ 439 release_sock(sk_); \ 440 *(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \ 441 sched_annotate_sleep(); \ 442 lock_sock(sk_); \ 443 remove_wait_queue(sk_sleep(sk_), &wait_); \ 444 } \ 445 rc_; \ 446 }) 447 448 /** 449 * tipc_sk_create - create a TIPC socket 450 * @net: network namespace (must be default network) 451 * @sock: pre-allocated socket structure 452 * @protocol: protocol indicator (must be 0) 453 * @kern: caused by kernel or by userspace? 454 * 455 * This routine creates additional data structures used by the TIPC socket, 456 * initializes them, and links them together. 457 * 458 * Return: 0 on success, errno otherwise 459 */ 460 static int tipc_sk_create(struct net *net, struct socket *sock, 461 int protocol, int kern) 462 { 463 const struct proto_ops *ops; 464 struct sock *sk; 465 struct tipc_sock *tsk; 466 struct tipc_msg *msg; 467 468 /* Validate arguments */ 469 if (unlikely(protocol != 0)) 470 return -EPROTONOSUPPORT; 471 472 switch (sock->type) { 473 case SOCK_STREAM: 474 ops = &stream_ops; 475 break; 476 case SOCK_SEQPACKET: 477 ops = &packet_ops; 478 break; 479 case SOCK_DGRAM: 480 case SOCK_RDM: 481 ops = &msg_ops; 482 break; 483 default: 484 return -EPROTOTYPE; 485 } 486 487 /* Allocate socket's protocol area */ 488 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern); 489 if (sk == NULL) 490 return -ENOMEM; 491 492 tsk = tipc_sk(sk); 493 tsk->max_pkt = MAX_PKT_DEFAULT; 494 tsk->maxnagle = 0; 495 tsk->nagle_start = NAGLE_START_INIT; 496 INIT_LIST_HEAD(&tsk->publications); 497 INIT_LIST_HEAD(&tsk->cong_links); 498 msg = &tsk->phdr; 499 500 /* Finish initializing socket data structures */ 501 sock->ops = ops; 502 sock_init_data(sock, sk); 503 tipc_set_sk_state(sk, TIPC_OPEN); 504 if (tipc_sk_insert(tsk)) { 505 sk_free(sk); 506 pr_warn("Socket create failed; port number exhausted\n"); 507 return -EINVAL; 508 } 509 510 /* Ensure tsk is visible before we read own_addr. */ 511 smp_mb(); 512 513 tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE, 514 TIPC_NAMED_MSG, NAMED_H_SIZE, 0); 515 516 msg_set_origport(msg, tsk->portid); 517 timer_setup(&sk->sk_timer, tipc_sk_timeout, 0); 518 sk->sk_shutdown = 0; 519 sk->sk_backlog_rcv = tipc_sk_backlog_rcv; 520 sk->sk_rcvbuf = READ_ONCE(sysctl_tipc_rmem[1]); 521 sk->sk_data_ready = tipc_data_ready; 522 sk->sk_write_space = tipc_write_space; 523 sk->sk_destruct = tipc_sock_destruct; 524 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT; 525 tsk->group_is_open = true; 526 atomic_set(&tsk->dupl_rcvcnt, 0); 527 528 /* Start out with safe limits until we receive an advertised window */ 529 tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN); 530 tsk->rcv_win = tsk->snd_win; 531 532 if (tipc_sk_type_connectionless(sk)) { 533 tsk_set_unreturnable(tsk, true); 534 if (sock->type == SOCK_DGRAM) 535 tsk_set_unreliable(tsk, true); 536 } 537 __skb_queue_head_init(&tsk->mc_method.deferredq); 538 trace_tipc_sk_create(sk, NULL, TIPC_DUMP_NONE, " "); 539 return 0; 540 } 541 542 static void tipc_sk_callback(struct rcu_head *head) 543 { 544 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu); 545 546 sock_put(&tsk->sk); 547 } 548 549 /* Caller should hold socket lock for the socket. */ 550 static void __tipc_shutdown(struct socket *sock, int error) 551 { 552 struct sock *sk = sock->sk; 553 struct tipc_sock *tsk = tipc_sk(sk); 554 struct net *net = sock_net(sk); 555 long timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT); 556 u32 dnode = tsk_peer_node(tsk); 557 struct sk_buff *skb; 558 559 /* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */ 560 tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt && 561 !tsk_conn_cong(tsk))); 562 563 /* Push out delayed messages if in Nagle mode */ 564 tipc_sk_push_backlog(tsk, false); 565 /* Remove pending SYN */ 566 __skb_queue_purge(&sk->sk_write_queue); 567 568 /* Remove partially received buffer if any */ 569 skb = skb_peek(&sk->sk_receive_queue); 570 if (skb && TIPC_SKB_CB(skb)->bytes_read) { 571 __skb_unlink(skb, &sk->sk_receive_queue); 572 kfree_skb(skb); 573 } 574 575 /* Reject all unreceived messages if connectionless */ 576 if (tipc_sk_type_connectionless(sk)) { 577 tsk_rej_rx_queue(sk, error); 578 return; 579 } 580 581 switch (sk->sk_state) { 582 case TIPC_CONNECTING: 583 case TIPC_ESTABLISHED: 584 tipc_set_sk_state(sk, TIPC_DISCONNECTING); 585 tipc_node_remove_conn(net, dnode, tsk->portid); 586 /* Send a FIN+/- to its peer */ 587 skb = __skb_dequeue(&sk->sk_receive_queue); 588 if (skb) { 589 __skb_queue_purge(&sk->sk_receive_queue); 590 tipc_sk_respond(sk, skb, error); 591 break; 592 } 593 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE, 594 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode, 595 tsk_own_node(tsk), tsk_peer_port(tsk), 596 tsk->portid, error); 597 if (skb) 598 tipc_node_xmit_skb(net, skb, dnode, tsk->portid); 599 break; 600 case TIPC_LISTEN: 601 /* Reject all SYN messages */ 602 tsk_rej_rx_queue(sk, error); 603 break; 604 default: 605 __skb_queue_purge(&sk->sk_receive_queue); 606 break; 607 } 608 } 609 610 /** 611 * tipc_release - destroy a TIPC socket 612 * @sock: socket to destroy 613 * 614 * This routine cleans up any messages that are still queued on the socket. 615 * For DGRAM and RDM socket types, all queued messages are rejected. 616 * For SEQPACKET and STREAM socket types, the first message is rejected 617 * and any others are discarded. (If the first message on a STREAM socket 618 * is partially-read, it is discarded and the next one is rejected instead.) 619 * 620 * NOTE: Rejected messages are not necessarily returned to the sender! They 621 * are returned or discarded according to the "destination droppable" setting 622 * specified for the message by the sender. 623 * 624 * Return: 0 on success, errno otherwise 625 */ 626 static int tipc_release(struct socket *sock) 627 { 628 struct sock *sk = sock->sk; 629 struct tipc_sock *tsk; 630 631 /* 632 * Exit if socket isn't fully initialized (occurs when a failed accept() 633 * releases a pre-allocated child socket that was never used) 634 */ 635 if (sk == NULL) 636 return 0; 637 638 tsk = tipc_sk(sk); 639 lock_sock(sk); 640 641 trace_tipc_sk_release(sk, NULL, TIPC_DUMP_ALL, " "); 642 __tipc_shutdown(sock, TIPC_ERR_NO_PORT); 643 sk->sk_shutdown = SHUTDOWN_MASK; 644 tipc_sk_leave(tsk); 645 tipc_sk_withdraw(tsk, NULL); 646 __skb_queue_purge(&tsk->mc_method.deferredq); 647 sk_stop_timer(sk, &sk->sk_timer); 648 tipc_sk_remove(tsk); 649 650 sock_orphan(sk); 651 /* Reject any messages that accumulated in backlog queue */ 652 release_sock(sk); 653 tipc_dest_list_purge(&tsk->cong_links); 654 tsk->cong_link_cnt = 0; 655 call_rcu(&tsk->rcu, tipc_sk_callback); 656 sock->sk = NULL; 657 658 return 0; 659 } 660 661 /** 662 * __tipc_bind - associate or disassocate TIPC name(s) with a socket 663 * @sock: socket structure 664 * @skaddr: socket address describing name(s) and desired operation 665 * @alen: size of socket address data structure 666 * 667 * Name and name sequence binding are indicated using a positive scope value; 668 * a negative scope value unbinds the specified name. Specifying no name 669 * (i.e. a socket address length of 0) unbinds all names from the socket. 670 * 671 * Return: 0 on success, errno otherwise 672 * 673 * NOTE: This routine doesn't need to take the socket lock since it doesn't 674 * access any non-constant socket information. 675 */ 676 static int __tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen) 677 { 678 struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr; 679 struct tipc_sock *tsk = tipc_sk(sock->sk); 680 bool unbind = false; 681 682 if (unlikely(!alen)) 683 return tipc_sk_withdraw(tsk, NULL); 684 685 if (ua->addrtype == TIPC_SERVICE_ADDR) { 686 ua->addrtype = TIPC_SERVICE_RANGE; 687 ua->sr.upper = ua->sr.lower; 688 } 689 if (ua->scope < 0) { 690 unbind = true; 691 ua->scope = -ua->scope; 692 } 693 /* Users may still use deprecated TIPC_ZONE_SCOPE */ 694 if (ua->scope != TIPC_NODE_SCOPE) 695 ua->scope = TIPC_CLUSTER_SCOPE; 696 697 if (tsk->group) 698 return -EACCES; 699 700 if (unbind) 701 return tipc_sk_withdraw(tsk, ua); 702 return tipc_sk_publish(tsk, ua); 703 } 704 705 int tipc_sk_bind(struct socket *sock, struct sockaddr *skaddr, int alen) 706 { 707 int res; 708 709 lock_sock(sock->sk); 710 res = __tipc_bind(sock, skaddr, alen); 711 release_sock(sock->sk); 712 return res; 713 } 714 715 static int tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen) 716 { 717 struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr; 718 u32 atype = ua->addrtype; 719 720 if (alen) { 721 if (!tipc_uaddr_valid(ua, alen)) 722 return -EINVAL; 723 if (atype == TIPC_SOCKET_ADDR) 724 return -EAFNOSUPPORT; 725 if (ua->sr.type < TIPC_RESERVED_TYPES) { 726 pr_warn_once("Can't bind to reserved service type %u\n", 727 ua->sr.type); 728 return -EACCES; 729 } 730 } 731 return tipc_sk_bind(sock, skaddr, alen); 732 } 733 734 /** 735 * tipc_getname - get port ID of socket or peer socket 736 * @sock: socket structure 737 * @uaddr: area for returned socket address 738 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID 739 * 740 * Return: 0 on success, errno otherwise 741 * 742 * NOTE: This routine doesn't need to take the socket lock since it only 743 * accesses socket information that is unchanging (or which changes in 744 * a completely predictable manner). 745 */ 746 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr, 747 int peer) 748 { 749 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr; 750 struct sock *sk = sock->sk; 751 struct tipc_sock *tsk = tipc_sk(sk); 752 753 memset(addr, 0, sizeof(*addr)); 754 if (peer) { 755 if ((!tipc_sk_connected(sk)) && 756 ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING))) 757 return -ENOTCONN; 758 addr->addr.id.ref = tsk_peer_port(tsk); 759 addr->addr.id.node = tsk_peer_node(tsk); 760 } else { 761 addr->addr.id.ref = tsk->portid; 762 addr->addr.id.node = tipc_own_addr(sock_net(sk)); 763 } 764 765 addr->addrtype = TIPC_SOCKET_ADDR; 766 addr->family = AF_TIPC; 767 addr->scope = 0; 768 addr->addr.name.domain = 0; 769 770 return sizeof(*addr); 771 } 772 773 /** 774 * tipc_poll - read and possibly block on pollmask 775 * @file: file structure associated with the socket 776 * @sock: socket for which to calculate the poll bits 777 * @wait: ??? 778 * 779 * Return: pollmask value 780 * 781 * COMMENTARY: 782 * It appears that the usual socket locking mechanisms are not useful here 783 * since the pollmask info is potentially out-of-date the moment this routine 784 * exits. TCP and other protocols seem to rely on higher level poll routines 785 * to handle any preventable race conditions, so TIPC will do the same ... 786 * 787 * IMPORTANT: The fact that a read or write operation is indicated does NOT 788 * imply that the operation will succeed, merely that it should be performed 789 * and will not block. 790 */ 791 static __poll_t tipc_poll(struct file *file, struct socket *sock, 792 poll_table *wait) 793 { 794 struct sock *sk = sock->sk; 795 struct tipc_sock *tsk = tipc_sk(sk); 796 __poll_t revents = 0; 797 798 sock_poll_wait(file, sock, wait); 799 trace_tipc_sk_poll(sk, NULL, TIPC_DUMP_ALL, " "); 800 801 if (sk->sk_shutdown & RCV_SHUTDOWN) 802 revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM; 803 if (sk->sk_shutdown == SHUTDOWN_MASK) 804 revents |= EPOLLHUP; 805 806 switch (sk->sk_state) { 807 case TIPC_ESTABLISHED: 808 if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk)) 809 revents |= EPOLLOUT; 810 fallthrough; 811 case TIPC_LISTEN: 812 case TIPC_CONNECTING: 813 if (!skb_queue_empty_lockless(&sk->sk_receive_queue)) 814 revents |= EPOLLIN | EPOLLRDNORM; 815 break; 816 case TIPC_OPEN: 817 if (tsk->group_is_open && !tsk->cong_link_cnt) 818 revents |= EPOLLOUT; 819 if (!tipc_sk_type_connectionless(sk)) 820 break; 821 if (skb_queue_empty_lockless(&sk->sk_receive_queue)) 822 break; 823 revents |= EPOLLIN | EPOLLRDNORM; 824 break; 825 case TIPC_DISCONNECTING: 826 revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP; 827 break; 828 } 829 return revents; 830 } 831 832 /** 833 * tipc_sendmcast - send multicast message 834 * @sock: socket structure 835 * @ua: destination address struct 836 * @msg: message to send 837 * @dlen: length of data to send 838 * @timeout: timeout to wait for wakeup 839 * 840 * Called from function tipc_sendmsg(), which has done all sanity checks 841 * Return: the number of bytes sent on success, or errno 842 */ 843 static int tipc_sendmcast(struct socket *sock, struct tipc_uaddr *ua, 844 struct msghdr *msg, size_t dlen, long timeout) 845 { 846 struct sock *sk = sock->sk; 847 struct tipc_sock *tsk = tipc_sk(sk); 848 struct tipc_msg *hdr = &tsk->phdr; 849 struct net *net = sock_net(sk); 850 int mtu = tipc_bcast_get_mtu(net); 851 struct sk_buff_head pkts; 852 struct tipc_nlist dsts; 853 int rc; 854 855 if (tsk->group) 856 return -EACCES; 857 858 /* Block or return if any destination link is congested */ 859 rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt); 860 if (unlikely(rc)) 861 return rc; 862 863 /* Lookup destination nodes */ 864 tipc_nlist_init(&dsts, tipc_own_addr(net)); 865 tipc_nametbl_lookup_mcast_nodes(net, ua, &dsts); 866 if (!dsts.local && !dsts.remote) 867 return -EHOSTUNREACH; 868 869 /* Build message header */ 870 msg_set_type(hdr, TIPC_MCAST_MSG); 871 msg_set_hdr_sz(hdr, MCAST_H_SIZE); 872 msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE); 873 msg_set_destport(hdr, 0); 874 msg_set_destnode(hdr, 0); 875 msg_set_nametype(hdr, ua->sr.type); 876 msg_set_namelower(hdr, ua->sr.lower); 877 msg_set_nameupper(hdr, ua->sr.upper); 878 879 /* Build message as chain of buffers */ 880 __skb_queue_head_init(&pkts); 881 rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts); 882 883 /* Send message if build was successful */ 884 if (unlikely(rc == dlen)) { 885 trace_tipc_sk_sendmcast(sk, skb_peek(&pkts), 886 TIPC_DUMP_SK_SNDQ, " "); 887 rc = tipc_mcast_xmit(net, &pkts, &tsk->mc_method, &dsts, 888 &tsk->cong_link_cnt); 889 } 890 891 tipc_nlist_purge(&dsts); 892 893 return rc ? rc : dlen; 894 } 895 896 /** 897 * tipc_send_group_msg - send a message to a member in the group 898 * @net: network namespace 899 * @tsk: tipc socket 900 * @m: message to send 901 * @mb: group member 902 * @dnode: destination node 903 * @dport: destination port 904 * @dlen: total length of message data 905 */ 906 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk, 907 struct msghdr *m, struct tipc_member *mb, 908 u32 dnode, u32 dport, int dlen) 909 { 910 u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group); 911 struct tipc_mc_method *method = &tsk->mc_method; 912 int blks = tsk_blocks(GROUP_H_SIZE + dlen); 913 struct tipc_msg *hdr = &tsk->phdr; 914 struct sk_buff_head pkts; 915 int mtu, rc; 916 917 /* Complete message header */ 918 msg_set_type(hdr, TIPC_GRP_UCAST_MSG); 919 msg_set_hdr_sz(hdr, GROUP_H_SIZE); 920 msg_set_destport(hdr, dport); 921 msg_set_destnode(hdr, dnode); 922 msg_set_grp_bc_seqno(hdr, bc_snd_nxt); 923 924 /* Build message as chain of buffers */ 925 __skb_queue_head_init(&pkts); 926 mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false); 927 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts); 928 if (unlikely(rc != dlen)) 929 return rc; 930 931 /* Send message */ 932 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid); 933 if (unlikely(rc == -ELINKCONG)) { 934 tipc_dest_push(&tsk->cong_links, dnode, 0); 935 tsk->cong_link_cnt++; 936 } 937 938 /* Update send window */ 939 tipc_group_update_member(mb, blks); 940 941 /* A broadcast sent within next EXPIRE period must follow same path */ 942 method->rcast = true; 943 method->mandatory = true; 944 return dlen; 945 } 946 947 /** 948 * tipc_send_group_unicast - send message to a member in the group 949 * @sock: socket structure 950 * @m: message to send 951 * @dlen: total length of message data 952 * @timeout: timeout to wait for wakeup 953 * 954 * Called from function tipc_sendmsg(), which has done all sanity checks 955 * Return: the number of bytes sent on success, or errno 956 */ 957 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m, 958 int dlen, long timeout) 959 { 960 struct sock *sk = sock->sk; 961 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name; 962 int blks = tsk_blocks(GROUP_H_SIZE + dlen); 963 struct tipc_sock *tsk = tipc_sk(sk); 964 struct net *net = sock_net(sk); 965 struct tipc_member *mb = NULL; 966 u32 node, port; 967 int rc; 968 969 node = ua->sk.node; 970 port = ua->sk.ref; 971 if (!port && !node) 972 return -EHOSTUNREACH; 973 974 /* Block or return if destination link or member is congested */ 975 rc = tipc_wait_for_cond(sock, &timeout, 976 !tipc_dest_find(&tsk->cong_links, node, 0) && 977 tsk->group && 978 !tipc_group_cong(tsk->group, node, port, blks, 979 &mb)); 980 if (unlikely(rc)) 981 return rc; 982 983 if (unlikely(!mb)) 984 return -EHOSTUNREACH; 985 986 rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen); 987 988 return rc ? rc : dlen; 989 } 990 991 /** 992 * tipc_send_group_anycast - send message to any member with given identity 993 * @sock: socket structure 994 * @m: message to send 995 * @dlen: total length of message data 996 * @timeout: timeout to wait for wakeup 997 * 998 * Called from function tipc_sendmsg(), which has done all sanity checks 999 * Return: the number of bytes sent on success, or errno 1000 */ 1001 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m, 1002 int dlen, long timeout) 1003 { 1004 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name; 1005 struct sock *sk = sock->sk; 1006 struct tipc_sock *tsk = tipc_sk(sk); 1007 struct list_head *cong_links = &tsk->cong_links; 1008 int blks = tsk_blocks(GROUP_H_SIZE + dlen); 1009 struct tipc_msg *hdr = &tsk->phdr; 1010 struct tipc_member *first = NULL; 1011 struct tipc_member *mbr = NULL; 1012 struct net *net = sock_net(sk); 1013 u32 node, port, exclude; 1014 struct list_head dsts; 1015 int lookups = 0; 1016 int dstcnt, rc; 1017 bool cong; 1018 1019 INIT_LIST_HEAD(&dsts); 1020 ua->sa.type = msg_nametype(hdr); 1021 ua->scope = msg_lookup_scope(hdr); 1022 1023 while (++lookups < 4) { 1024 exclude = tipc_group_exclude(tsk->group); 1025 1026 first = NULL; 1027 1028 /* Look for a non-congested destination member, if any */ 1029 while (1) { 1030 if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt, 1031 exclude, false)) 1032 return -EHOSTUNREACH; 1033 tipc_dest_pop(&dsts, &node, &port); 1034 cong = tipc_group_cong(tsk->group, node, port, blks, 1035 &mbr); 1036 if (!cong) 1037 break; 1038 if (mbr == first) 1039 break; 1040 if (!first) 1041 first = mbr; 1042 } 1043 1044 /* Start over if destination was not in member list */ 1045 if (unlikely(!mbr)) 1046 continue; 1047 1048 if (likely(!cong && !tipc_dest_find(cong_links, node, 0))) 1049 break; 1050 1051 /* Block or return if destination link or member is congested */ 1052 rc = tipc_wait_for_cond(sock, &timeout, 1053 !tipc_dest_find(cong_links, node, 0) && 1054 tsk->group && 1055 !tipc_group_cong(tsk->group, node, port, 1056 blks, &mbr)); 1057 if (unlikely(rc)) 1058 return rc; 1059 1060 /* Send, unless destination disappeared while waiting */ 1061 if (likely(mbr)) 1062 break; 1063 } 1064 1065 if (unlikely(lookups >= 4)) 1066 return -EHOSTUNREACH; 1067 1068 rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen); 1069 1070 return rc ? rc : dlen; 1071 } 1072 1073 /** 1074 * tipc_send_group_bcast - send message to all members in communication group 1075 * @sock: socket structure 1076 * @m: message to send 1077 * @dlen: total length of message data 1078 * @timeout: timeout to wait for wakeup 1079 * 1080 * Called from function tipc_sendmsg(), which has done all sanity checks 1081 * Return: the number of bytes sent on success, or errno 1082 */ 1083 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m, 1084 int dlen, long timeout) 1085 { 1086 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name; 1087 struct sock *sk = sock->sk; 1088 struct net *net = sock_net(sk); 1089 struct tipc_sock *tsk = tipc_sk(sk); 1090 struct tipc_nlist *dsts; 1091 struct tipc_mc_method *method = &tsk->mc_method; 1092 bool ack = method->mandatory && method->rcast; 1093 int blks = tsk_blocks(MCAST_H_SIZE + dlen); 1094 struct tipc_msg *hdr = &tsk->phdr; 1095 int mtu = tipc_bcast_get_mtu(net); 1096 struct sk_buff_head pkts; 1097 int rc = -EHOSTUNREACH; 1098 1099 /* Block or return if any destination link or member is congested */ 1100 rc = tipc_wait_for_cond(sock, &timeout, 1101 !tsk->cong_link_cnt && tsk->group && 1102 !tipc_group_bc_cong(tsk->group, blks)); 1103 if (unlikely(rc)) 1104 return rc; 1105 1106 dsts = tipc_group_dests(tsk->group); 1107 if (!dsts->local && !dsts->remote) 1108 return -EHOSTUNREACH; 1109 1110 /* Complete message header */ 1111 if (ua) { 1112 msg_set_type(hdr, TIPC_GRP_MCAST_MSG); 1113 msg_set_nameinst(hdr, ua->sa.instance); 1114 } else { 1115 msg_set_type(hdr, TIPC_GRP_BCAST_MSG); 1116 msg_set_nameinst(hdr, 0); 1117 } 1118 msg_set_hdr_sz(hdr, GROUP_H_SIZE); 1119 msg_set_destport(hdr, 0); 1120 msg_set_destnode(hdr, 0); 1121 msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group)); 1122 1123 /* Avoid getting stuck with repeated forced replicasts */ 1124 msg_set_grp_bc_ack_req(hdr, ack); 1125 1126 /* Build message as chain of buffers */ 1127 __skb_queue_head_init(&pkts); 1128 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts); 1129 if (unlikely(rc != dlen)) 1130 return rc; 1131 1132 /* Send message */ 1133 rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt); 1134 if (unlikely(rc)) 1135 return rc; 1136 1137 /* Update broadcast sequence number and send windows */ 1138 tipc_group_update_bc_members(tsk->group, blks, ack); 1139 1140 /* Broadcast link is now free to choose method for next broadcast */ 1141 method->mandatory = false; 1142 method->expires = jiffies; 1143 1144 return dlen; 1145 } 1146 1147 /** 1148 * tipc_send_group_mcast - send message to all members with given identity 1149 * @sock: socket structure 1150 * @m: message to send 1151 * @dlen: total length of message data 1152 * @timeout: timeout to wait for wakeup 1153 * 1154 * Called from function tipc_sendmsg(), which has done all sanity checks 1155 * Return: the number of bytes sent on success, or errno 1156 */ 1157 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m, 1158 int dlen, long timeout) 1159 { 1160 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name; 1161 struct sock *sk = sock->sk; 1162 struct tipc_sock *tsk = tipc_sk(sk); 1163 struct tipc_group *grp = tsk->group; 1164 struct tipc_msg *hdr = &tsk->phdr; 1165 struct net *net = sock_net(sk); 1166 struct list_head dsts; 1167 u32 dstcnt, exclude; 1168 1169 INIT_LIST_HEAD(&dsts); 1170 ua->sa.type = msg_nametype(hdr); 1171 ua->scope = msg_lookup_scope(hdr); 1172 exclude = tipc_group_exclude(grp); 1173 1174 if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt, exclude, true)) 1175 return -EHOSTUNREACH; 1176 1177 if (dstcnt == 1) { 1178 tipc_dest_pop(&dsts, &ua->sk.node, &ua->sk.ref); 1179 return tipc_send_group_unicast(sock, m, dlen, timeout); 1180 } 1181 1182 tipc_dest_list_purge(&dsts); 1183 return tipc_send_group_bcast(sock, m, dlen, timeout); 1184 } 1185 1186 /** 1187 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets 1188 * @net: the associated network namespace 1189 * @arrvq: queue with arriving messages, to be cloned after destination lookup 1190 * @inputq: queue with cloned messages, delivered to socket after dest lookup 1191 * 1192 * Multi-threaded: parallel calls with reference to same queues may occur 1193 */ 1194 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq, 1195 struct sk_buff_head *inputq) 1196 { 1197 u32 self = tipc_own_addr(net); 1198 struct sk_buff *skb, *_skb; 1199 u32 portid, onode; 1200 struct sk_buff_head tmpq; 1201 struct list_head dports; 1202 struct tipc_msg *hdr; 1203 struct tipc_uaddr ua; 1204 int user, mtyp, hlen; 1205 1206 __skb_queue_head_init(&tmpq); 1207 INIT_LIST_HEAD(&dports); 1208 ua.addrtype = TIPC_SERVICE_RANGE; 1209 1210 /* tipc_skb_peek() increments the head skb's reference counter */ 1211 skb = tipc_skb_peek(arrvq, &inputq->lock); 1212 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) { 1213 hdr = buf_msg(skb); 1214 user = msg_user(hdr); 1215 mtyp = msg_type(hdr); 1216 hlen = skb_headroom(skb) + msg_hdr_sz(hdr); 1217 onode = msg_orignode(hdr); 1218 ua.sr.type = msg_nametype(hdr); 1219 ua.sr.lower = msg_namelower(hdr); 1220 ua.sr.upper = msg_nameupper(hdr); 1221 if (onode == self) 1222 ua.scope = TIPC_ANY_SCOPE; 1223 else 1224 ua.scope = TIPC_CLUSTER_SCOPE; 1225 1226 if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) { 1227 spin_lock_bh(&inputq->lock); 1228 if (skb_peek(arrvq) == skb) { 1229 __skb_dequeue(arrvq); 1230 __skb_queue_tail(inputq, skb); 1231 } 1232 kfree_skb(skb); 1233 spin_unlock_bh(&inputq->lock); 1234 continue; 1235 } 1236 1237 /* Group messages require exact scope match */ 1238 if (msg_in_group(hdr)) { 1239 ua.sr.lower = 0; 1240 ua.sr.upper = ~0; 1241 ua.scope = msg_lookup_scope(hdr); 1242 } 1243 1244 /* Create destination port list: */ 1245 tipc_nametbl_lookup_mcast_sockets(net, &ua, &dports); 1246 1247 /* Clone message per destination */ 1248 while (tipc_dest_pop(&dports, NULL, &portid)) { 1249 _skb = __pskb_copy(skb, hlen, GFP_ATOMIC); 1250 if (_skb) { 1251 msg_set_destport(buf_msg(_skb), portid); 1252 __skb_queue_tail(&tmpq, _skb); 1253 continue; 1254 } 1255 pr_warn("Failed to clone mcast rcv buffer\n"); 1256 } 1257 /* Append clones to inputq only if skb is still head of arrvq */ 1258 spin_lock_bh(&inputq->lock); 1259 if (skb_peek(arrvq) == skb) { 1260 skb_queue_splice_tail_init(&tmpq, inputq); 1261 /* Decrement the skb's refcnt */ 1262 kfree_skb(__skb_dequeue(arrvq)); 1263 } 1264 spin_unlock_bh(&inputq->lock); 1265 __skb_queue_purge(&tmpq); 1266 kfree_skb(skb); 1267 } 1268 tipc_sk_rcv(net, inputq); 1269 } 1270 1271 /* tipc_sk_push_backlog(): send accumulated buffers in socket write queue 1272 * when socket is in Nagle mode 1273 */ 1274 static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack) 1275 { 1276 struct sk_buff_head *txq = &tsk->sk.sk_write_queue; 1277 struct sk_buff *skb = skb_peek_tail(txq); 1278 struct net *net = sock_net(&tsk->sk); 1279 u32 dnode = tsk_peer_node(tsk); 1280 int rc; 1281 1282 if (nagle_ack) { 1283 tsk->pkt_cnt += skb_queue_len(txq); 1284 if (!tsk->pkt_cnt || tsk->msg_acc / tsk->pkt_cnt < 2) { 1285 tsk->oneway = 0; 1286 if (tsk->nagle_start < NAGLE_START_MAX) 1287 tsk->nagle_start *= 2; 1288 tsk->expect_ack = false; 1289 pr_debug("tsk %10u: bad nagle %u -> %u, next start %u!\n", 1290 tsk->portid, tsk->msg_acc, tsk->pkt_cnt, 1291 tsk->nagle_start); 1292 } else { 1293 tsk->nagle_start = NAGLE_START_INIT; 1294 if (skb) { 1295 msg_set_ack_required(buf_msg(skb)); 1296 tsk->expect_ack = true; 1297 } else { 1298 tsk->expect_ack = false; 1299 } 1300 } 1301 tsk->msg_acc = 0; 1302 tsk->pkt_cnt = 0; 1303 } 1304 1305 if (!skb || tsk->cong_link_cnt) 1306 return; 1307 1308 /* Do not send SYN again after congestion */ 1309 if (msg_is_syn(buf_msg(skb))) 1310 return; 1311 1312 if (tsk->msg_acc) 1313 tsk->pkt_cnt += skb_queue_len(txq); 1314 tsk->snt_unacked += tsk->snd_backlog; 1315 tsk->snd_backlog = 0; 1316 rc = tipc_node_xmit(net, txq, dnode, tsk->portid); 1317 if (rc == -ELINKCONG) 1318 tsk->cong_link_cnt = 1; 1319 } 1320 1321 /** 1322 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message 1323 * @tsk: receiving socket 1324 * @skb: pointer to message buffer. 1325 * @inputq: buffer list containing the buffers 1326 * @xmitq: output message area 1327 */ 1328 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb, 1329 struct sk_buff_head *inputq, 1330 struct sk_buff_head *xmitq) 1331 { 1332 struct tipc_msg *hdr = buf_msg(skb); 1333 u32 onode = tsk_own_node(tsk); 1334 struct sock *sk = &tsk->sk; 1335 int mtyp = msg_type(hdr); 1336 bool was_cong; 1337 1338 /* Ignore if connection cannot be validated: */ 1339 if (!tsk_peer_msg(tsk, hdr)) { 1340 trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!"); 1341 goto exit; 1342 } 1343 1344 if (unlikely(msg_errcode(hdr))) { 1345 tipc_set_sk_state(sk, TIPC_DISCONNECTING); 1346 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk), 1347 tsk_peer_port(tsk)); 1348 sk->sk_state_change(sk); 1349 1350 /* State change is ignored if socket already awake, 1351 * - convert msg to abort msg and add to inqueue 1352 */ 1353 msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE); 1354 msg_set_type(hdr, TIPC_CONN_MSG); 1355 msg_set_size(hdr, BASIC_H_SIZE); 1356 msg_set_hdr_sz(hdr, BASIC_H_SIZE); 1357 __skb_queue_tail(inputq, skb); 1358 return; 1359 } 1360 1361 tsk->probe_unacked = false; 1362 1363 if (mtyp == CONN_PROBE) { 1364 msg_set_type(hdr, CONN_PROBE_REPLY); 1365 if (tipc_msg_reverse(onode, &skb, TIPC_OK)) 1366 __skb_queue_tail(xmitq, skb); 1367 return; 1368 } else if (mtyp == CONN_ACK) { 1369 was_cong = tsk_conn_cong(tsk); 1370 tipc_sk_push_backlog(tsk, msg_nagle_ack(hdr)); 1371 tsk->snt_unacked -= msg_conn_ack(hdr); 1372 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) 1373 tsk->snd_win = msg_adv_win(hdr); 1374 if (was_cong && !tsk_conn_cong(tsk)) 1375 sk->sk_write_space(sk); 1376 } else if (mtyp != CONN_PROBE_REPLY) { 1377 pr_warn("Received unknown CONN_PROTO msg\n"); 1378 } 1379 exit: 1380 kfree_skb(skb); 1381 } 1382 1383 /** 1384 * tipc_sendmsg - send message in connectionless manner 1385 * @sock: socket structure 1386 * @m: message to send 1387 * @dsz: amount of user data to be sent 1388 * 1389 * Message must have an destination specified explicitly. 1390 * Used for SOCK_RDM and SOCK_DGRAM messages, 1391 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections. 1392 * (Note: 'SYN+' is prohibited on SOCK_STREAM.) 1393 * 1394 * Return: the number of bytes sent on success, or errno otherwise 1395 */ 1396 static int tipc_sendmsg(struct socket *sock, 1397 struct msghdr *m, size_t dsz) 1398 { 1399 struct sock *sk = sock->sk; 1400 int ret; 1401 1402 lock_sock(sk); 1403 ret = __tipc_sendmsg(sock, m, dsz); 1404 release_sock(sk); 1405 1406 return ret; 1407 } 1408 1409 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen) 1410 { 1411 struct sock *sk = sock->sk; 1412 struct net *net = sock_net(sk); 1413 struct tipc_sock *tsk = tipc_sk(sk); 1414 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name; 1415 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT); 1416 struct list_head *clinks = &tsk->cong_links; 1417 bool syn = !tipc_sk_type_connectionless(sk); 1418 struct tipc_group *grp = tsk->group; 1419 struct tipc_msg *hdr = &tsk->phdr; 1420 struct tipc_socket_addr skaddr; 1421 struct sk_buff_head pkts; 1422 int atype, mtu, rc; 1423 1424 if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE)) 1425 return -EMSGSIZE; 1426 1427 if (ua) { 1428 if (!tipc_uaddr_valid(ua, m->msg_namelen)) 1429 return -EINVAL; 1430 atype = ua->addrtype; 1431 } 1432 1433 /* If socket belongs to a communication group follow other paths */ 1434 if (grp) { 1435 if (!ua) 1436 return tipc_send_group_bcast(sock, m, dlen, timeout); 1437 if (atype == TIPC_SERVICE_ADDR) 1438 return tipc_send_group_anycast(sock, m, dlen, timeout); 1439 if (atype == TIPC_SOCKET_ADDR) 1440 return tipc_send_group_unicast(sock, m, dlen, timeout); 1441 if (atype == TIPC_SERVICE_RANGE) 1442 return tipc_send_group_mcast(sock, m, dlen, timeout); 1443 return -EINVAL; 1444 } 1445 1446 if (!ua) { 1447 ua = (struct tipc_uaddr *)&tsk->peer; 1448 if (!syn && ua->family != AF_TIPC) 1449 return -EDESTADDRREQ; 1450 atype = ua->addrtype; 1451 } 1452 1453 if (unlikely(syn)) { 1454 if (sk->sk_state == TIPC_LISTEN) 1455 return -EPIPE; 1456 if (sk->sk_state != TIPC_OPEN) 1457 return -EISCONN; 1458 if (tsk->published) 1459 return -EOPNOTSUPP; 1460 if (atype == TIPC_SERVICE_ADDR) 1461 tsk->conn_addrtype = atype; 1462 msg_set_syn(hdr, 1); 1463 } 1464 1465 memset(&skaddr, 0, sizeof(skaddr)); 1466 1467 /* Determine destination */ 1468 if (atype == TIPC_SERVICE_RANGE) { 1469 return tipc_sendmcast(sock, ua, m, dlen, timeout); 1470 } else if (atype == TIPC_SERVICE_ADDR) { 1471 skaddr.node = ua->lookup_node; 1472 ua->scope = tipc_node2scope(skaddr.node); 1473 if (!tipc_nametbl_lookup_anycast(net, ua, &skaddr)) 1474 return -EHOSTUNREACH; 1475 } else if (atype == TIPC_SOCKET_ADDR) { 1476 skaddr = ua->sk; 1477 } else { 1478 return -EINVAL; 1479 } 1480 1481 /* Block or return if destination link is congested */ 1482 rc = tipc_wait_for_cond(sock, &timeout, 1483 !tipc_dest_find(clinks, skaddr.node, 0)); 1484 if (unlikely(rc)) 1485 return rc; 1486 1487 /* Finally build message header */ 1488 msg_set_destnode(hdr, skaddr.node); 1489 msg_set_destport(hdr, skaddr.ref); 1490 if (atype == TIPC_SERVICE_ADDR) { 1491 msg_set_type(hdr, TIPC_NAMED_MSG); 1492 msg_set_hdr_sz(hdr, NAMED_H_SIZE); 1493 msg_set_nametype(hdr, ua->sa.type); 1494 msg_set_nameinst(hdr, ua->sa.instance); 1495 msg_set_lookup_scope(hdr, ua->scope); 1496 } else { /* TIPC_SOCKET_ADDR */ 1497 msg_set_type(hdr, TIPC_DIRECT_MSG); 1498 msg_set_lookup_scope(hdr, 0); 1499 msg_set_hdr_sz(hdr, BASIC_H_SIZE); 1500 } 1501 1502 /* Add message body */ 1503 __skb_queue_head_init(&pkts); 1504 mtu = tipc_node_get_mtu(net, skaddr.node, tsk->portid, true); 1505 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts); 1506 if (unlikely(rc != dlen)) 1507 return rc; 1508 if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue))) { 1509 __skb_queue_purge(&pkts); 1510 return -ENOMEM; 1511 } 1512 1513 /* Send message */ 1514 trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " "); 1515 rc = tipc_node_xmit(net, &pkts, skaddr.node, tsk->portid); 1516 if (unlikely(rc == -ELINKCONG)) { 1517 tipc_dest_push(clinks, skaddr.node, 0); 1518 tsk->cong_link_cnt++; 1519 rc = 0; 1520 } 1521 1522 if (unlikely(syn && !rc)) { 1523 tipc_set_sk_state(sk, TIPC_CONNECTING); 1524 if (dlen && timeout) { 1525 timeout = msecs_to_jiffies(timeout); 1526 tipc_wait_for_connect(sock, &timeout); 1527 } 1528 } 1529 1530 return rc ? rc : dlen; 1531 } 1532 1533 /** 1534 * tipc_sendstream - send stream-oriented data 1535 * @sock: socket structure 1536 * @m: data to send 1537 * @dsz: total length of data to be transmitted 1538 * 1539 * Used for SOCK_STREAM data. 1540 * 1541 * Return: the number of bytes sent on success (or partial success), 1542 * or errno if no data sent 1543 */ 1544 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz) 1545 { 1546 struct sock *sk = sock->sk; 1547 int ret; 1548 1549 lock_sock(sk); 1550 ret = __tipc_sendstream(sock, m, dsz); 1551 release_sock(sk); 1552 1553 return ret; 1554 } 1555 1556 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen) 1557 { 1558 struct sock *sk = sock->sk; 1559 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name); 1560 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT); 1561 struct sk_buff_head *txq = &sk->sk_write_queue; 1562 struct tipc_sock *tsk = tipc_sk(sk); 1563 struct tipc_msg *hdr = &tsk->phdr; 1564 struct net *net = sock_net(sk); 1565 struct sk_buff *skb; 1566 u32 dnode = tsk_peer_node(tsk); 1567 int maxnagle = tsk->maxnagle; 1568 int maxpkt = tsk->max_pkt; 1569 int send, sent = 0; 1570 int blocks, rc = 0; 1571 1572 if (unlikely(dlen > INT_MAX)) 1573 return -EMSGSIZE; 1574 1575 /* Handle implicit connection setup */ 1576 if (unlikely(dest && sk->sk_state == TIPC_OPEN)) { 1577 rc = __tipc_sendmsg(sock, m, dlen); 1578 if (dlen && dlen == rc) { 1579 tsk->peer_caps = tipc_node_get_capabilities(net, dnode); 1580 tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr)); 1581 } 1582 return rc; 1583 } 1584 1585 do { 1586 rc = tipc_wait_for_cond(sock, &timeout, 1587 (!tsk->cong_link_cnt && 1588 !tsk_conn_cong(tsk) && 1589 tipc_sk_connected(sk))); 1590 if (unlikely(rc)) 1591 break; 1592 send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE); 1593 blocks = tsk->snd_backlog; 1594 if (tsk->oneway++ >= tsk->nagle_start && maxnagle && 1595 send <= maxnagle) { 1596 rc = tipc_msg_append(hdr, m, send, maxnagle, txq); 1597 if (unlikely(rc < 0)) 1598 break; 1599 blocks += rc; 1600 tsk->msg_acc++; 1601 if (blocks <= 64 && tsk->expect_ack) { 1602 tsk->snd_backlog = blocks; 1603 sent += send; 1604 break; 1605 } else if (blocks > 64) { 1606 tsk->pkt_cnt += skb_queue_len(txq); 1607 } else { 1608 skb = skb_peek_tail(txq); 1609 if (skb) { 1610 msg_set_ack_required(buf_msg(skb)); 1611 tsk->expect_ack = true; 1612 } else { 1613 tsk->expect_ack = false; 1614 } 1615 tsk->msg_acc = 0; 1616 tsk->pkt_cnt = 0; 1617 } 1618 } else { 1619 rc = tipc_msg_build(hdr, m, sent, send, maxpkt, txq); 1620 if (unlikely(rc != send)) 1621 break; 1622 blocks += tsk_inc(tsk, send + MIN_H_SIZE); 1623 } 1624 trace_tipc_sk_sendstream(sk, skb_peek(txq), 1625 TIPC_DUMP_SK_SNDQ, " "); 1626 rc = tipc_node_xmit(net, txq, dnode, tsk->portid); 1627 if (unlikely(rc == -ELINKCONG)) { 1628 tsk->cong_link_cnt = 1; 1629 rc = 0; 1630 } 1631 if (likely(!rc)) { 1632 tsk->snt_unacked += blocks; 1633 tsk->snd_backlog = 0; 1634 sent += send; 1635 } 1636 } while (sent < dlen && !rc); 1637 1638 return sent ? sent : rc; 1639 } 1640 1641 /** 1642 * tipc_send_packet - send a connection-oriented message 1643 * @sock: socket structure 1644 * @m: message to send 1645 * @dsz: length of data to be transmitted 1646 * 1647 * Used for SOCK_SEQPACKET messages. 1648 * 1649 * Return: the number of bytes sent on success, or errno otherwise 1650 */ 1651 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz) 1652 { 1653 if (dsz > TIPC_MAX_USER_MSG_SIZE) 1654 return -EMSGSIZE; 1655 1656 return tipc_sendstream(sock, m, dsz); 1657 } 1658 1659 /* tipc_sk_finish_conn - complete the setup of a connection 1660 */ 1661 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port, 1662 u32 peer_node) 1663 { 1664 struct sock *sk = &tsk->sk; 1665 struct net *net = sock_net(sk); 1666 struct tipc_msg *msg = &tsk->phdr; 1667 1668 msg_set_syn(msg, 0); 1669 msg_set_destnode(msg, peer_node); 1670 msg_set_destport(msg, peer_port); 1671 msg_set_type(msg, TIPC_CONN_MSG); 1672 msg_set_lookup_scope(msg, 0); 1673 msg_set_hdr_sz(msg, SHORT_H_SIZE); 1674 1675 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV); 1676 tipc_set_sk_state(sk, TIPC_ESTABLISHED); 1677 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port); 1678 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid, true); 1679 tsk->peer_caps = tipc_node_get_capabilities(net, peer_node); 1680 tsk_set_nagle(tsk); 1681 __skb_queue_purge(&sk->sk_write_queue); 1682 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) 1683 return; 1684 1685 /* Fall back to message based flow control */ 1686 tsk->rcv_win = FLOWCTL_MSG_WIN; 1687 tsk->snd_win = FLOWCTL_MSG_WIN; 1688 } 1689 1690 /** 1691 * tipc_sk_set_orig_addr - capture sender's address for received message 1692 * @m: descriptor for message info 1693 * @skb: received message 1694 * 1695 * Note: Address is not captured if not requested by receiver. 1696 */ 1697 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb) 1698 { 1699 DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name); 1700 struct tipc_msg *hdr = buf_msg(skb); 1701 1702 if (!srcaddr) 1703 return; 1704 1705 srcaddr->sock.family = AF_TIPC; 1706 srcaddr->sock.addrtype = TIPC_SOCKET_ADDR; 1707 srcaddr->sock.scope = 0; 1708 srcaddr->sock.addr.id.ref = msg_origport(hdr); 1709 srcaddr->sock.addr.id.node = msg_orignode(hdr); 1710 srcaddr->sock.addr.name.domain = 0; 1711 m->msg_namelen = sizeof(struct sockaddr_tipc); 1712 1713 if (!msg_in_group(hdr)) 1714 return; 1715 1716 /* Group message users may also want to know sending member's id */ 1717 srcaddr->member.family = AF_TIPC; 1718 srcaddr->member.addrtype = TIPC_SERVICE_ADDR; 1719 srcaddr->member.scope = 0; 1720 srcaddr->member.addr.name.name.type = msg_nametype(hdr); 1721 srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member; 1722 srcaddr->member.addr.name.domain = 0; 1723 m->msg_namelen = sizeof(*srcaddr); 1724 } 1725 1726 /** 1727 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message 1728 * @m: descriptor for message info 1729 * @skb: received message buffer 1730 * @tsk: TIPC port associated with message 1731 * 1732 * Note: Ancillary data is not captured if not requested by receiver. 1733 * 1734 * Return: 0 if successful, otherwise errno 1735 */ 1736 static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb, 1737 struct tipc_sock *tsk) 1738 { 1739 struct tipc_msg *hdr; 1740 u32 data[3] = {0,}; 1741 bool has_addr; 1742 int dlen, rc; 1743 1744 if (likely(m->msg_controllen == 0)) 1745 return 0; 1746 1747 hdr = buf_msg(skb); 1748 dlen = msg_data_sz(hdr); 1749 1750 /* Capture errored message object, if any */ 1751 if (msg_errcode(hdr)) { 1752 if (skb_linearize(skb)) 1753 return -ENOMEM; 1754 hdr = buf_msg(skb); 1755 data[0] = msg_errcode(hdr); 1756 data[1] = dlen; 1757 rc = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, data); 1758 if (rc || !dlen) 1759 return rc; 1760 rc = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, dlen, msg_data(hdr)); 1761 if (rc) 1762 return rc; 1763 } 1764 1765 /* Capture TIPC_SERVICE_ADDR/RANGE destination address, if any */ 1766 switch (msg_type(hdr)) { 1767 case TIPC_NAMED_MSG: 1768 has_addr = true; 1769 data[0] = msg_nametype(hdr); 1770 data[1] = msg_namelower(hdr); 1771 data[2] = data[1]; 1772 break; 1773 case TIPC_MCAST_MSG: 1774 has_addr = true; 1775 data[0] = msg_nametype(hdr); 1776 data[1] = msg_namelower(hdr); 1777 data[2] = msg_nameupper(hdr); 1778 break; 1779 case TIPC_CONN_MSG: 1780 has_addr = !!tsk->conn_addrtype; 1781 data[0] = msg_nametype(&tsk->phdr); 1782 data[1] = msg_nameinst(&tsk->phdr); 1783 data[2] = data[1]; 1784 break; 1785 default: 1786 has_addr = false; 1787 } 1788 if (!has_addr) 1789 return 0; 1790 return put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, data); 1791 } 1792 1793 static struct sk_buff *tipc_sk_build_ack(struct tipc_sock *tsk) 1794 { 1795 struct sock *sk = &tsk->sk; 1796 struct sk_buff *skb = NULL; 1797 struct tipc_msg *msg; 1798 u32 peer_port = tsk_peer_port(tsk); 1799 u32 dnode = tsk_peer_node(tsk); 1800 1801 if (!tipc_sk_connected(sk)) 1802 return NULL; 1803 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0, 1804 dnode, tsk_own_node(tsk), peer_port, 1805 tsk->portid, TIPC_OK); 1806 if (!skb) 1807 return NULL; 1808 msg = buf_msg(skb); 1809 msg_set_conn_ack(msg, tsk->rcv_unacked); 1810 tsk->rcv_unacked = 0; 1811 1812 /* Adjust to and advertize the correct window limit */ 1813 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) { 1814 tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf); 1815 msg_set_adv_win(msg, tsk->rcv_win); 1816 } 1817 return skb; 1818 } 1819 1820 static void tipc_sk_send_ack(struct tipc_sock *tsk) 1821 { 1822 struct sk_buff *skb; 1823 1824 skb = tipc_sk_build_ack(tsk); 1825 if (!skb) 1826 return; 1827 1828 tipc_node_xmit_skb(sock_net(&tsk->sk), skb, tsk_peer_node(tsk), 1829 msg_link_selector(buf_msg(skb))); 1830 } 1831 1832 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop) 1833 { 1834 struct sock *sk = sock->sk; 1835 DEFINE_WAIT_FUNC(wait, woken_wake_function); 1836 long timeo = *timeop; 1837 int err = sock_error(sk); 1838 1839 if (err) 1840 return err; 1841 1842 for (;;) { 1843 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) { 1844 if (sk->sk_shutdown & RCV_SHUTDOWN) { 1845 err = -ENOTCONN; 1846 break; 1847 } 1848 add_wait_queue(sk_sleep(sk), &wait); 1849 release_sock(sk); 1850 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo); 1851 sched_annotate_sleep(); 1852 lock_sock(sk); 1853 remove_wait_queue(sk_sleep(sk), &wait); 1854 } 1855 err = 0; 1856 if (!skb_queue_empty(&sk->sk_receive_queue)) 1857 break; 1858 err = -EAGAIN; 1859 if (!timeo) 1860 break; 1861 err = sock_intr_errno(timeo); 1862 if (signal_pending(current)) 1863 break; 1864 1865 err = sock_error(sk); 1866 if (err) 1867 break; 1868 } 1869 *timeop = timeo; 1870 return err; 1871 } 1872 1873 /** 1874 * tipc_recvmsg - receive packet-oriented message 1875 * @sock: network socket 1876 * @m: descriptor for message info 1877 * @buflen: length of user buffer area 1878 * @flags: receive flags 1879 * 1880 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages. 1881 * If the complete message doesn't fit in user area, truncate it. 1882 * 1883 * Return: size of returned message data, errno otherwise 1884 */ 1885 static int tipc_recvmsg(struct socket *sock, struct msghdr *m, 1886 size_t buflen, int flags) 1887 { 1888 struct sock *sk = sock->sk; 1889 bool connected = !tipc_sk_type_connectionless(sk); 1890 struct tipc_sock *tsk = tipc_sk(sk); 1891 int rc, err, hlen, dlen, copy; 1892 struct tipc_skb_cb *skb_cb; 1893 struct sk_buff_head xmitq; 1894 struct tipc_msg *hdr; 1895 struct sk_buff *skb; 1896 bool grp_evt; 1897 long timeout; 1898 1899 /* Catch invalid receive requests */ 1900 if (unlikely(!buflen)) 1901 return -EINVAL; 1902 1903 lock_sock(sk); 1904 if (unlikely(connected && sk->sk_state == TIPC_OPEN)) { 1905 rc = -ENOTCONN; 1906 goto exit; 1907 } 1908 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT); 1909 1910 /* Step rcv queue to first msg with data or error; wait if necessary */ 1911 do { 1912 rc = tipc_wait_for_rcvmsg(sock, &timeout); 1913 if (unlikely(rc)) 1914 goto exit; 1915 skb = skb_peek(&sk->sk_receive_queue); 1916 skb_cb = TIPC_SKB_CB(skb); 1917 hdr = buf_msg(skb); 1918 dlen = msg_data_sz(hdr); 1919 hlen = msg_hdr_sz(hdr); 1920 err = msg_errcode(hdr); 1921 grp_evt = msg_is_grp_evt(hdr); 1922 if (likely(dlen || err)) 1923 break; 1924 tsk_advance_rx_queue(sk); 1925 } while (1); 1926 1927 /* Collect msg meta data, including error code and rejected data */ 1928 tipc_sk_set_orig_addr(m, skb); 1929 rc = tipc_sk_anc_data_recv(m, skb, tsk); 1930 if (unlikely(rc)) 1931 goto exit; 1932 hdr = buf_msg(skb); 1933 1934 /* Capture data if non-error msg, otherwise just set return value */ 1935 if (likely(!err)) { 1936 int offset = skb_cb->bytes_read; 1937 1938 copy = min_t(int, dlen - offset, buflen); 1939 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy); 1940 if (unlikely(rc)) 1941 goto exit; 1942 if (unlikely(offset + copy < dlen)) { 1943 if (flags & MSG_EOR) { 1944 if (!(flags & MSG_PEEK)) 1945 skb_cb->bytes_read = offset + copy; 1946 } else { 1947 m->msg_flags |= MSG_TRUNC; 1948 skb_cb->bytes_read = 0; 1949 } 1950 } else { 1951 if (flags & MSG_EOR) 1952 m->msg_flags |= MSG_EOR; 1953 skb_cb->bytes_read = 0; 1954 } 1955 } else { 1956 copy = 0; 1957 rc = 0; 1958 if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control) { 1959 rc = -ECONNRESET; 1960 goto exit; 1961 } 1962 } 1963 1964 /* Mark message as group event if applicable */ 1965 if (unlikely(grp_evt)) { 1966 if (msg_grp_evt(hdr) == TIPC_WITHDRAWN) 1967 m->msg_flags |= MSG_EOR; 1968 m->msg_flags |= MSG_OOB; 1969 copy = 0; 1970 } 1971 1972 /* Caption of data or error code/rejected data was successful */ 1973 if (unlikely(flags & MSG_PEEK)) 1974 goto exit; 1975 1976 /* Send group flow control advertisement when applicable */ 1977 if (tsk->group && msg_in_group(hdr) && !grp_evt) { 1978 __skb_queue_head_init(&xmitq); 1979 tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen), 1980 msg_orignode(hdr), msg_origport(hdr), 1981 &xmitq); 1982 tipc_node_distr_xmit(sock_net(sk), &xmitq); 1983 } 1984 1985 if (skb_cb->bytes_read) 1986 goto exit; 1987 1988 tsk_advance_rx_queue(sk); 1989 1990 if (likely(!connected)) 1991 goto exit; 1992 1993 /* Send connection flow control advertisement when applicable */ 1994 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen); 1995 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE) 1996 tipc_sk_send_ack(tsk); 1997 exit: 1998 release_sock(sk); 1999 return rc ? rc : copy; 2000 } 2001 2002 /** 2003 * tipc_recvstream - receive stream-oriented data 2004 * @sock: network socket 2005 * @m: descriptor for message info 2006 * @buflen: total size of user buffer area 2007 * @flags: receive flags 2008 * 2009 * Used for SOCK_STREAM messages only. If not enough data is available 2010 * will optionally wait for more; never truncates data. 2011 * 2012 * Return: size of returned message data, errno otherwise 2013 */ 2014 static int tipc_recvstream(struct socket *sock, struct msghdr *m, 2015 size_t buflen, int flags) 2016 { 2017 struct sock *sk = sock->sk; 2018 struct tipc_sock *tsk = tipc_sk(sk); 2019 struct sk_buff *skb; 2020 struct tipc_msg *hdr; 2021 struct tipc_skb_cb *skb_cb; 2022 bool peek = flags & MSG_PEEK; 2023 int offset, required, copy, copied = 0; 2024 int hlen, dlen, err, rc; 2025 long timeout; 2026 2027 /* Catch invalid receive attempts */ 2028 if (unlikely(!buflen)) 2029 return -EINVAL; 2030 2031 lock_sock(sk); 2032 2033 if (unlikely(sk->sk_state == TIPC_OPEN)) { 2034 rc = -ENOTCONN; 2035 goto exit; 2036 } 2037 required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen); 2038 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT); 2039 2040 do { 2041 /* Look at first msg in receive queue; wait if necessary */ 2042 rc = tipc_wait_for_rcvmsg(sock, &timeout); 2043 if (unlikely(rc)) 2044 break; 2045 skb = skb_peek(&sk->sk_receive_queue); 2046 skb_cb = TIPC_SKB_CB(skb); 2047 hdr = buf_msg(skb); 2048 dlen = msg_data_sz(hdr); 2049 hlen = msg_hdr_sz(hdr); 2050 err = msg_errcode(hdr); 2051 2052 /* Discard any empty non-errored (SYN-) message */ 2053 if (unlikely(!dlen && !err)) { 2054 tsk_advance_rx_queue(sk); 2055 continue; 2056 } 2057 2058 /* Collect msg meta data, incl. error code and rejected data */ 2059 if (!copied) { 2060 tipc_sk_set_orig_addr(m, skb); 2061 rc = tipc_sk_anc_data_recv(m, skb, tsk); 2062 if (rc) 2063 break; 2064 hdr = buf_msg(skb); 2065 } 2066 2067 /* Copy data if msg ok, otherwise return error/partial data */ 2068 if (likely(!err)) { 2069 offset = skb_cb->bytes_read; 2070 copy = min_t(int, dlen - offset, buflen - copied); 2071 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy); 2072 if (unlikely(rc)) 2073 break; 2074 copied += copy; 2075 offset += copy; 2076 if (unlikely(offset < dlen)) { 2077 if (!peek) 2078 skb_cb->bytes_read = offset; 2079 break; 2080 } 2081 } else { 2082 rc = 0; 2083 if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control) 2084 rc = -ECONNRESET; 2085 if (copied || rc) 2086 break; 2087 } 2088 2089 if (unlikely(peek)) 2090 break; 2091 2092 tsk_advance_rx_queue(sk); 2093 2094 /* Send connection flow control advertisement when applicable */ 2095 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen); 2096 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE) 2097 tipc_sk_send_ack(tsk); 2098 2099 /* Exit if all requested data or FIN/error received */ 2100 if (copied == buflen || err) 2101 break; 2102 2103 } while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required); 2104 exit: 2105 release_sock(sk); 2106 return copied ? copied : rc; 2107 } 2108 2109 /** 2110 * tipc_write_space - wake up thread if port congestion is released 2111 * @sk: socket 2112 */ 2113 static void tipc_write_space(struct sock *sk) 2114 { 2115 struct socket_wq *wq; 2116 2117 rcu_read_lock(); 2118 wq = rcu_dereference(sk->sk_wq); 2119 if (skwq_has_sleeper(wq)) 2120 wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT | 2121 EPOLLWRNORM | EPOLLWRBAND); 2122 rcu_read_unlock(); 2123 } 2124 2125 /** 2126 * tipc_data_ready - wake up threads to indicate messages have been received 2127 * @sk: socket 2128 */ 2129 static void tipc_data_ready(struct sock *sk) 2130 { 2131 struct socket_wq *wq; 2132 2133 trace_sk_data_ready(sk); 2134 2135 rcu_read_lock(); 2136 wq = rcu_dereference(sk->sk_wq); 2137 if (skwq_has_sleeper(wq)) 2138 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN | 2139 EPOLLRDNORM | EPOLLRDBAND); 2140 rcu_read_unlock(); 2141 } 2142 2143 static void tipc_sock_destruct(struct sock *sk) 2144 { 2145 __skb_queue_purge(&sk->sk_receive_queue); 2146 } 2147 2148 static void tipc_sk_proto_rcv(struct sock *sk, 2149 struct sk_buff_head *inputq, 2150 struct sk_buff_head *xmitq) 2151 { 2152 struct sk_buff *skb = __skb_dequeue(inputq); 2153 struct tipc_sock *tsk = tipc_sk(sk); 2154 struct tipc_msg *hdr = buf_msg(skb); 2155 struct tipc_group *grp = tsk->group; 2156 bool wakeup = false; 2157 2158 switch (msg_user(hdr)) { 2159 case CONN_MANAGER: 2160 tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq); 2161 return; 2162 case SOCK_WAKEUP: 2163 tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0); 2164 /* coupled with smp_rmb() in tipc_wait_for_cond() */ 2165 smp_wmb(); 2166 tsk->cong_link_cnt--; 2167 wakeup = true; 2168 tipc_sk_push_backlog(tsk, false); 2169 break; 2170 case GROUP_PROTOCOL: 2171 tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq); 2172 break; 2173 case TOP_SRV: 2174 tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf, 2175 hdr, inputq, xmitq); 2176 break; 2177 default: 2178 break; 2179 } 2180 2181 if (wakeup) 2182 sk->sk_write_space(sk); 2183 2184 kfree_skb(skb); 2185 } 2186 2187 /** 2188 * tipc_sk_filter_connect - check incoming message for a connection-based socket 2189 * @tsk: TIPC socket 2190 * @skb: pointer to message buffer. 2191 * @xmitq: for Nagle ACK if any 2192 * Return: true if message should be added to receive queue, false otherwise 2193 */ 2194 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb, 2195 struct sk_buff_head *xmitq) 2196 { 2197 struct sock *sk = &tsk->sk; 2198 struct net *net = sock_net(sk); 2199 struct tipc_msg *hdr = buf_msg(skb); 2200 bool con_msg = msg_connected(hdr); 2201 u32 pport = tsk_peer_port(tsk); 2202 u32 pnode = tsk_peer_node(tsk); 2203 u32 oport = msg_origport(hdr); 2204 u32 onode = msg_orignode(hdr); 2205 int err = msg_errcode(hdr); 2206 unsigned long delay; 2207 2208 if (unlikely(msg_mcast(hdr))) 2209 return false; 2210 tsk->oneway = 0; 2211 2212 switch (sk->sk_state) { 2213 case TIPC_CONNECTING: 2214 /* Setup ACK */ 2215 if (likely(con_msg)) { 2216 if (err) 2217 break; 2218 tipc_sk_finish_conn(tsk, oport, onode); 2219 msg_set_importance(&tsk->phdr, msg_importance(hdr)); 2220 /* ACK+ message with data is added to receive queue */ 2221 if (msg_data_sz(hdr)) 2222 return true; 2223 /* Empty ACK-, - wake up sleeping connect() and drop */ 2224 sk->sk_state_change(sk); 2225 msg_set_dest_droppable(hdr, 1); 2226 return false; 2227 } 2228 /* Ignore connectionless message if not from listening socket */ 2229 if (oport != pport || onode != pnode) 2230 return false; 2231 2232 /* Rejected SYN */ 2233 if (err != TIPC_ERR_OVERLOAD) 2234 break; 2235 2236 /* Prepare for new setup attempt if we have a SYN clone */ 2237 if (skb_queue_empty(&sk->sk_write_queue)) 2238 break; 2239 get_random_bytes(&delay, 2); 2240 delay %= (tsk->conn_timeout / 4); 2241 delay = msecs_to_jiffies(delay + 100); 2242 sk_reset_timer(sk, &sk->sk_timer, jiffies + delay); 2243 return false; 2244 case TIPC_OPEN: 2245 case TIPC_DISCONNECTING: 2246 return false; 2247 case TIPC_LISTEN: 2248 /* Accept only SYN message */ 2249 if (!msg_is_syn(hdr) && 2250 tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT) 2251 return false; 2252 if (!con_msg && !err) 2253 return true; 2254 return false; 2255 case TIPC_ESTABLISHED: 2256 if (!skb_queue_empty(&sk->sk_write_queue)) 2257 tipc_sk_push_backlog(tsk, false); 2258 /* Accept only connection-based messages sent by peer */ 2259 if (likely(con_msg && !err && pport == oport && 2260 pnode == onode)) { 2261 if (msg_ack_required(hdr)) { 2262 struct sk_buff *skb; 2263 2264 skb = tipc_sk_build_ack(tsk); 2265 if (skb) { 2266 msg_set_nagle_ack(buf_msg(skb)); 2267 __skb_queue_tail(xmitq, skb); 2268 } 2269 } 2270 return true; 2271 } 2272 if (!tsk_peer_msg(tsk, hdr)) 2273 return false; 2274 if (!err) 2275 return true; 2276 tipc_set_sk_state(sk, TIPC_DISCONNECTING); 2277 tipc_node_remove_conn(net, pnode, tsk->portid); 2278 sk->sk_state_change(sk); 2279 return true; 2280 default: 2281 pr_err("Unknown sk_state %u\n", sk->sk_state); 2282 } 2283 /* Abort connection setup attempt */ 2284 tipc_set_sk_state(sk, TIPC_DISCONNECTING); 2285 sk->sk_err = ECONNREFUSED; 2286 sk->sk_state_change(sk); 2287 return true; 2288 } 2289 2290 /** 2291 * rcvbuf_limit - get proper overload limit of socket receive queue 2292 * @sk: socket 2293 * @skb: message 2294 * 2295 * For connection oriented messages, irrespective of importance, 2296 * default queue limit is 2 MB. 2297 * 2298 * For connectionless messages, queue limits are based on message 2299 * importance as follows: 2300 * 2301 * TIPC_LOW_IMPORTANCE (2 MB) 2302 * TIPC_MEDIUM_IMPORTANCE (4 MB) 2303 * TIPC_HIGH_IMPORTANCE (8 MB) 2304 * TIPC_CRITICAL_IMPORTANCE (16 MB) 2305 * 2306 * Return: overload limit according to corresponding message importance 2307 */ 2308 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb) 2309 { 2310 struct tipc_sock *tsk = tipc_sk(sk); 2311 struct tipc_msg *hdr = buf_msg(skb); 2312 2313 if (unlikely(msg_in_group(hdr))) 2314 return READ_ONCE(sk->sk_rcvbuf); 2315 2316 if (unlikely(!msg_connected(hdr))) 2317 return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr); 2318 2319 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL)) 2320 return READ_ONCE(sk->sk_rcvbuf); 2321 2322 return FLOWCTL_MSG_LIM; 2323 } 2324 2325 /** 2326 * tipc_sk_filter_rcv - validate incoming message 2327 * @sk: socket 2328 * @skb: pointer to message. 2329 * @xmitq: output message area (FIXME) 2330 * 2331 * Enqueues message on receive queue if acceptable; optionally handles 2332 * disconnect indication for a connected socket. 2333 * 2334 * Called with socket lock already taken 2335 */ 2336 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb, 2337 struct sk_buff_head *xmitq) 2338 { 2339 bool sk_conn = !tipc_sk_type_connectionless(sk); 2340 struct tipc_sock *tsk = tipc_sk(sk); 2341 struct tipc_group *grp = tsk->group; 2342 struct tipc_msg *hdr = buf_msg(skb); 2343 struct net *net = sock_net(sk); 2344 struct sk_buff_head inputq; 2345 int mtyp = msg_type(hdr); 2346 int limit, err = TIPC_OK; 2347 2348 trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " "); 2349 TIPC_SKB_CB(skb)->bytes_read = 0; 2350 __skb_queue_head_init(&inputq); 2351 __skb_queue_tail(&inputq, skb); 2352 2353 if (unlikely(!msg_isdata(hdr))) 2354 tipc_sk_proto_rcv(sk, &inputq, xmitq); 2355 2356 if (unlikely(grp)) 2357 tipc_group_filter_msg(grp, &inputq, xmitq); 2358 2359 if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG) 2360 tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq); 2361 2362 /* Validate and add to receive buffer if there is space */ 2363 while ((skb = __skb_dequeue(&inputq))) { 2364 hdr = buf_msg(skb); 2365 limit = rcvbuf_limit(sk, skb); 2366 if ((sk_conn && !tipc_sk_filter_connect(tsk, skb, xmitq)) || 2367 (!sk_conn && msg_connected(hdr)) || 2368 (!grp && msg_in_group(hdr))) 2369 err = TIPC_ERR_NO_PORT; 2370 else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) { 2371 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, 2372 "err_overload2!"); 2373 atomic_inc(&sk->sk_drops); 2374 err = TIPC_ERR_OVERLOAD; 2375 } 2376 2377 if (unlikely(err)) { 2378 if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) { 2379 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, 2380 "@filter_rcv!"); 2381 __skb_queue_tail(xmitq, skb); 2382 } 2383 err = TIPC_OK; 2384 continue; 2385 } 2386 __skb_queue_tail(&sk->sk_receive_queue, skb); 2387 skb_set_owner_r(skb, sk); 2388 trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL, 2389 "rcvq >90% allocated!"); 2390 sk->sk_data_ready(sk); 2391 } 2392 } 2393 2394 /** 2395 * tipc_sk_backlog_rcv - handle incoming message from backlog queue 2396 * @sk: socket 2397 * @skb: message 2398 * 2399 * Caller must hold socket lock 2400 */ 2401 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb) 2402 { 2403 unsigned int before = sk_rmem_alloc_get(sk); 2404 struct sk_buff_head xmitq; 2405 unsigned int added; 2406 2407 __skb_queue_head_init(&xmitq); 2408 2409 tipc_sk_filter_rcv(sk, skb, &xmitq); 2410 added = sk_rmem_alloc_get(sk) - before; 2411 atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt); 2412 2413 /* Send pending response/rejected messages, if any */ 2414 tipc_node_distr_xmit(sock_net(sk), &xmitq); 2415 return 0; 2416 } 2417 2418 /** 2419 * tipc_sk_enqueue - extract all buffers with destination 'dport' from 2420 * inputq and try adding them to socket or backlog queue 2421 * @inputq: list of incoming buffers with potentially different destinations 2422 * @sk: socket where the buffers should be enqueued 2423 * @dport: port number for the socket 2424 * @xmitq: output queue 2425 * 2426 * Caller must hold socket lock 2427 */ 2428 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk, 2429 u32 dport, struct sk_buff_head *xmitq) 2430 { 2431 unsigned long time_limit = jiffies + usecs_to_jiffies(20000); 2432 struct sk_buff *skb; 2433 unsigned int lim; 2434 atomic_t *dcnt; 2435 u32 onode; 2436 2437 while (skb_queue_len(inputq)) { 2438 if (unlikely(time_after_eq(jiffies, time_limit))) 2439 return; 2440 2441 skb = tipc_skb_dequeue(inputq, dport); 2442 if (unlikely(!skb)) 2443 return; 2444 2445 /* Add message directly to receive queue if possible */ 2446 if (!sock_owned_by_user(sk)) { 2447 tipc_sk_filter_rcv(sk, skb, xmitq); 2448 continue; 2449 } 2450 2451 /* Try backlog, compensating for double-counted bytes */ 2452 dcnt = &tipc_sk(sk)->dupl_rcvcnt; 2453 if (!sk->sk_backlog.len) 2454 atomic_set(dcnt, 0); 2455 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt); 2456 if (likely(!sk_add_backlog(sk, skb, lim))) { 2457 trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL, 2458 "bklg & rcvq >90% allocated!"); 2459 continue; 2460 } 2461 2462 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!"); 2463 /* Overload => reject message back to sender */ 2464 onode = tipc_own_addr(sock_net(sk)); 2465 atomic_inc(&sk->sk_drops); 2466 if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) { 2467 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL, 2468 "@sk_enqueue!"); 2469 __skb_queue_tail(xmitq, skb); 2470 } 2471 break; 2472 } 2473 } 2474 2475 /** 2476 * tipc_sk_rcv - handle a chain of incoming buffers 2477 * @net: the associated network namespace 2478 * @inputq: buffer list containing the buffers 2479 * Consumes all buffers in list until inputq is empty 2480 * Note: may be called in multiple threads referring to the same queue 2481 */ 2482 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq) 2483 { 2484 struct sk_buff_head xmitq; 2485 u32 dnode, dport = 0; 2486 int err; 2487 struct tipc_sock *tsk; 2488 struct sock *sk; 2489 struct sk_buff *skb; 2490 2491 __skb_queue_head_init(&xmitq); 2492 while (skb_queue_len(inputq)) { 2493 dport = tipc_skb_peek_port(inputq, dport); 2494 tsk = tipc_sk_lookup(net, dport); 2495 2496 if (likely(tsk)) { 2497 sk = &tsk->sk; 2498 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) { 2499 tipc_sk_enqueue(inputq, sk, dport, &xmitq); 2500 spin_unlock_bh(&sk->sk_lock.slock); 2501 } 2502 /* Send pending response/rejected messages, if any */ 2503 tipc_node_distr_xmit(sock_net(sk), &xmitq); 2504 sock_put(sk); 2505 continue; 2506 } 2507 /* No destination socket => dequeue skb if still there */ 2508 skb = tipc_skb_dequeue(inputq, dport); 2509 if (!skb) 2510 return; 2511 2512 /* Try secondary lookup if unresolved named message */ 2513 err = TIPC_ERR_NO_PORT; 2514 if (tipc_msg_lookup_dest(net, skb, &err)) 2515 goto xmit; 2516 2517 /* Prepare for message rejection */ 2518 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err)) 2519 continue; 2520 2521 trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!"); 2522 xmit: 2523 dnode = msg_destnode(buf_msg(skb)); 2524 tipc_node_xmit_skb(net, skb, dnode, dport); 2525 } 2526 } 2527 2528 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p) 2529 { 2530 DEFINE_WAIT_FUNC(wait, woken_wake_function); 2531 struct sock *sk = sock->sk; 2532 int done; 2533 2534 do { 2535 int err = sock_error(sk); 2536 if (err) 2537 return err; 2538 if (!*timeo_p) 2539 return -ETIMEDOUT; 2540 if (signal_pending(current)) 2541 return sock_intr_errno(*timeo_p); 2542 if (sk->sk_state == TIPC_DISCONNECTING) 2543 break; 2544 2545 add_wait_queue(sk_sleep(sk), &wait); 2546 done = sk_wait_event(sk, timeo_p, tipc_sk_connected(sk), 2547 &wait); 2548 remove_wait_queue(sk_sleep(sk), &wait); 2549 } while (!done); 2550 return 0; 2551 } 2552 2553 static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr) 2554 { 2555 if (addr->family != AF_TIPC) 2556 return false; 2557 if (addr->addrtype == TIPC_SERVICE_RANGE) 2558 return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper); 2559 return (addr->addrtype == TIPC_SERVICE_ADDR || 2560 addr->addrtype == TIPC_SOCKET_ADDR); 2561 } 2562 2563 /** 2564 * tipc_connect - establish a connection to another TIPC port 2565 * @sock: socket structure 2566 * @dest: socket address for destination port 2567 * @destlen: size of socket address data structure 2568 * @flags: file-related flags associated with socket 2569 * 2570 * Return: 0 on success, errno otherwise 2571 */ 2572 static int tipc_connect(struct socket *sock, struct sockaddr *dest, 2573 int destlen, int flags) 2574 { 2575 struct sock *sk = sock->sk; 2576 struct tipc_sock *tsk = tipc_sk(sk); 2577 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest; 2578 struct msghdr m = {NULL,}; 2579 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout; 2580 int previous; 2581 int res = 0; 2582 2583 if (destlen != sizeof(struct sockaddr_tipc)) 2584 return -EINVAL; 2585 2586 lock_sock(sk); 2587 2588 if (tsk->group) { 2589 res = -EINVAL; 2590 goto exit; 2591 } 2592 2593 if (dst->family == AF_UNSPEC) { 2594 memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc)); 2595 if (!tipc_sk_type_connectionless(sk)) 2596 res = -EINVAL; 2597 goto exit; 2598 } 2599 if (!tipc_sockaddr_is_sane(dst)) { 2600 res = -EINVAL; 2601 goto exit; 2602 } 2603 /* DGRAM/RDM connect(), just save the destaddr */ 2604 if (tipc_sk_type_connectionless(sk)) { 2605 memcpy(&tsk->peer, dest, destlen); 2606 goto exit; 2607 } else if (dst->addrtype == TIPC_SERVICE_RANGE) { 2608 res = -EINVAL; 2609 goto exit; 2610 } 2611 2612 previous = sk->sk_state; 2613 2614 switch (sk->sk_state) { 2615 case TIPC_OPEN: 2616 /* Send a 'SYN-' to destination */ 2617 m.msg_name = dest; 2618 m.msg_namelen = destlen; 2619 iov_iter_kvec(&m.msg_iter, ITER_SOURCE, NULL, 0, 0); 2620 2621 /* If connect is in non-blocking case, set MSG_DONTWAIT to 2622 * indicate send_msg() is never blocked. 2623 */ 2624 if (!timeout) 2625 m.msg_flags = MSG_DONTWAIT; 2626 2627 res = __tipc_sendmsg(sock, &m, 0); 2628 if ((res < 0) && (res != -EWOULDBLOCK)) 2629 goto exit; 2630 2631 /* Just entered TIPC_CONNECTING state; the only 2632 * difference is that return value in non-blocking 2633 * case is EINPROGRESS, rather than EALREADY. 2634 */ 2635 res = -EINPROGRESS; 2636 fallthrough; 2637 case TIPC_CONNECTING: 2638 if (!timeout) { 2639 if (previous == TIPC_CONNECTING) 2640 res = -EALREADY; 2641 goto exit; 2642 } 2643 timeout = msecs_to_jiffies(timeout); 2644 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */ 2645 res = tipc_wait_for_connect(sock, &timeout); 2646 break; 2647 case TIPC_ESTABLISHED: 2648 res = -EISCONN; 2649 break; 2650 default: 2651 res = -EINVAL; 2652 } 2653 2654 exit: 2655 release_sock(sk); 2656 return res; 2657 } 2658 2659 /** 2660 * tipc_listen - allow socket to listen for incoming connections 2661 * @sock: socket structure 2662 * @len: (unused) 2663 * 2664 * Return: 0 on success, errno otherwise 2665 */ 2666 static int tipc_listen(struct socket *sock, int len) 2667 { 2668 struct sock *sk = sock->sk; 2669 int res; 2670 2671 lock_sock(sk); 2672 res = tipc_set_sk_state(sk, TIPC_LISTEN); 2673 release_sock(sk); 2674 2675 return res; 2676 } 2677 2678 static int tipc_wait_for_accept(struct socket *sock, long timeo) 2679 { 2680 struct sock *sk = sock->sk; 2681 DEFINE_WAIT_FUNC(wait, woken_wake_function); 2682 int err; 2683 2684 /* True wake-one mechanism for incoming connections: only 2685 * one process gets woken up, not the 'whole herd'. 2686 * Since we do not 'race & poll' for established sockets 2687 * anymore, the common case will execute the loop only once. 2688 */ 2689 for (;;) { 2690 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) { 2691 add_wait_queue(sk_sleep(sk), &wait); 2692 release_sock(sk); 2693 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo); 2694 lock_sock(sk); 2695 remove_wait_queue(sk_sleep(sk), &wait); 2696 } 2697 err = 0; 2698 if (!skb_queue_empty(&sk->sk_receive_queue)) 2699 break; 2700 err = -EAGAIN; 2701 if (!timeo) 2702 break; 2703 err = sock_intr_errno(timeo); 2704 if (signal_pending(current)) 2705 break; 2706 } 2707 return err; 2708 } 2709 2710 /** 2711 * tipc_accept - wait for connection request 2712 * @sock: listening socket 2713 * @new_sock: new socket that is to be connected 2714 * @flags: file-related flags associated with socket 2715 * @kern: caused by kernel or by userspace? 2716 * 2717 * Return: 0 on success, errno otherwise 2718 */ 2719 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags, 2720 bool kern) 2721 { 2722 struct sock *new_sk, *sk = sock->sk; 2723 struct tipc_sock *new_tsock; 2724 struct msghdr m = {NULL,}; 2725 struct tipc_msg *msg; 2726 struct sk_buff *buf; 2727 long timeo; 2728 int res; 2729 2730 lock_sock(sk); 2731 2732 if (sk->sk_state != TIPC_LISTEN) { 2733 res = -EINVAL; 2734 goto exit; 2735 } 2736 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK); 2737 res = tipc_wait_for_accept(sock, timeo); 2738 if (res) 2739 goto exit; 2740 2741 buf = skb_peek(&sk->sk_receive_queue); 2742 2743 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern); 2744 if (res) 2745 goto exit; 2746 security_sk_clone(sock->sk, new_sock->sk); 2747 2748 new_sk = new_sock->sk; 2749 new_tsock = tipc_sk(new_sk); 2750 msg = buf_msg(buf); 2751 2752 /* we lock on new_sk; but lockdep sees the lock on sk */ 2753 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING); 2754 2755 /* 2756 * Reject any stray messages received by new socket 2757 * before the socket lock was taken (very, very unlikely) 2758 */ 2759 tsk_rej_rx_queue(new_sk, TIPC_ERR_NO_PORT); 2760 2761 /* Connect new socket to it's peer */ 2762 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg)); 2763 2764 tsk_set_importance(new_sk, msg_importance(msg)); 2765 if (msg_named(msg)) { 2766 new_tsock->conn_addrtype = TIPC_SERVICE_ADDR; 2767 msg_set_nametype(&new_tsock->phdr, msg_nametype(msg)); 2768 msg_set_nameinst(&new_tsock->phdr, msg_nameinst(msg)); 2769 } 2770 2771 /* 2772 * Respond to 'SYN-' by discarding it & returning 'ACK'. 2773 * Respond to 'SYN+' by queuing it on new socket & returning 'ACK'. 2774 */ 2775 if (!msg_data_sz(msg)) { 2776 tsk_advance_rx_queue(sk); 2777 } else { 2778 __skb_dequeue(&sk->sk_receive_queue); 2779 __skb_queue_head(&new_sk->sk_receive_queue, buf); 2780 skb_set_owner_r(buf, new_sk); 2781 } 2782 iov_iter_kvec(&m.msg_iter, ITER_SOURCE, NULL, 0, 0); 2783 __tipc_sendstream(new_sock, &m, 0); 2784 release_sock(new_sk); 2785 exit: 2786 release_sock(sk); 2787 return res; 2788 } 2789 2790 /** 2791 * tipc_shutdown - shutdown socket connection 2792 * @sock: socket structure 2793 * @how: direction to close (must be SHUT_RDWR) 2794 * 2795 * Terminates connection (if necessary), then purges socket's receive queue. 2796 * 2797 * Return: 0 on success, errno otherwise 2798 */ 2799 static int tipc_shutdown(struct socket *sock, int how) 2800 { 2801 struct sock *sk = sock->sk; 2802 int res; 2803 2804 if (how != SHUT_RDWR) 2805 return -EINVAL; 2806 2807 lock_sock(sk); 2808 2809 trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " "); 2810 __tipc_shutdown(sock, TIPC_CONN_SHUTDOWN); 2811 sk->sk_shutdown = SHUTDOWN_MASK; 2812 2813 if (sk->sk_state == TIPC_DISCONNECTING) { 2814 /* Discard any unreceived messages */ 2815 __skb_queue_purge(&sk->sk_receive_queue); 2816 2817 res = 0; 2818 } else { 2819 res = -ENOTCONN; 2820 } 2821 /* Wake up anyone sleeping in poll. */ 2822 sk->sk_state_change(sk); 2823 2824 release_sock(sk); 2825 return res; 2826 } 2827 2828 static void tipc_sk_check_probing_state(struct sock *sk, 2829 struct sk_buff_head *list) 2830 { 2831 struct tipc_sock *tsk = tipc_sk(sk); 2832 u32 pnode = tsk_peer_node(tsk); 2833 u32 pport = tsk_peer_port(tsk); 2834 u32 self = tsk_own_node(tsk); 2835 u32 oport = tsk->portid; 2836 struct sk_buff *skb; 2837 2838 if (tsk->probe_unacked) { 2839 tipc_set_sk_state(sk, TIPC_DISCONNECTING); 2840 sk->sk_err = ECONNABORTED; 2841 tipc_node_remove_conn(sock_net(sk), pnode, pport); 2842 sk->sk_state_change(sk); 2843 return; 2844 } 2845 /* Prepare new probe */ 2846 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0, 2847 pnode, self, pport, oport, TIPC_OK); 2848 if (skb) 2849 __skb_queue_tail(list, skb); 2850 tsk->probe_unacked = true; 2851 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV); 2852 } 2853 2854 static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list) 2855 { 2856 struct tipc_sock *tsk = tipc_sk(sk); 2857 2858 /* Try again later if dest link is congested */ 2859 if (tsk->cong_link_cnt) { 2860 sk_reset_timer(sk, &sk->sk_timer, 2861 jiffies + msecs_to_jiffies(100)); 2862 return; 2863 } 2864 /* Prepare SYN for retransmit */ 2865 tipc_msg_skb_clone(&sk->sk_write_queue, list); 2866 } 2867 2868 static void tipc_sk_timeout(struct timer_list *t) 2869 { 2870 struct sock *sk = from_timer(sk, t, sk_timer); 2871 struct tipc_sock *tsk = tipc_sk(sk); 2872 u32 pnode = tsk_peer_node(tsk); 2873 struct sk_buff_head list; 2874 int rc = 0; 2875 2876 __skb_queue_head_init(&list); 2877 bh_lock_sock(sk); 2878 2879 /* Try again later if socket is busy */ 2880 if (sock_owned_by_user(sk)) { 2881 sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20); 2882 bh_unlock_sock(sk); 2883 sock_put(sk); 2884 return; 2885 } 2886 2887 if (sk->sk_state == TIPC_ESTABLISHED) 2888 tipc_sk_check_probing_state(sk, &list); 2889 else if (sk->sk_state == TIPC_CONNECTING) 2890 tipc_sk_retry_connect(sk, &list); 2891 2892 bh_unlock_sock(sk); 2893 2894 if (!skb_queue_empty(&list)) 2895 rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid); 2896 2897 /* SYN messages may cause link congestion */ 2898 if (rc == -ELINKCONG) { 2899 tipc_dest_push(&tsk->cong_links, pnode, 0); 2900 tsk->cong_link_cnt = 1; 2901 } 2902 sock_put(sk); 2903 } 2904 2905 static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua) 2906 { 2907 struct sock *sk = &tsk->sk; 2908 struct net *net = sock_net(sk); 2909 struct tipc_socket_addr skaddr; 2910 struct publication *p; 2911 u32 key; 2912 2913 if (tipc_sk_connected(sk)) 2914 return -EINVAL; 2915 key = tsk->portid + tsk->pub_count + 1; 2916 if (key == tsk->portid) 2917 return -EADDRINUSE; 2918 skaddr.ref = tsk->portid; 2919 skaddr.node = tipc_own_addr(net); 2920 p = tipc_nametbl_publish(net, ua, &skaddr, key); 2921 if (unlikely(!p)) 2922 return -EINVAL; 2923 2924 list_add(&p->binding_sock, &tsk->publications); 2925 tsk->pub_count++; 2926 tsk->published = true; 2927 return 0; 2928 } 2929 2930 static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua) 2931 { 2932 struct net *net = sock_net(&tsk->sk); 2933 struct publication *safe, *p; 2934 struct tipc_uaddr _ua; 2935 int rc = -EINVAL; 2936 2937 list_for_each_entry_safe(p, safe, &tsk->publications, binding_sock) { 2938 if (!ua) { 2939 tipc_uaddr(&_ua, TIPC_SERVICE_RANGE, p->scope, 2940 p->sr.type, p->sr.lower, p->sr.upper); 2941 tipc_nametbl_withdraw(net, &_ua, &p->sk, p->key); 2942 continue; 2943 } 2944 /* Unbind specific publication */ 2945 if (p->scope != ua->scope) 2946 continue; 2947 if (p->sr.type != ua->sr.type) 2948 continue; 2949 if (p->sr.lower != ua->sr.lower) 2950 continue; 2951 if (p->sr.upper != ua->sr.upper) 2952 break; 2953 tipc_nametbl_withdraw(net, ua, &p->sk, p->key); 2954 rc = 0; 2955 break; 2956 } 2957 if (list_empty(&tsk->publications)) { 2958 tsk->published = 0; 2959 rc = 0; 2960 } 2961 return rc; 2962 } 2963 2964 /* tipc_sk_reinit: set non-zero address in all existing sockets 2965 * when we go from standalone to network mode. 2966 */ 2967 void tipc_sk_reinit(struct net *net) 2968 { 2969 struct tipc_net *tn = net_generic(net, tipc_net_id); 2970 struct rhashtable_iter iter; 2971 struct tipc_sock *tsk; 2972 struct tipc_msg *msg; 2973 2974 rhashtable_walk_enter(&tn->sk_rht, &iter); 2975 2976 do { 2977 rhashtable_walk_start(&iter); 2978 2979 while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) { 2980 sock_hold(&tsk->sk); 2981 rhashtable_walk_stop(&iter); 2982 lock_sock(&tsk->sk); 2983 msg = &tsk->phdr; 2984 msg_set_prevnode(msg, tipc_own_addr(net)); 2985 msg_set_orignode(msg, tipc_own_addr(net)); 2986 release_sock(&tsk->sk); 2987 rhashtable_walk_start(&iter); 2988 sock_put(&tsk->sk); 2989 } 2990 2991 rhashtable_walk_stop(&iter); 2992 } while (tsk == ERR_PTR(-EAGAIN)); 2993 2994 rhashtable_walk_exit(&iter); 2995 } 2996 2997 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid) 2998 { 2999 struct tipc_net *tn = net_generic(net, tipc_net_id); 3000 struct tipc_sock *tsk; 3001 3002 rcu_read_lock(); 3003 tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params); 3004 if (tsk) 3005 sock_hold(&tsk->sk); 3006 rcu_read_unlock(); 3007 3008 return tsk; 3009 } 3010 3011 static int tipc_sk_insert(struct tipc_sock *tsk) 3012 { 3013 struct sock *sk = &tsk->sk; 3014 struct net *net = sock_net(sk); 3015 struct tipc_net *tn = net_generic(net, tipc_net_id); 3016 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1; 3017 u32 portid = get_random_u32_below(remaining) + TIPC_MIN_PORT; 3018 3019 while (remaining--) { 3020 portid++; 3021 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT)) 3022 portid = TIPC_MIN_PORT; 3023 tsk->portid = portid; 3024 sock_hold(&tsk->sk); 3025 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node, 3026 tsk_rht_params)) 3027 return 0; 3028 sock_put(&tsk->sk); 3029 } 3030 3031 return -1; 3032 } 3033 3034 static void tipc_sk_remove(struct tipc_sock *tsk) 3035 { 3036 struct sock *sk = &tsk->sk; 3037 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id); 3038 3039 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) { 3040 WARN_ON(refcount_read(&sk->sk_refcnt) == 1); 3041 __sock_put(sk); 3042 } 3043 } 3044 3045 static const struct rhashtable_params tsk_rht_params = { 3046 .nelem_hint = 192, 3047 .head_offset = offsetof(struct tipc_sock, node), 3048 .key_offset = offsetof(struct tipc_sock, portid), 3049 .key_len = sizeof(u32), /* portid */ 3050 .max_size = 1048576, 3051 .min_size = 256, 3052 .automatic_shrinking = true, 3053 }; 3054 3055 int tipc_sk_rht_init(struct net *net) 3056 { 3057 struct tipc_net *tn = net_generic(net, tipc_net_id); 3058 3059 return rhashtable_init(&tn->sk_rht, &tsk_rht_params); 3060 } 3061 3062 void tipc_sk_rht_destroy(struct net *net) 3063 { 3064 struct tipc_net *tn = net_generic(net, tipc_net_id); 3065 3066 /* Wait for socket readers to complete */ 3067 synchronize_net(); 3068 3069 rhashtable_destroy(&tn->sk_rht); 3070 } 3071 3072 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq) 3073 { 3074 struct net *net = sock_net(&tsk->sk); 3075 struct tipc_group *grp = tsk->group; 3076 struct tipc_msg *hdr = &tsk->phdr; 3077 struct tipc_uaddr ua; 3078 int rc; 3079 3080 if (mreq->type < TIPC_RESERVED_TYPES) 3081 return -EACCES; 3082 if (mreq->scope > TIPC_NODE_SCOPE) 3083 return -EINVAL; 3084 if (mreq->scope != TIPC_NODE_SCOPE) 3085 mreq->scope = TIPC_CLUSTER_SCOPE; 3086 if (grp) 3087 return -EACCES; 3088 grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open); 3089 if (!grp) 3090 return -ENOMEM; 3091 tsk->group = grp; 3092 msg_set_lookup_scope(hdr, mreq->scope); 3093 msg_set_nametype(hdr, mreq->type); 3094 msg_set_dest_droppable(hdr, true); 3095 tipc_uaddr(&ua, TIPC_SERVICE_RANGE, mreq->scope, 3096 mreq->type, mreq->instance, mreq->instance); 3097 tipc_nametbl_build_group(net, grp, &ua); 3098 rc = tipc_sk_publish(tsk, &ua); 3099 if (rc) { 3100 tipc_group_delete(net, grp); 3101 tsk->group = NULL; 3102 return rc; 3103 } 3104 /* Eliminate any risk that a broadcast overtakes sent JOINs */ 3105 tsk->mc_method.rcast = true; 3106 tsk->mc_method.mandatory = true; 3107 tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf); 3108 return rc; 3109 } 3110 3111 static int tipc_sk_leave(struct tipc_sock *tsk) 3112 { 3113 struct net *net = sock_net(&tsk->sk); 3114 struct tipc_group *grp = tsk->group; 3115 struct tipc_uaddr ua; 3116 int scope; 3117 3118 if (!grp) 3119 return -EINVAL; 3120 ua.addrtype = TIPC_SERVICE_RANGE; 3121 tipc_group_self(grp, &ua.sr, &scope); 3122 ua.scope = scope; 3123 tipc_group_delete(net, grp); 3124 tsk->group = NULL; 3125 tipc_sk_withdraw(tsk, &ua); 3126 return 0; 3127 } 3128 3129 /** 3130 * tipc_setsockopt - set socket option 3131 * @sock: socket structure 3132 * @lvl: option level 3133 * @opt: option identifier 3134 * @ov: pointer to new option value 3135 * @ol: length of option value 3136 * 3137 * For stream sockets only, accepts and ignores all IPPROTO_TCP options 3138 * (to ease compatibility). 3139 * 3140 * Return: 0 on success, errno otherwise 3141 */ 3142 static int tipc_setsockopt(struct socket *sock, int lvl, int opt, 3143 sockptr_t ov, unsigned int ol) 3144 { 3145 struct sock *sk = sock->sk; 3146 struct tipc_sock *tsk = tipc_sk(sk); 3147 struct tipc_group_req mreq; 3148 u32 value = 0; 3149 int res = 0; 3150 3151 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM)) 3152 return 0; 3153 if (lvl != SOL_TIPC) 3154 return -ENOPROTOOPT; 3155 3156 switch (opt) { 3157 case TIPC_IMPORTANCE: 3158 case TIPC_SRC_DROPPABLE: 3159 case TIPC_DEST_DROPPABLE: 3160 case TIPC_CONN_TIMEOUT: 3161 case TIPC_NODELAY: 3162 if (ol < sizeof(value)) 3163 return -EINVAL; 3164 if (copy_from_sockptr(&value, ov, sizeof(u32))) 3165 return -EFAULT; 3166 break; 3167 case TIPC_GROUP_JOIN: 3168 if (ol < sizeof(mreq)) 3169 return -EINVAL; 3170 if (copy_from_sockptr(&mreq, ov, sizeof(mreq))) 3171 return -EFAULT; 3172 break; 3173 default: 3174 if (!sockptr_is_null(ov) || ol) 3175 return -EINVAL; 3176 } 3177 3178 lock_sock(sk); 3179 3180 switch (opt) { 3181 case TIPC_IMPORTANCE: 3182 res = tsk_set_importance(sk, value); 3183 break; 3184 case TIPC_SRC_DROPPABLE: 3185 if (sock->type != SOCK_STREAM) 3186 tsk_set_unreliable(tsk, value); 3187 else 3188 res = -ENOPROTOOPT; 3189 break; 3190 case TIPC_DEST_DROPPABLE: 3191 tsk_set_unreturnable(tsk, value); 3192 break; 3193 case TIPC_CONN_TIMEOUT: 3194 tipc_sk(sk)->conn_timeout = value; 3195 break; 3196 case TIPC_MCAST_BROADCAST: 3197 tsk->mc_method.rcast = false; 3198 tsk->mc_method.mandatory = true; 3199 break; 3200 case TIPC_MCAST_REPLICAST: 3201 tsk->mc_method.rcast = true; 3202 tsk->mc_method.mandatory = true; 3203 break; 3204 case TIPC_GROUP_JOIN: 3205 res = tipc_sk_join(tsk, &mreq); 3206 break; 3207 case TIPC_GROUP_LEAVE: 3208 res = tipc_sk_leave(tsk); 3209 break; 3210 case TIPC_NODELAY: 3211 tsk->nodelay = !!value; 3212 tsk_set_nagle(tsk); 3213 break; 3214 default: 3215 res = -EINVAL; 3216 } 3217 3218 release_sock(sk); 3219 3220 return res; 3221 } 3222 3223 /** 3224 * tipc_getsockopt - get socket option 3225 * @sock: socket structure 3226 * @lvl: option level 3227 * @opt: option identifier 3228 * @ov: receptacle for option value 3229 * @ol: receptacle for length of option value 3230 * 3231 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options 3232 * (to ease compatibility). 3233 * 3234 * Return: 0 on success, errno otherwise 3235 */ 3236 static int tipc_getsockopt(struct socket *sock, int lvl, int opt, 3237 char __user *ov, int __user *ol) 3238 { 3239 struct sock *sk = sock->sk; 3240 struct tipc_sock *tsk = tipc_sk(sk); 3241 struct tipc_service_range seq; 3242 int len, scope; 3243 u32 value; 3244 int res; 3245 3246 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM)) 3247 return put_user(0, ol); 3248 if (lvl != SOL_TIPC) 3249 return -ENOPROTOOPT; 3250 res = get_user(len, ol); 3251 if (res) 3252 return res; 3253 3254 lock_sock(sk); 3255 3256 switch (opt) { 3257 case TIPC_IMPORTANCE: 3258 value = tsk_importance(tsk); 3259 break; 3260 case TIPC_SRC_DROPPABLE: 3261 value = tsk_unreliable(tsk); 3262 break; 3263 case TIPC_DEST_DROPPABLE: 3264 value = tsk_unreturnable(tsk); 3265 break; 3266 case TIPC_CONN_TIMEOUT: 3267 value = tsk->conn_timeout; 3268 /* no need to set "res", since already 0 at this point */ 3269 break; 3270 case TIPC_NODE_RECVQ_DEPTH: 3271 value = 0; /* was tipc_queue_size, now obsolete */ 3272 break; 3273 case TIPC_SOCK_RECVQ_DEPTH: 3274 value = skb_queue_len(&sk->sk_receive_queue); 3275 break; 3276 case TIPC_SOCK_RECVQ_USED: 3277 value = sk_rmem_alloc_get(sk); 3278 break; 3279 case TIPC_GROUP_JOIN: 3280 seq.type = 0; 3281 if (tsk->group) 3282 tipc_group_self(tsk->group, &seq, &scope); 3283 value = seq.type; 3284 break; 3285 default: 3286 res = -EINVAL; 3287 } 3288 3289 release_sock(sk); 3290 3291 if (res) 3292 return res; /* "get" failed */ 3293 3294 if (len < sizeof(value)) 3295 return -EINVAL; 3296 3297 if (copy_to_user(ov, &value, sizeof(value))) 3298 return -EFAULT; 3299 3300 return put_user(sizeof(value), ol); 3301 } 3302 3303 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 3304 { 3305 struct net *net = sock_net(sock->sk); 3306 struct tipc_sioc_nodeid_req nr = {0}; 3307 struct tipc_sioc_ln_req lnr; 3308 void __user *argp = (void __user *)arg; 3309 3310 switch (cmd) { 3311 case SIOCGETLINKNAME: 3312 if (copy_from_user(&lnr, argp, sizeof(lnr))) 3313 return -EFAULT; 3314 if (!tipc_node_get_linkname(net, 3315 lnr.bearer_id & 0xffff, lnr.peer, 3316 lnr.linkname, TIPC_MAX_LINK_NAME)) { 3317 if (copy_to_user(argp, &lnr, sizeof(lnr))) 3318 return -EFAULT; 3319 return 0; 3320 } 3321 return -EADDRNOTAVAIL; 3322 case SIOCGETNODEID: 3323 if (copy_from_user(&nr, argp, sizeof(nr))) 3324 return -EFAULT; 3325 if (!tipc_node_get_id(net, nr.peer, nr.node_id)) 3326 return -EADDRNOTAVAIL; 3327 if (copy_to_user(argp, &nr, sizeof(nr))) 3328 return -EFAULT; 3329 return 0; 3330 default: 3331 return -ENOIOCTLCMD; 3332 } 3333 } 3334 3335 static int tipc_socketpair(struct socket *sock1, struct socket *sock2) 3336 { 3337 struct tipc_sock *tsk2 = tipc_sk(sock2->sk); 3338 struct tipc_sock *tsk1 = tipc_sk(sock1->sk); 3339 u32 onode = tipc_own_addr(sock_net(sock1->sk)); 3340 3341 tsk1->peer.family = AF_TIPC; 3342 tsk1->peer.addrtype = TIPC_SOCKET_ADDR; 3343 tsk1->peer.scope = TIPC_NODE_SCOPE; 3344 tsk1->peer.addr.id.ref = tsk2->portid; 3345 tsk1->peer.addr.id.node = onode; 3346 tsk2->peer.family = AF_TIPC; 3347 tsk2->peer.addrtype = TIPC_SOCKET_ADDR; 3348 tsk2->peer.scope = TIPC_NODE_SCOPE; 3349 tsk2->peer.addr.id.ref = tsk1->portid; 3350 tsk2->peer.addr.id.node = onode; 3351 3352 tipc_sk_finish_conn(tsk1, tsk2->portid, onode); 3353 tipc_sk_finish_conn(tsk2, tsk1->portid, onode); 3354 return 0; 3355 } 3356 3357 /* Protocol switches for the various types of TIPC sockets */ 3358 3359 static const struct proto_ops msg_ops = { 3360 .owner = THIS_MODULE, 3361 .family = AF_TIPC, 3362 .release = tipc_release, 3363 .bind = tipc_bind, 3364 .connect = tipc_connect, 3365 .socketpair = tipc_socketpair, 3366 .accept = sock_no_accept, 3367 .getname = tipc_getname, 3368 .poll = tipc_poll, 3369 .ioctl = tipc_ioctl, 3370 .listen = sock_no_listen, 3371 .shutdown = tipc_shutdown, 3372 .setsockopt = tipc_setsockopt, 3373 .getsockopt = tipc_getsockopt, 3374 .sendmsg = tipc_sendmsg, 3375 .recvmsg = tipc_recvmsg, 3376 .mmap = sock_no_mmap, 3377 }; 3378 3379 static const struct proto_ops packet_ops = { 3380 .owner = THIS_MODULE, 3381 .family = AF_TIPC, 3382 .release = tipc_release, 3383 .bind = tipc_bind, 3384 .connect = tipc_connect, 3385 .socketpair = tipc_socketpair, 3386 .accept = tipc_accept, 3387 .getname = tipc_getname, 3388 .poll = tipc_poll, 3389 .ioctl = tipc_ioctl, 3390 .listen = tipc_listen, 3391 .shutdown = tipc_shutdown, 3392 .setsockopt = tipc_setsockopt, 3393 .getsockopt = tipc_getsockopt, 3394 .sendmsg = tipc_send_packet, 3395 .recvmsg = tipc_recvmsg, 3396 .mmap = sock_no_mmap, 3397 }; 3398 3399 static const struct proto_ops stream_ops = { 3400 .owner = THIS_MODULE, 3401 .family = AF_TIPC, 3402 .release = tipc_release, 3403 .bind = tipc_bind, 3404 .connect = tipc_connect, 3405 .socketpair = tipc_socketpair, 3406 .accept = tipc_accept, 3407 .getname = tipc_getname, 3408 .poll = tipc_poll, 3409 .ioctl = tipc_ioctl, 3410 .listen = tipc_listen, 3411 .shutdown = tipc_shutdown, 3412 .setsockopt = tipc_setsockopt, 3413 .getsockopt = tipc_getsockopt, 3414 .sendmsg = tipc_sendstream, 3415 .recvmsg = tipc_recvstream, 3416 .mmap = sock_no_mmap, 3417 }; 3418 3419 static const struct net_proto_family tipc_family_ops = { 3420 .owner = THIS_MODULE, 3421 .family = AF_TIPC, 3422 .create = tipc_sk_create 3423 }; 3424 3425 static struct proto tipc_proto = { 3426 .name = "TIPC", 3427 .owner = THIS_MODULE, 3428 .obj_size = sizeof(struct tipc_sock), 3429 .sysctl_rmem = sysctl_tipc_rmem 3430 }; 3431 3432 /** 3433 * tipc_socket_init - initialize TIPC socket interface 3434 * 3435 * Return: 0 on success, errno otherwise 3436 */ 3437 int tipc_socket_init(void) 3438 { 3439 int res; 3440 3441 res = proto_register(&tipc_proto, 1); 3442 if (res) { 3443 pr_err("Failed to register TIPC protocol type\n"); 3444 goto out; 3445 } 3446 3447 res = sock_register(&tipc_family_ops); 3448 if (res) { 3449 pr_err("Failed to register TIPC socket type\n"); 3450 proto_unregister(&tipc_proto); 3451 goto out; 3452 } 3453 out: 3454 return res; 3455 } 3456 3457 /** 3458 * tipc_socket_stop - stop TIPC socket interface 3459 */ 3460 void tipc_socket_stop(void) 3461 { 3462 sock_unregister(tipc_family_ops.family); 3463 proto_unregister(&tipc_proto); 3464 } 3465 3466 /* Caller should hold socket lock for the passed tipc socket. */ 3467 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk) 3468 { 3469 u32 peer_node, peer_port; 3470 u32 conn_type, conn_instance; 3471 struct nlattr *nest; 3472 3473 peer_node = tsk_peer_node(tsk); 3474 peer_port = tsk_peer_port(tsk); 3475 conn_type = msg_nametype(&tsk->phdr); 3476 conn_instance = msg_nameinst(&tsk->phdr); 3477 nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON); 3478 if (!nest) 3479 return -EMSGSIZE; 3480 3481 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node)) 3482 goto msg_full; 3483 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port)) 3484 goto msg_full; 3485 3486 if (tsk->conn_addrtype != 0) { 3487 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG)) 3488 goto msg_full; 3489 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, conn_type)) 3490 goto msg_full; 3491 if (nla_put_u32(skb, TIPC_NLA_CON_INST, conn_instance)) 3492 goto msg_full; 3493 } 3494 nla_nest_end(skb, nest); 3495 3496 return 0; 3497 3498 msg_full: 3499 nla_nest_cancel(skb, nest); 3500 3501 return -EMSGSIZE; 3502 } 3503 3504 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock 3505 *tsk) 3506 { 3507 struct net *net = sock_net(skb->sk); 3508 struct sock *sk = &tsk->sk; 3509 3510 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) || 3511 nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net))) 3512 return -EMSGSIZE; 3513 3514 if (tipc_sk_connected(sk)) { 3515 if (__tipc_nl_add_sk_con(skb, tsk)) 3516 return -EMSGSIZE; 3517 } else if (!list_empty(&tsk->publications)) { 3518 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL)) 3519 return -EMSGSIZE; 3520 } 3521 return 0; 3522 } 3523 3524 /* Caller should hold socket lock for the passed tipc socket. */ 3525 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb, 3526 struct tipc_sock *tsk) 3527 { 3528 struct nlattr *attrs; 3529 void *hdr; 3530 3531 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq, 3532 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET); 3533 if (!hdr) 3534 goto msg_cancel; 3535 3536 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK); 3537 if (!attrs) 3538 goto genlmsg_cancel; 3539 3540 if (__tipc_nl_add_sk_info(skb, tsk)) 3541 goto attr_msg_cancel; 3542 3543 nla_nest_end(skb, attrs); 3544 genlmsg_end(skb, hdr); 3545 3546 return 0; 3547 3548 attr_msg_cancel: 3549 nla_nest_cancel(skb, attrs); 3550 genlmsg_cancel: 3551 genlmsg_cancel(skb, hdr); 3552 msg_cancel: 3553 return -EMSGSIZE; 3554 } 3555 3556 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb, 3557 int (*skb_handler)(struct sk_buff *skb, 3558 struct netlink_callback *cb, 3559 struct tipc_sock *tsk)) 3560 { 3561 struct rhashtable_iter *iter = (void *)cb->args[4]; 3562 struct tipc_sock *tsk; 3563 int err; 3564 3565 rhashtable_walk_start(iter); 3566 while ((tsk = rhashtable_walk_next(iter)) != NULL) { 3567 if (IS_ERR(tsk)) { 3568 err = PTR_ERR(tsk); 3569 if (err == -EAGAIN) { 3570 err = 0; 3571 continue; 3572 } 3573 break; 3574 } 3575 3576 sock_hold(&tsk->sk); 3577 rhashtable_walk_stop(iter); 3578 lock_sock(&tsk->sk); 3579 err = skb_handler(skb, cb, tsk); 3580 if (err) { 3581 release_sock(&tsk->sk); 3582 sock_put(&tsk->sk); 3583 goto out; 3584 } 3585 release_sock(&tsk->sk); 3586 rhashtable_walk_start(iter); 3587 sock_put(&tsk->sk); 3588 } 3589 rhashtable_walk_stop(iter); 3590 out: 3591 return skb->len; 3592 } 3593 EXPORT_SYMBOL(tipc_nl_sk_walk); 3594 3595 int tipc_dump_start(struct netlink_callback *cb) 3596 { 3597 return __tipc_dump_start(cb, sock_net(cb->skb->sk)); 3598 } 3599 EXPORT_SYMBOL(tipc_dump_start); 3600 3601 int __tipc_dump_start(struct netlink_callback *cb, struct net *net) 3602 { 3603 /* tipc_nl_name_table_dump() uses cb->args[0...3]. */ 3604 struct rhashtable_iter *iter = (void *)cb->args[4]; 3605 struct tipc_net *tn = tipc_net(net); 3606 3607 if (!iter) { 3608 iter = kmalloc(sizeof(*iter), GFP_KERNEL); 3609 if (!iter) 3610 return -ENOMEM; 3611 3612 cb->args[4] = (long)iter; 3613 } 3614 3615 rhashtable_walk_enter(&tn->sk_rht, iter); 3616 return 0; 3617 } 3618 3619 int tipc_dump_done(struct netlink_callback *cb) 3620 { 3621 struct rhashtable_iter *hti = (void *)cb->args[4]; 3622 3623 rhashtable_walk_exit(hti); 3624 kfree(hti); 3625 return 0; 3626 } 3627 EXPORT_SYMBOL(tipc_dump_done); 3628 3629 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb, 3630 struct tipc_sock *tsk, u32 sk_filter_state, 3631 u64 (*tipc_diag_gen_cookie)(struct sock *sk)) 3632 { 3633 struct sock *sk = &tsk->sk; 3634 struct nlattr *attrs; 3635 struct nlattr *stat; 3636 3637 /*filter response w.r.t sk_state*/ 3638 if (!(sk_filter_state & (1 << sk->sk_state))) 3639 return 0; 3640 3641 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK); 3642 if (!attrs) 3643 goto msg_cancel; 3644 3645 if (__tipc_nl_add_sk_info(skb, tsk)) 3646 goto attr_msg_cancel; 3647 3648 if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) || 3649 nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) || 3650 nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) || 3651 nla_put_u32(skb, TIPC_NLA_SOCK_UID, 3652 from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk), 3653 sock_i_uid(sk))) || 3654 nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE, 3655 tipc_diag_gen_cookie(sk), 3656 TIPC_NLA_SOCK_PAD)) 3657 goto attr_msg_cancel; 3658 3659 stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT); 3660 if (!stat) 3661 goto attr_msg_cancel; 3662 3663 if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ, 3664 skb_queue_len(&sk->sk_receive_queue)) || 3665 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ, 3666 skb_queue_len(&sk->sk_write_queue)) || 3667 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP, 3668 atomic_read(&sk->sk_drops))) 3669 goto stat_msg_cancel; 3670 3671 if (tsk->cong_link_cnt && 3672 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG)) 3673 goto stat_msg_cancel; 3674 3675 if (tsk_conn_cong(tsk) && 3676 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG)) 3677 goto stat_msg_cancel; 3678 3679 nla_nest_end(skb, stat); 3680 3681 if (tsk->group) 3682 if (tipc_group_fill_sock_diag(tsk->group, skb)) 3683 goto stat_msg_cancel; 3684 3685 nla_nest_end(skb, attrs); 3686 3687 return 0; 3688 3689 stat_msg_cancel: 3690 nla_nest_cancel(skb, stat); 3691 attr_msg_cancel: 3692 nla_nest_cancel(skb, attrs); 3693 msg_cancel: 3694 return -EMSGSIZE; 3695 } 3696 EXPORT_SYMBOL(tipc_sk_fill_sock_diag); 3697 3698 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb) 3699 { 3700 return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk); 3701 } 3702 3703 /* Caller should hold socket lock for the passed tipc socket. */ 3704 static int __tipc_nl_add_sk_publ(struct sk_buff *skb, 3705 struct netlink_callback *cb, 3706 struct publication *publ) 3707 { 3708 void *hdr; 3709 struct nlattr *attrs; 3710 3711 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq, 3712 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET); 3713 if (!hdr) 3714 goto msg_cancel; 3715 3716 attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL); 3717 if (!attrs) 3718 goto genlmsg_cancel; 3719 3720 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key)) 3721 goto attr_msg_cancel; 3722 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->sr.type)) 3723 goto attr_msg_cancel; 3724 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->sr.lower)) 3725 goto attr_msg_cancel; 3726 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->sr.upper)) 3727 goto attr_msg_cancel; 3728 3729 nla_nest_end(skb, attrs); 3730 genlmsg_end(skb, hdr); 3731 3732 return 0; 3733 3734 attr_msg_cancel: 3735 nla_nest_cancel(skb, attrs); 3736 genlmsg_cancel: 3737 genlmsg_cancel(skb, hdr); 3738 msg_cancel: 3739 return -EMSGSIZE; 3740 } 3741 3742 /* Caller should hold socket lock for the passed tipc socket. */ 3743 static int __tipc_nl_list_sk_publ(struct sk_buff *skb, 3744 struct netlink_callback *cb, 3745 struct tipc_sock *tsk, u32 *last_publ) 3746 { 3747 int err; 3748 struct publication *p; 3749 3750 if (*last_publ) { 3751 list_for_each_entry(p, &tsk->publications, binding_sock) { 3752 if (p->key == *last_publ) 3753 break; 3754 } 3755 if (list_entry_is_head(p, &tsk->publications, binding_sock)) { 3756 /* We never set seq or call nl_dump_check_consistent() 3757 * this means that setting prev_seq here will cause the 3758 * consistence check to fail in the netlink callback 3759 * handler. Resulting in the last NLMSG_DONE message 3760 * having the NLM_F_DUMP_INTR flag set. 3761 */ 3762 cb->prev_seq = 1; 3763 *last_publ = 0; 3764 return -EPIPE; 3765 } 3766 } else { 3767 p = list_first_entry(&tsk->publications, struct publication, 3768 binding_sock); 3769 } 3770 3771 list_for_each_entry_from(p, &tsk->publications, binding_sock) { 3772 err = __tipc_nl_add_sk_publ(skb, cb, p); 3773 if (err) { 3774 *last_publ = p->key; 3775 return err; 3776 } 3777 } 3778 *last_publ = 0; 3779 3780 return 0; 3781 } 3782 3783 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb) 3784 { 3785 int err; 3786 u32 tsk_portid = cb->args[0]; 3787 u32 last_publ = cb->args[1]; 3788 u32 done = cb->args[2]; 3789 struct net *net = sock_net(skb->sk); 3790 struct tipc_sock *tsk; 3791 3792 if (!tsk_portid) { 3793 struct nlattr **attrs = genl_dumpit_info(cb)->info.attrs; 3794 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1]; 3795 3796 if (!attrs[TIPC_NLA_SOCK]) 3797 return -EINVAL; 3798 3799 err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX, 3800 attrs[TIPC_NLA_SOCK], 3801 tipc_nl_sock_policy, NULL); 3802 if (err) 3803 return err; 3804 3805 if (!sock[TIPC_NLA_SOCK_REF]) 3806 return -EINVAL; 3807 3808 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]); 3809 } 3810 3811 if (done) 3812 return 0; 3813 3814 tsk = tipc_sk_lookup(net, tsk_portid); 3815 if (!tsk) 3816 return -EINVAL; 3817 3818 lock_sock(&tsk->sk); 3819 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ); 3820 if (!err) 3821 done = 1; 3822 release_sock(&tsk->sk); 3823 sock_put(&tsk->sk); 3824 3825 cb->args[0] = tsk_portid; 3826 cb->args[1] = last_publ; 3827 cb->args[2] = done; 3828 3829 return skb->len; 3830 } 3831 3832 /** 3833 * tipc_sk_filtering - check if a socket should be traced 3834 * @sk: the socket to be examined 3835 * 3836 * @sysctl_tipc_sk_filter is used as the socket tuple for filtering: 3837 * (portid, sock type, name type, name lower, name upper) 3838 * 3839 * Return: true if the socket meets the socket tuple data 3840 * (value 0 = 'any') or when there is no tuple set (all = 0), 3841 * otherwise false 3842 */ 3843 bool tipc_sk_filtering(struct sock *sk) 3844 { 3845 struct tipc_sock *tsk; 3846 struct publication *p; 3847 u32 _port, _sktype, _type, _lower, _upper; 3848 u32 type = 0, lower = 0, upper = 0; 3849 3850 if (!sk) 3851 return true; 3852 3853 tsk = tipc_sk(sk); 3854 3855 _port = sysctl_tipc_sk_filter[0]; 3856 _sktype = sysctl_tipc_sk_filter[1]; 3857 _type = sysctl_tipc_sk_filter[2]; 3858 _lower = sysctl_tipc_sk_filter[3]; 3859 _upper = sysctl_tipc_sk_filter[4]; 3860 3861 if (!_port && !_sktype && !_type && !_lower && !_upper) 3862 return true; 3863 3864 if (_port) 3865 return (_port == tsk->portid); 3866 3867 if (_sktype && _sktype != sk->sk_type) 3868 return false; 3869 3870 if (tsk->published) { 3871 p = list_first_entry_or_null(&tsk->publications, 3872 struct publication, binding_sock); 3873 if (p) { 3874 type = p->sr.type; 3875 lower = p->sr.lower; 3876 upper = p->sr.upper; 3877 } 3878 } 3879 3880 if (!tipc_sk_type_connectionless(sk)) { 3881 type = msg_nametype(&tsk->phdr); 3882 lower = msg_nameinst(&tsk->phdr); 3883 upper = lower; 3884 } 3885 3886 if ((_type && _type != type) || (_lower && _lower != lower) || 3887 (_upper && _upper != upper)) 3888 return false; 3889 3890 return true; 3891 } 3892 3893 u32 tipc_sock_get_portid(struct sock *sk) 3894 { 3895 return (sk) ? (tipc_sk(sk))->portid : 0; 3896 } 3897 3898 /** 3899 * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded, 3900 * both the rcv and backlog queues are considered 3901 * @sk: tipc sk to be checked 3902 * @skb: tipc msg to be checked 3903 * 3904 * Return: true if the socket rx queue allocation is > 90%, otherwise false 3905 */ 3906 3907 bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb) 3908 { 3909 atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt; 3910 unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt); 3911 unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk); 3912 3913 return (qsize > lim * 90 / 100); 3914 } 3915 3916 /** 3917 * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded, 3918 * only the rcv queue is considered 3919 * @sk: tipc sk to be checked 3920 * @skb: tipc msg to be checked 3921 * 3922 * Return: true if the socket rx queue allocation is > 90%, otherwise false 3923 */ 3924 3925 bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb) 3926 { 3927 unsigned int lim = rcvbuf_limit(sk, skb); 3928 unsigned int qsize = sk_rmem_alloc_get(sk); 3929 3930 return (qsize > lim * 90 / 100); 3931 } 3932 3933 /** 3934 * tipc_sk_dump - dump TIPC socket 3935 * @sk: tipc sk to be dumped 3936 * @dqueues: bitmask to decide if any socket queue to be dumped? 3937 * - TIPC_DUMP_NONE: don't dump socket queues 3938 * - TIPC_DUMP_SK_SNDQ: dump socket send queue 3939 * - TIPC_DUMP_SK_RCVQ: dump socket rcv queue 3940 * - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue 3941 * - TIPC_DUMP_ALL: dump all the socket queues above 3942 * @buf: returned buffer of dump data in format 3943 */ 3944 int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf) 3945 { 3946 int i = 0; 3947 size_t sz = (dqueues) ? SK_LMAX : SK_LMIN; 3948 u32 conn_type, conn_instance; 3949 struct tipc_sock *tsk; 3950 struct publication *p; 3951 bool tsk_connected; 3952 3953 if (!sk) { 3954 i += scnprintf(buf, sz, "sk data: (null)\n"); 3955 return i; 3956 } 3957 3958 tsk = tipc_sk(sk); 3959 tsk_connected = !tipc_sk_type_connectionless(sk); 3960 3961 i += scnprintf(buf, sz, "sk data: %u", sk->sk_type); 3962 i += scnprintf(buf + i, sz - i, " %d", sk->sk_state); 3963 i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk)); 3964 i += scnprintf(buf + i, sz - i, " %u", tsk->portid); 3965 i += scnprintf(buf + i, sz - i, " | %u", tsk_connected); 3966 if (tsk_connected) { 3967 i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk)); 3968 i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk)); 3969 conn_type = msg_nametype(&tsk->phdr); 3970 conn_instance = msg_nameinst(&tsk->phdr); 3971 i += scnprintf(buf + i, sz - i, " %u", conn_type); 3972 i += scnprintf(buf + i, sz - i, " %u", conn_instance); 3973 } 3974 i += scnprintf(buf + i, sz - i, " | %u", tsk->published); 3975 if (tsk->published) { 3976 p = list_first_entry_or_null(&tsk->publications, 3977 struct publication, binding_sock); 3978 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.type : 0); 3979 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.lower : 0); 3980 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.upper : 0); 3981 } 3982 i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win); 3983 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win); 3984 i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt); 3985 i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps); 3986 i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt); 3987 i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked); 3988 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked); 3989 i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt)); 3990 i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown); 3991 i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk)); 3992 i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf); 3993 i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk)); 3994 i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf); 3995 i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len)); 3996 3997 if (dqueues & TIPC_DUMP_SK_SNDQ) { 3998 i += scnprintf(buf + i, sz - i, "sk_write_queue: "); 3999 i += tipc_list_dump(&sk->sk_write_queue, false, buf + i); 4000 } 4001 4002 if (dqueues & TIPC_DUMP_SK_RCVQ) { 4003 i += scnprintf(buf + i, sz - i, "sk_receive_queue: "); 4004 i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i); 4005 } 4006 4007 if (dqueues & TIPC_DUMP_SK_BKLGQ) { 4008 i += scnprintf(buf + i, sz - i, "sk_backlog:\n head "); 4009 i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i); 4010 if (sk->sk_backlog.tail != sk->sk_backlog.head) { 4011 i += scnprintf(buf + i, sz - i, " tail "); 4012 i += tipc_skb_dump(sk->sk_backlog.tail, false, 4013 buf + i); 4014 } 4015 } 4016 4017 return i; 4018 } 4019