1 /* 2 * net/tipc/socket.c: TIPC socket API 3 * 4 * Copyright (c) 2001-2007, 2012-2019, Ericsson AB 5 * Copyright (c) 2004-2008, 2010-2013, Wind River Systems 6 * Copyright (c) 2020-2021, Red Hat Inc 7 * All rights reserved. 8 * 9 * Redistribution and use in source and binary forms, with or without 10 * modification, are permitted provided that the following conditions are met: 11 * 12 * 1. Redistributions of source code must retain the above copyright 13 * notice, this list of conditions and the following disclaimer. 14 * 2. Redistributions in binary form must reproduce the above copyright 15 * notice, this list of conditions and the following disclaimer in the 16 * documentation and/or other materials provided with the distribution. 17 * 3. Neither the names of the copyright holders nor the names of its 18 * contributors may be used to endorse or promote products derived from 19 * this software without specific prior written permission. 20 * 21 * Alternatively, this software may be distributed under the terms of the 22 * GNU General Public License ("GPL") version 2 as published by the Free 23 * Software Foundation. 24 * 25 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 26 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 28 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE 29 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 30 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 31 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 32 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 33 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 34 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 35 * POSSIBILITY OF SUCH DAMAGE. 36 */ 37 38 #include <linux/rhashtable.h> 39 #include <linux/sched/signal.h> 40 #include <linux/uio.h> 41 #include <trace/events/sock.h> 42 43 #include "core.h" 44 #include "name_table.h" 45 #include "node.h" 46 #include "link.h" 47 #include "name_distr.h" 48 #include "socket.h" 49 #include "bcast.h" 50 #include "netlink.h" 51 #include "group.h" 52 #include "trace.h" 53 54 #define NAGLE_START_INIT 4 55 #define NAGLE_START_MAX 1024 56 #define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */ 57 #define CONN_PROBING_INTV msecs_to_jiffies(3600000) /* [ms] => 1 h */ 58 #define TIPC_MAX_PORT 0xffffffff 59 #define TIPC_MIN_PORT 1 60 #define TIPC_ACK_RATE 4 /* ACK at 1/4 of rcv window size */ 61 62 enum { 63 TIPC_LISTEN = TCP_LISTEN, 64 TIPC_ESTABLISHED = TCP_ESTABLISHED, 65 TIPC_OPEN = TCP_CLOSE, 66 TIPC_DISCONNECTING = TCP_CLOSE_WAIT, 67 TIPC_CONNECTING = TCP_SYN_SENT, 68 }; 69 70 struct sockaddr_pair { 71 struct sockaddr_tipc sock; 72 struct sockaddr_tipc member; 73 }; 74 75 /** 76 * struct tipc_sock - TIPC socket structure 77 * @sk: socket - interacts with 'port' and with user via the socket API 78 * @max_pkt: maximum packet size "hint" used when building messages sent by port 79 * @maxnagle: maximum size of msg which can be subject to nagle 80 * @portid: unique port identity in TIPC socket hash table 81 * @phdr: preformatted message header used when sending messages 82 * @cong_links: list of congested links 83 * @publications: list of publications for port 84 * @pub_count: total # of publications port has made during its lifetime 85 * @conn_timeout: the time we can wait for an unresponded setup request 86 * @probe_unacked: probe has not received ack yet 87 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue 88 * @cong_link_cnt: number of congested links 89 * @snt_unacked: # messages sent by socket, and not yet acked by peer 90 * @snd_win: send window size 91 * @peer_caps: peer capabilities mask 92 * @rcv_unacked: # messages read by user, but not yet acked back to peer 93 * @rcv_win: receive window size 94 * @peer: 'connected' peer for dgram/rdm 95 * @node: hash table node 96 * @mc_method: cookie for use between socket and broadcast layer 97 * @rcu: rcu struct for tipc_sock 98 * @group: TIPC communications group 99 * @oneway: message count in one direction (FIXME) 100 * @nagle_start: current nagle value 101 * @snd_backlog: send backlog count 102 * @msg_acc: messages accepted; used in managing backlog and nagle 103 * @pkt_cnt: TIPC socket packet count 104 * @expect_ack: whether this TIPC socket is expecting an ack 105 * @nodelay: setsockopt() TIPC_NODELAY setting 106 * @group_is_open: TIPC socket group is fully open (FIXME) 107 * @published: true if port has one or more associated names 108 * @conn_addrtype: address type used when establishing connection 109 */ 110 struct tipc_sock { 111 struct sock sk; 112 u32 max_pkt; 113 u32 maxnagle; 114 u32 portid; 115 struct tipc_msg phdr; 116 struct list_head cong_links; 117 struct list_head publications; 118 u32 pub_count; 119 atomic_t dupl_rcvcnt; 120 u16 conn_timeout; 121 bool probe_unacked; 122 u16 cong_link_cnt; 123 u16 snt_unacked; 124 u16 snd_win; 125 u16 peer_caps; 126 u16 rcv_unacked; 127 u16 rcv_win; 128 struct sockaddr_tipc peer; 129 struct rhash_head node; 130 struct tipc_mc_method mc_method; 131 struct rcu_head rcu; 132 struct tipc_group *group; 133 u32 oneway; 134 u32 nagle_start; 135 u16 snd_backlog; 136 u16 msg_acc; 137 u16 pkt_cnt; 138 bool expect_ack; 139 bool nodelay; 140 bool group_is_open; 141 bool published; 142 u8 conn_addrtype; 143 }; 144 145 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb); 146 static void tipc_data_ready(struct sock *sk); 147 static void tipc_write_space(struct sock *sk); 148 static void tipc_sock_destruct(struct sock *sk); 149 static int tipc_release(struct socket *sock); 150 static void tipc_sk_timeout(struct timer_list *t); 151 static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua); 152 static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua); 153 static int tipc_sk_leave(struct tipc_sock *tsk); 154 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid); 155 static int tipc_sk_insert(struct tipc_sock *tsk); 156 static void tipc_sk_remove(struct tipc_sock *tsk); 157 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz); 158 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz); 159 static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack); 160 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p); 161 162 static const struct proto_ops packet_ops; 163 static const struct proto_ops stream_ops; 164 static const struct proto_ops msg_ops; 165 static struct proto tipc_proto; 166 static const struct rhashtable_params tsk_rht_params; 167 168 static u32 tsk_own_node(struct tipc_sock *tsk) 169 { 170 return msg_prevnode(&tsk->phdr); 171 } 172 173 static u32 tsk_peer_node(struct tipc_sock *tsk) 174 { 175 return msg_destnode(&tsk->phdr); 176 } 177 178 static u32 tsk_peer_port(struct tipc_sock *tsk) 179 { 180 return msg_destport(&tsk->phdr); 181 } 182 183 static bool tsk_unreliable(struct tipc_sock *tsk) 184 { 185 return msg_src_droppable(&tsk->phdr) != 0; 186 } 187 188 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable) 189 { 190 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0); 191 } 192 193 static bool tsk_unreturnable(struct tipc_sock *tsk) 194 { 195 return msg_dest_droppable(&tsk->phdr) != 0; 196 } 197 198 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable) 199 { 200 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0); 201 } 202 203 static int tsk_importance(struct tipc_sock *tsk) 204 { 205 return msg_importance(&tsk->phdr); 206 } 207 208 static struct tipc_sock *tipc_sk(const struct sock *sk) 209 { 210 return container_of(sk, struct tipc_sock, sk); 211 } 212 213 int tsk_set_importance(struct sock *sk, int imp) 214 { 215 if (imp > TIPC_CRITICAL_IMPORTANCE) 216 return -EINVAL; 217 msg_set_importance(&tipc_sk(sk)->phdr, (u32)imp); 218 return 0; 219 } 220 221 static bool tsk_conn_cong(struct tipc_sock *tsk) 222 { 223 return tsk->snt_unacked > tsk->snd_win; 224 } 225 226 static u16 tsk_blocks(int len) 227 { 228 return ((len / FLOWCTL_BLK_SZ) + 1); 229 } 230 231 /* tsk_blocks(): translate a buffer size in bytes to number of 232 * advertisable blocks, taking into account the ratio truesize(len)/len 233 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ 234 */ 235 static u16 tsk_adv_blocks(int len) 236 { 237 return len / FLOWCTL_BLK_SZ / 4; 238 } 239 240 /* tsk_inc(): increment counter for sent or received data 241 * - If block based flow control is not supported by peer we 242 * fall back to message based ditto, incrementing the counter 243 */ 244 static u16 tsk_inc(struct tipc_sock *tsk, int msglen) 245 { 246 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL)) 247 return ((msglen / FLOWCTL_BLK_SZ) + 1); 248 return 1; 249 } 250 251 /* tsk_set_nagle - enable/disable nagle property by manipulating maxnagle 252 */ 253 static void tsk_set_nagle(struct tipc_sock *tsk) 254 { 255 struct sock *sk = &tsk->sk; 256 257 tsk->maxnagle = 0; 258 if (sk->sk_type != SOCK_STREAM) 259 return; 260 if (tsk->nodelay) 261 return; 262 if (!(tsk->peer_caps & TIPC_NAGLE)) 263 return; 264 /* Limit node local buffer size to avoid receive queue overflow */ 265 if (tsk->max_pkt == MAX_MSG_SIZE) 266 tsk->maxnagle = 1500; 267 else 268 tsk->maxnagle = tsk->max_pkt; 269 } 270 271 /** 272 * tsk_advance_rx_queue - discard first buffer in socket receive queue 273 * @sk: network socket 274 * 275 * Caller must hold socket lock 276 */ 277 static void tsk_advance_rx_queue(struct sock *sk) 278 { 279 trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " "); 280 kfree_skb(__skb_dequeue(&sk->sk_receive_queue)); 281 } 282 283 /* tipc_sk_respond() : send response message back to sender 284 */ 285 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err) 286 { 287 u32 selector; 288 u32 dnode; 289 u32 onode = tipc_own_addr(sock_net(sk)); 290 291 if (!tipc_msg_reverse(onode, &skb, err)) 292 return; 293 294 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!"); 295 dnode = msg_destnode(buf_msg(skb)); 296 selector = msg_origport(buf_msg(skb)); 297 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector); 298 } 299 300 /** 301 * tsk_rej_rx_queue - reject all buffers in socket receive queue 302 * @sk: network socket 303 * @error: response error code 304 * 305 * Caller must hold socket lock 306 */ 307 static void tsk_rej_rx_queue(struct sock *sk, int error) 308 { 309 struct sk_buff *skb; 310 311 while ((skb = __skb_dequeue(&sk->sk_receive_queue))) 312 tipc_sk_respond(sk, skb, error); 313 } 314 315 static bool tipc_sk_connected(const struct sock *sk) 316 { 317 return READ_ONCE(sk->sk_state) == TIPC_ESTABLISHED; 318 } 319 320 /* tipc_sk_type_connectionless - check if the socket is datagram socket 321 * @sk: socket 322 * 323 * Returns true if connection less, false otherwise 324 */ 325 static bool tipc_sk_type_connectionless(struct sock *sk) 326 { 327 return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM; 328 } 329 330 /* tsk_peer_msg - verify if message was sent by connected port's peer 331 * 332 * Handles cases where the node's network address has changed from 333 * the default of <0.0.0> to its configured setting. 334 */ 335 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg) 336 { 337 struct sock *sk = &tsk->sk; 338 u32 self = tipc_own_addr(sock_net(sk)); 339 u32 peer_port = tsk_peer_port(tsk); 340 u32 orig_node, peer_node; 341 342 if (unlikely(!tipc_sk_connected(sk))) 343 return false; 344 345 if (unlikely(msg_origport(msg) != peer_port)) 346 return false; 347 348 orig_node = msg_orignode(msg); 349 peer_node = tsk_peer_node(tsk); 350 351 if (likely(orig_node == peer_node)) 352 return true; 353 354 if (!orig_node && peer_node == self) 355 return true; 356 357 if (!peer_node && orig_node == self) 358 return true; 359 360 return false; 361 } 362 363 /* tipc_set_sk_state - set the sk_state of the socket 364 * @sk: socket 365 * 366 * Caller must hold socket lock 367 * 368 * Returns 0 on success, errno otherwise 369 */ 370 static int tipc_set_sk_state(struct sock *sk, int state) 371 { 372 int oldsk_state = sk->sk_state; 373 int res = -EINVAL; 374 375 switch (state) { 376 case TIPC_OPEN: 377 res = 0; 378 break; 379 case TIPC_LISTEN: 380 case TIPC_CONNECTING: 381 if (oldsk_state == TIPC_OPEN) 382 res = 0; 383 break; 384 case TIPC_ESTABLISHED: 385 if (oldsk_state == TIPC_CONNECTING || 386 oldsk_state == TIPC_OPEN) 387 res = 0; 388 break; 389 case TIPC_DISCONNECTING: 390 if (oldsk_state == TIPC_CONNECTING || 391 oldsk_state == TIPC_ESTABLISHED) 392 res = 0; 393 break; 394 } 395 396 if (!res) 397 sk->sk_state = state; 398 399 return res; 400 } 401 402 static int tipc_sk_sock_err(struct socket *sock, long *timeout) 403 { 404 struct sock *sk = sock->sk; 405 int err = sock_error(sk); 406 int typ = sock->type; 407 408 if (err) 409 return err; 410 if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) { 411 if (sk->sk_state == TIPC_DISCONNECTING) 412 return -EPIPE; 413 else if (!tipc_sk_connected(sk)) 414 return -ENOTCONN; 415 } 416 if (!*timeout) 417 return -EAGAIN; 418 if (signal_pending(current)) 419 return sock_intr_errno(*timeout); 420 421 return 0; 422 } 423 424 #define tipc_wait_for_cond(sock_, timeo_, condition_) \ 425 ({ \ 426 DEFINE_WAIT_FUNC(wait_, woken_wake_function); \ 427 struct sock *sk_; \ 428 int rc_; \ 429 \ 430 while ((rc_ = !(condition_))) { \ 431 /* coupled with smp_wmb() in tipc_sk_proto_rcv() */ \ 432 smp_rmb(); \ 433 sk_ = (sock_)->sk; \ 434 rc_ = tipc_sk_sock_err((sock_), timeo_); \ 435 if (rc_) \ 436 break; \ 437 add_wait_queue(sk_sleep(sk_), &wait_); \ 438 release_sock(sk_); \ 439 *(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \ 440 sched_annotate_sleep(); \ 441 lock_sock(sk_); \ 442 remove_wait_queue(sk_sleep(sk_), &wait_); \ 443 } \ 444 rc_; \ 445 }) 446 447 /** 448 * tipc_sk_create - create a TIPC socket 449 * @net: network namespace (must be default network) 450 * @sock: pre-allocated socket structure 451 * @protocol: protocol indicator (must be 0) 452 * @kern: caused by kernel or by userspace? 453 * 454 * This routine creates additional data structures used by the TIPC socket, 455 * initializes them, and links them together. 456 * 457 * Return: 0 on success, errno otherwise 458 */ 459 static int tipc_sk_create(struct net *net, struct socket *sock, 460 int protocol, int kern) 461 { 462 const struct proto_ops *ops; 463 struct sock *sk; 464 struct tipc_sock *tsk; 465 struct tipc_msg *msg; 466 467 /* Validate arguments */ 468 if (unlikely(protocol != 0)) 469 return -EPROTONOSUPPORT; 470 471 switch (sock->type) { 472 case SOCK_STREAM: 473 ops = &stream_ops; 474 break; 475 case SOCK_SEQPACKET: 476 ops = &packet_ops; 477 break; 478 case SOCK_DGRAM: 479 case SOCK_RDM: 480 ops = &msg_ops; 481 break; 482 default: 483 return -EPROTOTYPE; 484 } 485 486 /* Allocate socket's protocol area */ 487 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern); 488 if (sk == NULL) 489 return -ENOMEM; 490 491 tsk = tipc_sk(sk); 492 tsk->max_pkt = MAX_PKT_DEFAULT; 493 tsk->maxnagle = 0; 494 tsk->nagle_start = NAGLE_START_INIT; 495 INIT_LIST_HEAD(&tsk->publications); 496 INIT_LIST_HEAD(&tsk->cong_links); 497 msg = &tsk->phdr; 498 499 /* Finish initializing socket data structures */ 500 sock->ops = ops; 501 sock_init_data(sock, sk); 502 tipc_set_sk_state(sk, TIPC_OPEN); 503 if (tipc_sk_insert(tsk)) { 504 sk_free(sk); 505 sock->sk = NULL; 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 disassociate 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_unsized *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, (struct sockaddr *)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_NONE, " "); 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 LIST_HEAD(dsts); 1015 int lookups = 0; 1016 int dstcnt, rc; 1017 bool cong; 1018 1019 ua->sa.type = msg_nametype(hdr); 1020 ua->scope = msg_lookup_scope(hdr); 1021 1022 while (++lookups < 4) { 1023 exclude = tipc_group_exclude(tsk->group); 1024 1025 first = NULL; 1026 1027 /* Look for a non-congested destination member, if any */ 1028 while (1) { 1029 if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt, 1030 exclude, false)) 1031 return -EHOSTUNREACH; 1032 tipc_dest_pop(&dsts, &node, &port); 1033 cong = tipc_group_cong(tsk->group, node, port, blks, 1034 &mbr); 1035 if (!cong) 1036 break; 1037 if (mbr == first) 1038 break; 1039 if (!first) 1040 first = mbr; 1041 } 1042 1043 /* Start over if destination was not in member list */ 1044 if (unlikely(!mbr)) 1045 continue; 1046 1047 if (likely(!cong && !tipc_dest_find(cong_links, node, 0))) 1048 break; 1049 1050 /* Block or return if destination link or member is congested */ 1051 rc = tipc_wait_for_cond(sock, &timeout, 1052 !tipc_dest_find(cong_links, node, 0) && 1053 tsk->group && 1054 !tipc_group_cong(tsk->group, node, port, 1055 blks, &mbr)); 1056 if (unlikely(rc)) 1057 return rc; 1058 1059 /* Send, unless destination disappeared while waiting */ 1060 if (likely(mbr)) 1061 break; 1062 } 1063 1064 if (unlikely(lookups >= 4)) 1065 return -EHOSTUNREACH; 1066 1067 rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen); 1068 1069 return rc ? rc : dlen; 1070 } 1071 1072 /** 1073 * tipc_send_group_bcast - send message to all members in communication group 1074 * @sock: socket structure 1075 * @m: message to send 1076 * @dlen: total length of message data 1077 * @timeout: timeout to wait for wakeup 1078 * 1079 * Called from function tipc_sendmsg(), which has done all sanity checks 1080 * Return: the number of bytes sent on success, or errno 1081 */ 1082 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m, 1083 int dlen, long timeout) 1084 { 1085 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name; 1086 struct sock *sk = sock->sk; 1087 struct net *net = sock_net(sk); 1088 struct tipc_sock *tsk = tipc_sk(sk); 1089 struct tipc_nlist *dsts; 1090 struct tipc_mc_method *method = &tsk->mc_method; 1091 bool ack = method->mandatory && method->rcast; 1092 int blks = tsk_blocks(MCAST_H_SIZE + dlen); 1093 struct tipc_msg *hdr = &tsk->phdr; 1094 int mtu = tipc_bcast_get_mtu(net); 1095 struct sk_buff_head pkts; 1096 int rc = -EHOSTUNREACH; 1097 1098 /* Block or return if any destination link or member is congested */ 1099 rc = tipc_wait_for_cond(sock, &timeout, 1100 !tsk->cong_link_cnt && tsk->group && 1101 !tipc_group_bc_cong(tsk->group, blks)); 1102 if (unlikely(rc)) 1103 return rc; 1104 1105 dsts = tipc_group_dests(tsk->group); 1106 if (!dsts->local && !dsts->remote) 1107 return -EHOSTUNREACH; 1108 1109 /* Complete message header */ 1110 if (ua) { 1111 msg_set_type(hdr, TIPC_GRP_MCAST_MSG); 1112 msg_set_nameinst(hdr, ua->sa.instance); 1113 } else { 1114 msg_set_type(hdr, TIPC_GRP_BCAST_MSG); 1115 msg_set_nameinst(hdr, 0); 1116 } 1117 msg_set_hdr_sz(hdr, GROUP_H_SIZE); 1118 msg_set_destport(hdr, 0); 1119 msg_set_destnode(hdr, 0); 1120 msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group)); 1121 1122 /* Avoid getting stuck with repeated forced replicasts */ 1123 msg_set_grp_bc_ack_req(hdr, ack); 1124 1125 /* Build message as chain of buffers */ 1126 __skb_queue_head_init(&pkts); 1127 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts); 1128 if (unlikely(rc != dlen)) 1129 return rc; 1130 1131 /* Send message */ 1132 rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt); 1133 if (unlikely(rc)) 1134 return rc; 1135 1136 /* Update broadcast sequence number and send windows */ 1137 tipc_group_update_bc_members(tsk->group, blks, ack); 1138 1139 /* Broadcast link is now free to choose method for next broadcast */ 1140 method->mandatory = false; 1141 method->expires = jiffies; 1142 1143 return dlen; 1144 } 1145 1146 /** 1147 * tipc_send_group_mcast - send message to all members with given identity 1148 * @sock: socket structure 1149 * @m: message to send 1150 * @dlen: total length of message data 1151 * @timeout: timeout to wait for wakeup 1152 * 1153 * Called from function tipc_sendmsg(), which has done all sanity checks 1154 * Return: the number of bytes sent on success, or errno 1155 */ 1156 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m, 1157 int dlen, long timeout) 1158 { 1159 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name; 1160 struct sock *sk = sock->sk; 1161 struct tipc_sock *tsk = tipc_sk(sk); 1162 struct tipc_group *grp = tsk->group; 1163 struct tipc_msg *hdr = &tsk->phdr; 1164 struct net *net = sock_net(sk); 1165 u32 dstcnt, exclude; 1166 LIST_HEAD(dsts); 1167 1168 ua->sa.type = msg_nametype(hdr); 1169 ua->scope = msg_lookup_scope(hdr); 1170 exclude = tipc_group_exclude(grp); 1171 1172 if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt, exclude, true)) 1173 return -EHOSTUNREACH; 1174 1175 if (dstcnt == 1) { 1176 tipc_dest_pop(&dsts, &ua->sk.node, &ua->sk.ref); 1177 return tipc_send_group_unicast(sock, m, dlen, timeout); 1178 } 1179 1180 tipc_dest_list_purge(&dsts); 1181 return tipc_send_group_bcast(sock, m, dlen, timeout); 1182 } 1183 1184 /** 1185 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets 1186 * @net: the associated network namespace 1187 * @arrvq: queue with arriving messages, to be cloned after destination lookup 1188 * @inputq: queue with cloned messages, delivered to socket after dest lookup 1189 * 1190 * Multi-threaded: parallel calls with reference to same queues may occur 1191 */ 1192 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq, 1193 struct sk_buff_head *inputq) 1194 { 1195 u32 self = tipc_own_addr(net); 1196 struct sk_buff *skb, *_skb; 1197 u32 portid, onode; 1198 struct sk_buff_head tmpq; 1199 struct list_head dports; 1200 struct tipc_msg *hdr; 1201 struct tipc_uaddr ua; 1202 int user, mtyp, hlen; 1203 1204 __skb_queue_head_init(&tmpq); 1205 INIT_LIST_HEAD(&dports); 1206 ua.addrtype = TIPC_SERVICE_RANGE; 1207 1208 /* tipc_skb_peek() increments the head skb's reference counter */ 1209 skb = tipc_skb_peek(arrvq, &inputq->lock); 1210 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) { 1211 hdr = buf_msg(skb); 1212 user = msg_user(hdr); 1213 mtyp = msg_type(hdr); 1214 hlen = skb_headroom(skb) + msg_hdr_sz(hdr); 1215 onode = msg_orignode(hdr); 1216 ua.sr.type = msg_nametype(hdr); 1217 ua.sr.lower = msg_namelower(hdr); 1218 ua.sr.upper = msg_nameupper(hdr); 1219 if (onode == self) 1220 ua.scope = TIPC_ANY_SCOPE; 1221 else 1222 ua.scope = TIPC_CLUSTER_SCOPE; 1223 1224 if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) { 1225 spin_lock_bh(&inputq->lock); 1226 if (skb_peek(arrvq) == skb) { 1227 __skb_dequeue(arrvq); 1228 __skb_queue_tail(inputq, skb); 1229 } 1230 kfree_skb(skb); 1231 spin_unlock_bh(&inputq->lock); 1232 continue; 1233 } 1234 1235 /* Group messages require exact scope match */ 1236 if (msg_in_group(hdr)) { 1237 ua.sr.lower = 0; 1238 ua.sr.upper = ~0; 1239 ua.scope = msg_lookup_scope(hdr); 1240 } 1241 1242 /* Create destination port list: */ 1243 tipc_nametbl_lookup_mcast_sockets(net, &ua, &dports); 1244 1245 /* Clone message per destination */ 1246 while (tipc_dest_pop(&dports, NULL, &portid)) { 1247 _skb = __pskb_copy(skb, hlen, GFP_ATOMIC); 1248 if (_skb) { 1249 msg_set_destport(buf_msg(_skb), portid); 1250 __skb_queue_tail(&tmpq, _skb); 1251 continue; 1252 } 1253 pr_warn("Failed to clone mcast rcv buffer\n"); 1254 } 1255 /* Append clones to inputq only if skb is still head of arrvq */ 1256 spin_lock_bh(&inputq->lock); 1257 if (skb_peek(arrvq) == skb) { 1258 skb_queue_splice_tail_init(&tmpq, inputq); 1259 /* Decrement the skb's refcnt */ 1260 kfree_skb(__skb_dequeue(arrvq)); 1261 } 1262 spin_unlock_bh(&inputq->lock); 1263 __skb_queue_purge(&tmpq); 1264 kfree_skb(skb); 1265 } 1266 tipc_sk_rcv(net, inputq); 1267 } 1268 1269 /* tipc_sk_push_backlog(): send accumulated buffers in socket write queue 1270 * when socket is in Nagle mode 1271 */ 1272 static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack) 1273 { 1274 struct sk_buff_head *txq = &tsk->sk.sk_write_queue; 1275 struct sk_buff *skb = skb_peek_tail(txq); 1276 struct net *net = sock_net(&tsk->sk); 1277 u32 dnode = tsk_peer_node(tsk); 1278 int rc; 1279 1280 if (nagle_ack) { 1281 tsk->pkt_cnt += skb_queue_len(txq); 1282 if (!tsk->pkt_cnt || tsk->msg_acc / tsk->pkt_cnt < 2) { 1283 tsk->oneway = 0; 1284 if (tsk->nagle_start < NAGLE_START_MAX) 1285 tsk->nagle_start *= 2; 1286 tsk->expect_ack = false; 1287 pr_debug("tsk %10u: bad nagle %u -> %u, next start %u!\n", 1288 tsk->portid, tsk->msg_acc, tsk->pkt_cnt, 1289 tsk->nagle_start); 1290 } else { 1291 tsk->nagle_start = NAGLE_START_INIT; 1292 if (skb) { 1293 msg_set_ack_required(buf_msg(skb)); 1294 tsk->expect_ack = true; 1295 } else { 1296 tsk->expect_ack = false; 1297 } 1298 } 1299 tsk->msg_acc = 0; 1300 tsk->pkt_cnt = 0; 1301 } 1302 1303 if (!skb || tsk->cong_link_cnt) 1304 return; 1305 1306 /* Do not send SYN again after congestion */ 1307 if (msg_is_syn(buf_msg(skb))) 1308 return; 1309 1310 if (tsk->msg_acc) 1311 tsk->pkt_cnt += skb_queue_len(txq); 1312 tsk->snt_unacked += tsk->snd_backlog; 1313 tsk->snd_backlog = 0; 1314 rc = tipc_node_xmit(net, txq, dnode, tsk->portid); 1315 if (rc == -ELINKCONG) 1316 tsk->cong_link_cnt = 1; 1317 } 1318 1319 /** 1320 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message 1321 * @tsk: receiving socket 1322 * @skb: pointer to message buffer. 1323 * @inputq: buffer list containing the buffers 1324 * @xmitq: output message area 1325 */ 1326 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb, 1327 struct sk_buff_head *inputq, 1328 struct sk_buff_head *xmitq) 1329 { 1330 struct tipc_msg *hdr = buf_msg(skb); 1331 u32 onode = tsk_own_node(tsk); 1332 struct sock *sk = &tsk->sk; 1333 int mtyp = msg_type(hdr); 1334 bool was_cong; 1335 1336 /* Ignore if connection cannot be validated: */ 1337 if (!tsk_peer_msg(tsk, hdr)) { 1338 trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!"); 1339 goto exit; 1340 } 1341 1342 if (unlikely(msg_errcode(hdr))) { 1343 tipc_set_sk_state(sk, TIPC_DISCONNECTING); 1344 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk), 1345 tsk_peer_port(tsk)); 1346 sk->sk_state_change(sk); 1347 1348 /* State change is ignored if socket already awake, 1349 * - convert msg to abort msg and add to inqueue 1350 */ 1351 msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE); 1352 msg_set_type(hdr, TIPC_CONN_MSG); 1353 msg_set_size(hdr, BASIC_H_SIZE); 1354 msg_set_hdr_sz(hdr, BASIC_H_SIZE); 1355 __skb_queue_tail(inputq, skb); 1356 return; 1357 } 1358 1359 tsk->probe_unacked = false; 1360 1361 if (mtyp == CONN_PROBE) { 1362 msg_set_type(hdr, CONN_PROBE_REPLY); 1363 if (tipc_msg_reverse(onode, &skb, TIPC_OK)) 1364 __skb_queue_tail(xmitq, skb); 1365 return; 1366 } else if (mtyp == CONN_ACK) { 1367 if (tsk->snt_unacked < msg_conn_ack(hdr)) 1368 goto exit; 1369 1370 was_cong = tsk_conn_cong(tsk); 1371 tipc_sk_push_backlog(tsk, msg_nagle_ack(hdr)); 1372 tsk->snt_unacked -= msg_conn_ack(hdr); 1373 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) 1374 tsk->snd_win = msg_adv_win(hdr); 1375 if (was_cong && !tsk_conn_cong(tsk)) 1376 sk->sk_write_space(sk); 1377 } else if (mtyp != CONN_PROBE_REPLY) { 1378 pr_warn("Received unknown CONN_PROTO msg\n"); 1379 } 1380 exit: 1381 kfree_skb(skb); 1382 } 1383 1384 /** 1385 * tipc_sendmsg - send message in connectionless manner 1386 * @sock: socket structure 1387 * @m: message to send 1388 * @dsz: amount of user data to be sent 1389 * 1390 * Message must have an destination specified explicitly. 1391 * Used for SOCK_RDM and SOCK_DGRAM messages, 1392 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections. 1393 * (Note: 'SYN+' is prohibited on SOCK_STREAM.) 1394 * 1395 * Return: the number of bytes sent on success, or errno otherwise 1396 */ 1397 static int tipc_sendmsg(struct socket *sock, 1398 struct msghdr *m, size_t dsz) 1399 { 1400 struct sock *sk = sock->sk; 1401 int ret; 1402 1403 lock_sock(sk); 1404 ret = __tipc_sendmsg(sock, m, dsz); 1405 release_sock(sk); 1406 1407 return ret; 1408 } 1409 1410 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen) 1411 { 1412 struct sock *sk = sock->sk; 1413 struct net *net = sock_net(sk); 1414 struct tipc_sock *tsk = tipc_sk(sk); 1415 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name; 1416 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT); 1417 struct list_head *clinks = &tsk->cong_links; 1418 bool syn = !tipc_sk_type_connectionless(sk); 1419 struct tipc_group *grp = tsk->group; 1420 struct tipc_msg *hdr = &tsk->phdr; 1421 struct tipc_socket_addr skaddr; 1422 struct sk_buff_head pkts; 1423 int atype, mtu, rc; 1424 1425 if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE)) 1426 return -EMSGSIZE; 1427 1428 if (ua) { 1429 if (!tipc_uaddr_valid(ua, m->msg_namelen)) 1430 return -EINVAL; 1431 atype = ua->addrtype; 1432 } 1433 1434 /* If socket belongs to a communication group follow other paths */ 1435 if (grp) { 1436 if (!ua) 1437 return tipc_send_group_bcast(sock, m, dlen, timeout); 1438 if (atype == TIPC_SERVICE_ADDR) 1439 return tipc_send_group_anycast(sock, m, dlen, timeout); 1440 if (atype == TIPC_SOCKET_ADDR) 1441 return tipc_send_group_unicast(sock, m, dlen, timeout); 1442 if (atype == TIPC_SERVICE_RANGE) 1443 return tipc_send_group_mcast(sock, m, dlen, timeout); 1444 return -EINVAL; 1445 } 1446 1447 if (!ua) { 1448 ua = (struct tipc_uaddr *)&tsk->peer; 1449 if (!syn && ua->family != AF_TIPC) 1450 return -EDESTADDRREQ; 1451 atype = ua->addrtype; 1452 } 1453 1454 if (unlikely(syn)) { 1455 if (sk->sk_state == TIPC_LISTEN) 1456 return -EPIPE; 1457 if (sk->sk_state != TIPC_OPEN) 1458 return -EISCONN; 1459 if (tsk->published) 1460 return -EOPNOTSUPP; 1461 if (atype == TIPC_SERVICE_ADDR) 1462 tsk->conn_addrtype = atype; 1463 msg_set_syn(hdr, 1); 1464 } 1465 1466 memset(&skaddr, 0, sizeof(skaddr)); 1467 1468 /* Determine destination */ 1469 if (atype == TIPC_SERVICE_RANGE) { 1470 return tipc_sendmcast(sock, ua, m, dlen, timeout); 1471 } else if (atype == TIPC_SERVICE_ADDR) { 1472 skaddr.node = ua->lookup_node; 1473 ua->scope = tipc_node2scope(skaddr.node); 1474 if (!tipc_nametbl_lookup_anycast(net, ua, &skaddr)) 1475 return -EHOSTUNREACH; 1476 } else if (atype == TIPC_SOCKET_ADDR) { 1477 skaddr = ua->sk; 1478 } else { 1479 return -EINVAL; 1480 } 1481 1482 /* Block or return if destination link is congested */ 1483 rc = tipc_wait_for_cond(sock, &timeout, 1484 !tipc_dest_find(clinks, skaddr.node, 0)); 1485 if (unlikely(rc)) 1486 return rc; 1487 1488 /* Finally build message header */ 1489 msg_set_destnode(hdr, skaddr.node); 1490 msg_set_destport(hdr, skaddr.ref); 1491 if (atype == TIPC_SERVICE_ADDR) { 1492 msg_set_type(hdr, TIPC_NAMED_MSG); 1493 msg_set_hdr_sz(hdr, NAMED_H_SIZE); 1494 msg_set_nametype(hdr, ua->sa.type); 1495 msg_set_nameinst(hdr, ua->sa.instance); 1496 msg_set_lookup_scope(hdr, ua->scope); 1497 } else { /* TIPC_SOCKET_ADDR */ 1498 msg_set_type(hdr, TIPC_DIRECT_MSG); 1499 msg_set_lookup_scope(hdr, 0); 1500 msg_set_hdr_sz(hdr, BASIC_H_SIZE); 1501 } 1502 1503 /* Add message body */ 1504 __skb_queue_head_init(&pkts); 1505 mtu = tipc_node_get_mtu(net, skaddr.node, tsk->portid, true); 1506 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts); 1507 if (unlikely(rc != dlen)) 1508 return rc; 1509 if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue))) { 1510 __skb_queue_purge(&pkts); 1511 return -ENOMEM; 1512 } 1513 1514 /* Send message */ 1515 trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " "); 1516 rc = tipc_node_xmit(net, &pkts, skaddr.node, tsk->portid); 1517 if (unlikely(rc == -ELINKCONG)) { 1518 tipc_dest_push(clinks, skaddr.node, 0); 1519 tsk->cong_link_cnt++; 1520 rc = 0; 1521 } 1522 1523 if (unlikely(syn && !rc)) { 1524 tipc_set_sk_state(sk, TIPC_CONNECTING); 1525 if (dlen && timeout) { 1526 timeout = msecs_to_jiffies(timeout); 1527 tipc_wait_for_connect(sock, &timeout); 1528 } 1529 } 1530 1531 return rc ? rc : dlen; 1532 } 1533 1534 /** 1535 * tipc_sendstream - send stream-oriented data 1536 * @sock: socket structure 1537 * @m: data to send 1538 * @dsz: total length of data to be transmitted 1539 * 1540 * Used for SOCK_STREAM data. 1541 * 1542 * Return: the number of bytes sent on success (or partial success), 1543 * or errno if no data sent 1544 */ 1545 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz) 1546 { 1547 struct sock *sk = sock->sk; 1548 int ret; 1549 1550 lock_sock(sk); 1551 ret = __tipc_sendstream(sock, m, dsz); 1552 release_sock(sk); 1553 1554 return ret; 1555 } 1556 1557 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen) 1558 { 1559 struct sock *sk = sock->sk; 1560 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name); 1561 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT); 1562 struct sk_buff_head *txq = &sk->sk_write_queue; 1563 struct tipc_sock *tsk = tipc_sk(sk); 1564 struct tipc_msg *hdr = &tsk->phdr; 1565 struct net *net = sock_net(sk); 1566 struct sk_buff *skb; 1567 u32 dnode = tsk_peer_node(tsk); 1568 int maxnagle = tsk->maxnagle; 1569 int maxpkt = tsk->max_pkt; 1570 int send, sent = 0; 1571 int blocks, rc = 0; 1572 1573 if (unlikely(dlen > INT_MAX)) 1574 return -EMSGSIZE; 1575 1576 /* Handle implicit connection setup */ 1577 if (unlikely(dest && sk->sk_state == TIPC_OPEN)) { 1578 rc = __tipc_sendmsg(sock, m, dlen); 1579 if (dlen && dlen == rc) { 1580 tsk->peer_caps = tipc_node_get_capabilities(net, dnode); 1581 tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr)); 1582 } 1583 return rc; 1584 } 1585 1586 do { 1587 rc = tipc_wait_for_cond(sock, &timeout, 1588 (!tsk->cong_link_cnt && 1589 !tsk_conn_cong(tsk) && 1590 tipc_sk_connected(sk))); 1591 if (unlikely(rc)) 1592 break; 1593 send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE); 1594 blocks = tsk->snd_backlog; 1595 if (tsk->oneway++ >= tsk->nagle_start && maxnagle && 1596 send <= maxnagle) { 1597 rc = tipc_msg_append(hdr, m, send, maxnagle, txq); 1598 if (unlikely(rc < 0)) 1599 break; 1600 blocks += rc; 1601 tsk->msg_acc++; 1602 if (blocks <= 64 && tsk->expect_ack) { 1603 tsk->snd_backlog = blocks; 1604 sent += send; 1605 break; 1606 } else if (blocks > 64) { 1607 tsk->pkt_cnt += skb_queue_len(txq); 1608 } else { 1609 skb = skb_peek_tail(txq); 1610 if (skb) { 1611 msg_set_ack_required(buf_msg(skb)); 1612 tsk->expect_ack = true; 1613 } else { 1614 tsk->expect_ack = false; 1615 } 1616 tsk->msg_acc = 0; 1617 tsk->pkt_cnt = 0; 1618 } 1619 } else { 1620 rc = tipc_msg_build(hdr, m, sent, send, maxpkt, txq); 1621 if (unlikely(rc != send)) 1622 break; 1623 blocks += tsk_inc(tsk, send + MIN_H_SIZE); 1624 } 1625 trace_tipc_sk_sendstream(sk, skb_peek(txq), 1626 TIPC_DUMP_SK_SNDQ, " "); 1627 rc = tipc_node_xmit(net, txq, dnode, tsk->portid); 1628 if (unlikely(rc == -ELINKCONG)) { 1629 tsk->cong_link_cnt = 1; 1630 rc = 0; 1631 } 1632 if (likely(!rc)) { 1633 tsk->snt_unacked += blocks; 1634 tsk->snd_backlog = 0; 1635 sent += send; 1636 } 1637 } while (sent < dlen && !rc); 1638 1639 return sent ? sent : rc; 1640 } 1641 1642 /** 1643 * tipc_send_packet - send a connection-oriented message 1644 * @sock: socket structure 1645 * @m: message to send 1646 * @dsz: length of data to be transmitted 1647 * 1648 * Used for SOCK_SEQPACKET messages. 1649 * 1650 * Return: the number of bytes sent on success, or errno otherwise 1651 */ 1652 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz) 1653 { 1654 if (dsz > TIPC_MAX_USER_MSG_SIZE) 1655 return -EMSGSIZE; 1656 1657 return tipc_sendstream(sock, m, dsz); 1658 } 1659 1660 /* tipc_sk_finish_conn - complete the setup of a connection 1661 */ 1662 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port, 1663 u32 peer_node) 1664 { 1665 struct sock *sk = &tsk->sk; 1666 struct net *net = sock_net(sk); 1667 struct tipc_msg *msg = &tsk->phdr; 1668 1669 msg_set_syn(msg, 0); 1670 msg_set_destnode(msg, peer_node); 1671 msg_set_destport(msg, peer_port); 1672 msg_set_type(msg, TIPC_CONN_MSG); 1673 msg_set_lookup_scope(msg, 0); 1674 msg_set_hdr_sz(msg, SHORT_H_SIZE); 1675 1676 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV); 1677 tipc_set_sk_state(sk, TIPC_ESTABLISHED); 1678 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port); 1679 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid, true); 1680 tsk->peer_caps = tipc_node_get_capabilities(net, peer_node); 1681 tsk_set_nagle(tsk); 1682 __skb_queue_purge(&sk->sk_write_queue); 1683 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) 1684 return; 1685 1686 /* Fall back to message based flow control */ 1687 tsk->rcv_win = FLOWCTL_MSG_WIN; 1688 tsk->snd_win = FLOWCTL_MSG_WIN; 1689 } 1690 1691 /** 1692 * tipc_sk_set_orig_addr - capture sender's address for received message 1693 * @m: descriptor for message info 1694 * @skb: received message 1695 * 1696 * Note: Address is not captured if not requested by receiver. 1697 */ 1698 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb) 1699 { 1700 DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name); 1701 struct tipc_msg *hdr = buf_msg(skb); 1702 1703 if (!srcaddr) 1704 return; 1705 1706 srcaddr->sock.family = AF_TIPC; 1707 srcaddr->sock.addrtype = TIPC_SOCKET_ADDR; 1708 srcaddr->sock.scope = 0; 1709 srcaddr->sock.addr.id.ref = msg_origport(hdr); 1710 srcaddr->sock.addr.id.node = msg_orignode(hdr); 1711 srcaddr->sock.addr.name.domain = 0; 1712 m->msg_namelen = sizeof(struct sockaddr_tipc); 1713 1714 if (!msg_in_group(hdr)) 1715 return; 1716 1717 /* Group message users may also want to know sending member's id */ 1718 srcaddr->member.family = AF_TIPC; 1719 srcaddr->member.addrtype = TIPC_SERVICE_ADDR; 1720 srcaddr->member.scope = 0; 1721 srcaddr->member.addr.name.name.type = msg_nametype(hdr); 1722 srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member; 1723 srcaddr->member.addr.name.domain = 0; 1724 m->msg_namelen = sizeof(*srcaddr); 1725 } 1726 1727 /** 1728 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message 1729 * @m: descriptor for message info 1730 * @skb: received message buffer 1731 * @tsk: TIPC port associated with message 1732 * 1733 * Note: Ancillary data is not captured if not requested by receiver. 1734 * 1735 * Return: 0 if successful, otherwise errno 1736 */ 1737 static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb, 1738 struct tipc_sock *tsk) 1739 { 1740 struct tipc_msg *hdr; 1741 u32 data[3] = {0,}; 1742 bool has_addr; 1743 int dlen, rc; 1744 1745 if (likely(m->msg_controllen == 0)) 1746 return 0; 1747 1748 hdr = buf_msg(skb); 1749 dlen = msg_data_sz(hdr); 1750 1751 /* Capture errored message object, if any */ 1752 if (msg_errcode(hdr)) { 1753 if (skb_linearize(skb)) 1754 return -ENOMEM; 1755 hdr = buf_msg(skb); 1756 data[0] = msg_errcode(hdr); 1757 data[1] = dlen; 1758 rc = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, data); 1759 if (rc || !dlen) 1760 return rc; 1761 rc = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, dlen, msg_data(hdr)); 1762 if (rc) 1763 return rc; 1764 } 1765 1766 /* Capture TIPC_SERVICE_ADDR/RANGE destination address, if any */ 1767 switch (msg_type(hdr)) { 1768 case TIPC_NAMED_MSG: 1769 has_addr = true; 1770 data[0] = msg_nametype(hdr); 1771 data[1] = msg_namelower(hdr); 1772 data[2] = data[1]; 1773 break; 1774 case TIPC_MCAST_MSG: 1775 has_addr = true; 1776 data[0] = msg_nametype(hdr); 1777 data[1] = msg_namelower(hdr); 1778 data[2] = msg_nameupper(hdr); 1779 break; 1780 case TIPC_CONN_MSG: 1781 has_addr = !!tsk->conn_addrtype; 1782 data[0] = msg_nametype(&tsk->phdr); 1783 data[1] = msg_nameinst(&tsk->phdr); 1784 data[2] = data[1]; 1785 break; 1786 default: 1787 has_addr = false; 1788 } 1789 if (!has_addr) 1790 return 0; 1791 return put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, data); 1792 } 1793 1794 static struct sk_buff *tipc_sk_build_ack(struct tipc_sock *tsk) 1795 { 1796 struct sock *sk = &tsk->sk; 1797 struct sk_buff *skb = NULL; 1798 struct tipc_msg *msg; 1799 u32 peer_port = tsk_peer_port(tsk); 1800 u32 dnode = tsk_peer_node(tsk); 1801 1802 if (!tipc_sk_connected(sk)) 1803 return NULL; 1804 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0, 1805 dnode, tsk_own_node(tsk), peer_port, 1806 tsk->portid, TIPC_OK); 1807 if (!skb) 1808 return NULL; 1809 msg = buf_msg(skb); 1810 msg_set_conn_ack(msg, tsk->rcv_unacked); 1811 tsk->rcv_unacked = 0; 1812 1813 /* Adjust to and advertize the correct window limit */ 1814 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) { 1815 tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf); 1816 msg_set_adv_win(msg, tsk->rcv_win); 1817 } 1818 return skb; 1819 } 1820 1821 static void tipc_sk_send_ack(struct tipc_sock *tsk) 1822 { 1823 struct sk_buff *skb; 1824 1825 skb = tipc_sk_build_ack(tsk); 1826 if (!skb) 1827 return; 1828 1829 tipc_node_xmit_skb(sock_net(&tsk->sk), skb, tsk_peer_node(tsk), 1830 msg_link_selector(buf_msg(skb))); 1831 } 1832 1833 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop) 1834 { 1835 struct sock *sk = sock->sk; 1836 DEFINE_WAIT_FUNC(wait, woken_wake_function); 1837 long timeo = *timeop; 1838 int err = sock_error(sk); 1839 1840 if (err) 1841 return err; 1842 1843 for (;;) { 1844 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) { 1845 if (sk->sk_shutdown & RCV_SHUTDOWN) { 1846 err = -ENOTCONN; 1847 break; 1848 } 1849 add_wait_queue(sk_sleep(sk), &wait); 1850 release_sock(sk); 1851 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo); 1852 sched_annotate_sleep(); 1853 lock_sock(sk); 1854 remove_wait_queue(sk_sleep(sk), &wait); 1855 } 1856 err = 0; 1857 if (!skb_queue_empty(&sk->sk_receive_queue)) 1858 break; 1859 err = -EAGAIN; 1860 if (!timeo) 1861 break; 1862 err = sock_intr_errno(timeo); 1863 if (signal_pending(current)) 1864 break; 1865 1866 err = sock_error(sk); 1867 if (err) 1868 break; 1869 } 1870 *timeop = timeo; 1871 return err; 1872 } 1873 1874 /** 1875 * tipc_recvmsg - receive packet-oriented message 1876 * @sock: network socket 1877 * @m: descriptor for message info 1878 * @buflen: length of user buffer area 1879 * @flags: receive flags 1880 * 1881 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages. 1882 * If the complete message doesn't fit in user area, truncate it. 1883 * 1884 * Return: size of returned message data, errno otherwise 1885 */ 1886 static int tipc_recvmsg(struct socket *sock, struct msghdr *m, 1887 size_t buflen, int flags) 1888 { 1889 struct sock *sk = sock->sk; 1890 bool connected = !tipc_sk_type_connectionless(sk); 1891 struct tipc_sock *tsk = tipc_sk(sk); 1892 int rc, err, hlen, dlen, copy; 1893 struct tipc_skb_cb *skb_cb; 1894 struct sk_buff_head xmitq; 1895 struct tipc_msg *hdr; 1896 struct sk_buff *skb; 1897 bool grp_evt; 1898 long timeout; 1899 1900 /* Catch invalid receive requests */ 1901 if (unlikely(!buflen)) 1902 return -EINVAL; 1903 1904 lock_sock(sk); 1905 if (unlikely(connected && sk->sk_state == TIPC_OPEN)) { 1906 rc = -ENOTCONN; 1907 goto exit; 1908 } 1909 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT); 1910 1911 /* Step rcv queue to first msg with data or error; wait if necessary */ 1912 do { 1913 rc = tipc_wait_for_rcvmsg(sock, &timeout); 1914 if (unlikely(rc)) 1915 goto exit; 1916 skb = skb_peek(&sk->sk_receive_queue); 1917 skb_cb = TIPC_SKB_CB(skb); 1918 hdr = buf_msg(skb); 1919 dlen = msg_data_sz(hdr); 1920 hlen = msg_hdr_sz(hdr); 1921 err = msg_errcode(hdr); 1922 grp_evt = msg_is_grp_evt(hdr); 1923 if (likely(dlen || err)) 1924 break; 1925 tsk_advance_rx_queue(sk); 1926 } while (1); 1927 1928 /* Collect msg meta data, including error code and rejected data */ 1929 tipc_sk_set_orig_addr(m, skb); 1930 rc = tipc_sk_anc_data_recv(m, skb, tsk); 1931 if (unlikely(rc)) 1932 goto exit; 1933 hdr = buf_msg(skb); 1934 1935 /* Capture data if non-error msg, otherwise just set return value */ 1936 if (likely(!err)) { 1937 int offset = skb_cb->bytes_read; 1938 1939 copy = min_t(size_t, dlen - offset, buflen); 1940 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy); 1941 if (unlikely(rc)) 1942 goto exit; 1943 if (unlikely(offset + copy < dlen)) { 1944 if (flags & MSG_EOR) { 1945 if (!(flags & MSG_PEEK)) 1946 skb_cb->bytes_read = offset + copy; 1947 } else { 1948 m->msg_flags |= MSG_TRUNC; 1949 skb_cb->bytes_read = 0; 1950 } 1951 } else { 1952 if (flags & MSG_EOR) 1953 m->msg_flags |= MSG_EOR; 1954 skb_cb->bytes_read = 0; 1955 } 1956 } else { 1957 copy = 0; 1958 rc = 0; 1959 if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control) { 1960 rc = -ECONNRESET; 1961 goto exit; 1962 } 1963 } 1964 1965 /* Mark message as group event if applicable */ 1966 if (unlikely(grp_evt)) { 1967 if (msg_grp_evt(hdr) == TIPC_WITHDRAWN) 1968 m->msg_flags |= MSG_EOR; 1969 m->msg_flags |= MSG_OOB; 1970 copy = 0; 1971 } 1972 1973 /* Caption of data or error code/rejected data was successful */ 1974 if (unlikely(flags & MSG_PEEK)) 1975 goto exit; 1976 1977 /* Send group flow control advertisement when applicable */ 1978 if (tsk->group && msg_in_group(hdr) && !grp_evt) { 1979 __skb_queue_head_init(&xmitq); 1980 tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen), 1981 msg_orignode(hdr), msg_origport(hdr), 1982 &xmitq); 1983 tipc_node_distr_xmit(sock_net(sk), &xmitq); 1984 } 1985 1986 if (skb_cb->bytes_read) 1987 goto exit; 1988 1989 tsk_advance_rx_queue(sk); 1990 1991 if (likely(!connected)) 1992 goto exit; 1993 1994 /* Send connection flow control advertisement when applicable */ 1995 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen); 1996 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE) 1997 tipc_sk_send_ack(tsk); 1998 exit: 1999 release_sock(sk); 2000 return rc ? rc : copy; 2001 } 2002 2003 /** 2004 * tipc_recvstream - receive stream-oriented data 2005 * @sock: network socket 2006 * @m: descriptor for message info 2007 * @buflen: total size of user buffer area 2008 * @flags: receive flags 2009 * 2010 * Used for SOCK_STREAM messages only. If not enough data is available 2011 * will optionally wait for more; never truncates data. 2012 * 2013 * Return: size of returned message data, errno otherwise 2014 */ 2015 static int tipc_recvstream(struct socket *sock, struct msghdr *m, 2016 size_t buflen, int flags) 2017 { 2018 struct sock *sk = sock->sk; 2019 struct tipc_sock *tsk = tipc_sk(sk); 2020 struct sk_buff *skb; 2021 struct tipc_msg *hdr; 2022 struct tipc_skb_cb *skb_cb; 2023 bool peek = flags & MSG_PEEK; 2024 int offset, required, copy, copied = 0; 2025 int hlen, dlen, err, rc; 2026 long timeout; 2027 2028 /* Catch invalid receive attempts */ 2029 if (unlikely(!buflen)) 2030 return -EINVAL; 2031 2032 lock_sock(sk); 2033 2034 if (unlikely(sk->sk_state == TIPC_OPEN)) { 2035 rc = -ENOTCONN; 2036 goto exit; 2037 } 2038 required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen); 2039 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT); 2040 2041 do { 2042 /* Look at first msg in receive queue; wait if necessary */ 2043 rc = tipc_wait_for_rcvmsg(sock, &timeout); 2044 if (unlikely(rc)) 2045 break; 2046 skb = skb_peek(&sk->sk_receive_queue); 2047 skb_cb = TIPC_SKB_CB(skb); 2048 hdr = buf_msg(skb); 2049 dlen = msg_data_sz(hdr); 2050 hlen = msg_hdr_sz(hdr); 2051 err = msg_errcode(hdr); 2052 2053 /* Discard any empty non-errored (SYN-) message */ 2054 if (unlikely(!dlen && !err)) { 2055 tsk_advance_rx_queue(sk); 2056 continue; 2057 } 2058 2059 /* Collect msg meta data, incl. error code and rejected data */ 2060 if (!copied) { 2061 tipc_sk_set_orig_addr(m, skb); 2062 rc = tipc_sk_anc_data_recv(m, skb, tsk); 2063 if (rc) 2064 break; 2065 hdr = buf_msg(skb); 2066 } 2067 2068 /* Copy data if msg ok, otherwise return error/partial data */ 2069 if (likely(!err)) { 2070 offset = skb_cb->bytes_read; 2071 copy = min_t(size_t, dlen - offset, buflen - copied); 2072 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy); 2073 if (unlikely(rc)) 2074 break; 2075 copied += copy; 2076 offset += copy; 2077 if (unlikely(offset < dlen)) { 2078 if (!peek) 2079 skb_cb->bytes_read = offset; 2080 break; 2081 } 2082 } else { 2083 rc = 0; 2084 if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control) 2085 rc = -ECONNRESET; 2086 if (copied || rc) 2087 break; 2088 } 2089 2090 if (unlikely(peek)) 2091 break; 2092 2093 tsk_advance_rx_queue(sk); 2094 2095 /* Send connection flow control advertisement when applicable */ 2096 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen); 2097 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE) 2098 tipc_sk_send_ack(tsk); 2099 2100 /* Exit if all requested data or FIN/error received */ 2101 if (copied == buflen || err) 2102 break; 2103 2104 } while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required); 2105 exit: 2106 release_sock(sk); 2107 return copied ? copied : rc; 2108 } 2109 2110 /** 2111 * tipc_write_space - wake up thread if port congestion is released 2112 * @sk: socket 2113 */ 2114 static void tipc_write_space(struct sock *sk) 2115 { 2116 struct socket_wq *wq; 2117 2118 rcu_read_lock(); 2119 wq = rcu_dereference(sk->sk_wq); 2120 if (skwq_has_sleeper(wq)) 2121 wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT | 2122 EPOLLWRNORM | EPOLLWRBAND); 2123 rcu_read_unlock(); 2124 } 2125 2126 /** 2127 * tipc_data_ready - wake up threads to indicate messages have been received 2128 * @sk: socket 2129 */ 2130 static void tipc_data_ready(struct sock *sk) 2131 { 2132 struct socket_wq *wq; 2133 2134 trace_sk_data_ready(sk); 2135 2136 rcu_read_lock(); 2137 wq = rcu_dereference(sk->sk_wq); 2138 if (skwq_has_sleeper(wq)) 2139 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN | 2140 EPOLLRDNORM | EPOLLRDBAND); 2141 rcu_read_unlock(); 2142 } 2143 2144 static void tipc_sock_destruct(struct sock *sk) 2145 { 2146 __skb_queue_purge(&sk->sk_receive_queue); 2147 } 2148 2149 static void tipc_sk_proto_rcv(struct sock *sk, 2150 struct sk_buff_head *inputq, 2151 struct sk_buff_head *xmitq) 2152 { 2153 struct sk_buff *skb = __skb_dequeue(inputq); 2154 struct tipc_sock *tsk = tipc_sk(sk); 2155 struct tipc_msg *hdr = buf_msg(skb); 2156 struct tipc_group *grp = tsk->group; 2157 bool wakeup = false; 2158 2159 switch (msg_user(hdr)) { 2160 case CONN_MANAGER: 2161 tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq); 2162 return; 2163 case SOCK_WAKEUP: 2164 tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0); 2165 /* coupled with smp_rmb() in tipc_wait_for_cond() */ 2166 smp_wmb(); 2167 tsk->cong_link_cnt--; 2168 wakeup = true; 2169 tipc_sk_push_backlog(tsk, false); 2170 break; 2171 case GROUP_PROTOCOL: 2172 tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq); 2173 break; 2174 case TOP_SRV: 2175 tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf, 2176 hdr, inputq, xmitq); 2177 break; 2178 default: 2179 break; 2180 } 2181 2182 if (wakeup) 2183 sk->sk_write_space(sk); 2184 2185 kfree_skb(skb); 2186 } 2187 2188 /** 2189 * tipc_sk_filter_connect - check incoming message for a connection-based socket 2190 * @tsk: TIPC socket 2191 * @skb: pointer to message buffer. 2192 * @xmitq: for Nagle ACK if any 2193 * Return: true if message should be added to receive queue, false otherwise 2194 */ 2195 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb, 2196 struct sk_buff_head *xmitq) 2197 { 2198 struct sock *sk = &tsk->sk; 2199 struct net *net = sock_net(sk); 2200 struct tipc_msg *hdr = buf_msg(skb); 2201 bool con_msg = msg_connected(hdr); 2202 u32 pport = tsk_peer_port(tsk); 2203 u32 pnode = tsk_peer_node(tsk); 2204 u32 oport = msg_origport(hdr); 2205 u32 onode = msg_orignode(hdr); 2206 int err = msg_errcode(hdr); 2207 unsigned long delay; 2208 2209 if (unlikely(msg_mcast(hdr))) 2210 return false; 2211 tsk->oneway = 0; 2212 2213 switch (sk->sk_state) { 2214 case TIPC_CONNECTING: 2215 /* Setup ACK */ 2216 if (likely(con_msg)) { 2217 if (err) 2218 break; 2219 tipc_sk_finish_conn(tsk, oport, onode); 2220 msg_set_importance(&tsk->phdr, msg_importance(hdr)); 2221 /* ACK+ message with data is added to receive queue */ 2222 if (msg_data_sz(hdr)) 2223 return true; 2224 /* Empty ACK-, - wake up sleeping connect() and drop */ 2225 sk->sk_state_change(sk); 2226 msg_set_dest_droppable(hdr, 1); 2227 return false; 2228 } 2229 /* Ignore connectionless message if not from listening socket */ 2230 if (oport != pport || onode != pnode) 2231 return false; 2232 2233 /* Rejected SYN */ 2234 if (err != TIPC_ERR_OVERLOAD) 2235 break; 2236 2237 /* Prepare for new setup attempt if we have a SYN clone */ 2238 if (skb_queue_empty(&sk->sk_write_queue)) 2239 break; 2240 get_random_bytes(&delay, 2); 2241 if (tsk->conn_timeout < 4) 2242 tsk->conn_timeout = 4; 2243 delay %= (tsk->conn_timeout / 4); 2244 delay = msecs_to_jiffies(delay + 100); 2245 sk_reset_timer(sk, &sk->sk_timer, jiffies + delay); 2246 return false; 2247 case TIPC_OPEN: 2248 case TIPC_DISCONNECTING: 2249 return false; 2250 case TIPC_LISTEN: 2251 /* Accept only SYN message */ 2252 if (!msg_is_syn(hdr) && 2253 tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT) 2254 return false; 2255 if (!con_msg && !err) 2256 return true; 2257 return false; 2258 case TIPC_ESTABLISHED: 2259 if (!skb_queue_empty(&sk->sk_write_queue)) 2260 tipc_sk_push_backlog(tsk, false); 2261 /* Accept only connection-based messages sent by peer */ 2262 if (likely(con_msg && !err && pport == oport && 2263 pnode == onode)) { 2264 if (msg_ack_required(hdr)) { 2265 struct sk_buff *skb; 2266 2267 skb = tipc_sk_build_ack(tsk); 2268 if (skb) { 2269 msg_set_nagle_ack(buf_msg(skb)); 2270 __skb_queue_tail(xmitq, skb); 2271 } 2272 } 2273 return true; 2274 } 2275 if (!tsk_peer_msg(tsk, hdr)) 2276 return false; 2277 if (!err) 2278 return true; 2279 tipc_set_sk_state(sk, TIPC_DISCONNECTING); 2280 tipc_node_remove_conn(net, pnode, tsk->portid); 2281 sk->sk_state_change(sk); 2282 return true; 2283 default: 2284 pr_err("Unknown sk_state %u\n", sk->sk_state); 2285 } 2286 /* Abort connection setup attempt */ 2287 tipc_set_sk_state(sk, TIPC_DISCONNECTING); 2288 sk->sk_err = ECONNREFUSED; 2289 sk->sk_state_change(sk); 2290 return true; 2291 } 2292 2293 /** 2294 * rcvbuf_limit - get proper overload limit of socket receive queue 2295 * @sk: socket 2296 * @skb: message 2297 * 2298 * For connection oriented messages, irrespective of importance, 2299 * default queue limit is 2 MB. 2300 * 2301 * For connectionless messages, queue limits are based on message 2302 * importance as follows: 2303 * 2304 * TIPC_LOW_IMPORTANCE (2 MB) 2305 * TIPC_MEDIUM_IMPORTANCE (4 MB) 2306 * TIPC_HIGH_IMPORTANCE (8 MB) 2307 * TIPC_CRITICAL_IMPORTANCE (16 MB) 2308 * 2309 * Return: overload limit according to corresponding message importance 2310 */ 2311 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb) 2312 { 2313 struct tipc_sock *tsk = tipc_sk(sk); 2314 struct tipc_msg *hdr = buf_msg(skb); 2315 2316 if (unlikely(msg_in_group(hdr))) 2317 return READ_ONCE(sk->sk_rcvbuf); 2318 2319 if (unlikely(!msg_connected(hdr))) 2320 return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr); 2321 2322 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL)) 2323 return READ_ONCE(sk->sk_rcvbuf); 2324 2325 return FLOWCTL_MSG_LIM; 2326 } 2327 2328 /** 2329 * tipc_sk_filter_rcv - validate incoming message 2330 * @sk: socket 2331 * @skb: pointer to message. 2332 * @xmitq: output message area (FIXME) 2333 * 2334 * Enqueues message on receive queue if acceptable; optionally handles 2335 * disconnect indication for a connected socket. 2336 * 2337 * Called with socket lock already taken 2338 */ 2339 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb, 2340 struct sk_buff_head *xmitq) 2341 { 2342 bool sk_conn = !tipc_sk_type_connectionless(sk); 2343 struct tipc_sock *tsk = tipc_sk(sk); 2344 struct tipc_group *grp = tsk->group; 2345 struct tipc_msg *hdr = buf_msg(skb); 2346 struct net *net = sock_net(sk); 2347 struct sk_buff_head inputq; 2348 int mtyp = msg_type(hdr); 2349 int limit, err = TIPC_OK; 2350 2351 trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " "); 2352 TIPC_SKB_CB(skb)->bytes_read = 0; 2353 __skb_queue_head_init(&inputq); 2354 __skb_queue_tail(&inputq, skb); 2355 2356 if (unlikely(!msg_isdata(hdr))) 2357 tipc_sk_proto_rcv(sk, &inputq, xmitq); 2358 2359 if (unlikely(grp)) { 2360 bool was_open = tsk->group_is_open; 2361 2362 tipc_group_filter_msg(grp, &inputq, xmitq); 2363 /* A departing member may have held the last outstanding 2364 * replicast ack, which can restore the open state inside 2365 * tipc_group_filter_msg(); wake up senders blocked in 2366 * poll()/write() if the group just became writable. 2367 */ 2368 if (!was_open && tsk->group_is_open) 2369 sk->sk_write_space(sk); 2370 } 2371 2372 if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG) 2373 tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq); 2374 2375 /* Validate and add to receive buffer if there is space */ 2376 while ((skb = __skb_dequeue(&inputq))) { 2377 hdr = buf_msg(skb); 2378 limit = rcvbuf_limit(sk, skb); 2379 if ((sk_conn && !tipc_sk_filter_connect(tsk, skb, xmitq)) || 2380 (!sk_conn && msg_connected(hdr)) || 2381 (!grp && msg_in_group(hdr))) 2382 err = TIPC_ERR_NO_PORT; 2383 else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) { 2384 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, 2385 "err_overload2!"); 2386 sk_drops_inc(sk); 2387 err = TIPC_ERR_OVERLOAD; 2388 } 2389 2390 if (unlikely(err)) { 2391 if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) { 2392 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, 2393 "@filter_rcv!"); 2394 __skb_queue_tail(xmitq, skb); 2395 } 2396 err = TIPC_OK; 2397 continue; 2398 } 2399 __skb_queue_tail(&sk->sk_receive_queue, skb); 2400 skb_set_owner_r(skb, sk); 2401 trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL, 2402 "rcvq >90% allocated!"); 2403 sk->sk_data_ready(sk); 2404 } 2405 } 2406 2407 /** 2408 * tipc_sk_backlog_rcv - handle incoming message from backlog queue 2409 * @sk: socket 2410 * @skb: message 2411 * 2412 * Caller must hold socket lock 2413 */ 2414 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb) 2415 { 2416 unsigned int before = sk_rmem_alloc_get(sk); 2417 struct sk_buff_head xmitq; 2418 unsigned int added; 2419 2420 __skb_queue_head_init(&xmitq); 2421 2422 tipc_sk_filter_rcv(sk, skb, &xmitq); 2423 added = sk_rmem_alloc_get(sk) - before; 2424 atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt); 2425 2426 /* Send pending response/rejected messages, if any */ 2427 tipc_node_distr_xmit(sock_net(sk), &xmitq); 2428 return 0; 2429 } 2430 2431 /** 2432 * tipc_sk_enqueue - extract all buffers with destination 'dport' from 2433 * inputq and try adding them to socket or backlog queue 2434 * @inputq: list of incoming buffers with potentially different destinations 2435 * @sk: socket where the buffers should be enqueued 2436 * @dport: port number for the socket 2437 * @xmitq: output queue 2438 * 2439 * Caller must hold socket lock 2440 */ 2441 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk, 2442 u32 dport, struct sk_buff_head *xmitq) 2443 { 2444 unsigned long time_limit = jiffies + usecs_to_jiffies(20000); 2445 struct sk_buff *skb; 2446 unsigned int lim; 2447 atomic_t *dcnt; 2448 u32 onode; 2449 2450 while (skb_queue_len(inputq)) { 2451 if (unlikely(time_after_eq(jiffies, time_limit))) 2452 return; 2453 2454 skb = tipc_skb_dequeue(inputq, dport); 2455 if (unlikely(!skb)) 2456 return; 2457 2458 /* Add message directly to receive queue if possible */ 2459 if (!sock_owned_by_user(sk)) { 2460 tipc_sk_filter_rcv(sk, skb, xmitq); 2461 continue; 2462 } 2463 2464 /* Try backlog, compensating for double-counted bytes */ 2465 dcnt = &tipc_sk(sk)->dupl_rcvcnt; 2466 if (!sk->sk_backlog.len) 2467 atomic_set(dcnt, 0); 2468 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt); 2469 if (likely(!sk_add_backlog(sk, skb, lim))) { 2470 trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_SK_BKLGQ, 2471 "bklg & rcvq >90% allocated!"); 2472 continue; 2473 } 2474 2475 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_SK_BKLGQ, "err_overload!"); 2476 /* Overload => reject message back to sender */ 2477 onode = tipc_own_addr(sock_net(sk)); 2478 sk_drops_inc(sk); 2479 if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) { 2480 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_SK_BKLGQ, 2481 "@sk_enqueue!"); 2482 __skb_queue_tail(xmitq, skb); 2483 } 2484 break; 2485 } 2486 } 2487 2488 /** 2489 * tipc_sk_rcv - handle a chain of incoming buffers 2490 * @net: the associated network namespace 2491 * @inputq: buffer list containing the buffers 2492 * Consumes all buffers in list until inputq is empty 2493 * Note: may be called in multiple threads referring to the same queue 2494 */ 2495 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq) 2496 { 2497 struct sk_buff_head xmitq; 2498 u32 dnode, dport = 0; 2499 int err; 2500 struct tipc_sock *tsk; 2501 struct sock *sk; 2502 struct sk_buff *skb; 2503 2504 __skb_queue_head_init(&xmitq); 2505 while (skb_queue_len(inputq)) { 2506 dport = tipc_skb_peek_port(inputq, dport); 2507 tsk = tipc_sk_lookup(net, dport); 2508 2509 if (likely(tsk)) { 2510 sk = &tsk->sk; 2511 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) { 2512 tipc_sk_enqueue(inputq, sk, dport, &xmitq); 2513 spin_unlock_bh(&sk->sk_lock.slock); 2514 } 2515 /* Send pending response/rejected messages, if any */ 2516 tipc_node_distr_xmit(sock_net(sk), &xmitq); 2517 sock_put(sk); 2518 continue; 2519 } 2520 /* No destination socket => dequeue skb if still there */ 2521 skb = tipc_skb_dequeue(inputq, dport); 2522 if (!skb) 2523 return; 2524 2525 /* Try secondary lookup if unresolved named message */ 2526 err = TIPC_ERR_NO_PORT; 2527 if (tipc_msg_lookup_dest(net, skb, &err)) 2528 goto xmit; 2529 2530 /* Prepare for message rejection */ 2531 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err)) 2532 continue; 2533 2534 trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!"); 2535 xmit: 2536 dnode = msg_destnode(buf_msg(skb)); 2537 tipc_node_xmit_skb(net, skb, dnode, dport); 2538 } 2539 } 2540 2541 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p) 2542 { 2543 DEFINE_WAIT_FUNC(wait, woken_wake_function); 2544 struct sock *sk = sock->sk; 2545 int done; 2546 2547 do { 2548 int err = sock_error(sk); 2549 if (err) 2550 return err; 2551 if (!*timeo_p) 2552 return -ETIMEDOUT; 2553 if (signal_pending(current)) 2554 return sock_intr_errno(*timeo_p); 2555 if (sk->sk_state == TIPC_DISCONNECTING) 2556 break; 2557 2558 add_wait_queue(sk_sleep(sk), &wait); 2559 done = sk_wait_event(sk, timeo_p, tipc_sk_connected(sk), 2560 &wait); 2561 remove_wait_queue(sk_sleep(sk), &wait); 2562 } while (!done); 2563 return 0; 2564 } 2565 2566 static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr) 2567 { 2568 if (addr->family != AF_TIPC) 2569 return false; 2570 if (addr->addrtype == TIPC_SERVICE_RANGE) 2571 return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper); 2572 return (addr->addrtype == TIPC_SERVICE_ADDR || 2573 addr->addrtype == TIPC_SOCKET_ADDR); 2574 } 2575 2576 /** 2577 * tipc_connect - establish a connection to another TIPC port 2578 * @sock: socket structure 2579 * @dest: socket address for destination port 2580 * @destlen: size of socket address data structure 2581 * @flags: file-related flags associated with socket 2582 * 2583 * Return: 0 on success, errno otherwise 2584 */ 2585 static int tipc_connect(struct socket *sock, struct sockaddr_unsized *dest, 2586 int destlen, int flags) 2587 { 2588 struct sock *sk = sock->sk; 2589 struct tipc_sock *tsk = tipc_sk(sk); 2590 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest; 2591 struct msghdr m = {NULL,}; 2592 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout; 2593 int previous; 2594 int res = 0; 2595 2596 if (destlen != sizeof(struct sockaddr_tipc)) 2597 return -EINVAL; 2598 2599 lock_sock(sk); 2600 2601 if (tsk->group) { 2602 res = -EINVAL; 2603 goto exit; 2604 } 2605 2606 if (dst->family == AF_UNSPEC) { 2607 memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc)); 2608 if (!tipc_sk_type_connectionless(sk)) 2609 res = -EINVAL; 2610 goto exit; 2611 } 2612 if (!tipc_sockaddr_is_sane(dst)) { 2613 res = -EINVAL; 2614 goto exit; 2615 } 2616 /* DGRAM/RDM connect(), just save the destaddr */ 2617 if (tipc_sk_type_connectionless(sk)) { 2618 memcpy(&tsk->peer, dest, destlen); 2619 goto exit; 2620 } else if (dst->addrtype == TIPC_SERVICE_RANGE) { 2621 res = -EINVAL; 2622 goto exit; 2623 } 2624 2625 previous = sk->sk_state; 2626 2627 switch (sk->sk_state) { 2628 case TIPC_OPEN: 2629 /* Send a 'SYN-' to destination */ 2630 m.msg_name = dest; 2631 m.msg_namelen = destlen; 2632 iov_iter_kvec(&m.msg_iter, ITER_SOURCE, NULL, 0, 0); 2633 2634 /* If connect is in non-blocking case, set MSG_DONTWAIT to 2635 * indicate send_msg() is never blocked. 2636 */ 2637 if (!timeout) 2638 m.msg_flags = MSG_DONTWAIT; 2639 2640 res = __tipc_sendmsg(sock, &m, 0); 2641 if ((res < 0) && (res != -EWOULDBLOCK)) 2642 goto exit; 2643 2644 /* Just entered TIPC_CONNECTING state; the only 2645 * difference is that return value in non-blocking 2646 * case is EINPROGRESS, rather than EALREADY. 2647 */ 2648 res = -EINPROGRESS; 2649 fallthrough; 2650 case TIPC_CONNECTING: 2651 if (!timeout) { 2652 if (previous == TIPC_CONNECTING) 2653 res = -EALREADY; 2654 goto exit; 2655 } 2656 timeout = msecs_to_jiffies(timeout); 2657 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */ 2658 res = tipc_wait_for_connect(sock, &timeout); 2659 break; 2660 case TIPC_ESTABLISHED: 2661 res = -EISCONN; 2662 break; 2663 default: 2664 res = -EINVAL; 2665 } 2666 2667 exit: 2668 release_sock(sk); 2669 return res; 2670 } 2671 2672 /** 2673 * tipc_listen - allow socket to listen for incoming connections 2674 * @sock: socket structure 2675 * @len: (unused) 2676 * 2677 * Return: 0 on success, errno otherwise 2678 */ 2679 static int tipc_listen(struct socket *sock, int len) 2680 { 2681 struct sock *sk = sock->sk; 2682 int res; 2683 2684 lock_sock(sk); 2685 res = tipc_set_sk_state(sk, TIPC_LISTEN); 2686 release_sock(sk); 2687 2688 return res; 2689 } 2690 2691 static int tipc_wait_for_accept(struct socket *sock, long timeo) 2692 { 2693 struct sock *sk = sock->sk; 2694 DEFINE_WAIT_FUNC(wait, woken_wake_function); 2695 int err; 2696 2697 /* True wake-one mechanism for incoming connections: only 2698 * one process gets woken up, not the 'whole herd'. 2699 * Since we do not 'race & poll' for established sockets 2700 * anymore, the common case will execute the loop only once. 2701 */ 2702 for (;;) { 2703 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) { 2704 add_wait_queue(sk_sleep(sk), &wait); 2705 release_sock(sk); 2706 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo); 2707 lock_sock(sk); 2708 remove_wait_queue(sk_sleep(sk), &wait); 2709 } 2710 err = 0; 2711 if (!skb_queue_empty(&sk->sk_receive_queue)) 2712 break; 2713 err = -EAGAIN; 2714 if (!timeo) 2715 break; 2716 err = sock_intr_errno(timeo); 2717 if (signal_pending(current)) 2718 break; 2719 } 2720 return err; 2721 } 2722 2723 /** 2724 * tipc_accept - wait for connection request 2725 * @sock: listening socket 2726 * @new_sock: new socket that is to be connected 2727 * @arg: arguments for accept 2728 * 2729 * Return: 0 on success, errno otherwise 2730 */ 2731 static int tipc_accept(struct socket *sock, struct socket *new_sock, 2732 struct proto_accept_arg *arg) 2733 { 2734 struct sock *new_sk, *sk = sock->sk; 2735 struct tipc_sock *new_tsock; 2736 struct msghdr m = {NULL,}; 2737 struct tipc_msg *msg; 2738 struct sk_buff *buf; 2739 long timeo; 2740 int res; 2741 2742 lock_sock(sk); 2743 2744 if (sk->sk_state != TIPC_LISTEN) { 2745 res = -EINVAL; 2746 goto exit; 2747 } 2748 timeo = sock_rcvtimeo(sk, arg->flags & O_NONBLOCK); 2749 res = tipc_wait_for_accept(sock, timeo); 2750 if (res) 2751 goto exit; 2752 2753 buf = skb_peek(&sk->sk_receive_queue); 2754 2755 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, arg->kern); 2756 if (res) 2757 goto exit; 2758 security_sk_clone(sock->sk, new_sock->sk); 2759 2760 new_sk = new_sock->sk; 2761 new_tsock = tipc_sk(new_sk); 2762 msg = buf_msg(buf); 2763 2764 /* we lock on new_sk; but lockdep sees the lock on sk */ 2765 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING); 2766 2767 /* 2768 * Reject any stray messages received by new socket 2769 * before the socket lock was taken (very, very unlikely) 2770 */ 2771 tsk_rej_rx_queue(new_sk, TIPC_ERR_NO_PORT); 2772 2773 /* Connect new socket to it's peer */ 2774 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg)); 2775 2776 tsk_set_importance(new_sk, msg_importance(msg)); 2777 if (msg_named(msg)) { 2778 new_tsock->conn_addrtype = TIPC_SERVICE_ADDR; 2779 msg_set_nametype(&new_tsock->phdr, msg_nametype(msg)); 2780 msg_set_nameinst(&new_tsock->phdr, msg_nameinst(msg)); 2781 } 2782 2783 /* 2784 * Respond to 'SYN-' by discarding it & returning 'ACK'. 2785 * Respond to 'SYN+' by queuing it on new socket & returning 'ACK'. 2786 */ 2787 if (!msg_data_sz(msg)) { 2788 tsk_advance_rx_queue(sk); 2789 } else { 2790 __skb_dequeue(&sk->sk_receive_queue); 2791 __skb_queue_head(&new_sk->sk_receive_queue, buf); 2792 skb_set_owner_r(buf, new_sk); 2793 } 2794 iov_iter_kvec(&m.msg_iter, ITER_SOURCE, NULL, 0, 0); 2795 __tipc_sendstream(new_sock, &m, 0); 2796 release_sock(new_sk); 2797 exit: 2798 release_sock(sk); 2799 return res; 2800 } 2801 2802 /** 2803 * tipc_shutdown - shutdown socket connection 2804 * @sock: socket structure 2805 * @how: direction to close (must be SHUT_RDWR) 2806 * 2807 * Terminates connection (if necessary), then purges socket's receive queue. 2808 * 2809 * Return: 0 on success, errno otherwise 2810 */ 2811 static int tipc_shutdown(struct socket *sock, int how) 2812 { 2813 struct sock *sk = sock->sk; 2814 int res; 2815 2816 if (how != SHUT_RDWR) 2817 return -EINVAL; 2818 2819 lock_sock(sk); 2820 2821 trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " "); 2822 __tipc_shutdown(sock, TIPC_CONN_SHUTDOWN); 2823 sk->sk_shutdown = SHUTDOWN_MASK; 2824 2825 if (sk->sk_state == TIPC_DISCONNECTING) { 2826 /* Discard any unreceived messages */ 2827 __skb_queue_purge(&sk->sk_receive_queue); 2828 2829 res = 0; 2830 } else { 2831 res = -ENOTCONN; 2832 } 2833 /* Wake up anyone sleeping in poll. */ 2834 sk->sk_state_change(sk); 2835 2836 release_sock(sk); 2837 return res; 2838 } 2839 2840 static void tipc_sk_check_probing_state(struct sock *sk, 2841 struct sk_buff_head *list) 2842 { 2843 struct tipc_sock *tsk = tipc_sk(sk); 2844 u32 pnode = tsk_peer_node(tsk); 2845 u32 pport = tsk_peer_port(tsk); 2846 u32 self = tsk_own_node(tsk); 2847 u32 oport = tsk->portid; 2848 struct sk_buff *skb; 2849 2850 if (tsk->probe_unacked) { 2851 tipc_set_sk_state(sk, TIPC_DISCONNECTING); 2852 sk->sk_err = ECONNABORTED; 2853 tipc_node_remove_conn(sock_net(sk), pnode, pport); 2854 sk->sk_state_change(sk); 2855 return; 2856 } 2857 /* Prepare new probe */ 2858 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0, 2859 pnode, self, pport, oport, TIPC_OK); 2860 if (skb) 2861 __skb_queue_tail(list, skb); 2862 tsk->probe_unacked = true; 2863 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV); 2864 } 2865 2866 static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list) 2867 { 2868 struct tipc_sock *tsk = tipc_sk(sk); 2869 2870 /* Try again later if dest link is congested */ 2871 if (tsk->cong_link_cnt) { 2872 sk_reset_timer(sk, &sk->sk_timer, 2873 jiffies + msecs_to_jiffies(100)); 2874 return; 2875 } 2876 /* Prepare SYN for retransmit */ 2877 tipc_msg_skb_clone(&sk->sk_write_queue, list); 2878 } 2879 2880 static void tipc_sk_timeout(struct timer_list *t) 2881 { 2882 struct sock *sk = timer_container_of(sk, t, sk_timer); 2883 struct tipc_sock *tsk = tipc_sk(sk); 2884 u32 pnode = tsk_peer_node(tsk); 2885 struct sk_buff_head list; 2886 int rc = 0; 2887 2888 __skb_queue_head_init(&list); 2889 bh_lock_sock(sk); 2890 2891 /* Try again later if socket is busy */ 2892 if (sock_owned_by_user(sk)) { 2893 sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20); 2894 bh_unlock_sock(sk); 2895 sock_put(sk); 2896 return; 2897 } 2898 2899 if (sk->sk_state == TIPC_ESTABLISHED) 2900 tipc_sk_check_probing_state(sk, &list); 2901 else if (sk->sk_state == TIPC_CONNECTING) 2902 tipc_sk_retry_connect(sk, &list); 2903 2904 bh_unlock_sock(sk); 2905 2906 if (!skb_queue_empty(&list)) 2907 rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid); 2908 2909 /* SYN messages may cause link congestion */ 2910 if (rc == -ELINKCONG) { 2911 tipc_dest_push(&tsk->cong_links, pnode, 0); 2912 tsk->cong_link_cnt = 1; 2913 } 2914 sock_put(sk); 2915 } 2916 2917 static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua) 2918 { 2919 struct sock *sk = &tsk->sk; 2920 struct net *net = sock_net(sk); 2921 struct tipc_socket_addr skaddr; 2922 struct publication *p; 2923 u32 key; 2924 2925 if (tipc_sk_connected(sk)) 2926 return -EINVAL; 2927 key = tsk->portid + tsk->pub_count + 1; 2928 if (key == tsk->portid) 2929 return -EADDRINUSE; 2930 skaddr.ref = tsk->portid; 2931 skaddr.node = tipc_own_addr(net); 2932 p = tipc_nametbl_publish(net, ua, &skaddr, key); 2933 if (unlikely(!p)) 2934 return -EINVAL; 2935 2936 list_add(&p->binding_sock, &tsk->publications); 2937 tsk->pub_count++; 2938 tsk->published = true; 2939 return 0; 2940 } 2941 2942 static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua) 2943 { 2944 struct net *net = sock_net(&tsk->sk); 2945 struct publication *safe, *p; 2946 struct tipc_uaddr _ua; 2947 int rc = -EINVAL; 2948 2949 list_for_each_entry_safe(p, safe, &tsk->publications, binding_sock) { 2950 if (!ua) { 2951 tipc_uaddr(&_ua, TIPC_SERVICE_RANGE, p->scope, 2952 p->sr.type, p->sr.lower, p->sr.upper); 2953 tipc_nametbl_withdraw(net, &_ua, &p->sk, p->key); 2954 continue; 2955 } 2956 /* Unbind specific publication */ 2957 if (p->scope != ua->scope) 2958 continue; 2959 if (p->sr.type != ua->sr.type) 2960 continue; 2961 if (p->sr.lower != ua->sr.lower) 2962 continue; 2963 if (p->sr.upper != ua->sr.upper) 2964 break; 2965 tipc_nametbl_withdraw(net, ua, &p->sk, p->key); 2966 rc = 0; 2967 break; 2968 } 2969 if (list_empty(&tsk->publications)) { 2970 tsk->published = 0; 2971 rc = 0; 2972 } 2973 return rc; 2974 } 2975 2976 /* tipc_sk_reinit: set non-zero address in all existing sockets 2977 * when we go from standalone to network mode. 2978 */ 2979 void tipc_sk_reinit(struct net *net) 2980 { 2981 struct tipc_net *tn = net_generic(net, tipc_net_id); 2982 struct rhashtable_iter iter; 2983 struct tipc_sock *tsk; 2984 struct tipc_msg *msg; 2985 2986 rhashtable_walk_enter(&tn->sk_rht, &iter); 2987 2988 do { 2989 rhashtable_walk_start(&iter); 2990 2991 while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) { 2992 sock_hold(&tsk->sk); 2993 rhashtable_walk_stop(&iter); 2994 lock_sock(&tsk->sk); 2995 msg = &tsk->phdr; 2996 msg_set_prevnode(msg, tipc_own_addr(net)); 2997 msg_set_orignode(msg, tipc_own_addr(net)); 2998 release_sock(&tsk->sk); 2999 rhashtable_walk_start(&iter); 3000 sock_put(&tsk->sk); 3001 } 3002 3003 rhashtable_walk_stop(&iter); 3004 } while (tsk == ERR_PTR(-EAGAIN)); 3005 3006 rhashtable_walk_exit(&iter); 3007 } 3008 3009 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid) 3010 { 3011 struct tipc_net *tn = net_generic(net, tipc_net_id); 3012 struct tipc_sock *tsk; 3013 3014 rcu_read_lock(); 3015 tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params); 3016 if (tsk) 3017 sock_hold(&tsk->sk); 3018 rcu_read_unlock(); 3019 3020 return tsk; 3021 } 3022 3023 static int tipc_sk_insert(struct tipc_sock *tsk) 3024 { 3025 struct sock *sk = &tsk->sk; 3026 struct net *net = sock_net(sk); 3027 struct tipc_net *tn = net_generic(net, tipc_net_id); 3028 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1; 3029 u32 portid = get_random_u32_below(remaining) + TIPC_MIN_PORT; 3030 3031 while (remaining--) { 3032 portid++; 3033 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT)) 3034 portid = TIPC_MIN_PORT; 3035 tsk->portid = portid; 3036 sock_hold(&tsk->sk); 3037 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node, 3038 tsk_rht_params)) 3039 return 0; 3040 sock_put(&tsk->sk); 3041 } 3042 3043 return -1; 3044 } 3045 3046 static void tipc_sk_remove(struct tipc_sock *tsk) 3047 { 3048 struct sock *sk = &tsk->sk; 3049 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id); 3050 3051 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) 3052 __sock_put(sk); 3053 } 3054 3055 static const struct rhashtable_params tsk_rht_params = { 3056 .nelem_hint = 192, 3057 .head_offset = offsetof(struct tipc_sock, node), 3058 .key_offset = offsetof(struct tipc_sock, portid), 3059 .key_len = sizeof(u32), /* portid */ 3060 .max_size = 1048576, 3061 .min_size = 256, 3062 .automatic_shrinking = true, 3063 }; 3064 3065 int tipc_sk_rht_init(struct net *net) 3066 { 3067 struct tipc_net *tn = net_generic(net, tipc_net_id); 3068 3069 return rhashtable_init(&tn->sk_rht, &tsk_rht_params); 3070 } 3071 3072 void tipc_sk_rht_destroy(struct net *net) 3073 { 3074 struct tipc_net *tn = net_generic(net, tipc_net_id); 3075 3076 /* Wait for socket readers to complete */ 3077 synchronize_net(); 3078 3079 rhashtable_destroy(&tn->sk_rht); 3080 } 3081 3082 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq) 3083 { 3084 struct net *net = sock_net(&tsk->sk); 3085 struct tipc_group *grp = tsk->group; 3086 struct tipc_msg *hdr = &tsk->phdr; 3087 struct tipc_uaddr ua; 3088 int rc; 3089 3090 if (mreq->type < TIPC_RESERVED_TYPES) 3091 return -EACCES; 3092 if (mreq->scope > TIPC_NODE_SCOPE) 3093 return -EINVAL; 3094 if (mreq->scope != TIPC_NODE_SCOPE) 3095 mreq->scope = TIPC_CLUSTER_SCOPE; 3096 if (grp) 3097 return -EACCES; 3098 grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open); 3099 if (!grp) 3100 return -ENOMEM; 3101 tsk->group = grp; 3102 msg_set_lookup_scope(hdr, mreq->scope); 3103 msg_set_nametype(hdr, mreq->type); 3104 msg_set_dest_droppable(hdr, true); 3105 tipc_uaddr(&ua, TIPC_SERVICE_RANGE, mreq->scope, 3106 mreq->type, mreq->instance, mreq->instance); 3107 tipc_nametbl_build_group(net, grp, &ua); 3108 rc = tipc_sk_publish(tsk, &ua); 3109 if (rc) { 3110 tipc_group_delete(net, grp); 3111 tsk->group = NULL; 3112 return rc; 3113 } 3114 /* Eliminate any risk that a broadcast overtakes sent JOINs */ 3115 tsk->mc_method.rcast = true; 3116 tsk->mc_method.mandatory = true; 3117 tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf); 3118 return rc; 3119 } 3120 3121 static int tipc_sk_leave(struct tipc_sock *tsk) 3122 { 3123 struct net *net = sock_net(&tsk->sk); 3124 struct tipc_group *grp = tsk->group; 3125 struct tipc_uaddr ua; 3126 int scope; 3127 3128 if (!grp) 3129 return -EINVAL; 3130 ua.addrtype = TIPC_SERVICE_RANGE; 3131 tipc_group_self(grp, &ua.sr, &scope); 3132 ua.scope = scope; 3133 tipc_group_delete(net, grp); 3134 tsk->group = NULL; 3135 tipc_sk_withdraw(tsk, &ua); 3136 return 0; 3137 } 3138 3139 /** 3140 * tipc_setsockopt - set socket option 3141 * @sock: socket structure 3142 * @lvl: option level 3143 * @opt: option identifier 3144 * @ov: pointer to new option value 3145 * @ol: length of option value 3146 * 3147 * For stream sockets only, accepts and ignores all IPPROTO_TCP options 3148 * (to ease compatibility). 3149 * 3150 * Return: 0 on success, errno otherwise 3151 */ 3152 static int tipc_setsockopt(struct socket *sock, int lvl, int opt, 3153 sockptr_t ov, unsigned int ol) 3154 { 3155 struct sock *sk = sock->sk; 3156 struct tipc_sock *tsk = tipc_sk(sk); 3157 struct tipc_group_req mreq; 3158 u32 value = 0; 3159 int res = 0; 3160 3161 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM)) 3162 return 0; 3163 if (lvl != SOL_TIPC) 3164 return -ENOPROTOOPT; 3165 3166 switch (opt) { 3167 case TIPC_IMPORTANCE: 3168 case TIPC_SRC_DROPPABLE: 3169 case TIPC_DEST_DROPPABLE: 3170 case TIPC_CONN_TIMEOUT: 3171 case TIPC_NODELAY: 3172 if (ol < sizeof(value)) 3173 return -EINVAL; 3174 if (copy_from_sockptr(&value, ov, sizeof(u32))) 3175 return -EFAULT; 3176 break; 3177 case TIPC_GROUP_JOIN: 3178 if (ol < sizeof(mreq)) 3179 return -EINVAL; 3180 if (copy_from_sockptr(&mreq, ov, sizeof(mreq))) 3181 return -EFAULT; 3182 break; 3183 default: 3184 if (!sockptr_is_null(ov) || ol) 3185 return -EINVAL; 3186 } 3187 3188 lock_sock(sk); 3189 3190 switch (opt) { 3191 case TIPC_IMPORTANCE: 3192 res = tsk_set_importance(sk, value); 3193 break; 3194 case TIPC_SRC_DROPPABLE: 3195 if (sock->type != SOCK_STREAM) 3196 tsk_set_unreliable(tsk, value); 3197 else 3198 res = -ENOPROTOOPT; 3199 break; 3200 case TIPC_DEST_DROPPABLE: 3201 tsk_set_unreturnable(tsk, value); 3202 break; 3203 case TIPC_CONN_TIMEOUT: 3204 tipc_sk(sk)->conn_timeout = value; 3205 break; 3206 case TIPC_MCAST_BROADCAST: 3207 tsk->mc_method.rcast = false; 3208 tsk->mc_method.mandatory = true; 3209 break; 3210 case TIPC_MCAST_REPLICAST: 3211 tsk->mc_method.rcast = true; 3212 tsk->mc_method.mandatory = true; 3213 break; 3214 case TIPC_GROUP_JOIN: 3215 res = tipc_sk_join(tsk, &mreq); 3216 break; 3217 case TIPC_GROUP_LEAVE: 3218 res = tipc_sk_leave(tsk); 3219 break; 3220 case TIPC_NODELAY: 3221 tsk->nodelay = !!value; 3222 tsk_set_nagle(tsk); 3223 break; 3224 default: 3225 res = -EINVAL; 3226 } 3227 3228 release_sock(sk); 3229 3230 return res; 3231 } 3232 3233 /** 3234 * tipc_getsockopt - get socket option 3235 * @sock: socket structure 3236 * @lvl: option level 3237 * @opt: option identifier 3238 * @sopt: socket option container (input buffer length, output value/length) 3239 * 3240 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options 3241 * (to ease compatibility). 3242 * 3243 * Return: 0 on success, errno otherwise 3244 */ 3245 static int tipc_getsockopt(struct socket *sock, int lvl, int opt, 3246 sockopt_t *sopt) 3247 { 3248 struct sock *sk = sock->sk; 3249 struct tipc_sock *tsk = tipc_sk(sk); 3250 struct tipc_service_range seq; 3251 int len, scope; 3252 int res = 0; 3253 u32 value; 3254 3255 if (lvl == IPPROTO_TCP && sock->type == SOCK_STREAM) { 3256 sopt->optlen = 0; 3257 return 0; 3258 } 3259 if (lvl != SOL_TIPC) 3260 return -ENOPROTOOPT; 3261 len = sopt->optlen; 3262 3263 lock_sock(sk); 3264 3265 switch (opt) { 3266 case TIPC_IMPORTANCE: 3267 value = tsk_importance(tsk); 3268 break; 3269 case TIPC_SRC_DROPPABLE: 3270 value = tsk_unreliable(tsk); 3271 break; 3272 case TIPC_DEST_DROPPABLE: 3273 value = tsk_unreturnable(tsk); 3274 break; 3275 case TIPC_CONN_TIMEOUT: 3276 value = tsk->conn_timeout; 3277 break; 3278 case TIPC_NODE_RECVQ_DEPTH: 3279 value = 0; /* was tipc_queue_size, now obsolete */ 3280 break; 3281 case TIPC_SOCK_RECVQ_DEPTH: 3282 value = skb_queue_len(&sk->sk_receive_queue); 3283 break; 3284 case TIPC_SOCK_RECVQ_USED: 3285 value = sk_rmem_alloc_get(sk); 3286 break; 3287 case TIPC_GROUP_JOIN: 3288 seq.type = 0; 3289 if (tsk->group) 3290 tipc_group_self(tsk->group, &seq, &scope); 3291 value = seq.type; 3292 break; 3293 default: 3294 res = -EINVAL; 3295 } 3296 3297 release_sock(sk); 3298 3299 if (res) 3300 return res; 3301 3302 if (len < sizeof(value)) 3303 return -EINVAL; 3304 3305 if (copy_to_iter(&value, sizeof(value), &sopt->iter_out) != 3306 sizeof(value)) 3307 return -EFAULT; 3308 sopt->optlen = sizeof(value); 3309 3310 return 0; 3311 } 3312 3313 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 3314 { 3315 struct net *net = sock_net(sock->sk); 3316 struct tipc_sioc_nodeid_req nr = {0}; 3317 struct tipc_sioc_ln_req lnr; 3318 void __user *argp = (void __user *)arg; 3319 3320 switch (cmd) { 3321 case SIOCGETLINKNAME: 3322 if (copy_from_user(&lnr, argp, sizeof(lnr))) 3323 return -EFAULT; 3324 if (!tipc_node_get_linkname(net, 3325 lnr.bearer_id & 0xffff, lnr.peer, 3326 lnr.linkname, TIPC_MAX_LINK_NAME)) { 3327 if (copy_to_user(argp, &lnr, sizeof(lnr))) 3328 return -EFAULT; 3329 return 0; 3330 } 3331 return -EADDRNOTAVAIL; 3332 case SIOCGETNODEID: 3333 if (copy_from_user(&nr, argp, sizeof(nr))) 3334 return -EFAULT; 3335 if (!tipc_node_get_id(net, nr.peer, nr.node_id)) 3336 return -EADDRNOTAVAIL; 3337 if (copy_to_user(argp, &nr, sizeof(nr))) 3338 return -EFAULT; 3339 return 0; 3340 default: 3341 return -ENOIOCTLCMD; 3342 } 3343 } 3344 3345 static int tipc_socketpair(struct socket *sock1, struct socket *sock2) 3346 { 3347 struct tipc_sock *tsk2 = tipc_sk(sock2->sk); 3348 struct tipc_sock *tsk1 = tipc_sk(sock1->sk); 3349 u32 onode = tipc_own_addr(sock_net(sock1->sk)); 3350 3351 tsk1->peer.family = AF_TIPC; 3352 tsk1->peer.addrtype = TIPC_SOCKET_ADDR; 3353 tsk1->peer.scope = TIPC_NODE_SCOPE; 3354 tsk1->peer.addr.id.ref = tsk2->portid; 3355 tsk1->peer.addr.id.node = onode; 3356 tsk2->peer.family = AF_TIPC; 3357 tsk2->peer.addrtype = TIPC_SOCKET_ADDR; 3358 tsk2->peer.scope = TIPC_NODE_SCOPE; 3359 tsk2->peer.addr.id.ref = tsk1->portid; 3360 tsk2->peer.addr.id.node = onode; 3361 3362 tipc_sk_finish_conn(tsk1, tsk2->portid, onode); 3363 tipc_sk_finish_conn(tsk2, tsk1->portid, onode); 3364 return 0; 3365 } 3366 3367 /* Protocol switches for the various types of TIPC sockets */ 3368 3369 static const struct proto_ops msg_ops = { 3370 .owner = THIS_MODULE, 3371 .family = AF_TIPC, 3372 .release = tipc_release, 3373 .bind = tipc_bind, 3374 .connect = tipc_connect, 3375 .socketpair = tipc_socketpair, 3376 .accept = sock_no_accept, 3377 .getname = tipc_getname, 3378 .poll = tipc_poll, 3379 .ioctl = tipc_ioctl, 3380 .listen = sock_no_listen, 3381 .shutdown = tipc_shutdown, 3382 .setsockopt = tipc_setsockopt, 3383 .getsockopt_iter = tipc_getsockopt, 3384 .sendmsg = tipc_sendmsg, 3385 .recvmsg = tipc_recvmsg, 3386 .mmap = sock_no_mmap, 3387 }; 3388 3389 static const struct proto_ops packet_ops = { 3390 .owner = THIS_MODULE, 3391 .family = AF_TIPC, 3392 .release = tipc_release, 3393 .bind = tipc_bind, 3394 .connect = tipc_connect, 3395 .socketpair = tipc_socketpair, 3396 .accept = tipc_accept, 3397 .getname = tipc_getname, 3398 .poll = tipc_poll, 3399 .ioctl = tipc_ioctl, 3400 .listen = tipc_listen, 3401 .shutdown = tipc_shutdown, 3402 .setsockopt = tipc_setsockopt, 3403 .getsockopt_iter = tipc_getsockopt, 3404 .sendmsg = tipc_send_packet, 3405 .recvmsg = tipc_recvmsg, 3406 .mmap = sock_no_mmap, 3407 }; 3408 3409 static const struct proto_ops stream_ops = { 3410 .owner = THIS_MODULE, 3411 .family = AF_TIPC, 3412 .release = tipc_release, 3413 .bind = tipc_bind, 3414 .connect = tipc_connect, 3415 .socketpair = tipc_socketpair, 3416 .accept = tipc_accept, 3417 .getname = tipc_getname, 3418 .poll = tipc_poll, 3419 .ioctl = tipc_ioctl, 3420 .listen = tipc_listen, 3421 .shutdown = tipc_shutdown, 3422 .setsockopt = tipc_setsockopt, 3423 .getsockopt_iter = tipc_getsockopt, 3424 .sendmsg = tipc_sendstream, 3425 .recvmsg = tipc_recvstream, 3426 .mmap = sock_no_mmap, 3427 }; 3428 3429 static const struct net_proto_family tipc_family_ops = { 3430 .owner = THIS_MODULE, 3431 .family = AF_TIPC, 3432 .create = tipc_sk_create 3433 }; 3434 3435 static struct proto tipc_proto = { 3436 .name = "TIPC", 3437 .owner = THIS_MODULE, 3438 .obj_size = sizeof(struct tipc_sock), 3439 .sysctl_rmem = sysctl_tipc_rmem 3440 }; 3441 3442 /** 3443 * tipc_socket_init - initialize TIPC socket interface 3444 * 3445 * Return: 0 on success, errno otherwise 3446 */ 3447 int tipc_socket_init(void) 3448 { 3449 int res; 3450 3451 res = proto_register(&tipc_proto, 1); 3452 if (res) { 3453 pr_err("Failed to register TIPC protocol type\n"); 3454 goto out; 3455 } 3456 3457 res = sock_register(&tipc_family_ops); 3458 if (res) { 3459 pr_err("Failed to register TIPC socket type\n"); 3460 proto_unregister(&tipc_proto); 3461 goto out; 3462 } 3463 out: 3464 return res; 3465 } 3466 3467 /** 3468 * tipc_socket_stop - stop TIPC socket interface 3469 */ 3470 void tipc_socket_stop(void) 3471 { 3472 sock_unregister(tipc_family_ops.family); 3473 proto_unregister(&tipc_proto); 3474 } 3475 3476 /* Caller should hold socket lock for the passed tipc socket. */ 3477 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk) 3478 { 3479 u32 peer_node, peer_port; 3480 u32 conn_type, conn_instance; 3481 struct nlattr *nest; 3482 3483 peer_node = tsk_peer_node(tsk); 3484 peer_port = tsk_peer_port(tsk); 3485 conn_type = msg_nametype(&tsk->phdr); 3486 conn_instance = msg_nameinst(&tsk->phdr); 3487 nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON); 3488 if (!nest) 3489 return -EMSGSIZE; 3490 3491 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node)) 3492 goto msg_full; 3493 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port)) 3494 goto msg_full; 3495 3496 if (tsk->conn_addrtype != 0) { 3497 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG)) 3498 goto msg_full; 3499 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, conn_type)) 3500 goto msg_full; 3501 if (nla_put_u32(skb, TIPC_NLA_CON_INST, conn_instance)) 3502 goto msg_full; 3503 } 3504 nla_nest_end(skb, nest); 3505 3506 return 0; 3507 3508 msg_full: 3509 nla_nest_cancel(skb, nest); 3510 3511 return -EMSGSIZE; 3512 } 3513 3514 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock 3515 *tsk) 3516 { 3517 struct net *net = sock_net(skb->sk); 3518 struct sock *sk = &tsk->sk; 3519 3520 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) || 3521 nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net))) 3522 return -EMSGSIZE; 3523 3524 if (tipc_sk_connected(sk)) { 3525 if (__tipc_nl_add_sk_con(skb, tsk)) 3526 return -EMSGSIZE; 3527 } else if (!list_empty(&tsk->publications)) { 3528 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL)) 3529 return -EMSGSIZE; 3530 } 3531 return 0; 3532 } 3533 3534 /* Caller should hold socket lock for the passed tipc socket. */ 3535 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb, 3536 struct tipc_sock *tsk) 3537 { 3538 struct nlattr *attrs; 3539 void *hdr; 3540 3541 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq, 3542 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET); 3543 if (!hdr) 3544 goto msg_cancel; 3545 3546 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK); 3547 if (!attrs) 3548 goto genlmsg_cancel; 3549 3550 if (__tipc_nl_add_sk_info(skb, tsk)) 3551 goto attr_msg_cancel; 3552 3553 nla_nest_end(skb, attrs); 3554 genlmsg_end(skb, hdr); 3555 3556 return 0; 3557 3558 attr_msg_cancel: 3559 nla_nest_cancel(skb, attrs); 3560 genlmsg_cancel: 3561 genlmsg_cancel(skb, hdr); 3562 msg_cancel: 3563 return -EMSGSIZE; 3564 } 3565 3566 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb, 3567 int (*skb_handler)(struct sk_buff *skb, 3568 struct netlink_callback *cb, 3569 struct tipc_sock *tsk)) 3570 { 3571 struct rhashtable_iter *iter = (void *)cb->args[4]; 3572 struct tipc_sock *tsk; 3573 int err; 3574 3575 rhashtable_walk_start(iter); 3576 while ((tsk = rhashtable_walk_next(iter)) != NULL) { 3577 if (IS_ERR(tsk)) { 3578 if (PTR_ERR(tsk) == -EAGAIN) 3579 continue; 3580 break; 3581 } 3582 3583 sock_hold(&tsk->sk); 3584 rhashtable_walk_stop(iter); 3585 lock_sock(&tsk->sk); 3586 err = skb_handler(skb, cb, tsk); 3587 if (err) { 3588 release_sock(&tsk->sk); 3589 sock_put(&tsk->sk); 3590 goto out; 3591 } 3592 release_sock(&tsk->sk); 3593 rhashtable_walk_start(iter); 3594 sock_put(&tsk->sk); 3595 } 3596 rhashtable_walk_stop(iter); 3597 out: 3598 return skb->len; 3599 } 3600 EXPORT_SYMBOL(tipc_nl_sk_walk); 3601 3602 int tipc_dump_start(struct netlink_callback *cb) 3603 { 3604 return __tipc_dump_start(cb, sock_net(cb->skb->sk)); 3605 } 3606 EXPORT_SYMBOL(tipc_dump_start); 3607 3608 int __tipc_dump_start(struct netlink_callback *cb, struct net *net) 3609 { 3610 /* tipc_nl_name_table_dump() uses cb->args[0...3]. */ 3611 struct rhashtable_iter *iter = (void *)cb->args[4]; 3612 struct tipc_net *tn = tipc_net(net); 3613 3614 if (!iter) { 3615 iter = kmalloc_obj(*iter); 3616 if (!iter) 3617 return -ENOMEM; 3618 3619 cb->args[4] = (long)iter; 3620 } 3621 3622 rhashtable_walk_enter(&tn->sk_rht, iter); 3623 return 0; 3624 } 3625 3626 int tipc_dump_done(struct netlink_callback *cb) 3627 { 3628 struct rhashtable_iter *hti = (void *)cb->args[4]; 3629 3630 rhashtable_walk_exit(hti); 3631 kfree(hti); 3632 return 0; 3633 } 3634 EXPORT_SYMBOL(tipc_dump_done); 3635 3636 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb, 3637 struct tipc_sock *tsk, u32 sk_filter_state, 3638 u64 (*tipc_diag_gen_cookie)(struct sock *sk)) 3639 { 3640 struct sock *sk = &tsk->sk; 3641 struct nlattr *attrs; 3642 struct nlattr *stat; 3643 3644 /*filter response w.r.t sk_state*/ 3645 if (!(sk_filter_state & (1 << sk->sk_state))) 3646 return 0; 3647 3648 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK); 3649 if (!attrs) 3650 goto msg_cancel; 3651 3652 if (__tipc_nl_add_sk_info(skb, tsk)) 3653 goto attr_msg_cancel; 3654 3655 if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) || 3656 nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) || 3657 nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) || 3658 nla_put_u32(skb, TIPC_NLA_SOCK_UID, 3659 from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk), 3660 sk_uid(sk))) || 3661 nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE, 3662 tipc_diag_gen_cookie(sk), 3663 TIPC_NLA_SOCK_PAD)) 3664 goto attr_msg_cancel; 3665 3666 stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT); 3667 if (!stat) 3668 goto attr_msg_cancel; 3669 3670 if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ, 3671 skb_queue_len(&sk->sk_receive_queue)) || 3672 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ, 3673 skb_queue_len(&sk->sk_write_queue)) || 3674 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP, 3675 sk_drops_read(sk))) 3676 goto stat_msg_cancel; 3677 3678 if (tsk->cong_link_cnt && 3679 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG)) 3680 goto stat_msg_cancel; 3681 3682 if (tsk_conn_cong(tsk) && 3683 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG)) 3684 goto stat_msg_cancel; 3685 3686 nla_nest_end(skb, stat); 3687 3688 if (tsk->group) 3689 if (tipc_group_fill_sock_diag(tsk->group, skb)) 3690 goto stat_msg_cancel; 3691 3692 nla_nest_end(skb, attrs); 3693 3694 return 0; 3695 3696 stat_msg_cancel: 3697 nla_nest_cancel(skb, stat); 3698 attr_msg_cancel: 3699 nla_nest_cancel(skb, attrs); 3700 msg_cancel: 3701 return -EMSGSIZE; 3702 } 3703 EXPORT_SYMBOL(tipc_sk_fill_sock_diag); 3704 3705 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb) 3706 { 3707 return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk); 3708 } 3709 3710 /* Caller should hold socket lock for the passed tipc socket. */ 3711 static int __tipc_nl_add_sk_publ(struct sk_buff *skb, 3712 struct netlink_callback *cb, 3713 struct publication *publ) 3714 { 3715 void *hdr; 3716 struct nlattr *attrs; 3717 3718 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq, 3719 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET); 3720 if (!hdr) 3721 goto msg_cancel; 3722 3723 attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL); 3724 if (!attrs) 3725 goto genlmsg_cancel; 3726 3727 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key)) 3728 goto attr_msg_cancel; 3729 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->sr.type)) 3730 goto attr_msg_cancel; 3731 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->sr.lower)) 3732 goto attr_msg_cancel; 3733 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->sr.upper)) 3734 goto attr_msg_cancel; 3735 3736 nla_nest_end(skb, attrs); 3737 genlmsg_end(skb, hdr); 3738 3739 return 0; 3740 3741 attr_msg_cancel: 3742 nla_nest_cancel(skb, attrs); 3743 genlmsg_cancel: 3744 genlmsg_cancel(skb, hdr); 3745 msg_cancel: 3746 return -EMSGSIZE; 3747 } 3748 3749 /* Caller should hold socket lock for the passed tipc socket. */ 3750 static int __tipc_nl_list_sk_publ(struct sk_buff *skb, 3751 struct netlink_callback *cb, 3752 struct tipc_sock *tsk, u32 *last_publ) 3753 { 3754 int err; 3755 struct publication *p; 3756 3757 if (*last_publ) { 3758 list_for_each_entry(p, &tsk->publications, binding_sock) { 3759 if (p->key == *last_publ) 3760 break; 3761 } 3762 if (list_entry_is_head(p, &tsk->publications, binding_sock)) { 3763 /* We never set seq or call nl_dump_check_consistent() 3764 * this means that setting prev_seq here will cause the 3765 * consistence check to fail in the netlink callback 3766 * handler. Resulting in the last NLMSG_DONE message 3767 * having the NLM_F_DUMP_INTR flag set. 3768 */ 3769 cb->prev_seq = 1; 3770 *last_publ = 0; 3771 return -EPIPE; 3772 } 3773 } else { 3774 p = list_first_entry(&tsk->publications, struct publication, 3775 binding_sock); 3776 } 3777 3778 list_for_each_entry_from(p, &tsk->publications, binding_sock) { 3779 err = __tipc_nl_add_sk_publ(skb, cb, p); 3780 if (err) { 3781 *last_publ = p->key; 3782 return err; 3783 } 3784 } 3785 *last_publ = 0; 3786 3787 return 0; 3788 } 3789 3790 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb) 3791 { 3792 int err; 3793 u32 tsk_portid = cb->args[0]; 3794 u32 last_publ = cb->args[1]; 3795 u32 done = cb->args[2]; 3796 struct net *net = sock_net(skb->sk); 3797 struct tipc_sock *tsk; 3798 3799 if (!tsk_portid) { 3800 struct nlattr **attrs = genl_dumpit_info(cb)->info.attrs; 3801 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1]; 3802 3803 if (!attrs[TIPC_NLA_SOCK]) 3804 return -EINVAL; 3805 3806 err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX, 3807 attrs[TIPC_NLA_SOCK], 3808 tipc_nl_sock_policy, NULL); 3809 if (err) 3810 return err; 3811 3812 if (!sock[TIPC_NLA_SOCK_REF]) 3813 return -EINVAL; 3814 3815 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]); 3816 } 3817 3818 if (done) 3819 return 0; 3820 3821 tsk = tipc_sk_lookup(net, tsk_portid); 3822 if (!tsk) 3823 return -EINVAL; 3824 3825 lock_sock(&tsk->sk); 3826 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ); 3827 if (!err) 3828 done = 1; 3829 release_sock(&tsk->sk); 3830 sock_put(&tsk->sk); 3831 3832 cb->args[0] = tsk_portid; 3833 cb->args[1] = last_publ; 3834 cb->args[2] = done; 3835 3836 return skb->len; 3837 } 3838 3839 /** 3840 * tipc_sk_filtering - check if a socket should be traced 3841 * @sk: the socket to be examined 3842 * 3843 * @sysctl_tipc_sk_filter is used as the socket tuple for filtering: 3844 * (portid, sock type, name type, name lower, name upper) 3845 * 3846 * Return: true if the socket meets the socket tuple data 3847 * (value 0 = 'any') or when there is no tuple set (all = 0), 3848 * otherwise false 3849 */ 3850 bool tipc_sk_filtering(struct sock *sk) 3851 { 3852 struct tipc_sock *tsk; 3853 struct publication *p; 3854 u32 _port, _sktype, _type, _lower, _upper; 3855 u32 type = 0, lower = 0, upper = 0; 3856 3857 if (!sk) 3858 return true; 3859 3860 tsk = tipc_sk(sk); 3861 3862 _port = sysctl_tipc_sk_filter[0]; 3863 _sktype = sysctl_tipc_sk_filter[1]; 3864 _type = sysctl_tipc_sk_filter[2]; 3865 _lower = sysctl_tipc_sk_filter[3]; 3866 _upper = sysctl_tipc_sk_filter[4]; 3867 3868 if (!_port && !_sktype && !_type && !_lower && !_upper) 3869 return true; 3870 3871 if (_port) 3872 return (_port == tsk->portid); 3873 3874 if (_sktype && _sktype != sk->sk_type) 3875 return false; 3876 3877 if (tsk->published) { 3878 p = list_first_entry_or_null(&tsk->publications, 3879 struct publication, binding_sock); 3880 if (p) { 3881 type = p->sr.type; 3882 lower = p->sr.lower; 3883 upper = p->sr.upper; 3884 } 3885 } 3886 3887 if (!tipc_sk_type_connectionless(sk)) { 3888 type = msg_nametype(&tsk->phdr); 3889 lower = msg_nameinst(&tsk->phdr); 3890 upper = lower; 3891 } 3892 3893 if ((_type && _type != type) || (_lower && _lower != lower) || 3894 (_upper && _upper != upper)) 3895 return false; 3896 3897 return true; 3898 } 3899 3900 u32 tipc_sock_get_portid(struct sock *sk) 3901 { 3902 return (sk) ? (tipc_sk(sk))->portid : 0; 3903 } 3904 3905 /** 3906 * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded, 3907 * both the rcv and backlog queues are considered 3908 * @sk: tipc sk to be checked 3909 * @skb: tipc msg to be checked 3910 * 3911 * Return: true if the socket rx queue allocation is > 90%, otherwise false 3912 */ 3913 3914 bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb) 3915 { 3916 atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt; 3917 unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt); 3918 unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk); 3919 3920 return (qsize > lim * 90 / 100); 3921 } 3922 3923 /** 3924 * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded, 3925 * only the rcv queue is considered 3926 * @sk: tipc sk to be checked 3927 * @skb: tipc msg to be checked 3928 * 3929 * Return: true if the socket rx queue allocation is > 90%, otherwise false 3930 */ 3931 3932 bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb) 3933 { 3934 unsigned int lim = rcvbuf_limit(sk, skb); 3935 unsigned int qsize = sk_rmem_alloc_get(sk); 3936 3937 return (qsize > lim * 90 / 100); 3938 } 3939 3940 /** 3941 * tipc_sk_dump - dump TIPC socket 3942 * @sk: tipc sk to be dumped 3943 * @dqueues: bitmask to decide if any socket queue to be dumped? 3944 * - TIPC_DUMP_NONE: don't dump socket queues 3945 * - TIPC_DUMP_SK_SNDQ: dump socket send queue 3946 * - TIPC_DUMP_SK_RCVQ: dump socket rcv queue 3947 * - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue 3948 * - TIPC_DUMP_ALL: dump all the socket queues above 3949 * @buf: returned buffer of dump data in format 3950 */ 3951 int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf) 3952 { 3953 int i = 0; 3954 size_t sz = (dqueues) ? SK_LMAX : SK_LMIN; 3955 u32 conn_type, conn_instance; 3956 struct tipc_sock *tsk; 3957 struct publication *p; 3958 bool tsk_connected; 3959 3960 if (!sk) { 3961 i += scnprintf(buf, sz, "sk data: (null)\n"); 3962 return i; 3963 } 3964 3965 tsk = tipc_sk(sk); 3966 tsk_connected = !tipc_sk_type_connectionless(sk); 3967 3968 i += scnprintf(buf, sz, "sk data: %u", sk->sk_type); 3969 i += scnprintf(buf + i, sz - i, " %d", sk->sk_state); 3970 i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk)); 3971 i += scnprintf(buf + i, sz - i, " %u", tsk->portid); 3972 i += scnprintf(buf + i, sz - i, " | %u", tsk_connected); 3973 if (tsk_connected) { 3974 i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk)); 3975 i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk)); 3976 conn_type = msg_nametype(&tsk->phdr); 3977 conn_instance = msg_nameinst(&tsk->phdr); 3978 i += scnprintf(buf + i, sz - i, " %u", conn_type); 3979 i += scnprintf(buf + i, sz - i, " %u", conn_instance); 3980 } 3981 i += scnprintf(buf + i, sz - i, " | %u", tsk->published); 3982 if (tsk->published) { 3983 p = list_first_entry_or_null(&tsk->publications, 3984 struct publication, binding_sock); 3985 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.type : 0); 3986 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.lower : 0); 3987 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.upper : 0); 3988 } 3989 i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win); 3990 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win); 3991 i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt); 3992 i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps); 3993 i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt); 3994 i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked); 3995 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked); 3996 i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt)); 3997 i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown); 3998 i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk)); 3999 i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf); 4000 i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk)); 4001 i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf); 4002 i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len)); 4003 4004 if (dqueues & TIPC_DUMP_SK_SNDQ) { 4005 i += scnprintf(buf + i, sz - i, "sk_write_queue: "); 4006 i += tipc_list_dump(&sk->sk_write_queue, false, buf + i); 4007 } 4008 4009 if (dqueues & TIPC_DUMP_SK_RCVQ) { 4010 i += scnprintf(buf + i, sz - i, "sk_receive_queue: "); 4011 i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i); 4012 } 4013 4014 if (dqueues & TIPC_DUMP_SK_BKLGQ) { 4015 i += scnprintf(buf + i, sz - i, "sk_backlog:\n head "); 4016 i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i); 4017 if (sk->sk_backlog.tail != sk->sk_backlog.head) { 4018 i += scnprintf(buf + i, sz - i, " tail "); 4019 i += tipc_skb_dump(sk->sk_backlog.tail, false, 4020 buf + i); 4021 } 4022 } 4023 4024 return i; 4025 } 4026