Linux Audio

Check our new training course

Loading...
v6.13.7
   1/*
   2 * net/tipc/socket.c: TIPC socket API
   3 *
   4 * Copyright (c) 2001-2007, 2012-2019, Ericsson AB
   5 * Copyright (c) 2004-2008, 2010-2013, Wind River Systems
   6 * Copyright (c) 2020-2021, Red Hat Inc
   7 * All rights reserved.
   8 *
   9 * Redistribution and use in source and binary forms, with or without
  10 * modification, are permitted provided that the following conditions are met:
  11 *
  12 * 1. Redistributions of source code must retain the above copyright
  13 *    notice, this list of conditions and the following disclaimer.
  14 * 2. Redistributions in binary form must reproduce the above copyright
  15 *    notice, this list of conditions and the following disclaimer in the
  16 *    documentation and/or other materials provided with the distribution.
  17 * 3. Neither the names of the copyright holders nor the names of its
  18 *    contributors may be used to endorse or promote products derived from
  19 *    this software without specific prior written permission.
  20 *
  21 * Alternatively, this software may be distributed under the terms of the
  22 * GNU General Public License ("GPL") version 2 as published by the Free
  23 * Software Foundation.
  24 *
  25 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
  26 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  28 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
  29 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  32 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  33 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  34 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  35 * POSSIBILITY OF SUCH DAMAGE.
  36 */
  37
  38#include <linux/rhashtable.h>
  39#include <linux/sched/signal.h>
  40#include <trace/events/sock.h>
  41
  42#include "core.h"
  43#include "name_table.h"
  44#include "node.h"
  45#include "link.h"
  46#include "name_distr.h"
  47#include "socket.h"
  48#include "bcast.h"
  49#include "netlink.h"
  50#include "group.h"
  51#include "trace.h"
  52
  53#define NAGLE_START_INIT	4
  54#define NAGLE_START_MAX		1024
  55#define CONN_TIMEOUT_DEFAULT    8000    /* default connect timeout = 8s */
  56#define CONN_PROBING_INTV	msecs_to_jiffies(3600000)  /* [ms] => 1 h */
 
  57#define TIPC_MAX_PORT		0xffffffff
  58#define TIPC_MIN_PORT		1
  59#define TIPC_ACK_RATE		4       /* ACK at 1/4 of rcv window size */
  60
  61enum {
  62	TIPC_LISTEN = TCP_LISTEN,
  63	TIPC_ESTABLISHED = TCP_ESTABLISHED,
  64	TIPC_OPEN = TCP_CLOSE,
  65	TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
  66	TIPC_CONNECTING = TCP_SYN_SENT,
  67};
  68
  69struct sockaddr_pair {
  70	struct sockaddr_tipc sock;
  71	struct sockaddr_tipc member;
  72};
  73
  74/**
  75 * struct tipc_sock - TIPC socket structure
  76 * @sk: socket - interacts with 'port' and with user via the socket API
 
 
 
  77 * @max_pkt: maximum packet size "hint" used when building messages sent by port
  78 * @maxnagle: maximum size of msg which can be subject to nagle
  79 * @portid: unique port identity in TIPC socket hash table
  80 * @phdr: preformatted message header used when sending messages
  81 * @cong_links: list of congested links
  82 * @publications: list of publications for port
 
  83 * @pub_count: total # of publications port has made during its lifetime
 
  84 * @conn_timeout: the time we can wait for an unresponded setup request
  85 * @probe_unacked: probe has not received ack yet
  86 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
  87 * @cong_link_cnt: number of congested links
  88 * @snt_unacked: # messages sent by socket, and not yet acked by peer
  89 * @snd_win: send window size
  90 * @peer_caps: peer capabilities mask
  91 * @rcv_unacked: # messages read by user, but not yet acked back to peer
  92 * @rcv_win: receive window size
  93 * @peer: 'connected' peer for dgram/rdm
  94 * @node: hash table node
  95 * @mc_method: cookie for use between socket and broadcast layer
  96 * @rcu: rcu struct for tipc_sock
  97 * @group: TIPC communications group
  98 * @oneway: message count in one direction (FIXME)
  99 * @nagle_start: current nagle value
 100 * @snd_backlog: send backlog count
 101 * @msg_acc: messages accepted; used in managing backlog and nagle
 102 * @pkt_cnt: TIPC socket packet count
 103 * @expect_ack: whether this TIPC socket is expecting an ack
 104 * @nodelay: setsockopt() TIPC_NODELAY setting
 105 * @group_is_open: TIPC socket group is fully open (FIXME)
 106 * @published: true if port has one or more associated names
 107 * @conn_addrtype: address type used when establishing connection
 108 */
 109struct tipc_sock {
 110	struct sock sk;
 
 
 
 111	u32 max_pkt;
 112	u32 maxnagle;
 113	u32 portid;
 114	struct tipc_msg phdr;
 115	struct list_head cong_links;
 116	struct list_head publications;
 117	u32 pub_count;
 
 118	atomic_t dupl_rcvcnt;
 119	u16 conn_timeout;
 120	bool probe_unacked;
 121	u16 cong_link_cnt;
 122	u16 snt_unacked;
 123	u16 snd_win;
 124	u16 peer_caps;
 125	u16 rcv_unacked;
 126	u16 rcv_win;
 127	struct sockaddr_tipc peer;
 128	struct rhash_head node;
 129	struct tipc_mc_method mc_method;
 130	struct rcu_head rcu;
 131	struct tipc_group *group;
 132	u32 oneway;
 133	u32 nagle_start;
 134	u16 snd_backlog;
 135	u16 msg_acc;
 136	u16 pkt_cnt;
 137	bool expect_ack;
 138	bool nodelay;
 139	bool group_is_open;
 140	bool published;
 141	u8 conn_addrtype;
 142};
 143
 144static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
 145static void tipc_data_ready(struct sock *sk);
 146static void tipc_write_space(struct sock *sk);
 147static void tipc_sock_destruct(struct sock *sk);
 148static int tipc_release(struct socket *sock);
 
 
 149static void tipc_sk_timeout(struct timer_list *t);
 150static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua);
 151static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua);
 
 
 152static int tipc_sk_leave(struct tipc_sock *tsk);
 153static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
 154static int tipc_sk_insert(struct tipc_sock *tsk);
 155static void tipc_sk_remove(struct tipc_sock *tsk);
 156static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
 157static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
 158static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack);
 159static int tipc_wait_for_connect(struct socket *sock, long *timeo_p);
 160
 161static const struct proto_ops packet_ops;
 162static const struct proto_ops stream_ops;
 163static const struct proto_ops msg_ops;
 164static struct proto tipc_proto;
 165static const struct rhashtable_params tsk_rht_params;
 166
 167static u32 tsk_own_node(struct tipc_sock *tsk)
 168{
 169	return msg_prevnode(&tsk->phdr);
 170}
 171
 172static u32 tsk_peer_node(struct tipc_sock *tsk)
 173{
 174	return msg_destnode(&tsk->phdr);
 175}
 176
 177static u32 tsk_peer_port(struct tipc_sock *tsk)
 178{
 179	return msg_destport(&tsk->phdr);
 180}
 181
 182static  bool tsk_unreliable(struct tipc_sock *tsk)
 183{
 184	return msg_src_droppable(&tsk->phdr) != 0;
 185}
 186
 187static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
 188{
 189	msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
 190}
 191
 192static bool tsk_unreturnable(struct tipc_sock *tsk)
 193{
 194	return msg_dest_droppable(&tsk->phdr) != 0;
 195}
 196
 197static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
 198{
 199	msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
 200}
 201
 202static int tsk_importance(struct tipc_sock *tsk)
 203{
 204	return msg_importance(&tsk->phdr);
 205}
 206
 207static struct tipc_sock *tipc_sk(const struct sock *sk)
 208{
 209	return container_of(sk, struct tipc_sock, sk);
 210}
 211
 212int tsk_set_importance(struct sock *sk, int imp)
 213{
 214	if (imp > TIPC_CRITICAL_IMPORTANCE)
 215		return -EINVAL;
 216	msg_set_importance(&tipc_sk(sk)->phdr, (u32)imp);
 217	return 0;
 218}
 219
 
 
 
 
 
 220static bool tsk_conn_cong(struct tipc_sock *tsk)
 221{
 222	return tsk->snt_unacked > tsk->snd_win;
 223}
 224
 225static u16 tsk_blocks(int len)
 226{
 227	return ((len / FLOWCTL_BLK_SZ) + 1);
 228}
 229
 230/* tsk_blocks(): translate a buffer size in bytes to number of
 231 * advertisable blocks, taking into account the ratio truesize(len)/len
 232 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
 233 */
 234static u16 tsk_adv_blocks(int len)
 235{
 236	return len / FLOWCTL_BLK_SZ / 4;
 237}
 238
 239/* tsk_inc(): increment counter for sent or received data
 240 * - If block based flow control is not supported by peer we
 241 *   fall back to message based ditto, incrementing the counter
 242 */
 243static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
 244{
 245	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
 246		return ((msglen / FLOWCTL_BLK_SZ) + 1);
 247	return 1;
 248}
 249
 250/* tsk_set_nagle - enable/disable nagle property by manipulating maxnagle
 251 */
 252static void tsk_set_nagle(struct tipc_sock *tsk)
 253{
 254	struct sock *sk = &tsk->sk;
 255
 256	tsk->maxnagle = 0;
 257	if (sk->sk_type != SOCK_STREAM)
 258		return;
 259	if (tsk->nodelay)
 260		return;
 261	if (!(tsk->peer_caps & TIPC_NAGLE))
 262		return;
 263	/* Limit node local buffer size to avoid receive queue overflow */
 264	if (tsk->max_pkt == MAX_MSG_SIZE)
 265		tsk->maxnagle = 1500;
 266	else
 267		tsk->maxnagle = tsk->max_pkt;
 268}
 269
 270/**
 271 * tsk_advance_rx_queue - discard first buffer in socket receive queue
 272 * @sk: network socket
 273 *
 274 * Caller must hold socket lock
 275 */
 276static void tsk_advance_rx_queue(struct sock *sk)
 277{
 278	trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " ");
 279	kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
 280}
 281
 282/* tipc_sk_respond() : send response message back to sender
 283 */
 284static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
 285{
 286	u32 selector;
 287	u32 dnode;
 288	u32 onode = tipc_own_addr(sock_net(sk));
 289
 290	if (!tipc_msg_reverse(onode, &skb, err))
 291		return;
 292
 293	trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!");
 294	dnode = msg_destnode(buf_msg(skb));
 295	selector = msg_origport(buf_msg(skb));
 296	tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
 297}
 298
 299/**
 300 * tsk_rej_rx_queue - reject all buffers in socket receive queue
 301 * @sk: network socket
 302 * @error: response error code
 303 *
 304 * Caller must hold socket lock
 305 */
 306static void tsk_rej_rx_queue(struct sock *sk, int error)
 307{
 308	struct sk_buff *skb;
 309
 310	while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
 311		tipc_sk_respond(sk, skb, error);
 312}
 313
 314static bool tipc_sk_connected(const struct sock *sk)
 315{
 316	return READ_ONCE(sk->sk_state) == TIPC_ESTABLISHED;
 317}
 318
 319/* tipc_sk_type_connectionless - check if the socket is datagram socket
 320 * @sk: socket
 321 *
 322 * Returns true if connection less, false otherwise
 323 */
 324static bool tipc_sk_type_connectionless(struct sock *sk)
 325{
 326	return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
 327}
 328
 329/* tsk_peer_msg - verify if message was sent by connected port's peer
 330 *
 331 * Handles cases where the node's network address has changed from
 332 * the default of <0.0.0> to its configured setting.
 333 */
 334static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
 335{
 336	struct sock *sk = &tsk->sk;
 337	u32 self = tipc_own_addr(sock_net(sk));
 338	u32 peer_port = tsk_peer_port(tsk);
 339	u32 orig_node, peer_node;
 340
 341	if (unlikely(!tipc_sk_connected(sk)))
 342		return false;
 343
 344	if (unlikely(msg_origport(msg) != peer_port))
 345		return false;
 346
 347	orig_node = msg_orignode(msg);
 348	peer_node = tsk_peer_node(tsk);
 349
 350	if (likely(orig_node == peer_node))
 351		return true;
 352
 353	if (!orig_node && peer_node == self)
 354		return true;
 355
 356	if (!peer_node && orig_node == self)
 357		return true;
 358
 359	return false;
 360}
 361
 362/* tipc_set_sk_state - set the sk_state of the socket
 363 * @sk: socket
 364 *
 365 * Caller must hold socket lock
 366 *
 367 * Returns 0 on success, errno otherwise
 368 */
 369static int tipc_set_sk_state(struct sock *sk, int state)
 370{
 371	int oldsk_state = sk->sk_state;
 372	int res = -EINVAL;
 373
 374	switch (state) {
 375	case TIPC_OPEN:
 376		res = 0;
 377		break;
 378	case TIPC_LISTEN:
 379	case TIPC_CONNECTING:
 380		if (oldsk_state == TIPC_OPEN)
 381			res = 0;
 382		break;
 383	case TIPC_ESTABLISHED:
 384		if (oldsk_state == TIPC_CONNECTING ||
 385		    oldsk_state == TIPC_OPEN)
 386			res = 0;
 387		break;
 388	case TIPC_DISCONNECTING:
 389		if (oldsk_state == TIPC_CONNECTING ||
 390		    oldsk_state == TIPC_ESTABLISHED)
 391			res = 0;
 392		break;
 393	}
 394
 395	if (!res)
 396		sk->sk_state = state;
 397
 398	return res;
 399}
 400
 401static int tipc_sk_sock_err(struct socket *sock, long *timeout)
 402{
 403	struct sock *sk = sock->sk;
 404	int err = sock_error(sk);
 405	int typ = sock->type;
 406
 407	if (err)
 408		return err;
 409	if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
 410		if (sk->sk_state == TIPC_DISCONNECTING)
 411			return -EPIPE;
 412		else if (!tipc_sk_connected(sk))
 413			return -ENOTCONN;
 414	}
 415	if (!*timeout)
 416		return -EAGAIN;
 417	if (signal_pending(current))
 418		return sock_intr_errno(*timeout);
 419
 420	return 0;
 421}
 422
 423#define tipc_wait_for_cond(sock_, timeo_, condition_)			       \
 424({                                                                             \
 425	DEFINE_WAIT_FUNC(wait_, woken_wake_function);                          \
 426	struct sock *sk_;						       \
 427	int rc_;							       \
 428									       \
 429	while ((rc_ = !(condition_))) {					       \
 430		/* coupled with smp_wmb() in tipc_sk_proto_rcv() */            \
 431		smp_rmb();                                                     \
 432		sk_ = (sock_)->sk;					       \
 433		rc_ = tipc_sk_sock_err((sock_), timeo_);		       \
 434		if (rc_)						       \
 435			break;						       \
 436		add_wait_queue(sk_sleep(sk_), &wait_);                         \
 437		release_sock(sk_);					       \
 438		*(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
 439		sched_annotate_sleep();				               \
 440		lock_sock(sk_);						       \
 441		remove_wait_queue(sk_sleep(sk_), &wait_);		       \
 442	}								       \
 443	rc_;								       \
 444})
 445
 446/**
 447 * tipc_sk_create - create a TIPC socket
 448 * @net: network namespace (must be default network)
 449 * @sock: pre-allocated socket structure
 450 * @protocol: protocol indicator (must be 0)
 451 * @kern: caused by kernel or by userspace?
 452 *
 453 * This routine creates additional data structures used by the TIPC socket,
 454 * initializes them, and links them together.
 455 *
 456 * Return: 0 on success, errno otherwise
 457 */
 458static int tipc_sk_create(struct net *net, struct socket *sock,
 459			  int protocol, int kern)
 460{
 
 461	const struct proto_ops *ops;
 462	struct sock *sk;
 463	struct tipc_sock *tsk;
 464	struct tipc_msg *msg;
 465
 466	/* Validate arguments */
 467	if (unlikely(protocol != 0))
 468		return -EPROTONOSUPPORT;
 469
 470	switch (sock->type) {
 471	case SOCK_STREAM:
 472		ops = &stream_ops;
 473		break;
 474	case SOCK_SEQPACKET:
 475		ops = &packet_ops;
 476		break;
 477	case SOCK_DGRAM:
 478	case SOCK_RDM:
 479		ops = &msg_ops;
 480		break;
 481	default:
 482		return -EPROTOTYPE;
 483	}
 484
 485	/* Allocate socket's protocol area */
 486	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
 487	if (sk == NULL)
 488		return -ENOMEM;
 489
 490	tsk = tipc_sk(sk);
 491	tsk->max_pkt = MAX_PKT_DEFAULT;
 492	tsk->maxnagle = 0;
 493	tsk->nagle_start = NAGLE_START_INIT;
 494	INIT_LIST_HEAD(&tsk->publications);
 495	INIT_LIST_HEAD(&tsk->cong_links);
 496	msg = &tsk->phdr;
 
 497
 498	/* Finish initializing socket data structures */
 499	sock->ops = ops;
 500	sock_init_data(sock, sk);
 501	tipc_set_sk_state(sk, TIPC_OPEN);
 502	if (tipc_sk_insert(tsk)) {
 503		sk_free(sk);
 504		pr_warn("Socket create failed; port number exhausted\n");
 505		return -EINVAL;
 506	}
 507
 508	/* Ensure tsk is visible before we read own_addr. */
 509	smp_mb();
 510
 511	tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
 512		      TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
 513
 514	msg_set_origport(msg, tsk->portid);
 515	timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
 516	sk->sk_shutdown = 0;
 517	sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
 518	sk->sk_rcvbuf = READ_ONCE(sysctl_tipc_rmem[1]);
 519	sk->sk_data_ready = tipc_data_ready;
 520	sk->sk_write_space = tipc_write_space;
 521	sk->sk_destruct = tipc_sock_destruct;
 522	tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
 523	tsk->group_is_open = true;
 524	atomic_set(&tsk->dupl_rcvcnt, 0);
 525
 526	/* Start out with safe limits until we receive an advertised window */
 527	tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
 528	tsk->rcv_win = tsk->snd_win;
 529
 530	if (tipc_sk_type_connectionless(sk)) {
 531		tsk_set_unreturnable(tsk, true);
 532		if (sock->type == SOCK_DGRAM)
 533			tsk_set_unreliable(tsk, true);
 534	}
 535	__skb_queue_head_init(&tsk->mc_method.deferredq);
 536	trace_tipc_sk_create(sk, NULL, TIPC_DUMP_NONE, " ");
 537	return 0;
 538}
 539
 540static void tipc_sk_callback(struct rcu_head *head)
 541{
 542	struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
 543
 544	sock_put(&tsk->sk);
 545}
 546
 547/* Caller should hold socket lock for the socket. */
 548static void __tipc_shutdown(struct socket *sock, int error)
 549{
 550	struct sock *sk = sock->sk;
 551	struct tipc_sock *tsk = tipc_sk(sk);
 552	struct net *net = sock_net(sk);
 553	long timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
 554	u32 dnode = tsk_peer_node(tsk);
 555	struct sk_buff *skb;
 556
 557	/* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
 558	tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
 559					    !tsk_conn_cong(tsk)));
 560
 561	/* Push out delayed messages if in Nagle mode */
 562	tipc_sk_push_backlog(tsk, false);
 563	/* Remove pending SYN */
 564	__skb_queue_purge(&sk->sk_write_queue);
 565
 566	/* Remove partially received buffer if any */
 567	skb = skb_peek(&sk->sk_receive_queue);
 568	if (skb && TIPC_SKB_CB(skb)->bytes_read) {
 569		__skb_unlink(skb, &sk->sk_receive_queue);
 570		kfree_skb(skb);
 
 
 
 
 571	}
 572
 573	/* Reject all unreceived messages if connectionless */
 574	if (tipc_sk_type_connectionless(sk)) {
 575		tsk_rej_rx_queue(sk, error);
 576		return;
 577	}
 578
 579	switch (sk->sk_state) {
 580	case TIPC_CONNECTING:
 581	case TIPC_ESTABLISHED:
 582		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
 583		tipc_node_remove_conn(net, dnode, tsk->portid);
 584		/* Send a FIN+/- to its peer */
 585		skb = __skb_dequeue(&sk->sk_receive_queue);
 586		if (skb) {
 587			__skb_queue_purge(&sk->sk_receive_queue);
 588			tipc_sk_respond(sk, skb, error);
 589			break;
 590		}
 591		skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
 592				      TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
 593				      tsk_own_node(tsk), tsk_peer_port(tsk),
 594				      tsk->portid, error);
 595		if (skb)
 596			tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
 597		break;
 598	case TIPC_LISTEN:
 599		/* Reject all SYN messages */
 600		tsk_rej_rx_queue(sk, error);
 601		break;
 602	default:
 603		__skb_queue_purge(&sk->sk_receive_queue);
 604		break;
 605	}
 606}
 607
 608/**
 609 * tipc_release - destroy a TIPC socket
 610 * @sock: socket to destroy
 611 *
 612 * This routine cleans up any messages that are still queued on the socket.
 613 * For DGRAM and RDM socket types, all queued messages are rejected.
 614 * For SEQPACKET and STREAM socket types, the first message is rejected
 615 * and any others are discarded.  (If the first message on a STREAM socket
 616 * is partially-read, it is discarded and the next one is rejected instead.)
 617 *
 618 * NOTE: Rejected messages are not necessarily returned to the sender!  They
 619 * are returned or discarded according to the "destination droppable" setting
 620 * specified for the message by the sender.
 621 *
 622 * Return: 0 on success, errno otherwise
 623 */
 624static int tipc_release(struct socket *sock)
 625{
 626	struct sock *sk = sock->sk;
 627	struct tipc_sock *tsk;
 628
 629	/*
 630	 * Exit if socket isn't fully initialized (occurs when a failed accept()
 631	 * releases a pre-allocated child socket that was never used)
 632	 */
 633	if (sk == NULL)
 634		return 0;
 635
 636	tsk = tipc_sk(sk);
 637	lock_sock(sk);
 638
 639	trace_tipc_sk_release(sk, NULL, TIPC_DUMP_ALL, " ");
 640	__tipc_shutdown(sock, TIPC_ERR_NO_PORT);
 641	sk->sk_shutdown = SHUTDOWN_MASK;
 642	tipc_sk_leave(tsk);
 643	tipc_sk_withdraw(tsk, NULL);
 644	__skb_queue_purge(&tsk->mc_method.deferredq);
 645	sk_stop_timer(sk, &sk->sk_timer);
 646	tipc_sk_remove(tsk);
 647
 648	sock_orphan(sk);
 649	/* Reject any messages that accumulated in backlog queue */
 650	release_sock(sk);
 651	tipc_dest_list_purge(&tsk->cong_links);
 652	tsk->cong_link_cnt = 0;
 653	call_rcu(&tsk->rcu, tipc_sk_callback);
 654	sock->sk = NULL;
 655
 656	return 0;
 657}
 658
 659/**
 660 * __tipc_bind - associate or disassociate TIPC name(s) with a socket
 661 * @sock: socket structure
 662 * @skaddr: socket address describing name(s) and desired operation
 663 * @alen: size of socket address data structure
 664 *
 665 * Name and name sequence binding are indicated using a positive scope value;
 666 * a negative scope value unbinds the specified name.  Specifying no name
 667 * (i.e. a socket address length of 0) unbinds all names from the socket.
 668 *
 669 * Return: 0 on success, errno otherwise
 670 *
 671 * NOTE: This routine doesn't need to take the socket lock since it doesn't
 672 *       access any non-constant socket information.
 673 */
 674static int __tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
 
 675{
 676	struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr;
 677	struct tipc_sock *tsk = tipc_sk(sock->sk);
 678	bool unbind = false;
 679
 680	if (unlikely(!alen))
 681		return tipc_sk_withdraw(tsk, NULL);
 682
 683	if (ua->addrtype == TIPC_SERVICE_ADDR) {
 684		ua->addrtype = TIPC_SERVICE_RANGE;
 685		ua->sr.upper = ua->sr.lower;
 686	}
 687	if (ua->scope < 0) {
 688		unbind = true;
 689		ua->scope = -ua->scope;
 690	}
 691	/* Users may still use deprecated TIPC_ZONE_SCOPE */
 692	if (ua->scope != TIPC_NODE_SCOPE)
 693		ua->scope = TIPC_CLUSTER_SCOPE;
 694
 695	if (tsk->group)
 696		return -EACCES;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 697
 698	if (unbind)
 699		return tipc_sk_withdraw(tsk, ua);
 700	return tipc_sk_publish(tsk, ua);
 701}
 
 
 702
 703int tipc_sk_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
 704{
 705	int res;
 
 
 
 706
 707	lock_sock(sock->sk);
 708	res = __tipc_bind(sock, skaddr, alen);
 709	release_sock(sock->sk);
 
 
 710	return res;
 711}
 712
 713static int tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
 714{
 715	struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr;
 716	u32 atype = ua->addrtype;
 717
 718	if (alen) {
 719		if (!tipc_uaddr_valid(ua, alen))
 720			return -EINVAL;
 721		if (atype == TIPC_SOCKET_ADDR)
 722			return -EAFNOSUPPORT;
 723		if (ua->sr.type < TIPC_RESERVED_TYPES) {
 724			pr_warn_once("Can't bind to reserved service type %u\n",
 725				     ua->sr.type);
 726			return -EACCES;
 727		}
 728	}
 729	return tipc_sk_bind(sock, skaddr, alen);
 730}
 731
 732/**
 733 * tipc_getname - get port ID of socket or peer socket
 734 * @sock: socket structure
 735 * @uaddr: area for returned socket address
 
 736 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
 737 *
 738 * Return: 0 on success, errno otherwise
 739 *
 740 * NOTE: This routine doesn't need to take the socket lock since it only
 741 *       accesses socket information that is unchanging (or which changes in
 742 *       a completely predictable manner).
 743 */
 744static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
 745			int peer)
 746{
 747	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
 748	struct sock *sk = sock->sk;
 749	struct tipc_sock *tsk = tipc_sk(sk);
 750
 751	memset(addr, 0, sizeof(*addr));
 752	if (peer) {
 753		if ((!tipc_sk_connected(sk)) &&
 754		    ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
 755			return -ENOTCONN;
 756		addr->addr.id.ref = tsk_peer_port(tsk);
 757		addr->addr.id.node = tsk_peer_node(tsk);
 758	} else {
 759		addr->addr.id.ref = tsk->portid;
 760		addr->addr.id.node = tipc_own_addr(sock_net(sk));
 761	}
 762
 763	addr->addrtype = TIPC_SOCKET_ADDR;
 764	addr->family = AF_TIPC;
 765	addr->scope = 0;
 766	addr->addr.name.domain = 0;
 767
 768	return sizeof(*addr);
 769}
 770
 771/**
 772 * tipc_poll - read and possibly block on pollmask
 773 * @file: file structure associated with the socket
 774 * @sock: socket for which to calculate the poll bits
 775 * @wait: ???
 776 *
 777 * Return: pollmask value
 778 *
 779 * COMMENTARY:
 780 * It appears that the usual socket locking mechanisms are not useful here
 781 * since the pollmask info is potentially out-of-date the moment this routine
 782 * exits.  TCP and other protocols seem to rely on higher level poll routines
 783 * to handle any preventable race conditions, so TIPC will do the same ...
 784 *
 785 * IMPORTANT: The fact that a read or write operation is indicated does NOT
 786 * imply that the operation will succeed, merely that it should be performed
 787 * and will not block.
 788 */
 789static __poll_t tipc_poll(struct file *file, struct socket *sock,
 790			      poll_table *wait)
 791{
 792	struct sock *sk = sock->sk;
 793	struct tipc_sock *tsk = tipc_sk(sk);
 794	__poll_t revents = 0;
 795
 796	sock_poll_wait(file, sock, wait);
 797	trace_tipc_sk_poll(sk, NULL, TIPC_DUMP_ALL, " ");
 798
 799	if (sk->sk_shutdown & RCV_SHUTDOWN)
 800		revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
 801	if (sk->sk_shutdown == SHUTDOWN_MASK)
 802		revents |= EPOLLHUP;
 803
 804	switch (sk->sk_state) {
 805	case TIPC_ESTABLISHED:
 
 806		if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
 807			revents |= EPOLLOUT;
 808		fallthrough;
 809	case TIPC_LISTEN:
 810	case TIPC_CONNECTING:
 811		if (!skb_queue_empty_lockless(&sk->sk_receive_queue))
 812			revents |= EPOLLIN | EPOLLRDNORM;
 813		break;
 814	case TIPC_OPEN:
 815		if (tsk->group_is_open && !tsk->cong_link_cnt)
 816			revents |= EPOLLOUT;
 817		if (!tipc_sk_type_connectionless(sk))
 818			break;
 819		if (skb_queue_empty_lockless(&sk->sk_receive_queue))
 820			break;
 821		revents |= EPOLLIN | EPOLLRDNORM;
 822		break;
 823	case TIPC_DISCONNECTING:
 824		revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
 825		break;
 826	}
 827	return revents;
 828}
 829
 830/**
 831 * tipc_sendmcast - send multicast message
 832 * @sock: socket structure
 833 * @ua: destination address struct
 834 * @msg: message to send
 835 * @dlen: length of data to send
 836 * @timeout: timeout to wait for wakeup
 837 *
 838 * Called from function tipc_sendmsg(), which has done all sanity checks
 839 * Return: the number of bytes sent on success, or errno
 840 */
 841static int tipc_sendmcast(struct  socket *sock, struct tipc_uaddr *ua,
 842			  struct msghdr *msg, size_t dlen, long timeout)
 843{
 844	struct sock *sk = sock->sk;
 845	struct tipc_sock *tsk = tipc_sk(sk);
 846	struct tipc_msg *hdr = &tsk->phdr;
 847	struct net *net = sock_net(sk);
 848	int mtu = tipc_bcast_get_mtu(net);
 
 849	struct sk_buff_head pkts;
 850	struct tipc_nlist dsts;
 851	int rc;
 852
 853	if (tsk->group)
 854		return -EACCES;
 855
 856	/* Block or return if any destination link is congested */
 857	rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
 858	if (unlikely(rc))
 859		return rc;
 860
 861	/* Lookup destination nodes */
 862	tipc_nlist_init(&dsts, tipc_own_addr(net));
 863	tipc_nametbl_lookup_mcast_nodes(net, ua, &dsts);
 
 864	if (!dsts.local && !dsts.remote)
 865		return -EHOSTUNREACH;
 866
 867	/* Build message header */
 868	msg_set_type(hdr, TIPC_MCAST_MSG);
 869	msg_set_hdr_sz(hdr, MCAST_H_SIZE);
 870	msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
 871	msg_set_destport(hdr, 0);
 872	msg_set_destnode(hdr, 0);
 873	msg_set_nametype(hdr, ua->sr.type);
 874	msg_set_namelower(hdr, ua->sr.lower);
 875	msg_set_nameupper(hdr, ua->sr.upper);
 876
 877	/* Build message as chain of buffers */
 878	__skb_queue_head_init(&pkts);
 879	rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
 880
 881	/* Send message if build was successful */
 882	if (unlikely(rc == dlen)) {
 883		trace_tipc_sk_sendmcast(sk, skb_peek(&pkts),
 884					TIPC_DUMP_SK_SNDQ, " ");
 885		rc = tipc_mcast_xmit(net, &pkts, &tsk->mc_method, &dsts,
 886				     &tsk->cong_link_cnt);
 887	}
 888
 889	tipc_nlist_purge(&dsts);
 890
 891	return rc ? rc : dlen;
 892}
 893
 894/**
 895 * tipc_send_group_msg - send a message to a member in the group
 896 * @net: network namespace
 897 * @tsk: tipc socket
 898 * @m: message to send
 899 * @mb: group member
 900 * @dnode: destination node
 901 * @dport: destination port
 902 * @dlen: total length of message data
 903 */
 904static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
 905			       struct msghdr *m, struct tipc_member *mb,
 906			       u32 dnode, u32 dport, int dlen)
 907{
 908	u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
 909	struct tipc_mc_method *method = &tsk->mc_method;
 910	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
 911	struct tipc_msg *hdr = &tsk->phdr;
 912	struct sk_buff_head pkts;
 913	int mtu, rc;
 914
 915	/* Complete message header */
 916	msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
 917	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
 918	msg_set_destport(hdr, dport);
 919	msg_set_destnode(hdr, dnode);
 920	msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
 921
 922	/* Build message as chain of buffers */
 923	__skb_queue_head_init(&pkts);
 924	mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false);
 925	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
 926	if (unlikely(rc != dlen))
 927		return rc;
 928
 929	/* Send message */
 930	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
 931	if (unlikely(rc == -ELINKCONG)) {
 932		tipc_dest_push(&tsk->cong_links, dnode, 0);
 933		tsk->cong_link_cnt++;
 934	}
 935
 936	/* Update send window */
 937	tipc_group_update_member(mb, blks);
 938
 939	/* A broadcast sent within next EXPIRE period must follow same path */
 940	method->rcast = true;
 941	method->mandatory = true;
 942	return dlen;
 943}
 944
 945/**
 946 * tipc_send_group_unicast - send message to a member in the group
 947 * @sock: socket structure
 948 * @m: message to send
 949 * @dlen: total length of message data
 950 * @timeout: timeout to wait for wakeup
 951 *
 952 * Called from function tipc_sendmsg(), which has done all sanity checks
 953 * Return: the number of bytes sent on success, or errno
 954 */
 955static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
 956				   int dlen, long timeout)
 957{
 958	struct sock *sk = sock->sk;
 959	struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
 960	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
 961	struct tipc_sock *tsk = tipc_sk(sk);
 
 962	struct net *net = sock_net(sk);
 963	struct tipc_member *mb = NULL;
 964	u32 node, port;
 965	int rc;
 966
 967	node = ua->sk.node;
 968	port = ua->sk.ref;
 969	if (!port && !node)
 970		return -EHOSTUNREACH;
 971
 972	/* Block or return if destination link or member is congested */
 973	rc = tipc_wait_for_cond(sock, &timeout,
 974				!tipc_dest_find(&tsk->cong_links, node, 0) &&
 975				tsk->group &&
 976				!tipc_group_cong(tsk->group, node, port, blks,
 977						 &mb));
 978	if (unlikely(rc))
 979		return rc;
 980
 981	if (unlikely(!mb))
 982		return -EHOSTUNREACH;
 983
 984	rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
 985
 986	return rc ? rc : dlen;
 987}
 988
 989/**
 990 * tipc_send_group_anycast - send message to any member with given identity
 991 * @sock: socket structure
 992 * @m: message to send
 993 * @dlen: total length of message data
 994 * @timeout: timeout to wait for wakeup
 995 *
 996 * Called from function tipc_sendmsg(), which has done all sanity checks
 997 * Return: the number of bytes sent on success, or errno
 998 */
 999static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
1000				   int dlen, long timeout)
1001{
1002	struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1003	struct sock *sk = sock->sk;
1004	struct tipc_sock *tsk = tipc_sk(sk);
1005	struct list_head *cong_links = &tsk->cong_links;
1006	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
 
1007	struct tipc_msg *hdr = &tsk->phdr;
1008	struct tipc_member *first = NULL;
1009	struct tipc_member *mbr = NULL;
1010	struct net *net = sock_net(sk);
1011	u32 node, port, exclude;
1012	LIST_HEAD(dsts);
 
1013	int lookups = 0;
1014	int dstcnt, rc;
1015	bool cong;
1016
1017	ua->sa.type = msg_nametype(hdr);
1018	ua->scope = msg_lookup_scope(hdr);
1019
1020	while (++lookups < 4) {
1021		exclude = tipc_group_exclude(tsk->group);
 
 
1022
 
1023		first = NULL;
1024
1025		/* Look for a non-congested destination member, if any */
1026		while (1) {
1027			if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt,
1028						       exclude, false))
1029				return -EHOSTUNREACH;
1030			tipc_dest_pop(&dsts, &node, &port);
1031			cong = tipc_group_cong(tsk->group, node, port, blks,
1032					       &mbr);
1033			if (!cong)
1034				break;
1035			if (mbr == first)
1036				break;
1037			if (!first)
1038				first = mbr;
1039		}
1040
1041		/* Start over if destination was not in member list */
1042		if (unlikely(!mbr))
1043			continue;
1044
1045		if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
1046			break;
1047
1048		/* Block or return if destination link or member is congested */
1049		rc = tipc_wait_for_cond(sock, &timeout,
1050					!tipc_dest_find(cong_links, node, 0) &&
1051					tsk->group &&
1052					!tipc_group_cong(tsk->group, node, port,
1053							 blks, &mbr));
1054		if (unlikely(rc))
1055			return rc;
1056
1057		/* Send, unless destination disappeared while waiting */
1058		if (likely(mbr))
1059			break;
1060	}
1061
1062	if (unlikely(lookups >= 4))
1063		return -EHOSTUNREACH;
1064
1065	rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
1066
1067	return rc ? rc : dlen;
1068}
1069
1070/**
1071 * tipc_send_group_bcast - send message to all members in communication group
1072 * @sock: socket structure
1073 * @m: message to send
1074 * @dlen: total length of message data
1075 * @timeout: timeout to wait for wakeup
1076 *
1077 * Called from function tipc_sendmsg(), which has done all sanity checks
1078 * Return: the number of bytes sent on success, or errno
1079 */
1080static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1081				 int dlen, long timeout)
1082{
1083	struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1084	struct sock *sk = sock->sk;
1085	struct net *net = sock_net(sk);
1086	struct tipc_sock *tsk = tipc_sk(sk);
1087	struct tipc_nlist *dsts;
 
1088	struct tipc_mc_method *method = &tsk->mc_method;
1089	bool ack = method->mandatory && method->rcast;
1090	int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1091	struct tipc_msg *hdr = &tsk->phdr;
1092	int mtu = tipc_bcast_get_mtu(net);
1093	struct sk_buff_head pkts;
1094	int rc = -EHOSTUNREACH;
1095
 
 
 
1096	/* Block or return if any destination link or member is congested */
1097	rc = tipc_wait_for_cond(sock, &timeout,
1098				!tsk->cong_link_cnt && tsk->group &&
1099				!tipc_group_bc_cong(tsk->group, blks));
1100	if (unlikely(rc))
1101		return rc;
1102
1103	dsts = tipc_group_dests(tsk->group);
1104	if (!dsts->local && !dsts->remote)
1105		return -EHOSTUNREACH;
1106
1107	/* Complete message header */
1108	if (ua) {
1109		msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1110		msg_set_nameinst(hdr, ua->sa.instance);
1111	} else {
1112		msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1113		msg_set_nameinst(hdr, 0);
1114	}
1115	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1116	msg_set_destport(hdr, 0);
1117	msg_set_destnode(hdr, 0);
1118	msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group));
1119
1120	/* Avoid getting stuck with repeated forced replicasts */
1121	msg_set_grp_bc_ack_req(hdr, ack);
1122
1123	/* Build message as chain of buffers */
1124	__skb_queue_head_init(&pkts);
1125	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1126	if (unlikely(rc != dlen))
1127		return rc;
1128
1129	/* Send message */
1130	rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1131	if (unlikely(rc))
1132		return rc;
1133
1134	/* Update broadcast sequence number and send windows */
1135	tipc_group_update_bc_members(tsk->group, blks, ack);
1136
1137	/* Broadcast link is now free to choose method for next broadcast */
1138	method->mandatory = false;
1139	method->expires = jiffies;
1140
1141	return dlen;
1142}
1143
1144/**
1145 * tipc_send_group_mcast - send message to all members with given identity
1146 * @sock: socket structure
1147 * @m: message to send
1148 * @dlen: total length of message data
1149 * @timeout: timeout to wait for wakeup
1150 *
1151 * Called from function tipc_sendmsg(), which has done all sanity checks
1152 * Return: the number of bytes sent on success, or errno
1153 */
1154static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1155				 int dlen, long timeout)
1156{
1157	struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1158	struct sock *sk = sock->sk;
 
1159	struct tipc_sock *tsk = tipc_sk(sk);
1160	struct tipc_group *grp = tsk->group;
1161	struct tipc_msg *hdr = &tsk->phdr;
1162	struct net *net = sock_net(sk);
1163	u32 dstcnt, exclude;
1164	LIST_HEAD(dsts);
1165
1166	ua->sa.type = msg_nametype(hdr);
1167	ua->scope = msg_lookup_scope(hdr);
 
 
 
 
1168	exclude = tipc_group_exclude(grp);
1169
1170	if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt, exclude, true))
 
1171		return -EHOSTUNREACH;
1172
1173	if (dstcnt == 1) {
1174		tipc_dest_pop(&dsts, &ua->sk.node, &ua->sk.ref);
1175		return tipc_send_group_unicast(sock, m, dlen, timeout);
1176	}
1177
1178	tipc_dest_list_purge(&dsts);
1179	return tipc_send_group_bcast(sock, m, dlen, timeout);
1180}
1181
1182/**
1183 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1184 * @net: the associated network namespace
1185 * @arrvq: queue with arriving messages, to be cloned after destination lookup
1186 * @inputq: queue with cloned messages, delivered to socket after dest lookup
1187 *
1188 * Multi-threaded: parallel calls with reference to same queues may occur
1189 */
1190void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1191		       struct sk_buff_head *inputq)
1192{
1193	u32 self = tipc_own_addr(net);
 
1194	struct sk_buff *skb, *_skb;
1195	u32 portid, onode;
1196	struct sk_buff_head tmpq;
1197	struct list_head dports;
1198	struct tipc_msg *hdr;
1199	struct tipc_uaddr ua;
1200	int user, mtyp, hlen;
 
1201
1202	__skb_queue_head_init(&tmpq);
1203	INIT_LIST_HEAD(&dports);
1204	ua.addrtype = TIPC_SERVICE_RANGE;
1205
1206	/* tipc_skb_peek() increments the head skb's reference counter */
1207	skb = tipc_skb_peek(arrvq, &inputq->lock);
1208	for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1209		hdr = buf_msg(skb);
1210		user = msg_user(hdr);
1211		mtyp = msg_type(hdr);
1212		hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
 
1213		onode = msg_orignode(hdr);
1214		ua.sr.type = msg_nametype(hdr);
1215		ua.sr.lower = msg_namelower(hdr);
1216		ua.sr.upper = msg_nameupper(hdr);
1217		if (onode == self)
1218			ua.scope = TIPC_ANY_SCOPE;
1219		else
1220			ua.scope = TIPC_CLUSTER_SCOPE;
1221
1222		if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1223			spin_lock_bh(&inputq->lock);
1224			if (skb_peek(arrvq) == skb) {
1225				__skb_dequeue(arrvq);
1226				__skb_queue_tail(inputq, skb);
1227			}
1228			kfree_skb(skb);
1229			spin_unlock_bh(&inputq->lock);
1230			continue;
1231		}
1232
1233		/* Group messages require exact scope match */
1234		if (msg_in_group(hdr)) {
1235			ua.sr.lower = 0;
1236			ua.sr.upper = ~0;
1237			ua.scope = msg_lookup_scope(hdr);
 
 
 
 
 
 
 
 
 
 
1238		}
1239
1240		/* Create destination port list: */
1241		tipc_nametbl_lookup_mcast_sockets(net, &ua, &dports);
 
1242
1243		/* Clone message per destination */
1244		while (tipc_dest_pop(&dports, NULL, &portid)) {
1245			_skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1246			if (_skb) {
1247				msg_set_destport(buf_msg(_skb), portid);
1248				__skb_queue_tail(&tmpq, _skb);
1249				continue;
1250			}
1251			pr_warn("Failed to clone mcast rcv buffer\n");
1252		}
1253		/* Append clones to inputq only if skb is still head of arrvq */
1254		spin_lock_bh(&inputq->lock);
1255		if (skb_peek(arrvq) == skb) {
1256			skb_queue_splice_tail_init(&tmpq, inputq);
1257			/* Decrement the skb's refcnt */
1258			kfree_skb(__skb_dequeue(arrvq));
1259		}
1260		spin_unlock_bh(&inputq->lock);
1261		__skb_queue_purge(&tmpq);
1262		kfree_skb(skb);
1263	}
1264	tipc_sk_rcv(net, inputq);
1265}
1266
1267/* tipc_sk_push_backlog(): send accumulated buffers in socket write queue
1268 *                         when socket is in Nagle mode
1269 */
1270static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack)
1271{
1272	struct sk_buff_head *txq = &tsk->sk.sk_write_queue;
1273	struct sk_buff *skb = skb_peek_tail(txq);
1274	struct net *net = sock_net(&tsk->sk);
1275	u32 dnode = tsk_peer_node(tsk);
1276	int rc;
1277
1278	if (nagle_ack) {
1279		tsk->pkt_cnt += skb_queue_len(txq);
1280		if (!tsk->pkt_cnt || tsk->msg_acc / tsk->pkt_cnt < 2) {
1281			tsk->oneway = 0;
1282			if (tsk->nagle_start < NAGLE_START_MAX)
1283				tsk->nagle_start *= 2;
1284			tsk->expect_ack = false;
1285			pr_debug("tsk %10u: bad nagle %u -> %u, next start %u!\n",
1286				 tsk->portid, tsk->msg_acc, tsk->pkt_cnt,
1287				 tsk->nagle_start);
1288		} else {
1289			tsk->nagle_start = NAGLE_START_INIT;
1290			if (skb) {
1291				msg_set_ack_required(buf_msg(skb));
1292				tsk->expect_ack = true;
1293			} else {
1294				tsk->expect_ack = false;
1295			}
1296		}
1297		tsk->msg_acc = 0;
1298		tsk->pkt_cnt = 0;
1299	}
1300
1301	if (!skb || tsk->cong_link_cnt)
1302		return;
1303
1304	/* Do not send SYN again after congestion */
1305	if (msg_is_syn(buf_msg(skb)))
1306		return;
1307
1308	if (tsk->msg_acc)
1309		tsk->pkt_cnt += skb_queue_len(txq);
1310	tsk->snt_unacked += tsk->snd_backlog;
1311	tsk->snd_backlog = 0;
1312	rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1313	if (rc == -ELINKCONG)
1314		tsk->cong_link_cnt = 1;
1315}
1316
1317/**
1318 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1319 * @tsk: receiving socket
1320 * @skb: pointer to message buffer.
1321 * @inputq: buffer list containing the buffers
1322 * @xmitq: output message area
1323 */
1324static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1325				   struct sk_buff_head *inputq,
1326				   struct sk_buff_head *xmitq)
1327{
1328	struct tipc_msg *hdr = buf_msg(skb);
1329	u32 onode = tsk_own_node(tsk);
1330	struct sock *sk = &tsk->sk;
1331	int mtyp = msg_type(hdr);
1332	bool was_cong;
1333
1334	/* Ignore if connection cannot be validated: */
1335	if (!tsk_peer_msg(tsk, hdr)) {
1336		trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!");
1337		goto exit;
1338	}
1339
1340	if (unlikely(msg_errcode(hdr))) {
1341		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1342		tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1343				      tsk_peer_port(tsk));
1344		sk->sk_state_change(sk);
1345
1346		/* State change is ignored if socket already awake,
1347		 * - convert msg to abort msg and add to inqueue
1348		 */
1349		msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1350		msg_set_type(hdr, TIPC_CONN_MSG);
1351		msg_set_size(hdr, BASIC_H_SIZE);
1352		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1353		__skb_queue_tail(inputq, skb);
1354		return;
1355	}
1356
1357	tsk->probe_unacked = false;
1358
1359	if (mtyp == CONN_PROBE) {
1360		msg_set_type(hdr, CONN_PROBE_REPLY);
1361		if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1362			__skb_queue_tail(xmitq, skb);
1363		return;
1364	} else if (mtyp == CONN_ACK) {
1365		was_cong = tsk_conn_cong(tsk);
1366		tipc_sk_push_backlog(tsk, msg_nagle_ack(hdr));
1367		tsk->snt_unacked -= msg_conn_ack(hdr);
1368		if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1369			tsk->snd_win = msg_adv_win(hdr);
1370		if (was_cong && !tsk_conn_cong(tsk))
1371			sk->sk_write_space(sk);
1372	} else if (mtyp != CONN_PROBE_REPLY) {
1373		pr_warn("Received unknown CONN_PROTO msg\n");
1374	}
1375exit:
1376	kfree_skb(skb);
1377}
1378
1379/**
1380 * tipc_sendmsg - send message in connectionless manner
1381 * @sock: socket structure
1382 * @m: message to send
1383 * @dsz: amount of user data to be sent
1384 *
1385 * Message must have an destination specified explicitly.
1386 * Used for SOCK_RDM and SOCK_DGRAM messages,
1387 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1388 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1389 *
1390 * Return: the number of bytes sent on success, or errno otherwise
1391 */
1392static int tipc_sendmsg(struct socket *sock,
1393			struct msghdr *m, size_t dsz)
1394{
1395	struct sock *sk = sock->sk;
1396	int ret;
1397
1398	lock_sock(sk);
1399	ret = __tipc_sendmsg(sock, m, dsz);
1400	release_sock(sk);
1401
1402	return ret;
1403}
1404
1405static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1406{
1407	struct sock *sk = sock->sk;
1408	struct net *net = sock_net(sk);
1409	struct tipc_sock *tsk = tipc_sk(sk);
1410	struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1411	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1412	struct list_head *clinks = &tsk->cong_links;
1413	bool syn = !tipc_sk_type_connectionless(sk);
1414	struct tipc_group *grp = tsk->group;
1415	struct tipc_msg *hdr = &tsk->phdr;
1416	struct tipc_socket_addr skaddr;
1417	struct sk_buff_head pkts;
1418	int atype, mtu, rc;
 
 
1419
1420	if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1421		return -EMSGSIZE;
1422
1423	if (ua) {
1424		if (!tipc_uaddr_valid(ua, m->msg_namelen))
 
 
1425			return -EINVAL;
1426		atype = ua->addrtype;
1427	}
1428
1429	/* If socket belongs to a communication group follow other paths */
1430	if (grp) {
1431		if (!ua)
1432			return tipc_send_group_bcast(sock, m, dlen, timeout);
1433		if (atype == TIPC_SERVICE_ADDR)
1434			return tipc_send_group_anycast(sock, m, dlen, timeout);
1435		if (atype == TIPC_SOCKET_ADDR)
1436			return tipc_send_group_unicast(sock, m, dlen, timeout);
1437		if (atype == TIPC_SERVICE_RANGE)
1438			return tipc_send_group_mcast(sock, m, dlen, timeout);
1439		return -EINVAL;
1440	}
1441
1442	if (!ua) {
1443		ua = (struct tipc_uaddr *)&tsk->peer;
1444		if (!syn && ua->family != AF_TIPC)
1445			return -EDESTADDRREQ;
1446		atype = ua->addrtype;
1447	}
1448
1449	if (unlikely(syn)) {
1450		if (sk->sk_state == TIPC_LISTEN)
1451			return -EPIPE;
1452		if (sk->sk_state != TIPC_OPEN)
1453			return -EISCONN;
1454		if (tsk->published)
1455			return -EOPNOTSUPP;
1456		if (atype == TIPC_SERVICE_ADDR)
1457			tsk->conn_addrtype = atype;
1458		msg_set_syn(hdr, 1);
 
1459	}
1460
1461	memset(&skaddr, 0, sizeof(skaddr));
1462
1463	/* Determine destination */
1464	if (atype == TIPC_SERVICE_RANGE) {
1465		return tipc_sendmcast(sock, ua, m, dlen, timeout);
1466	} else if (atype == TIPC_SERVICE_ADDR) {
1467		skaddr.node = ua->lookup_node;
1468		ua->scope = tipc_node2scope(skaddr.node);
1469		if (!tipc_nametbl_lookup_anycast(net, ua, &skaddr))
 
 
 
 
 
 
 
 
1470			return -EHOSTUNREACH;
1471	} else if (atype == TIPC_SOCKET_ADDR) {
1472		skaddr = ua->sk;
 
 
 
 
 
1473	} else {
1474		return -EINVAL;
1475	}
1476
1477	/* Block or return if destination link is congested */
1478	rc = tipc_wait_for_cond(sock, &timeout,
1479				!tipc_dest_find(clinks, skaddr.node, 0));
1480	if (unlikely(rc))
1481		return rc;
1482
1483	/* Finally build message header */
1484	msg_set_destnode(hdr, skaddr.node);
1485	msg_set_destport(hdr, skaddr.ref);
1486	if (atype == TIPC_SERVICE_ADDR) {
1487		msg_set_type(hdr, TIPC_NAMED_MSG);
1488		msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1489		msg_set_nametype(hdr, ua->sa.type);
1490		msg_set_nameinst(hdr, ua->sa.instance);
1491		msg_set_lookup_scope(hdr, ua->scope);
1492	} else { /* TIPC_SOCKET_ADDR */
1493		msg_set_type(hdr, TIPC_DIRECT_MSG);
1494		msg_set_lookup_scope(hdr, 0);
1495		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1496	}
1497
1498	/* Add message body */
1499	__skb_queue_head_init(&pkts);
1500	mtu = tipc_node_get_mtu(net, skaddr.node, tsk->portid, true);
1501	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1502	if (unlikely(rc != dlen))
1503		return rc;
1504	if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue))) {
1505		__skb_queue_purge(&pkts);
1506		return -ENOMEM;
1507	}
1508
1509	/* Send message */
1510	trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " ");
1511	rc = tipc_node_xmit(net, &pkts, skaddr.node, tsk->portid);
1512	if (unlikely(rc == -ELINKCONG)) {
1513		tipc_dest_push(clinks, skaddr.node, 0);
1514		tsk->cong_link_cnt++;
1515		rc = 0;
1516	}
1517
1518	if (unlikely(syn && !rc)) {
1519		tipc_set_sk_state(sk, TIPC_CONNECTING);
1520		if (dlen && timeout) {
1521			timeout = msecs_to_jiffies(timeout);
1522			tipc_wait_for_connect(sock, &timeout);
1523		}
1524	}
1525
1526	return rc ? rc : dlen;
1527}
1528
1529/**
1530 * tipc_sendstream - send stream-oriented data
1531 * @sock: socket structure
1532 * @m: data to send
1533 * @dsz: total length of data to be transmitted
1534 *
1535 * Used for SOCK_STREAM data.
1536 *
1537 * Return: the number of bytes sent on success (or partial success),
1538 * or errno if no data sent
1539 */
1540static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1541{
1542	struct sock *sk = sock->sk;
1543	int ret;
1544
1545	lock_sock(sk);
1546	ret = __tipc_sendstream(sock, m, dsz);
1547	release_sock(sk);
1548
1549	return ret;
1550}
1551
1552static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1553{
1554	struct sock *sk = sock->sk;
1555	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1556	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1557	struct sk_buff_head *txq = &sk->sk_write_queue;
1558	struct tipc_sock *tsk = tipc_sk(sk);
1559	struct tipc_msg *hdr = &tsk->phdr;
1560	struct net *net = sock_net(sk);
1561	struct sk_buff *skb;
1562	u32 dnode = tsk_peer_node(tsk);
1563	int maxnagle = tsk->maxnagle;
1564	int maxpkt = tsk->max_pkt;
1565	int send, sent = 0;
1566	int blocks, rc = 0;
 
 
1567
1568	if (unlikely(dlen > INT_MAX))
1569		return -EMSGSIZE;
1570
1571	/* Handle implicit connection setup */
1572	if (unlikely(dest && sk->sk_state == TIPC_OPEN)) {
1573		rc = __tipc_sendmsg(sock, m, dlen);
1574		if (dlen && dlen == rc) {
1575			tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1576			tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1577		}
1578		return rc;
1579	}
1580
1581	do {
1582		rc = tipc_wait_for_cond(sock, &timeout,
1583					(!tsk->cong_link_cnt &&
1584					 !tsk_conn_cong(tsk) &&
1585					 tipc_sk_connected(sk)));
1586		if (unlikely(rc))
1587			break;
 
1588		send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1589		blocks = tsk->snd_backlog;
1590		if (tsk->oneway++ >= tsk->nagle_start && maxnagle &&
1591		    send <= maxnagle) {
1592			rc = tipc_msg_append(hdr, m, send, maxnagle, txq);
1593			if (unlikely(rc < 0))
1594				break;
1595			blocks += rc;
1596			tsk->msg_acc++;
1597			if (blocks <= 64 && tsk->expect_ack) {
1598				tsk->snd_backlog = blocks;
1599				sent += send;
1600				break;
1601			} else if (blocks > 64) {
1602				tsk->pkt_cnt += skb_queue_len(txq);
1603			} else {
1604				skb = skb_peek_tail(txq);
1605				if (skb) {
1606					msg_set_ack_required(buf_msg(skb));
1607					tsk->expect_ack = true;
1608				} else {
1609					tsk->expect_ack = false;
1610				}
1611				tsk->msg_acc = 0;
1612				tsk->pkt_cnt = 0;
1613			}
1614		} else {
1615			rc = tipc_msg_build(hdr, m, sent, send, maxpkt, txq);
1616			if (unlikely(rc != send))
1617				break;
1618			blocks += tsk_inc(tsk, send + MIN_H_SIZE);
1619		}
1620		trace_tipc_sk_sendstream(sk, skb_peek(txq),
1621					 TIPC_DUMP_SK_SNDQ, " ");
1622		rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1623		if (unlikely(rc == -ELINKCONG)) {
1624			tsk->cong_link_cnt = 1;
1625			rc = 0;
1626		}
1627		if (likely(!rc)) {
1628			tsk->snt_unacked += blocks;
1629			tsk->snd_backlog = 0;
1630			sent += send;
1631		}
1632	} while (sent < dlen && !rc);
1633
1634	return sent ? sent : rc;
1635}
1636
1637/**
1638 * tipc_send_packet - send a connection-oriented message
1639 * @sock: socket structure
1640 * @m: message to send
1641 * @dsz: length of data to be transmitted
1642 *
1643 * Used for SOCK_SEQPACKET messages.
1644 *
1645 * Return: the number of bytes sent on success, or errno otherwise
1646 */
1647static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1648{
1649	if (dsz > TIPC_MAX_USER_MSG_SIZE)
1650		return -EMSGSIZE;
1651
1652	return tipc_sendstream(sock, m, dsz);
1653}
1654
1655/* tipc_sk_finish_conn - complete the setup of a connection
1656 */
1657static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1658				u32 peer_node)
1659{
1660	struct sock *sk = &tsk->sk;
1661	struct net *net = sock_net(sk);
1662	struct tipc_msg *msg = &tsk->phdr;
1663
1664	msg_set_syn(msg, 0);
1665	msg_set_destnode(msg, peer_node);
1666	msg_set_destport(msg, peer_port);
1667	msg_set_type(msg, TIPC_CONN_MSG);
1668	msg_set_lookup_scope(msg, 0);
1669	msg_set_hdr_sz(msg, SHORT_H_SIZE);
1670
1671	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1672	tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1673	tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1674	tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid, true);
1675	tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1676	tsk_set_nagle(tsk);
1677	__skb_queue_purge(&sk->sk_write_queue);
1678	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1679		return;
1680
1681	/* Fall back to message based flow control */
1682	tsk->rcv_win = FLOWCTL_MSG_WIN;
1683	tsk->snd_win = FLOWCTL_MSG_WIN;
1684}
1685
1686/**
1687 * tipc_sk_set_orig_addr - capture sender's address for received message
1688 * @m: descriptor for message info
1689 * @skb: received message
1690 *
1691 * Note: Address is not captured if not requested by receiver.
1692 */
1693static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1694{
1695	DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1696	struct tipc_msg *hdr = buf_msg(skb);
1697
1698	if (!srcaddr)
1699		return;
1700
1701	srcaddr->sock.family = AF_TIPC;
1702	srcaddr->sock.addrtype = TIPC_SOCKET_ADDR;
1703	srcaddr->sock.scope = 0;
1704	srcaddr->sock.addr.id.ref = msg_origport(hdr);
1705	srcaddr->sock.addr.id.node = msg_orignode(hdr);
1706	srcaddr->sock.addr.name.domain = 0;
1707	m->msg_namelen = sizeof(struct sockaddr_tipc);
1708
1709	if (!msg_in_group(hdr))
1710		return;
1711
1712	/* Group message users may also want to know sending member's id */
1713	srcaddr->member.family = AF_TIPC;
1714	srcaddr->member.addrtype = TIPC_SERVICE_ADDR;
1715	srcaddr->member.scope = 0;
1716	srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1717	srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1718	srcaddr->member.addr.name.domain = 0;
1719	m->msg_namelen = sizeof(*srcaddr);
1720}
1721
1722/**
1723 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1724 * @m: descriptor for message info
1725 * @skb: received message buffer
1726 * @tsk: TIPC port associated with message
1727 *
1728 * Note: Ancillary data is not captured if not requested by receiver.
1729 *
1730 * Return: 0 if successful, otherwise errno
1731 */
1732static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1733				 struct tipc_sock *tsk)
1734{
1735	struct tipc_msg *hdr;
1736	u32 data[3] = {0,};
1737	bool has_addr;
1738	int dlen, rc;
 
1739
1740	if (likely(m->msg_controllen == 0))
1741		return 0;
1742
1743	hdr = buf_msg(skb);
1744	dlen = msg_data_sz(hdr);
1745
1746	/* Capture errored message object, if any */
1747	if (msg_errcode(hdr)) {
1748		if (skb_linearize(skb))
1749			return -ENOMEM;
1750		hdr = buf_msg(skb);
1751		data[0] = msg_errcode(hdr);
1752		data[1] = dlen;
1753		rc = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, data);
1754		if (rc || !dlen)
1755			return rc;
1756		rc = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, dlen, msg_data(hdr));
1757		if (rc)
1758			return rc;
1759	}
1760
1761	/* Capture TIPC_SERVICE_ADDR/RANGE destination address, if any */
1762	switch (msg_type(hdr)) {
 
1763	case TIPC_NAMED_MSG:
1764		has_addr = true;
1765		data[0] = msg_nametype(hdr);
1766		data[1] = msg_namelower(hdr);
1767		data[2] = data[1];
1768		break;
1769	case TIPC_MCAST_MSG:
1770		has_addr = true;
1771		data[0] = msg_nametype(hdr);
1772		data[1] = msg_namelower(hdr);
1773		data[2] = msg_nameupper(hdr);
1774		break;
1775	case TIPC_CONN_MSG:
1776		has_addr = !!tsk->conn_addrtype;
1777		data[0] = msg_nametype(&tsk->phdr);
1778		data[1] = msg_nameinst(&tsk->phdr);
1779		data[2] = data[1];
1780		break;
1781	default:
1782		has_addr = false;
1783	}
1784	if (!has_addr)
1785		return 0;
1786	return put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, data);
 
 
 
 
1787}
1788
1789static struct sk_buff *tipc_sk_build_ack(struct tipc_sock *tsk)
1790{
1791	struct sock *sk = &tsk->sk;
 
1792	struct sk_buff *skb = NULL;
1793	struct tipc_msg *msg;
1794	u32 peer_port = tsk_peer_port(tsk);
1795	u32 dnode = tsk_peer_node(tsk);
1796
1797	if (!tipc_sk_connected(sk))
1798		return NULL;
1799	skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1800			      dnode, tsk_own_node(tsk), peer_port,
1801			      tsk->portid, TIPC_OK);
1802	if (!skb)
1803		return NULL;
1804	msg = buf_msg(skb);
1805	msg_set_conn_ack(msg, tsk->rcv_unacked);
1806	tsk->rcv_unacked = 0;
1807
1808	/* Adjust to and advertize the correct window limit */
1809	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1810		tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1811		msg_set_adv_win(msg, tsk->rcv_win);
1812	}
1813	return skb;
1814}
1815
1816static void tipc_sk_send_ack(struct tipc_sock *tsk)
1817{
1818	struct sk_buff *skb;
1819
1820	skb = tipc_sk_build_ack(tsk);
1821	if (!skb)
1822		return;
1823
1824	tipc_node_xmit_skb(sock_net(&tsk->sk), skb, tsk_peer_node(tsk),
1825			   msg_link_selector(buf_msg(skb)));
1826}
1827
1828static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1829{
1830	struct sock *sk = sock->sk;
1831	DEFINE_WAIT_FUNC(wait, woken_wake_function);
1832	long timeo = *timeop;
1833	int err = sock_error(sk);
1834
1835	if (err)
1836		return err;
1837
1838	for (;;) {
 
1839		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1840			if (sk->sk_shutdown & RCV_SHUTDOWN) {
1841				err = -ENOTCONN;
1842				break;
1843			}
1844			add_wait_queue(sk_sleep(sk), &wait);
1845			release_sock(sk);
1846			timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
1847			sched_annotate_sleep();
1848			lock_sock(sk);
1849			remove_wait_queue(sk_sleep(sk), &wait);
1850		}
1851		err = 0;
1852		if (!skb_queue_empty(&sk->sk_receive_queue))
1853			break;
1854		err = -EAGAIN;
1855		if (!timeo)
1856			break;
1857		err = sock_intr_errno(timeo);
1858		if (signal_pending(current))
1859			break;
1860
1861		err = sock_error(sk);
1862		if (err)
1863			break;
1864	}
 
1865	*timeop = timeo;
1866	return err;
1867}
1868
1869/**
1870 * tipc_recvmsg - receive packet-oriented message
1871 * @sock: network socket
1872 * @m: descriptor for message info
1873 * @buflen: length of user buffer area
1874 * @flags: receive flags
1875 *
1876 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1877 * If the complete message doesn't fit in user area, truncate it.
1878 *
1879 * Return: size of returned message data, errno otherwise
1880 */
1881static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1882			size_t buflen,	int flags)
1883{
1884	struct sock *sk = sock->sk;
1885	bool connected = !tipc_sk_type_connectionless(sk);
1886	struct tipc_sock *tsk = tipc_sk(sk);
1887	int rc, err, hlen, dlen, copy;
1888	struct tipc_skb_cb *skb_cb;
1889	struct sk_buff_head xmitq;
1890	struct tipc_msg *hdr;
1891	struct sk_buff *skb;
1892	bool grp_evt;
1893	long timeout;
1894
1895	/* Catch invalid receive requests */
1896	if (unlikely(!buflen))
1897		return -EINVAL;
1898
1899	lock_sock(sk);
1900	if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1901		rc = -ENOTCONN;
1902		goto exit;
1903	}
1904	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1905
1906	/* Step rcv queue to first msg with data or error; wait if necessary */
1907	do {
1908		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1909		if (unlikely(rc))
1910			goto exit;
1911		skb = skb_peek(&sk->sk_receive_queue);
1912		skb_cb = TIPC_SKB_CB(skb);
1913		hdr = buf_msg(skb);
1914		dlen = msg_data_sz(hdr);
1915		hlen = msg_hdr_sz(hdr);
1916		err = msg_errcode(hdr);
1917		grp_evt = msg_is_grp_evt(hdr);
1918		if (likely(dlen || err))
1919			break;
1920		tsk_advance_rx_queue(sk);
1921	} while (1);
1922
1923	/* Collect msg meta data, including error code and rejected data */
1924	tipc_sk_set_orig_addr(m, skb);
1925	rc = tipc_sk_anc_data_recv(m, skb, tsk);
1926	if (unlikely(rc))
1927		goto exit;
1928	hdr = buf_msg(skb);
1929
1930	/* Capture data if non-error msg, otherwise just set return value */
1931	if (likely(!err)) {
1932		int offset = skb_cb->bytes_read;
1933
1934		copy = min_t(int, dlen - offset, buflen);
1935		rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1936		if (unlikely(rc))
1937			goto exit;
1938		if (unlikely(offset + copy < dlen)) {
1939			if (flags & MSG_EOR) {
1940				if (!(flags & MSG_PEEK))
1941					skb_cb->bytes_read = offset + copy;
1942			} else {
1943				m->msg_flags |= MSG_TRUNC;
1944				skb_cb->bytes_read = 0;
1945			}
1946		} else {
1947			if (flags & MSG_EOR)
1948				m->msg_flags |= MSG_EOR;
1949			skb_cb->bytes_read = 0;
1950		}
1951	} else {
1952		copy = 0;
1953		rc = 0;
1954		if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control) {
1955			rc = -ECONNRESET;
1956			goto exit;
1957		}
1958	}
 
 
1959
1960	/* Mark message as group event if applicable */
1961	if (unlikely(grp_evt)) {
1962		if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1963			m->msg_flags |= MSG_EOR;
1964		m->msg_flags |= MSG_OOB;
1965		copy = 0;
1966	}
1967
1968	/* Caption of data or error code/rejected data was successful */
1969	if (unlikely(flags & MSG_PEEK))
1970		goto exit;
1971
1972	/* Send group flow control advertisement when applicable */
1973	if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1974		__skb_queue_head_init(&xmitq);
1975		tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1976					  msg_orignode(hdr), msg_origport(hdr),
1977					  &xmitq);
1978		tipc_node_distr_xmit(sock_net(sk), &xmitq);
1979	}
1980
1981	if (skb_cb->bytes_read)
1982		goto exit;
1983
1984	tsk_advance_rx_queue(sk);
1985
1986	if (likely(!connected))
1987		goto exit;
1988
1989	/* Send connection flow control advertisement when applicable */
1990	tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1991	if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1992		tipc_sk_send_ack(tsk);
1993exit:
1994	release_sock(sk);
1995	return rc ? rc : copy;
1996}
1997
1998/**
1999 * tipc_recvstream - receive stream-oriented data
2000 * @sock: network socket
2001 * @m: descriptor for message info
2002 * @buflen: total size of user buffer area
2003 * @flags: receive flags
2004 *
2005 * Used for SOCK_STREAM messages only.  If not enough data is available
2006 * will optionally wait for more; never truncates data.
2007 *
2008 * Return: size of returned message data, errno otherwise
2009 */
2010static int tipc_recvstream(struct socket *sock, struct msghdr *m,
2011			   size_t buflen, int flags)
2012{
2013	struct sock *sk = sock->sk;
2014	struct tipc_sock *tsk = tipc_sk(sk);
2015	struct sk_buff *skb;
2016	struct tipc_msg *hdr;
2017	struct tipc_skb_cb *skb_cb;
2018	bool peek = flags & MSG_PEEK;
2019	int offset, required, copy, copied = 0;
2020	int hlen, dlen, err, rc;
2021	long timeout;
2022
2023	/* Catch invalid receive attempts */
2024	if (unlikely(!buflen))
2025		return -EINVAL;
2026
2027	lock_sock(sk);
2028
2029	if (unlikely(sk->sk_state == TIPC_OPEN)) {
2030		rc = -ENOTCONN;
2031		goto exit;
2032	}
2033	required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
2034	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2035
2036	do {
2037		/* Look at first msg in receive queue; wait if necessary */
2038		rc = tipc_wait_for_rcvmsg(sock, &timeout);
2039		if (unlikely(rc))
2040			break;
2041		skb = skb_peek(&sk->sk_receive_queue);
2042		skb_cb = TIPC_SKB_CB(skb);
2043		hdr = buf_msg(skb);
2044		dlen = msg_data_sz(hdr);
2045		hlen = msg_hdr_sz(hdr);
2046		err = msg_errcode(hdr);
2047
2048		/* Discard any empty non-errored (SYN-) message */
2049		if (unlikely(!dlen && !err)) {
2050			tsk_advance_rx_queue(sk);
2051			continue;
2052		}
2053
2054		/* Collect msg meta data, incl. error code and rejected data */
2055		if (!copied) {
2056			tipc_sk_set_orig_addr(m, skb);
2057			rc = tipc_sk_anc_data_recv(m, skb, tsk);
2058			if (rc)
2059				break;
2060			hdr = buf_msg(skb);
2061		}
2062
2063		/* Copy data if msg ok, otherwise return error/partial data */
2064		if (likely(!err)) {
2065			offset = skb_cb->bytes_read;
2066			copy = min_t(int, dlen - offset, buflen - copied);
2067			rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
2068			if (unlikely(rc))
2069				break;
2070			copied += copy;
2071			offset += copy;
2072			if (unlikely(offset < dlen)) {
2073				if (!peek)
2074					skb_cb->bytes_read = offset;
2075				break;
2076			}
2077		} else {
2078			rc = 0;
2079			if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
2080				rc = -ECONNRESET;
2081			if (copied || rc)
2082				break;
2083		}
2084
2085		if (unlikely(peek))
2086			break;
2087
2088		tsk_advance_rx_queue(sk);
2089
2090		/* Send connection flow control advertisement when applicable */
2091		tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
2092		if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
2093			tipc_sk_send_ack(tsk);
2094
2095		/* Exit if all requested data or FIN/error received */
2096		if (copied == buflen || err)
2097			break;
2098
2099	} while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
2100exit:
2101	release_sock(sk);
2102	return copied ? copied : rc;
2103}
2104
2105/**
2106 * tipc_write_space - wake up thread if port congestion is released
2107 * @sk: socket
2108 */
2109static void tipc_write_space(struct sock *sk)
2110{
2111	struct socket_wq *wq;
2112
2113	rcu_read_lock();
2114	wq = rcu_dereference(sk->sk_wq);
2115	if (skwq_has_sleeper(wq))
2116		wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
2117						EPOLLWRNORM | EPOLLWRBAND);
2118	rcu_read_unlock();
2119}
2120
2121/**
2122 * tipc_data_ready - wake up threads to indicate messages have been received
2123 * @sk: socket
 
2124 */
2125static void tipc_data_ready(struct sock *sk)
2126{
2127	struct socket_wq *wq;
2128
2129	trace_sk_data_ready(sk);
2130
2131	rcu_read_lock();
2132	wq = rcu_dereference(sk->sk_wq);
2133	if (skwq_has_sleeper(wq))
2134		wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
2135						EPOLLRDNORM | EPOLLRDBAND);
2136	rcu_read_unlock();
2137}
2138
2139static void tipc_sock_destruct(struct sock *sk)
2140{
2141	__skb_queue_purge(&sk->sk_receive_queue);
2142}
2143
2144static void tipc_sk_proto_rcv(struct sock *sk,
2145			      struct sk_buff_head *inputq,
2146			      struct sk_buff_head *xmitq)
2147{
2148	struct sk_buff *skb = __skb_dequeue(inputq);
2149	struct tipc_sock *tsk = tipc_sk(sk);
2150	struct tipc_msg *hdr = buf_msg(skb);
2151	struct tipc_group *grp = tsk->group;
2152	bool wakeup = false;
2153
2154	switch (msg_user(hdr)) {
2155	case CONN_MANAGER:
2156		tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
2157		return;
2158	case SOCK_WAKEUP:
2159		tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
2160		/* coupled with smp_rmb() in tipc_wait_for_cond() */
2161		smp_wmb();
2162		tsk->cong_link_cnt--;
2163		wakeup = true;
2164		tipc_sk_push_backlog(tsk, false);
2165		break;
2166	case GROUP_PROTOCOL:
2167		tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
2168		break;
2169	case TOP_SRV:
2170		tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
2171				      hdr, inputq, xmitq);
2172		break;
2173	default:
2174		break;
2175	}
2176
2177	if (wakeup)
2178		sk->sk_write_space(sk);
2179
2180	kfree_skb(skb);
2181}
2182
2183/**
2184 * tipc_sk_filter_connect - check incoming message for a connection-based socket
2185 * @tsk: TIPC socket
2186 * @skb: pointer to message buffer.
2187 * @xmitq: for Nagle ACK if any
2188 * Return: true if message should be added to receive queue, false otherwise
2189 */
2190static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb,
2191				   struct sk_buff_head *xmitq)
2192{
2193	struct sock *sk = &tsk->sk;
2194	struct net *net = sock_net(sk);
2195	struct tipc_msg *hdr = buf_msg(skb);
2196	bool con_msg = msg_connected(hdr);
2197	u32 pport = tsk_peer_port(tsk);
2198	u32 pnode = tsk_peer_node(tsk);
2199	u32 oport = msg_origport(hdr);
2200	u32 onode = msg_orignode(hdr);
2201	int err = msg_errcode(hdr);
2202	unsigned long delay;
2203
2204	if (unlikely(msg_mcast(hdr)))
2205		return false;
2206	tsk->oneway = 0;
2207
2208	switch (sk->sk_state) {
2209	case TIPC_CONNECTING:
2210		/* Setup ACK */
2211		if (likely(con_msg)) {
2212			if (err)
2213				break;
2214			tipc_sk_finish_conn(tsk, oport, onode);
2215			msg_set_importance(&tsk->phdr, msg_importance(hdr));
2216			/* ACK+ message with data is added to receive queue */
2217			if (msg_data_sz(hdr))
2218				return true;
2219			/* Empty ACK-, - wake up sleeping connect() and drop */
2220			sk->sk_state_change(sk);
2221			msg_set_dest_droppable(hdr, 1);
2222			return false;
2223		}
2224		/* Ignore connectionless message if not from listening socket */
2225		if (oport != pport || onode != pnode)
2226			return false;
2227
2228		/* Rejected SYN */
2229		if (err != TIPC_ERR_OVERLOAD)
2230			break;
 
 
 
2231
2232		/* Prepare for new setup attempt if we have a SYN clone */
2233		if (skb_queue_empty(&sk->sk_write_queue))
2234			break;
2235		get_random_bytes(&delay, 2);
2236		delay %= (tsk->conn_timeout / 4);
2237		delay = msecs_to_jiffies(delay + 100);
2238		sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
 
 
 
 
 
 
 
 
 
 
 
 
2239		return false;
 
2240	case TIPC_OPEN:
2241	case TIPC_DISCONNECTING:
2242		return false;
2243	case TIPC_LISTEN:
2244		/* Accept only SYN message */
2245		if (!msg_is_syn(hdr) &&
2246		    tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2247			return false;
2248		if (!con_msg && !err)
2249			return true;
2250		return false;
2251	case TIPC_ESTABLISHED:
2252		if (!skb_queue_empty(&sk->sk_write_queue))
2253			tipc_sk_push_backlog(tsk, false);
2254		/* Accept only connection-based messages sent by peer */
2255		if (likely(con_msg && !err && pport == oport &&
2256			   pnode == onode)) {
2257			if (msg_ack_required(hdr)) {
2258				struct sk_buff *skb;
2259
2260				skb = tipc_sk_build_ack(tsk);
2261				if (skb) {
2262					msg_set_nagle_ack(buf_msg(skb));
2263					__skb_queue_tail(xmitq, skb);
2264				}
2265			}
2266			return true;
2267		}
2268		if (!tsk_peer_msg(tsk, hdr))
2269			return false;
2270		if (!err)
2271			return true;
2272		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2273		tipc_node_remove_conn(net, pnode, tsk->portid);
2274		sk->sk_state_change(sk);
 
 
 
2275		return true;
2276	default:
2277		pr_err("Unknown sk_state %u\n", sk->sk_state);
2278	}
2279	/* Abort connection setup attempt */
2280	tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2281	sk->sk_err = ECONNREFUSED;
2282	sk->sk_state_change(sk);
2283	return true;
2284}
2285
2286/**
2287 * rcvbuf_limit - get proper overload limit of socket receive queue
2288 * @sk: socket
2289 * @skb: message
2290 *
2291 * For connection oriented messages, irrespective of importance,
2292 * default queue limit is 2 MB.
2293 *
2294 * For connectionless messages, queue limits are based on message
2295 * importance as follows:
2296 *
2297 * TIPC_LOW_IMPORTANCE       (2 MB)
2298 * TIPC_MEDIUM_IMPORTANCE    (4 MB)
2299 * TIPC_HIGH_IMPORTANCE      (8 MB)
2300 * TIPC_CRITICAL_IMPORTANCE  (16 MB)
2301 *
2302 * Return: overload limit according to corresponding message importance
2303 */
2304static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2305{
2306	struct tipc_sock *tsk = tipc_sk(sk);
2307	struct tipc_msg *hdr = buf_msg(skb);
2308
2309	if (unlikely(msg_in_group(hdr)))
2310		return READ_ONCE(sk->sk_rcvbuf);
2311
2312	if (unlikely(!msg_connected(hdr)))
2313		return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr);
2314
2315	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2316		return READ_ONCE(sk->sk_rcvbuf);
2317
2318	return FLOWCTL_MSG_LIM;
2319}
2320
2321/**
2322 * tipc_sk_filter_rcv - validate incoming message
2323 * @sk: socket
2324 * @skb: pointer to message.
2325 * @xmitq: output message area (FIXME)
2326 *
2327 * Enqueues message on receive queue if acceptable; optionally handles
2328 * disconnect indication for a connected socket.
2329 *
2330 * Called with socket lock already taken
 
2331 */
2332static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2333			       struct sk_buff_head *xmitq)
2334{
2335	bool sk_conn = !tipc_sk_type_connectionless(sk);
2336	struct tipc_sock *tsk = tipc_sk(sk);
2337	struct tipc_group *grp = tsk->group;
2338	struct tipc_msg *hdr = buf_msg(skb);
2339	struct net *net = sock_net(sk);
2340	struct sk_buff_head inputq;
2341	int mtyp = msg_type(hdr);
2342	int limit, err = TIPC_OK;
2343
2344	trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " ");
2345	TIPC_SKB_CB(skb)->bytes_read = 0;
2346	__skb_queue_head_init(&inputq);
2347	__skb_queue_tail(&inputq, skb);
2348
2349	if (unlikely(!msg_isdata(hdr)))
2350		tipc_sk_proto_rcv(sk, &inputq, xmitq);
2351
2352	if (unlikely(grp))
2353		tipc_group_filter_msg(grp, &inputq, xmitq);
2354
2355	if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG)
2356		tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq);
2357
2358	/* Validate and add to receive buffer if there is space */
2359	while ((skb = __skb_dequeue(&inputq))) {
2360		hdr = buf_msg(skb);
2361		limit = rcvbuf_limit(sk, skb);
2362		if ((sk_conn && !tipc_sk_filter_connect(tsk, skb, xmitq)) ||
2363		    (!sk_conn && msg_connected(hdr)) ||
2364		    (!grp && msg_in_group(hdr)))
2365			err = TIPC_ERR_NO_PORT;
2366		else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2367			trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL,
2368					   "err_overload2!");
2369			atomic_inc(&sk->sk_drops);
2370			err = TIPC_ERR_OVERLOAD;
2371		}
2372
2373		if (unlikely(err)) {
2374			if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) {
2375				trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE,
2376						      "@filter_rcv!");
2377				__skb_queue_tail(xmitq, skb);
2378			}
2379			err = TIPC_OK;
2380			continue;
2381		}
2382		__skb_queue_tail(&sk->sk_receive_queue, skb);
2383		skb_set_owner_r(skb, sk);
2384		trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL,
2385					 "rcvq >90% allocated!");
2386		sk->sk_data_ready(sk);
2387	}
2388}
2389
2390/**
2391 * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2392 * @sk: socket
2393 * @skb: message
2394 *
2395 * Caller must hold socket lock
2396 */
2397static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2398{
2399	unsigned int before = sk_rmem_alloc_get(sk);
2400	struct sk_buff_head xmitq;
2401	unsigned int added;
2402
2403	__skb_queue_head_init(&xmitq);
2404
2405	tipc_sk_filter_rcv(sk, skb, &xmitq);
2406	added = sk_rmem_alloc_get(sk) - before;
2407	atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2408
2409	/* Send pending response/rejected messages, if any */
2410	tipc_node_distr_xmit(sock_net(sk), &xmitq);
2411	return 0;
2412}
2413
2414/**
2415 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2416 *                   inputq and try adding them to socket or backlog queue
2417 * @inputq: list of incoming buffers with potentially different destinations
2418 * @sk: socket where the buffers should be enqueued
2419 * @dport: port number for the socket
2420 * @xmitq: output queue
2421 *
2422 * Caller must hold socket lock
2423 */
2424static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2425			    u32 dport, struct sk_buff_head *xmitq)
2426{
2427	unsigned long time_limit = jiffies + usecs_to_jiffies(20000);
2428	struct sk_buff *skb;
2429	unsigned int lim;
2430	atomic_t *dcnt;
2431	u32 onode;
2432
2433	while (skb_queue_len(inputq)) {
2434		if (unlikely(time_after_eq(jiffies, time_limit)))
2435			return;
2436
2437		skb = tipc_skb_dequeue(inputq, dport);
2438		if (unlikely(!skb))
2439			return;
2440
2441		/* Add message directly to receive queue if possible */
2442		if (!sock_owned_by_user(sk)) {
2443			tipc_sk_filter_rcv(sk, skb, xmitq);
2444			continue;
2445		}
2446
2447		/* Try backlog, compensating for double-counted bytes */
2448		dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2449		if (!sk->sk_backlog.len)
2450			atomic_set(dcnt, 0);
2451		lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2452		if (likely(!sk_add_backlog(sk, skb, lim))) {
2453			trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL,
2454						 "bklg & rcvq >90% allocated!");
2455			continue;
2456		}
2457
2458		trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!");
2459		/* Overload => reject message back to sender */
2460		onode = tipc_own_addr(sock_net(sk));
2461		atomic_inc(&sk->sk_drops);
2462		if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) {
2463			trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL,
2464					      "@sk_enqueue!");
2465			__skb_queue_tail(xmitq, skb);
2466		}
2467		break;
2468	}
2469}
2470
2471/**
2472 * tipc_sk_rcv - handle a chain of incoming buffers
2473 * @net: the associated network namespace
2474 * @inputq: buffer list containing the buffers
2475 * Consumes all buffers in list until inputq is empty
2476 * Note: may be called in multiple threads referring to the same queue
2477 */
2478void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2479{
2480	struct sk_buff_head xmitq;
2481	u32 dnode, dport = 0;
2482	int err;
2483	struct tipc_sock *tsk;
2484	struct sock *sk;
2485	struct sk_buff *skb;
2486
2487	__skb_queue_head_init(&xmitq);
2488	while (skb_queue_len(inputq)) {
2489		dport = tipc_skb_peek_port(inputq, dport);
2490		tsk = tipc_sk_lookup(net, dport);
2491
2492		if (likely(tsk)) {
2493			sk = &tsk->sk;
2494			if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2495				tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2496				spin_unlock_bh(&sk->sk_lock.slock);
2497			}
2498			/* Send pending response/rejected messages, if any */
2499			tipc_node_distr_xmit(sock_net(sk), &xmitq);
2500			sock_put(sk);
2501			continue;
2502		}
2503		/* No destination socket => dequeue skb if still there */
2504		skb = tipc_skb_dequeue(inputq, dport);
2505		if (!skb)
2506			return;
2507
2508		/* Try secondary lookup if unresolved named message */
2509		err = TIPC_ERR_NO_PORT;
2510		if (tipc_msg_lookup_dest(net, skb, &err))
2511			goto xmit;
2512
2513		/* Prepare for message rejection */
2514		if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2515			continue;
2516
2517		trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!");
2518xmit:
2519		dnode = msg_destnode(buf_msg(skb));
2520		tipc_node_xmit_skb(net, skb, dnode, dport);
2521	}
2522}
2523
2524static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2525{
2526	DEFINE_WAIT_FUNC(wait, woken_wake_function);
2527	struct sock *sk = sock->sk;
2528	int done;
2529
2530	do {
2531		int err = sock_error(sk);
2532		if (err)
2533			return err;
2534		if (!*timeo_p)
2535			return -ETIMEDOUT;
2536		if (signal_pending(current))
2537			return sock_intr_errno(*timeo_p);
2538		if (sk->sk_state == TIPC_DISCONNECTING)
2539			break;
2540
2541		add_wait_queue(sk_sleep(sk), &wait);
2542		done = sk_wait_event(sk, timeo_p, tipc_sk_connected(sk),
2543				     &wait);
2544		remove_wait_queue(sk_sleep(sk), &wait);
2545	} while (!done);
2546	return 0;
2547}
2548
2549static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr)
2550{
2551	if (addr->family != AF_TIPC)
2552		return false;
2553	if (addr->addrtype == TIPC_SERVICE_RANGE)
2554		return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper);
2555	return (addr->addrtype == TIPC_SERVICE_ADDR ||
2556		addr->addrtype == TIPC_SOCKET_ADDR);
2557}
2558
2559/**
2560 * tipc_connect - establish a connection to another TIPC port
2561 * @sock: socket structure
2562 * @dest: socket address for destination port
2563 * @destlen: size of socket address data structure
2564 * @flags: file-related flags associated with socket
2565 *
2566 * Return: 0 on success, errno otherwise
2567 */
2568static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2569			int destlen, int flags)
2570{
2571	struct sock *sk = sock->sk;
2572	struct tipc_sock *tsk = tipc_sk(sk);
2573	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2574	struct msghdr m = {NULL,};
2575	long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2576	int previous;
2577	int res = 0;
2578
2579	if (destlen != sizeof(struct sockaddr_tipc))
2580		return -EINVAL;
2581
2582	lock_sock(sk);
2583
2584	if (tsk->group) {
2585		res = -EINVAL;
2586		goto exit;
2587	}
2588
2589	if (dst->family == AF_UNSPEC) {
2590		memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2591		if (!tipc_sk_type_connectionless(sk))
2592			res = -EINVAL;
2593		goto exit;
 
 
2594	}
2595	if (!tipc_sockaddr_is_sane(dst)) {
2596		res = -EINVAL;
 
2597		goto exit;
2598	}
2599	/* DGRAM/RDM connect(), just save the destaddr */
2600	if (tipc_sk_type_connectionless(sk)) {
2601		memcpy(&tsk->peer, dest, destlen);
2602		goto exit;
2603	} else if (dst->addrtype == TIPC_SERVICE_RANGE) {
2604		res = -EINVAL;
2605		goto exit;
2606	}
2607
2608	previous = sk->sk_state;
2609
2610	switch (sk->sk_state) {
2611	case TIPC_OPEN:
2612		/* Send a 'SYN-' to destination */
2613		m.msg_name = dest;
2614		m.msg_namelen = destlen;
2615		iov_iter_kvec(&m.msg_iter, ITER_SOURCE, NULL, 0, 0);
2616
2617		/* If connect is in non-blocking case, set MSG_DONTWAIT to
2618		 * indicate send_msg() is never blocked.
2619		 */
2620		if (!timeout)
2621			m.msg_flags = MSG_DONTWAIT;
2622
2623		res = __tipc_sendmsg(sock, &m, 0);
2624		if ((res < 0) && (res != -EWOULDBLOCK))
2625			goto exit;
2626
2627		/* Just entered TIPC_CONNECTING state; the only
2628		 * difference is that return value in non-blocking
2629		 * case is EINPROGRESS, rather than EALREADY.
2630		 */
2631		res = -EINPROGRESS;
2632		fallthrough;
2633	case TIPC_CONNECTING:
2634		if (!timeout) {
2635			if (previous == TIPC_CONNECTING)
2636				res = -EALREADY;
2637			goto exit;
2638		}
2639		timeout = msecs_to_jiffies(timeout);
2640		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2641		res = tipc_wait_for_connect(sock, &timeout);
2642		break;
2643	case TIPC_ESTABLISHED:
2644		res = -EISCONN;
2645		break;
2646	default:
2647		res = -EINVAL;
2648	}
2649
2650exit:
2651	release_sock(sk);
2652	return res;
2653}
2654
2655/**
2656 * tipc_listen - allow socket to listen for incoming connections
2657 * @sock: socket structure
2658 * @len: (unused)
2659 *
2660 * Return: 0 on success, errno otherwise
2661 */
2662static int tipc_listen(struct socket *sock, int len)
2663{
2664	struct sock *sk = sock->sk;
2665	int res;
2666
2667	lock_sock(sk);
2668	res = tipc_set_sk_state(sk, TIPC_LISTEN);
2669	release_sock(sk);
2670
2671	return res;
2672}
2673
2674static int tipc_wait_for_accept(struct socket *sock, long timeo)
2675{
2676	struct sock *sk = sock->sk;
2677	DEFINE_WAIT_FUNC(wait, woken_wake_function);
2678	int err;
2679
2680	/* True wake-one mechanism for incoming connections: only
2681	 * one process gets woken up, not the 'whole herd'.
2682	 * Since we do not 'race & poll' for established sockets
2683	 * anymore, the common case will execute the loop only once.
2684	*/
2685	for (;;) {
 
 
2686		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2687			add_wait_queue(sk_sleep(sk), &wait);
2688			release_sock(sk);
2689			timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
2690			lock_sock(sk);
2691			remove_wait_queue(sk_sleep(sk), &wait);
2692		}
2693		err = 0;
2694		if (!skb_queue_empty(&sk->sk_receive_queue))
2695			break;
2696		err = -EAGAIN;
2697		if (!timeo)
2698			break;
2699		err = sock_intr_errno(timeo);
2700		if (signal_pending(current))
2701			break;
2702	}
 
2703	return err;
2704}
2705
2706/**
2707 * tipc_accept - wait for connection request
2708 * @sock: listening socket
2709 * @new_sock: new socket that is to be connected
2710 * @arg: arguments for accept
2711 *
2712 * Return: 0 on success, errno otherwise
2713 */
2714static int tipc_accept(struct socket *sock, struct socket *new_sock,
2715		       struct proto_accept_arg *arg)
2716{
2717	struct sock *new_sk, *sk = sock->sk;
 
2718	struct tipc_sock *new_tsock;
2719	struct msghdr m = {NULL,};
2720	struct tipc_msg *msg;
2721	struct sk_buff *buf;
2722	long timeo;
2723	int res;
2724
2725	lock_sock(sk);
2726
2727	if (sk->sk_state != TIPC_LISTEN) {
2728		res = -EINVAL;
2729		goto exit;
2730	}
2731	timeo = sock_rcvtimeo(sk, arg->flags & O_NONBLOCK);
2732	res = tipc_wait_for_accept(sock, timeo);
2733	if (res)
2734		goto exit;
2735
2736	buf = skb_peek(&sk->sk_receive_queue);
2737
2738	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, arg->kern);
2739	if (res)
2740		goto exit;
2741	security_sk_clone(sock->sk, new_sock->sk);
2742
2743	new_sk = new_sock->sk;
2744	new_tsock = tipc_sk(new_sk);
2745	msg = buf_msg(buf);
2746
2747	/* we lock on new_sk; but lockdep sees the lock on sk */
2748	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2749
2750	/*
2751	 * Reject any stray messages received by new socket
2752	 * before the socket lock was taken (very, very unlikely)
2753	 */
2754	tsk_rej_rx_queue(new_sk, TIPC_ERR_NO_PORT);
2755
2756	/* Connect new socket to it's peer */
2757	tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2758
2759	tsk_set_importance(new_sk, msg_importance(msg));
2760	if (msg_named(msg)) {
2761		new_tsock->conn_addrtype = TIPC_SERVICE_ADDR;
2762		msg_set_nametype(&new_tsock->phdr, msg_nametype(msg));
2763		msg_set_nameinst(&new_tsock->phdr, msg_nameinst(msg));
2764	}
2765
2766	/*
2767	 * Respond to 'SYN-' by discarding it & returning 'ACK'.
2768	 * Respond to 'SYN+' by queuing it on new socket & returning 'ACK'.
2769	 */
2770	if (!msg_data_sz(msg)) {
 
 
2771		tsk_advance_rx_queue(sk);
 
2772	} else {
2773		__skb_dequeue(&sk->sk_receive_queue);
2774		__skb_queue_head(&new_sk->sk_receive_queue, buf);
2775		skb_set_owner_r(buf, new_sk);
2776	}
2777	iov_iter_kvec(&m.msg_iter, ITER_SOURCE, NULL, 0, 0);
2778	__tipc_sendstream(new_sock, &m, 0);
2779	release_sock(new_sk);
2780exit:
2781	release_sock(sk);
2782	return res;
2783}
2784
2785/**
2786 * tipc_shutdown - shutdown socket connection
2787 * @sock: socket structure
2788 * @how: direction to close (must be SHUT_RDWR)
2789 *
2790 * Terminates connection (if necessary), then purges socket's receive queue.
2791 *
2792 * Return: 0 on success, errno otherwise
2793 */
2794static int tipc_shutdown(struct socket *sock, int how)
2795{
2796	struct sock *sk = sock->sk;
2797	int res;
2798
2799	if (how != SHUT_RDWR)
2800		return -EINVAL;
2801
2802	lock_sock(sk);
2803
2804	trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " ");
2805	__tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2806	sk->sk_shutdown = SHUTDOWN_MASK;
2807
2808	if (sk->sk_state == TIPC_DISCONNECTING) {
2809		/* Discard any unreceived messages */
2810		__skb_queue_purge(&sk->sk_receive_queue);
2811
 
 
2812		res = 0;
2813	} else {
2814		res = -ENOTCONN;
2815	}
2816	/* Wake up anyone sleeping in poll. */
2817	sk->sk_state_change(sk);
2818
2819	release_sock(sk);
2820	return res;
2821}
2822
2823static void tipc_sk_check_probing_state(struct sock *sk,
2824					struct sk_buff_head *list)
2825{
2826	struct tipc_sock *tsk = tipc_sk(sk);
2827	u32 pnode = tsk_peer_node(tsk);
2828	u32 pport = tsk_peer_port(tsk);
2829	u32 self = tsk_own_node(tsk);
2830	u32 oport = tsk->portid;
2831	struct sk_buff *skb;
2832
2833	if (tsk->probe_unacked) {
2834		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2835		sk->sk_err = ECONNABORTED;
2836		tipc_node_remove_conn(sock_net(sk), pnode, pport);
2837		sk->sk_state_change(sk);
2838		return;
2839	}
2840	/* Prepare new probe */
2841	skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2842			      pnode, self, pport, oport, TIPC_OK);
2843	if (skb)
2844		__skb_queue_tail(list, skb);
2845	tsk->probe_unacked = true;
2846	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2847}
2848
2849static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2850{
2851	struct tipc_sock *tsk = tipc_sk(sk);
2852
2853	/* Try again later if dest link is congested */
2854	if (tsk->cong_link_cnt) {
2855		sk_reset_timer(sk, &sk->sk_timer,
2856			       jiffies + msecs_to_jiffies(100));
2857		return;
2858	}
2859	/* Prepare SYN for retransmit */
2860	tipc_msg_skb_clone(&sk->sk_write_queue, list);
2861}
2862
2863static void tipc_sk_timeout(struct timer_list *t)
2864{
2865	struct sock *sk = from_timer(sk, t, sk_timer);
2866	struct tipc_sock *tsk = tipc_sk(sk);
2867	u32 pnode = tsk_peer_node(tsk);
2868	struct sk_buff_head list;
2869	int rc = 0;
 
 
 
2870
2871	__skb_queue_head_init(&list);
2872	bh_lock_sock(sk);
 
 
2873
2874	/* Try again later if socket is busy */
2875	if (sock_owned_by_user(sk)) {
2876		sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2877		bh_unlock_sock(sk);
2878		sock_put(sk);
2879		return;
2880	}
2881
2882	if (sk->sk_state == TIPC_ESTABLISHED)
2883		tipc_sk_check_probing_state(sk, &list);
2884	else if (sk->sk_state == TIPC_CONNECTING)
2885		tipc_sk_retry_connect(sk, &list);
2886
2887	bh_unlock_sock(sk);
2888
2889	if (!skb_queue_empty(&list))
2890		rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2891
2892	/* SYN messages may cause link congestion */
2893	if (rc == -ELINKCONG) {
2894		tipc_dest_push(&tsk->cong_links, pnode, 0);
2895		tsk->cong_link_cnt = 1;
2896	}
 
 
 
 
 
 
 
 
 
 
2897	sock_put(sk);
2898}
2899
2900static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua)
 
2901{
2902	struct sock *sk = &tsk->sk;
2903	struct net *net = sock_net(sk);
2904	struct tipc_socket_addr skaddr;
2905	struct publication *p;
2906	u32 key;
2907
 
 
 
2908	if (tipc_sk_connected(sk))
2909		return -EINVAL;
2910	key = tsk->portid + tsk->pub_count + 1;
2911	if (key == tsk->portid)
2912		return -EADDRINUSE;
2913	skaddr.ref = tsk->portid;
2914	skaddr.node = tipc_own_addr(net);
2915	p = tipc_nametbl_publish(net, ua, &skaddr, key);
2916	if (unlikely(!p))
2917		return -EINVAL;
2918
2919	list_add(&p->binding_sock, &tsk->publications);
2920	tsk->pub_count++;
2921	tsk->published = true;
2922	return 0;
2923}
2924
2925static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua)
 
2926{
2927	struct net *net = sock_net(&tsk->sk);
2928	struct publication *safe, *p;
2929	struct tipc_uaddr _ua;
2930	int rc = -EINVAL;
2931
2932	list_for_each_entry_safe(p, safe, &tsk->publications, binding_sock) {
2933		if (!ua) {
2934			tipc_uaddr(&_ua, TIPC_SERVICE_RANGE, p->scope,
2935				   p->sr.type, p->sr.lower, p->sr.upper);
2936			tipc_nametbl_withdraw(net, &_ua, &p->sk, p->key);
2937			continue;
2938		}
2939		/* Unbind specific publication */
2940		if (p->scope != ua->scope)
2941			continue;
2942		if (p->sr.type != ua->sr.type)
2943			continue;
2944		if (p->sr.lower != ua->sr.lower)
2945			continue;
2946		if (p->sr.upper != ua->sr.upper)
 
2947			break;
2948		tipc_nametbl_withdraw(net, ua, &p->sk, p->key);
 
 
2949		rc = 0;
2950		break;
2951	}
2952	if (list_empty(&tsk->publications)) {
2953		tsk->published = 0;
2954		rc = 0;
2955	}
2956	return rc;
2957}
2958
2959/* tipc_sk_reinit: set non-zero address in all existing sockets
2960 *                 when we go from standalone to network mode.
2961 */
2962void tipc_sk_reinit(struct net *net)
2963{
2964	struct tipc_net *tn = net_generic(net, tipc_net_id);
2965	struct rhashtable_iter iter;
2966	struct tipc_sock *tsk;
2967	struct tipc_msg *msg;
2968
2969	rhashtable_walk_enter(&tn->sk_rht, &iter);
2970
2971	do {
2972		rhashtable_walk_start(&iter);
2973
2974		while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2975			sock_hold(&tsk->sk);
2976			rhashtable_walk_stop(&iter);
2977			lock_sock(&tsk->sk);
2978			msg = &tsk->phdr;
2979			msg_set_prevnode(msg, tipc_own_addr(net));
2980			msg_set_orignode(msg, tipc_own_addr(net));
2981			release_sock(&tsk->sk);
2982			rhashtable_walk_start(&iter);
2983			sock_put(&tsk->sk);
2984		}
2985
2986		rhashtable_walk_stop(&iter);
2987	} while (tsk == ERR_PTR(-EAGAIN));
2988
2989	rhashtable_walk_exit(&iter);
2990}
2991
2992static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2993{
2994	struct tipc_net *tn = net_generic(net, tipc_net_id);
2995	struct tipc_sock *tsk;
2996
2997	rcu_read_lock();
2998	tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params);
2999	if (tsk)
3000		sock_hold(&tsk->sk);
3001	rcu_read_unlock();
3002
3003	return tsk;
3004}
3005
3006static int tipc_sk_insert(struct tipc_sock *tsk)
3007{
3008	struct sock *sk = &tsk->sk;
3009	struct net *net = sock_net(sk);
3010	struct tipc_net *tn = net_generic(net, tipc_net_id);
3011	u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
3012	u32 portid = get_random_u32_below(remaining) + TIPC_MIN_PORT;
3013
3014	while (remaining--) {
3015		portid++;
3016		if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
3017			portid = TIPC_MIN_PORT;
3018		tsk->portid = portid;
3019		sock_hold(&tsk->sk);
3020		if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
3021						   tsk_rht_params))
3022			return 0;
3023		sock_put(&tsk->sk);
3024	}
3025
3026	return -1;
3027}
3028
3029static void tipc_sk_remove(struct tipc_sock *tsk)
3030{
3031	struct sock *sk = &tsk->sk;
3032	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
3033
3034	if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
3035		WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
3036		__sock_put(sk);
3037	}
3038}
3039
3040static const struct rhashtable_params tsk_rht_params = {
3041	.nelem_hint = 192,
3042	.head_offset = offsetof(struct tipc_sock, node),
3043	.key_offset = offsetof(struct tipc_sock, portid),
3044	.key_len = sizeof(u32), /* portid */
3045	.max_size = 1048576,
3046	.min_size = 256,
3047	.automatic_shrinking = true,
3048};
3049
3050int tipc_sk_rht_init(struct net *net)
3051{
3052	struct tipc_net *tn = net_generic(net, tipc_net_id);
3053
3054	return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
3055}
3056
3057void tipc_sk_rht_destroy(struct net *net)
3058{
3059	struct tipc_net *tn = net_generic(net, tipc_net_id);
3060
3061	/* Wait for socket readers to complete */
3062	synchronize_net();
3063
3064	rhashtable_destroy(&tn->sk_rht);
3065}
3066
3067static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
3068{
3069	struct net *net = sock_net(&tsk->sk);
3070	struct tipc_group *grp = tsk->group;
3071	struct tipc_msg *hdr = &tsk->phdr;
3072	struct tipc_uaddr ua;
3073	int rc;
3074
3075	if (mreq->type < TIPC_RESERVED_TYPES)
3076		return -EACCES;
3077	if (mreq->scope > TIPC_NODE_SCOPE)
3078		return -EINVAL;
3079	if (mreq->scope != TIPC_NODE_SCOPE)
3080		mreq->scope = TIPC_CLUSTER_SCOPE;
3081	if (grp)
3082		return -EACCES;
3083	grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
3084	if (!grp)
3085		return -ENOMEM;
3086	tsk->group = grp;
3087	msg_set_lookup_scope(hdr, mreq->scope);
3088	msg_set_nametype(hdr, mreq->type);
3089	msg_set_dest_droppable(hdr, true);
3090	tipc_uaddr(&ua, TIPC_SERVICE_RANGE, mreq->scope,
3091		   mreq->type, mreq->instance, mreq->instance);
3092	tipc_nametbl_build_group(net, grp, &ua);
3093	rc = tipc_sk_publish(tsk, &ua);
 
3094	if (rc) {
3095		tipc_group_delete(net, grp);
3096		tsk->group = NULL;
3097		return rc;
3098	}
3099	/* Eliminate any risk that a broadcast overtakes sent JOINs */
3100	tsk->mc_method.rcast = true;
3101	tsk->mc_method.mandatory = true;
3102	tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
3103	return rc;
3104}
3105
3106static int tipc_sk_leave(struct tipc_sock *tsk)
3107{
3108	struct net *net = sock_net(&tsk->sk);
3109	struct tipc_group *grp = tsk->group;
3110	struct tipc_uaddr ua;
3111	int scope;
3112
3113	if (!grp)
3114		return -EINVAL;
3115	ua.addrtype = TIPC_SERVICE_RANGE;
3116	tipc_group_self(grp, &ua.sr, &scope);
3117	ua.scope = scope;
3118	tipc_group_delete(net, grp);
3119	tsk->group = NULL;
3120	tipc_sk_withdraw(tsk, &ua);
3121	return 0;
3122}
3123
3124/**
3125 * tipc_setsockopt - set socket option
3126 * @sock: socket structure
3127 * @lvl: option level
3128 * @opt: option identifier
3129 * @ov: pointer to new option value
3130 * @ol: length of option value
3131 *
3132 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
3133 * (to ease compatibility).
3134 *
3135 * Return: 0 on success, errno otherwise
3136 */
3137static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
3138			   sockptr_t ov, unsigned int ol)
3139{
3140	struct sock *sk = sock->sk;
3141	struct tipc_sock *tsk = tipc_sk(sk);
3142	struct tipc_group_req mreq;
3143	u32 value = 0;
3144	int res = 0;
3145
3146	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3147		return 0;
3148	if (lvl != SOL_TIPC)
3149		return -ENOPROTOOPT;
3150
3151	switch (opt) {
3152	case TIPC_IMPORTANCE:
3153	case TIPC_SRC_DROPPABLE:
3154	case TIPC_DEST_DROPPABLE:
3155	case TIPC_CONN_TIMEOUT:
3156	case TIPC_NODELAY:
3157		if (ol < sizeof(value))
3158			return -EINVAL;
3159		if (copy_from_sockptr(&value, ov, sizeof(u32)))
3160			return -EFAULT;
3161		break;
3162	case TIPC_GROUP_JOIN:
3163		if (ol < sizeof(mreq))
3164			return -EINVAL;
3165		if (copy_from_sockptr(&mreq, ov, sizeof(mreq)))
3166			return -EFAULT;
3167		break;
3168	default:
3169		if (!sockptr_is_null(ov) || ol)
3170			return -EINVAL;
3171	}
3172
3173	lock_sock(sk);
3174
3175	switch (opt) {
3176	case TIPC_IMPORTANCE:
3177		res = tsk_set_importance(sk, value);
3178		break;
3179	case TIPC_SRC_DROPPABLE:
3180		if (sock->type != SOCK_STREAM)
3181			tsk_set_unreliable(tsk, value);
3182		else
3183			res = -ENOPROTOOPT;
3184		break;
3185	case TIPC_DEST_DROPPABLE:
3186		tsk_set_unreturnable(tsk, value);
3187		break;
3188	case TIPC_CONN_TIMEOUT:
3189		tipc_sk(sk)->conn_timeout = value;
3190		break;
3191	case TIPC_MCAST_BROADCAST:
3192		tsk->mc_method.rcast = false;
3193		tsk->mc_method.mandatory = true;
3194		break;
3195	case TIPC_MCAST_REPLICAST:
3196		tsk->mc_method.rcast = true;
3197		tsk->mc_method.mandatory = true;
3198		break;
3199	case TIPC_GROUP_JOIN:
3200		res = tipc_sk_join(tsk, &mreq);
3201		break;
3202	case TIPC_GROUP_LEAVE:
3203		res = tipc_sk_leave(tsk);
3204		break;
3205	case TIPC_NODELAY:
3206		tsk->nodelay = !!value;
3207		tsk_set_nagle(tsk);
3208		break;
3209	default:
3210		res = -EINVAL;
3211	}
3212
3213	release_sock(sk);
3214
3215	return res;
3216}
3217
3218/**
3219 * tipc_getsockopt - get socket option
3220 * @sock: socket structure
3221 * @lvl: option level
3222 * @opt: option identifier
3223 * @ov: receptacle for option value
3224 * @ol: receptacle for length of option value
3225 *
3226 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
3227 * (to ease compatibility).
3228 *
3229 * Return: 0 on success, errno otherwise
3230 */
3231static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
3232			   char __user *ov, int __user *ol)
3233{
3234	struct sock *sk = sock->sk;
3235	struct tipc_sock *tsk = tipc_sk(sk);
3236	struct tipc_service_range seq;
3237	int len, scope;
3238	u32 value;
3239	int res;
3240
3241	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3242		return put_user(0, ol);
3243	if (lvl != SOL_TIPC)
3244		return -ENOPROTOOPT;
3245	res = get_user(len, ol);
3246	if (res)
3247		return res;
3248
3249	lock_sock(sk);
3250
3251	switch (opt) {
3252	case TIPC_IMPORTANCE:
3253		value = tsk_importance(tsk);
3254		break;
3255	case TIPC_SRC_DROPPABLE:
3256		value = tsk_unreliable(tsk);
3257		break;
3258	case TIPC_DEST_DROPPABLE:
3259		value = tsk_unreturnable(tsk);
3260		break;
3261	case TIPC_CONN_TIMEOUT:
3262		value = tsk->conn_timeout;
3263		/* no need to set "res", since already 0 at this point */
3264		break;
3265	case TIPC_NODE_RECVQ_DEPTH:
3266		value = 0; /* was tipc_queue_size, now obsolete */
3267		break;
3268	case TIPC_SOCK_RECVQ_DEPTH:
3269		value = skb_queue_len(&sk->sk_receive_queue);
3270		break;
3271	case TIPC_SOCK_RECVQ_USED:
3272		value = sk_rmem_alloc_get(sk);
3273		break;
3274	case TIPC_GROUP_JOIN:
3275		seq.type = 0;
3276		if (tsk->group)
3277			tipc_group_self(tsk->group, &seq, &scope);
3278		value = seq.type;
3279		break;
3280	default:
3281		res = -EINVAL;
3282	}
3283
3284	release_sock(sk);
3285
3286	if (res)
3287		return res;	/* "get" failed */
3288
3289	if (len < sizeof(value))
3290		return -EINVAL;
3291
3292	if (copy_to_user(ov, &value, sizeof(value)))
3293		return -EFAULT;
3294
3295	return put_user(sizeof(value), ol);
3296}
3297
3298static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3299{
3300	struct net *net = sock_net(sock->sk);
3301	struct tipc_sioc_nodeid_req nr = {0};
3302	struct tipc_sioc_ln_req lnr;
3303	void __user *argp = (void __user *)arg;
3304
3305	switch (cmd) {
3306	case SIOCGETLINKNAME:
3307		if (copy_from_user(&lnr, argp, sizeof(lnr)))
3308			return -EFAULT;
3309		if (!tipc_node_get_linkname(net,
3310					    lnr.bearer_id & 0xffff, lnr.peer,
3311					    lnr.linkname, TIPC_MAX_LINK_NAME)) {
3312			if (copy_to_user(argp, &lnr, sizeof(lnr)))
3313				return -EFAULT;
3314			return 0;
3315		}
3316		return -EADDRNOTAVAIL;
3317	case SIOCGETNODEID:
3318		if (copy_from_user(&nr, argp, sizeof(nr)))
3319			return -EFAULT;
3320		if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3321			return -EADDRNOTAVAIL;
3322		if (copy_to_user(argp, &nr, sizeof(nr)))
3323			return -EFAULT;
3324		return 0;
3325	default:
3326		return -ENOIOCTLCMD;
3327	}
3328}
3329
3330static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3331{
3332	struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3333	struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3334	u32 onode = tipc_own_addr(sock_net(sock1->sk));
3335
3336	tsk1->peer.family = AF_TIPC;
3337	tsk1->peer.addrtype = TIPC_SOCKET_ADDR;
3338	tsk1->peer.scope = TIPC_NODE_SCOPE;
3339	tsk1->peer.addr.id.ref = tsk2->portid;
3340	tsk1->peer.addr.id.node = onode;
3341	tsk2->peer.family = AF_TIPC;
3342	tsk2->peer.addrtype = TIPC_SOCKET_ADDR;
3343	tsk2->peer.scope = TIPC_NODE_SCOPE;
3344	tsk2->peer.addr.id.ref = tsk1->portid;
3345	tsk2->peer.addr.id.node = onode;
3346
3347	tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3348	tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3349	return 0;
3350}
3351
3352/* Protocol switches for the various types of TIPC sockets */
3353
3354static const struct proto_ops msg_ops = {
3355	.owner		= THIS_MODULE,
3356	.family		= AF_TIPC,
3357	.release	= tipc_release,
3358	.bind		= tipc_bind,
3359	.connect	= tipc_connect,
3360	.socketpair	= tipc_socketpair,
3361	.accept		= sock_no_accept,
3362	.getname	= tipc_getname,
3363	.poll		= tipc_poll,
3364	.ioctl		= tipc_ioctl,
3365	.listen		= sock_no_listen,
3366	.shutdown	= tipc_shutdown,
3367	.setsockopt	= tipc_setsockopt,
3368	.getsockopt	= tipc_getsockopt,
3369	.sendmsg	= tipc_sendmsg,
3370	.recvmsg	= tipc_recvmsg,
3371	.mmap		= sock_no_mmap,
 
3372};
3373
3374static const struct proto_ops packet_ops = {
3375	.owner		= THIS_MODULE,
3376	.family		= AF_TIPC,
3377	.release	= tipc_release,
3378	.bind		= tipc_bind,
3379	.connect	= tipc_connect,
3380	.socketpair	= tipc_socketpair,
3381	.accept		= tipc_accept,
3382	.getname	= tipc_getname,
3383	.poll		= tipc_poll,
3384	.ioctl		= tipc_ioctl,
3385	.listen		= tipc_listen,
3386	.shutdown	= tipc_shutdown,
3387	.setsockopt	= tipc_setsockopt,
3388	.getsockopt	= tipc_getsockopt,
3389	.sendmsg	= tipc_send_packet,
3390	.recvmsg	= tipc_recvmsg,
3391	.mmap		= sock_no_mmap,
 
3392};
3393
3394static const struct proto_ops stream_ops = {
3395	.owner		= THIS_MODULE,
3396	.family		= AF_TIPC,
3397	.release	= tipc_release,
3398	.bind		= tipc_bind,
3399	.connect	= tipc_connect,
3400	.socketpair	= tipc_socketpair,
3401	.accept		= tipc_accept,
3402	.getname	= tipc_getname,
3403	.poll		= tipc_poll,
3404	.ioctl		= tipc_ioctl,
3405	.listen		= tipc_listen,
3406	.shutdown	= tipc_shutdown,
3407	.setsockopt	= tipc_setsockopt,
3408	.getsockopt	= tipc_getsockopt,
3409	.sendmsg	= tipc_sendstream,
3410	.recvmsg	= tipc_recvstream,
3411	.mmap		= sock_no_mmap,
 
3412};
3413
3414static const struct net_proto_family tipc_family_ops = {
3415	.owner		= THIS_MODULE,
3416	.family		= AF_TIPC,
3417	.create		= tipc_sk_create
3418};
3419
3420static struct proto tipc_proto = {
3421	.name		= "TIPC",
3422	.owner		= THIS_MODULE,
3423	.obj_size	= sizeof(struct tipc_sock),
3424	.sysctl_rmem	= sysctl_tipc_rmem
3425};
3426
3427/**
3428 * tipc_socket_init - initialize TIPC socket interface
3429 *
3430 * Return: 0 on success, errno otherwise
3431 */
3432int tipc_socket_init(void)
3433{
3434	int res;
3435
3436	res = proto_register(&tipc_proto, 1);
3437	if (res) {
3438		pr_err("Failed to register TIPC protocol type\n");
3439		goto out;
3440	}
3441
3442	res = sock_register(&tipc_family_ops);
3443	if (res) {
3444		pr_err("Failed to register TIPC socket type\n");
3445		proto_unregister(&tipc_proto);
3446		goto out;
3447	}
3448 out:
3449	return res;
3450}
3451
3452/**
3453 * tipc_socket_stop - stop TIPC socket interface
3454 */
3455void tipc_socket_stop(void)
3456{
3457	sock_unregister(tipc_family_ops.family);
3458	proto_unregister(&tipc_proto);
3459}
3460
3461/* Caller should hold socket lock for the passed tipc socket. */
3462static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3463{
3464	u32 peer_node, peer_port;
3465	u32 conn_type, conn_instance;
3466	struct nlattr *nest;
3467
3468	peer_node = tsk_peer_node(tsk);
3469	peer_port = tsk_peer_port(tsk);
3470	conn_type = msg_nametype(&tsk->phdr);
3471	conn_instance = msg_nameinst(&tsk->phdr);
3472	nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON);
3473	if (!nest)
3474		return -EMSGSIZE;
3475
3476	if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3477		goto msg_full;
3478	if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3479		goto msg_full;
3480
3481	if (tsk->conn_addrtype != 0) {
3482		if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3483			goto msg_full;
3484		if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, conn_type))
3485			goto msg_full;
3486		if (nla_put_u32(skb, TIPC_NLA_CON_INST, conn_instance))
3487			goto msg_full;
3488	}
3489	nla_nest_end(skb, nest);
3490
3491	return 0;
3492
3493msg_full:
3494	nla_nest_cancel(skb, nest);
3495
3496	return -EMSGSIZE;
3497}
3498
3499static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3500			  *tsk)
3501{
3502	struct net *net = sock_net(skb->sk);
3503	struct sock *sk = &tsk->sk;
3504
3505	if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3506	    nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3507		return -EMSGSIZE;
3508
3509	if (tipc_sk_connected(sk)) {
3510		if (__tipc_nl_add_sk_con(skb, tsk))
3511			return -EMSGSIZE;
3512	} else if (!list_empty(&tsk->publications)) {
3513		if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3514			return -EMSGSIZE;
3515	}
3516	return 0;
3517}
3518
3519/* Caller should hold socket lock for the passed tipc socket. */
3520static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3521			    struct tipc_sock *tsk)
3522{
3523	struct nlattr *attrs;
3524	void *hdr;
3525
3526	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3527			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3528	if (!hdr)
3529		goto msg_cancel;
3530
3531	attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3532	if (!attrs)
3533		goto genlmsg_cancel;
3534
3535	if (__tipc_nl_add_sk_info(skb, tsk))
3536		goto attr_msg_cancel;
3537
3538	nla_nest_end(skb, attrs);
3539	genlmsg_end(skb, hdr);
3540
3541	return 0;
3542
3543attr_msg_cancel:
3544	nla_nest_cancel(skb, attrs);
3545genlmsg_cancel:
3546	genlmsg_cancel(skb, hdr);
3547msg_cancel:
3548	return -EMSGSIZE;
3549}
3550
3551int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3552		    int (*skb_handler)(struct sk_buff *skb,
3553				       struct netlink_callback *cb,
3554				       struct tipc_sock *tsk))
3555{
3556	struct rhashtable_iter *iter = (void *)cb->args[4];
 
 
 
 
 
3557	struct tipc_sock *tsk;
3558	int err;
3559
3560	rhashtable_walk_start(iter);
3561	while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3562		if (IS_ERR(tsk)) {
3563			if (PTR_ERR(tsk) == -EAGAIN)
 
 
 
3564				continue;
3565			break;
3566		}
3567
3568		sock_hold(&tsk->sk);
3569		rhashtable_walk_stop(iter);
3570		lock_sock(&tsk->sk);
3571		err = skb_handler(skb, cb, tsk);
3572		if (err) {
3573			release_sock(&tsk->sk);
3574			sock_put(&tsk->sk);
3575			goto out;
 
3576		}
3577		release_sock(&tsk->sk);
3578		rhashtable_walk_start(iter);
3579		sock_put(&tsk->sk);
3580	}
3581	rhashtable_walk_stop(iter);
3582out:
 
 
 
 
3583	return skb->len;
3584}
3585EXPORT_SYMBOL(tipc_nl_sk_walk);
3586
3587int tipc_dump_start(struct netlink_callback *cb)
3588{
3589	return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3590}
3591EXPORT_SYMBOL(tipc_dump_start);
3592
3593int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3594{
3595	/* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3596	struct rhashtable_iter *iter = (void *)cb->args[4];
3597	struct tipc_net *tn = tipc_net(net);
3598
3599	if (!iter) {
3600		iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3601		if (!iter)
3602			return -ENOMEM;
3603
3604		cb->args[4] = (long)iter;
3605	}
3606
3607	rhashtable_walk_enter(&tn->sk_rht, iter);
3608	return 0;
3609}
3610
3611int tipc_dump_done(struct netlink_callback *cb)
3612{
3613	struct rhashtable_iter *hti = (void *)cb->args[4];
3614
3615	rhashtable_walk_exit(hti);
3616	kfree(hti);
3617	return 0;
3618}
3619EXPORT_SYMBOL(tipc_dump_done);
3620
3621int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3622			   struct tipc_sock *tsk, u32 sk_filter_state,
3623			   u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3624{
3625	struct sock *sk = &tsk->sk;
3626	struct nlattr *attrs;
3627	struct nlattr *stat;
3628
3629	/*filter response w.r.t sk_state*/
3630	if (!(sk_filter_state & (1 << sk->sk_state)))
3631		return 0;
3632
3633	attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3634	if (!attrs)
3635		goto msg_cancel;
3636
3637	if (__tipc_nl_add_sk_info(skb, tsk))
3638		goto attr_msg_cancel;
3639
3640	if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3641	    nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3642	    nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3643	    nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3644			from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3645					 sock_i_uid(sk))) ||
3646	    nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3647			      tipc_diag_gen_cookie(sk),
3648			      TIPC_NLA_SOCK_PAD))
3649		goto attr_msg_cancel;
3650
3651	stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT);
3652	if (!stat)
3653		goto attr_msg_cancel;
3654
3655	if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3656			skb_queue_len(&sk->sk_receive_queue)) ||
3657	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3658			skb_queue_len(&sk->sk_write_queue)) ||
3659	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3660			atomic_read(&sk->sk_drops)))
3661		goto stat_msg_cancel;
3662
3663	if (tsk->cong_link_cnt &&
3664	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3665		goto stat_msg_cancel;
3666
3667	if (tsk_conn_cong(tsk) &&
3668	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3669		goto stat_msg_cancel;
3670
3671	nla_nest_end(skb, stat);
3672
3673	if (tsk->group)
3674		if (tipc_group_fill_sock_diag(tsk->group, skb))
3675			goto stat_msg_cancel;
3676
3677	nla_nest_end(skb, attrs);
3678
3679	return 0;
3680
3681stat_msg_cancel:
3682	nla_nest_cancel(skb, stat);
3683attr_msg_cancel:
3684	nla_nest_cancel(skb, attrs);
3685msg_cancel:
3686	return -EMSGSIZE;
3687}
3688EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3689
3690int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3691{
3692	return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3693}
3694
3695/* Caller should hold socket lock for the passed tipc socket. */
3696static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3697				 struct netlink_callback *cb,
3698				 struct publication *publ)
3699{
3700	void *hdr;
3701	struct nlattr *attrs;
3702
3703	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3704			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3705	if (!hdr)
3706		goto msg_cancel;
3707
3708	attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL);
3709	if (!attrs)
3710		goto genlmsg_cancel;
3711
3712	if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3713		goto attr_msg_cancel;
3714	if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->sr.type))
3715		goto attr_msg_cancel;
3716	if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->sr.lower))
3717		goto attr_msg_cancel;
3718	if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->sr.upper))
3719		goto attr_msg_cancel;
3720
3721	nla_nest_end(skb, attrs);
3722	genlmsg_end(skb, hdr);
3723
3724	return 0;
3725
3726attr_msg_cancel:
3727	nla_nest_cancel(skb, attrs);
3728genlmsg_cancel:
3729	genlmsg_cancel(skb, hdr);
3730msg_cancel:
3731	return -EMSGSIZE;
3732}
3733
3734/* Caller should hold socket lock for the passed tipc socket. */
3735static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3736				  struct netlink_callback *cb,
3737				  struct tipc_sock *tsk, u32 *last_publ)
3738{
3739	int err;
3740	struct publication *p;
3741
3742	if (*last_publ) {
3743		list_for_each_entry(p, &tsk->publications, binding_sock) {
3744			if (p->key == *last_publ)
3745				break;
3746		}
3747		if (list_entry_is_head(p, &tsk->publications, binding_sock)) {
3748			/* We never set seq or call nl_dump_check_consistent()
3749			 * this means that setting prev_seq here will cause the
3750			 * consistence check to fail in the netlink callback
3751			 * handler. Resulting in the last NLMSG_DONE message
3752			 * having the NLM_F_DUMP_INTR flag set.
3753			 */
3754			cb->prev_seq = 1;
3755			*last_publ = 0;
3756			return -EPIPE;
3757		}
3758	} else {
3759		p = list_first_entry(&tsk->publications, struct publication,
3760				     binding_sock);
3761	}
3762
3763	list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3764		err = __tipc_nl_add_sk_publ(skb, cb, p);
3765		if (err) {
3766			*last_publ = p->key;
3767			return err;
3768		}
3769	}
3770	*last_publ = 0;
3771
3772	return 0;
3773}
3774
3775int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3776{
3777	int err;
3778	u32 tsk_portid = cb->args[0];
3779	u32 last_publ = cb->args[1];
3780	u32 done = cb->args[2];
3781	struct net *net = sock_net(skb->sk);
3782	struct tipc_sock *tsk;
3783
3784	if (!tsk_portid) {
3785		struct nlattr **attrs = genl_dumpit_info(cb)->info.attrs;
3786		struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3787
 
 
 
 
3788		if (!attrs[TIPC_NLA_SOCK])
3789			return -EINVAL;
3790
3791		err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX,
3792						  attrs[TIPC_NLA_SOCK],
3793						  tipc_nl_sock_policy, NULL);
3794		if (err)
3795			return err;
3796
3797		if (!sock[TIPC_NLA_SOCK_REF])
3798			return -EINVAL;
3799
3800		tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3801	}
3802
3803	if (done)
3804		return 0;
3805
3806	tsk = tipc_sk_lookup(net, tsk_portid);
3807	if (!tsk)
3808		return -EINVAL;
3809
3810	lock_sock(&tsk->sk);
3811	err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3812	if (!err)
3813		done = 1;
3814	release_sock(&tsk->sk);
3815	sock_put(&tsk->sk);
3816
3817	cb->args[0] = tsk_portid;
3818	cb->args[1] = last_publ;
3819	cb->args[2] = done;
3820
3821	return skb->len;
3822}
3823
3824/**
3825 * tipc_sk_filtering - check if a socket should be traced
3826 * @sk: the socket to be examined
3827 *
3828 * @sysctl_tipc_sk_filter is used as the socket tuple for filtering:
3829 * (portid, sock type, name type, name lower, name upper)
3830 *
3831 * Return: true if the socket meets the socket tuple data
3832 * (value 0 = 'any') or when there is no tuple set (all = 0),
3833 * otherwise false
3834 */
3835bool tipc_sk_filtering(struct sock *sk)
3836{
3837	struct tipc_sock *tsk;
3838	struct publication *p;
3839	u32 _port, _sktype, _type, _lower, _upper;
3840	u32 type = 0, lower = 0, upper = 0;
3841
3842	if (!sk)
3843		return true;
3844
3845	tsk = tipc_sk(sk);
3846
3847	_port = sysctl_tipc_sk_filter[0];
3848	_sktype = sysctl_tipc_sk_filter[1];
3849	_type = sysctl_tipc_sk_filter[2];
3850	_lower = sysctl_tipc_sk_filter[3];
3851	_upper = sysctl_tipc_sk_filter[4];
3852
3853	if (!_port && !_sktype && !_type && !_lower && !_upper)
3854		return true;
3855
3856	if (_port)
3857		return (_port == tsk->portid);
3858
3859	if (_sktype && _sktype != sk->sk_type)
3860		return false;
3861
3862	if (tsk->published) {
3863		p = list_first_entry_or_null(&tsk->publications,
3864					     struct publication, binding_sock);
3865		if (p) {
3866			type = p->sr.type;
3867			lower = p->sr.lower;
3868			upper = p->sr.upper;
3869		}
3870	}
3871
3872	if (!tipc_sk_type_connectionless(sk)) {
3873		type = msg_nametype(&tsk->phdr);
3874		lower = msg_nameinst(&tsk->phdr);
3875		upper = lower;
3876	}
3877
3878	if ((_type && _type != type) || (_lower && _lower != lower) ||
3879	    (_upper && _upper != upper))
3880		return false;
3881
3882	return true;
3883}
3884
3885u32 tipc_sock_get_portid(struct sock *sk)
3886{
3887	return (sk) ? (tipc_sk(sk))->portid : 0;
3888}
3889
3890/**
3891 * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded,
3892 *			both the rcv and backlog queues are considered
3893 * @sk: tipc sk to be checked
3894 * @skb: tipc msg to be checked
3895 *
3896 * Return: true if the socket rx queue allocation is > 90%, otherwise false
3897 */
3898
3899bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb)
3900{
3901	atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt;
3902	unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
3903	unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk);
3904
3905	return (qsize > lim * 90 / 100);
3906}
3907
3908/**
3909 * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded,
3910 *			only the rcv queue is considered
3911 * @sk: tipc sk to be checked
3912 * @skb: tipc msg to be checked
3913 *
3914 * Return: true if the socket rx queue allocation is > 90%, otherwise false
3915 */
3916
3917bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb)
3918{
3919	unsigned int lim = rcvbuf_limit(sk, skb);
3920	unsigned int qsize = sk_rmem_alloc_get(sk);
3921
3922	return (qsize > lim * 90 / 100);
3923}
3924
3925/**
3926 * tipc_sk_dump - dump TIPC socket
3927 * @sk: tipc sk to be dumped
3928 * @dqueues: bitmask to decide if any socket queue to be dumped?
3929 *           - TIPC_DUMP_NONE: don't dump socket queues
3930 *           - TIPC_DUMP_SK_SNDQ: dump socket send queue
3931 *           - TIPC_DUMP_SK_RCVQ: dump socket rcv queue
3932 *           - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue
3933 *           - TIPC_DUMP_ALL: dump all the socket queues above
3934 * @buf: returned buffer of dump data in format
3935 */
3936int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf)
3937{
3938	int i = 0;
3939	size_t sz = (dqueues) ? SK_LMAX : SK_LMIN;
3940	u32 conn_type, conn_instance;
3941	struct tipc_sock *tsk;
3942	struct publication *p;
3943	bool tsk_connected;
3944
3945	if (!sk) {
3946		i += scnprintf(buf, sz, "sk data: (null)\n");
3947		return i;
3948	}
3949
3950	tsk = tipc_sk(sk);
3951	tsk_connected = !tipc_sk_type_connectionless(sk);
3952
3953	i += scnprintf(buf, sz, "sk data: %u", sk->sk_type);
3954	i += scnprintf(buf + i, sz - i, " %d", sk->sk_state);
3955	i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk));
3956	i += scnprintf(buf + i, sz - i, " %u", tsk->portid);
3957	i += scnprintf(buf + i, sz - i, " | %u", tsk_connected);
3958	if (tsk_connected) {
3959		i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk));
3960		i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk));
3961		conn_type = msg_nametype(&tsk->phdr);
3962		conn_instance = msg_nameinst(&tsk->phdr);
3963		i += scnprintf(buf + i, sz - i, " %u", conn_type);
3964		i += scnprintf(buf + i, sz - i, " %u", conn_instance);
3965	}
3966	i += scnprintf(buf + i, sz - i, " | %u", tsk->published);
3967	if (tsk->published) {
3968		p = list_first_entry_or_null(&tsk->publications,
3969					     struct publication, binding_sock);
3970		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.type : 0);
3971		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.lower : 0);
3972		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.upper : 0);
3973	}
3974	i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win);
3975	i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win);
3976	i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt);
3977	i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps);
3978	i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt);
3979	i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked);
3980	i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked);
3981	i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt));
3982	i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown);
3983	i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk));
3984	i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf);
3985	i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk));
3986	i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf);
3987	i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len));
3988
3989	if (dqueues & TIPC_DUMP_SK_SNDQ) {
3990		i += scnprintf(buf + i, sz - i, "sk_write_queue: ");
3991		i += tipc_list_dump(&sk->sk_write_queue, false, buf + i);
3992	}
3993
3994	if (dqueues & TIPC_DUMP_SK_RCVQ) {
3995		i += scnprintf(buf + i, sz - i, "sk_receive_queue: ");
3996		i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i);
3997	}
3998
3999	if (dqueues & TIPC_DUMP_SK_BKLGQ) {
4000		i += scnprintf(buf + i, sz - i, "sk_backlog:\n  head ");
4001		i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i);
4002		if (sk->sk_backlog.tail != sk->sk_backlog.head) {
4003			i += scnprintf(buf + i, sz - i, "  tail ");
4004			i += tipc_skb_dump(sk->sk_backlog.tail, false,
4005					   buf + i);
4006		}
4007	}
4008
4009	return i;
4010}
v4.17
   1/*
   2 * net/tipc/socket.c: TIPC socket API
   3 *
   4 * Copyright (c) 2001-2007, 2012-2017, Ericsson AB
   5 * Copyright (c) 2004-2008, 2010-2013, Wind River Systems
 
   6 * All rights reserved.
   7 *
   8 * Redistribution and use in source and binary forms, with or without
   9 * modification, are permitted provided that the following conditions are met:
  10 *
  11 * 1. Redistributions of source code must retain the above copyright
  12 *    notice, this list of conditions and the following disclaimer.
  13 * 2. Redistributions in binary form must reproduce the above copyright
  14 *    notice, this list of conditions and the following disclaimer in the
  15 *    documentation and/or other materials provided with the distribution.
  16 * 3. Neither the names of the copyright holders nor the names of its
  17 *    contributors may be used to endorse or promote products derived from
  18 *    this software without specific prior written permission.
  19 *
  20 * Alternatively, this software may be distributed under the terms of the
  21 * GNU General Public License ("GPL") version 2 as published by the Free
  22 * Software Foundation.
  23 *
  24 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
  25 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  27 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
  28 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  31 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  32 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  33 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  34 * POSSIBILITY OF SUCH DAMAGE.
  35 */
  36
  37#include <linux/rhashtable.h>
  38#include <linux/sched/signal.h>
 
  39
  40#include "core.h"
  41#include "name_table.h"
  42#include "node.h"
  43#include "link.h"
  44#include "name_distr.h"
  45#include "socket.h"
  46#include "bcast.h"
  47#include "netlink.h"
  48#include "group.h"
 
  49
  50#define CONN_TIMEOUT_DEFAULT	8000	/* default connect timeout = 8s */
 
 
  51#define CONN_PROBING_INTV	msecs_to_jiffies(3600000)  /* [ms] => 1 h */
  52#define TIPC_FWD_MSG		1
  53#define TIPC_MAX_PORT		0xffffffff
  54#define TIPC_MIN_PORT		1
  55#define TIPC_ACK_RATE		4       /* ACK at 1/4 of of rcv window size */
  56
  57enum {
  58	TIPC_LISTEN = TCP_LISTEN,
  59	TIPC_ESTABLISHED = TCP_ESTABLISHED,
  60	TIPC_OPEN = TCP_CLOSE,
  61	TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
  62	TIPC_CONNECTING = TCP_SYN_SENT,
  63};
  64
  65struct sockaddr_pair {
  66	struct sockaddr_tipc sock;
  67	struct sockaddr_tipc member;
  68};
  69
  70/**
  71 * struct tipc_sock - TIPC socket structure
  72 * @sk: socket - interacts with 'port' and with user via the socket API
  73 * @conn_type: TIPC type used when connection was established
  74 * @conn_instance: TIPC instance used when connection was established
  75 * @published: non-zero if port has one or more associated names
  76 * @max_pkt: maximum packet size "hint" used when building messages sent by port
 
  77 * @portid: unique port identity in TIPC socket hash table
  78 * @phdr: preformatted message header used when sending messages
  79 * #cong_links: list of congested links
  80 * @publications: list of publications for port
  81 * @blocking_link: address of the congested link we are currently sleeping on
  82 * @pub_count: total # of publications port has made during its lifetime
  83 * @probing_state:
  84 * @conn_timeout: the time we can wait for an unresponded setup request
 
  85 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
  86 * @cong_link_cnt: number of congested links
  87 * @snt_unacked: # messages sent by socket, and not yet acked by peer
 
 
  88 * @rcv_unacked: # messages read by user, but not yet acked back to peer
 
  89 * @peer: 'connected' peer for dgram/rdm
  90 * @node: hash table node
  91 * @mc_method: cookie for use between socket and broadcast layer
  92 * @rcu: rcu struct for tipc_sock
 
 
 
 
 
 
 
 
 
 
 
  93 */
  94struct tipc_sock {
  95	struct sock sk;
  96	u32 conn_type;
  97	u32 conn_instance;
  98	int published;
  99	u32 max_pkt;
 
 100	u32 portid;
 101	struct tipc_msg phdr;
 102	struct list_head cong_links;
 103	struct list_head publications;
 104	u32 pub_count;
 105	uint conn_timeout;
 106	atomic_t dupl_rcvcnt;
 
 107	bool probe_unacked;
 108	u16 cong_link_cnt;
 109	u16 snt_unacked;
 110	u16 snd_win;
 111	u16 peer_caps;
 112	u16 rcv_unacked;
 113	u16 rcv_win;
 114	struct sockaddr_tipc peer;
 115	struct rhash_head node;
 116	struct tipc_mc_method mc_method;
 117	struct rcu_head rcu;
 118	struct tipc_group *group;
 
 
 
 
 
 
 
 119	bool group_is_open;
 
 
 120};
 121
 122static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
 123static void tipc_data_ready(struct sock *sk);
 124static void tipc_write_space(struct sock *sk);
 125static void tipc_sock_destruct(struct sock *sk);
 126static int tipc_release(struct socket *sock);
 127static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
 128		       bool kern);
 129static void tipc_sk_timeout(struct timer_list *t);
 130static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
 131			   struct tipc_name_seq const *seq);
 132static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
 133			    struct tipc_name_seq const *seq);
 134static int tipc_sk_leave(struct tipc_sock *tsk);
 135static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
 136static int tipc_sk_insert(struct tipc_sock *tsk);
 137static void tipc_sk_remove(struct tipc_sock *tsk);
 138static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
 139static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
 
 
 140
 141static const struct proto_ops packet_ops;
 142static const struct proto_ops stream_ops;
 143static const struct proto_ops msg_ops;
 144static struct proto tipc_proto;
 145static const struct rhashtable_params tsk_rht_params;
 146
 147static u32 tsk_own_node(struct tipc_sock *tsk)
 148{
 149	return msg_prevnode(&tsk->phdr);
 150}
 151
 152static u32 tsk_peer_node(struct tipc_sock *tsk)
 153{
 154	return msg_destnode(&tsk->phdr);
 155}
 156
 157static u32 tsk_peer_port(struct tipc_sock *tsk)
 158{
 159	return msg_destport(&tsk->phdr);
 160}
 161
 162static  bool tsk_unreliable(struct tipc_sock *tsk)
 163{
 164	return msg_src_droppable(&tsk->phdr) != 0;
 165}
 166
 167static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
 168{
 169	msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
 170}
 171
 172static bool tsk_unreturnable(struct tipc_sock *tsk)
 173{
 174	return msg_dest_droppable(&tsk->phdr) != 0;
 175}
 176
 177static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
 178{
 179	msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
 180}
 181
 182static int tsk_importance(struct tipc_sock *tsk)
 183{
 184	return msg_importance(&tsk->phdr);
 185}
 186
 187static int tsk_set_importance(struct tipc_sock *tsk, int imp)
 
 
 
 
 
 188{
 189	if (imp > TIPC_CRITICAL_IMPORTANCE)
 190		return -EINVAL;
 191	msg_set_importance(&tsk->phdr, (u32)imp);
 192	return 0;
 193}
 194
 195static struct tipc_sock *tipc_sk(const struct sock *sk)
 196{
 197	return container_of(sk, struct tipc_sock, sk);
 198}
 199
 200static bool tsk_conn_cong(struct tipc_sock *tsk)
 201{
 202	return tsk->snt_unacked > tsk->snd_win;
 203}
 204
 205static u16 tsk_blocks(int len)
 206{
 207	return ((len / FLOWCTL_BLK_SZ) + 1);
 208}
 209
 210/* tsk_blocks(): translate a buffer size in bytes to number of
 211 * advertisable blocks, taking into account the ratio truesize(len)/len
 212 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
 213 */
 214static u16 tsk_adv_blocks(int len)
 215{
 216	return len / FLOWCTL_BLK_SZ / 4;
 217}
 218
 219/* tsk_inc(): increment counter for sent or received data
 220 * - If block based flow control is not supported by peer we
 221 *   fall back to message based ditto, incrementing the counter
 222 */
 223static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
 224{
 225	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
 226		return ((msglen / FLOWCTL_BLK_SZ) + 1);
 227	return 1;
 228}
 229
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 230/**
 231 * tsk_advance_rx_queue - discard first buffer in socket receive queue
 
 232 *
 233 * Caller must hold socket lock
 234 */
 235static void tsk_advance_rx_queue(struct sock *sk)
 236{
 
 237	kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
 238}
 239
 240/* tipc_sk_respond() : send response message back to sender
 241 */
 242static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
 243{
 244	u32 selector;
 245	u32 dnode;
 246	u32 onode = tipc_own_addr(sock_net(sk));
 247
 248	if (!tipc_msg_reverse(onode, &skb, err))
 249		return;
 250
 
 251	dnode = msg_destnode(buf_msg(skb));
 252	selector = msg_origport(buf_msg(skb));
 253	tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
 254}
 255
 256/**
 257 * tsk_rej_rx_queue - reject all buffers in socket receive queue
 
 
 258 *
 259 * Caller must hold socket lock
 260 */
 261static void tsk_rej_rx_queue(struct sock *sk)
 262{
 263	struct sk_buff *skb;
 264
 265	while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
 266		tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
 267}
 268
 269static bool tipc_sk_connected(struct sock *sk)
 270{
 271	return sk->sk_state == TIPC_ESTABLISHED;
 272}
 273
 274/* tipc_sk_type_connectionless - check if the socket is datagram socket
 275 * @sk: socket
 276 *
 277 * Returns true if connection less, false otherwise
 278 */
 279static bool tipc_sk_type_connectionless(struct sock *sk)
 280{
 281	return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
 282}
 283
 284/* tsk_peer_msg - verify if message was sent by connected port's peer
 285 *
 286 * Handles cases where the node's network address has changed from
 287 * the default of <0.0.0> to its configured setting.
 288 */
 289static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
 290{
 291	struct sock *sk = &tsk->sk;
 292	u32 self = tipc_own_addr(sock_net(sk));
 293	u32 peer_port = tsk_peer_port(tsk);
 294	u32 orig_node, peer_node;
 295
 296	if (unlikely(!tipc_sk_connected(sk)))
 297		return false;
 298
 299	if (unlikely(msg_origport(msg) != peer_port))
 300		return false;
 301
 302	orig_node = msg_orignode(msg);
 303	peer_node = tsk_peer_node(tsk);
 304
 305	if (likely(orig_node == peer_node))
 306		return true;
 307
 308	if (!orig_node && peer_node == self)
 309		return true;
 310
 311	if (!peer_node && orig_node == self)
 312		return true;
 313
 314	return false;
 315}
 316
 317/* tipc_set_sk_state - set the sk_state of the socket
 318 * @sk: socket
 319 *
 320 * Caller must hold socket lock
 321 *
 322 * Returns 0 on success, errno otherwise
 323 */
 324static int tipc_set_sk_state(struct sock *sk, int state)
 325{
 326	int oldsk_state = sk->sk_state;
 327	int res = -EINVAL;
 328
 329	switch (state) {
 330	case TIPC_OPEN:
 331		res = 0;
 332		break;
 333	case TIPC_LISTEN:
 334	case TIPC_CONNECTING:
 335		if (oldsk_state == TIPC_OPEN)
 336			res = 0;
 337		break;
 338	case TIPC_ESTABLISHED:
 339		if (oldsk_state == TIPC_CONNECTING ||
 340		    oldsk_state == TIPC_OPEN)
 341			res = 0;
 342		break;
 343	case TIPC_DISCONNECTING:
 344		if (oldsk_state == TIPC_CONNECTING ||
 345		    oldsk_state == TIPC_ESTABLISHED)
 346			res = 0;
 347		break;
 348	}
 349
 350	if (!res)
 351		sk->sk_state = state;
 352
 353	return res;
 354}
 355
 356static int tipc_sk_sock_err(struct socket *sock, long *timeout)
 357{
 358	struct sock *sk = sock->sk;
 359	int err = sock_error(sk);
 360	int typ = sock->type;
 361
 362	if (err)
 363		return err;
 364	if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
 365		if (sk->sk_state == TIPC_DISCONNECTING)
 366			return -EPIPE;
 367		else if (!tipc_sk_connected(sk))
 368			return -ENOTCONN;
 369	}
 370	if (!*timeout)
 371		return -EAGAIN;
 372	if (signal_pending(current))
 373		return sock_intr_errno(*timeout);
 374
 375	return 0;
 376}
 377
 378#define tipc_wait_for_cond(sock_, timeo_, condition_)			       \
 379({                                                                             \
 
 380	struct sock *sk_;						       \
 381	int rc_;							       \
 382									       \
 383	while ((rc_ = !(condition_))) {					       \
 384		DEFINE_WAIT_FUNC(wait_, woken_wake_function);	               \
 
 385		sk_ = (sock_)->sk;					       \
 386		rc_ = tipc_sk_sock_err((sock_), timeo_);		       \
 387		if (rc_)						       \
 388			break;						       \
 389		prepare_to_wait(sk_sleep(sk_), &wait_, TASK_INTERRUPTIBLE);    \
 390		release_sock(sk_);					       \
 391		*(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
 392		sched_annotate_sleep();				               \
 393		lock_sock(sk_);						       \
 394		remove_wait_queue(sk_sleep(sk_), &wait_);		       \
 395	}								       \
 396	rc_;								       \
 397})
 398
 399/**
 400 * tipc_sk_create - create a TIPC socket
 401 * @net: network namespace (must be default network)
 402 * @sock: pre-allocated socket structure
 403 * @protocol: protocol indicator (must be 0)
 404 * @kern: caused by kernel or by userspace?
 405 *
 406 * This routine creates additional data structures used by the TIPC socket,
 407 * initializes them, and links them together.
 408 *
 409 * Returns 0 on success, errno otherwise
 410 */
 411static int tipc_sk_create(struct net *net, struct socket *sock,
 412			  int protocol, int kern)
 413{
 414	struct tipc_net *tn;
 415	const struct proto_ops *ops;
 416	struct sock *sk;
 417	struct tipc_sock *tsk;
 418	struct tipc_msg *msg;
 419
 420	/* Validate arguments */
 421	if (unlikely(protocol != 0))
 422		return -EPROTONOSUPPORT;
 423
 424	switch (sock->type) {
 425	case SOCK_STREAM:
 426		ops = &stream_ops;
 427		break;
 428	case SOCK_SEQPACKET:
 429		ops = &packet_ops;
 430		break;
 431	case SOCK_DGRAM:
 432	case SOCK_RDM:
 433		ops = &msg_ops;
 434		break;
 435	default:
 436		return -EPROTOTYPE;
 437	}
 438
 439	/* Allocate socket's protocol area */
 440	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
 441	if (sk == NULL)
 442		return -ENOMEM;
 443
 444	tsk = tipc_sk(sk);
 445	tsk->max_pkt = MAX_PKT_DEFAULT;
 
 
 446	INIT_LIST_HEAD(&tsk->publications);
 447	INIT_LIST_HEAD(&tsk->cong_links);
 448	msg = &tsk->phdr;
 449	tn = net_generic(sock_net(sk), tipc_net_id);
 450
 451	/* Finish initializing socket data structures */
 452	sock->ops = ops;
 453	sock_init_data(sock, sk);
 454	tipc_set_sk_state(sk, TIPC_OPEN);
 455	if (tipc_sk_insert(tsk)) {
 
 456		pr_warn("Socket create failed; port number exhausted\n");
 457		return -EINVAL;
 458	}
 459
 460	/* Ensure tsk is visible before we read own_addr. */
 461	smp_mb();
 462
 463	tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
 464		      TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
 465
 466	msg_set_origport(msg, tsk->portid);
 467	timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
 468	sk->sk_shutdown = 0;
 469	sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
 470	sk->sk_rcvbuf = sysctl_tipc_rmem[1];
 471	sk->sk_data_ready = tipc_data_ready;
 472	sk->sk_write_space = tipc_write_space;
 473	sk->sk_destruct = tipc_sock_destruct;
 474	tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
 475	tsk->group_is_open = true;
 476	atomic_set(&tsk->dupl_rcvcnt, 0);
 477
 478	/* Start out with safe limits until we receive an advertised window */
 479	tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
 480	tsk->rcv_win = tsk->snd_win;
 481
 482	if (tipc_sk_type_connectionless(sk)) {
 483		tsk_set_unreturnable(tsk, true);
 484		if (sock->type == SOCK_DGRAM)
 485			tsk_set_unreliable(tsk, true);
 486	}
 487
 
 488	return 0;
 489}
 490
 491static void tipc_sk_callback(struct rcu_head *head)
 492{
 493	struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
 494
 495	sock_put(&tsk->sk);
 496}
 497
 498/* Caller should hold socket lock for the socket. */
 499static void __tipc_shutdown(struct socket *sock, int error)
 500{
 501	struct sock *sk = sock->sk;
 502	struct tipc_sock *tsk = tipc_sk(sk);
 503	struct net *net = sock_net(sk);
 504	long timeout = CONN_TIMEOUT_DEFAULT;
 505	u32 dnode = tsk_peer_node(tsk);
 506	struct sk_buff *skb;
 507
 508	/* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
 509	tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
 510					    !tsk_conn_cong(tsk)));
 511
 512	/* Reject all unreceived messages, except on an active connection
 513	 * (which disconnects locally & sends a 'FIN+' to peer).
 514	 */
 515	while ((skb = __skb_dequeue(&sk->sk_receive_queue)) != NULL) {
 516		if (TIPC_SKB_CB(skb)->bytes_read) {
 517			kfree_skb(skb);
 518			continue;
 519		}
 520		if (!tipc_sk_type_connectionless(sk) &&
 521		    sk->sk_state != TIPC_DISCONNECTING) {
 522			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
 523			tipc_node_remove_conn(net, dnode, tsk->portid);
 524		}
 525		tipc_sk_respond(sk, skb, error);
 526	}
 527
 528	if (tipc_sk_type_connectionless(sk))
 
 
 529		return;
 
 530
 531	if (sk->sk_state != TIPC_DISCONNECTING) {
 
 
 
 
 
 
 
 
 
 
 
 532		skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
 533				      TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
 534				      tsk_own_node(tsk), tsk_peer_port(tsk),
 535				      tsk->portid, error);
 536		if (skb)
 537			tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
 538		tipc_node_remove_conn(net, dnode, tsk->portid);
 539		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
 
 
 
 
 
 
 540	}
 541}
 542
 543/**
 544 * tipc_release - destroy a TIPC socket
 545 * @sock: socket to destroy
 546 *
 547 * This routine cleans up any messages that are still queued on the socket.
 548 * For DGRAM and RDM socket types, all queued messages are rejected.
 549 * For SEQPACKET and STREAM socket types, the first message is rejected
 550 * and any others are discarded.  (If the first message on a STREAM socket
 551 * is partially-read, it is discarded and the next one is rejected instead.)
 552 *
 553 * NOTE: Rejected messages are not necessarily returned to the sender!  They
 554 * are returned or discarded according to the "destination droppable" setting
 555 * specified for the message by the sender.
 556 *
 557 * Returns 0 on success, errno otherwise
 558 */
 559static int tipc_release(struct socket *sock)
 560{
 561	struct sock *sk = sock->sk;
 562	struct tipc_sock *tsk;
 563
 564	/*
 565	 * Exit if socket isn't fully initialized (occurs when a failed accept()
 566	 * releases a pre-allocated child socket that was never used)
 567	 */
 568	if (sk == NULL)
 569		return 0;
 570
 571	tsk = tipc_sk(sk);
 572	lock_sock(sk);
 573
 
 574	__tipc_shutdown(sock, TIPC_ERR_NO_PORT);
 575	sk->sk_shutdown = SHUTDOWN_MASK;
 576	tipc_sk_leave(tsk);
 577	tipc_sk_withdraw(tsk, 0, NULL);
 
 578	sk_stop_timer(sk, &sk->sk_timer);
 579	tipc_sk_remove(tsk);
 580
 
 581	/* Reject any messages that accumulated in backlog queue */
 582	release_sock(sk);
 583	tipc_dest_list_purge(&tsk->cong_links);
 584	tsk->cong_link_cnt = 0;
 585	call_rcu(&tsk->rcu, tipc_sk_callback);
 586	sock->sk = NULL;
 587
 588	return 0;
 589}
 590
 591/**
 592 * tipc_bind - associate or disassocate TIPC name(s) with a socket
 593 * @sock: socket structure
 594 * @uaddr: socket address describing name(s) and desired operation
 595 * @uaddr_len: size of socket address data structure
 596 *
 597 * Name and name sequence binding is indicated using a positive scope value;
 598 * a negative scope value unbinds the specified name.  Specifying no name
 599 * (i.e. a socket address length of 0) unbinds all names from the socket.
 600 *
 601 * Returns 0 on success, errno otherwise
 602 *
 603 * NOTE: This routine doesn't need to take the socket lock since it doesn't
 604 *       access any non-constant socket information.
 605 */
 606static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
 607		     int uaddr_len)
 608{
 609	struct sock *sk = sock->sk;
 610	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
 611	struct tipc_sock *tsk = tipc_sk(sk);
 612	int res = -EINVAL;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 613
 614	lock_sock(sk);
 615	if (unlikely(!uaddr_len)) {
 616		res = tipc_sk_withdraw(tsk, 0, NULL);
 617		goto exit;
 618	}
 619	if (tsk->group) {
 620		res = -EACCES;
 621		goto exit;
 622	}
 623	if (uaddr_len < sizeof(struct sockaddr_tipc)) {
 624		res = -EINVAL;
 625		goto exit;
 626	}
 627	if (addr->family != AF_TIPC) {
 628		res = -EAFNOSUPPORT;
 629		goto exit;
 630	}
 631
 632	if (addr->addrtype == TIPC_ADDR_NAME)
 633		addr->addr.nameseq.upper = addr->addr.nameseq.lower;
 634	else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
 635		res = -EAFNOSUPPORT;
 636		goto exit;
 637	}
 638
 639	if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
 640	    (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
 641	    (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
 642		res = -EACCES;
 643		goto exit;
 644	}
 645
 646	res = (addr->scope >= 0) ?
 647		tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
 648		tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
 649exit:
 650	release_sock(sk);
 651	return res;
 652}
 653
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 654/**
 655 * tipc_getname - get port ID of socket or peer socket
 656 * @sock: socket structure
 657 * @uaddr: area for returned socket address
 658 * @uaddr_len: area for returned length of socket address
 659 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
 660 *
 661 * Returns 0 on success, errno otherwise
 662 *
 663 * NOTE: This routine doesn't need to take the socket lock since it only
 664 *       accesses socket information that is unchanging (or which changes in
 665 *       a completely predictable manner).
 666 */
 667static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
 668			int peer)
 669{
 670	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
 671	struct sock *sk = sock->sk;
 672	struct tipc_sock *tsk = tipc_sk(sk);
 673
 674	memset(addr, 0, sizeof(*addr));
 675	if (peer) {
 676		if ((!tipc_sk_connected(sk)) &&
 677		    ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
 678			return -ENOTCONN;
 679		addr->addr.id.ref = tsk_peer_port(tsk);
 680		addr->addr.id.node = tsk_peer_node(tsk);
 681	} else {
 682		addr->addr.id.ref = tsk->portid;
 683		addr->addr.id.node = tipc_own_addr(sock_net(sk));
 684	}
 685
 686	addr->addrtype = TIPC_ADDR_ID;
 687	addr->family = AF_TIPC;
 688	addr->scope = 0;
 689	addr->addr.name.domain = 0;
 690
 691	return sizeof(*addr);
 692}
 693
 694/**
 695 * tipc_poll - read and possibly block on pollmask
 696 * @file: file structure associated with the socket
 697 * @sock: socket for which to calculate the poll bits
 698 * @wait: ???
 699 *
 700 * Returns pollmask value
 701 *
 702 * COMMENTARY:
 703 * It appears that the usual socket locking mechanisms are not useful here
 704 * since the pollmask info is potentially out-of-date the moment this routine
 705 * exits.  TCP and other protocols seem to rely on higher level poll routines
 706 * to handle any preventable race conditions, so TIPC will do the same ...
 707 *
 708 * IMPORTANT: The fact that a read or write operation is indicated does NOT
 709 * imply that the operation will succeed, merely that it should be performed
 710 * and will not block.
 711 */
 712static __poll_t tipc_poll(struct file *file, struct socket *sock,
 713			      poll_table *wait)
 714{
 715	struct sock *sk = sock->sk;
 716	struct tipc_sock *tsk = tipc_sk(sk);
 717	__poll_t revents = 0;
 718
 719	sock_poll_wait(file, sk_sleep(sk), wait);
 
 720
 721	if (sk->sk_shutdown & RCV_SHUTDOWN)
 722		revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
 723	if (sk->sk_shutdown == SHUTDOWN_MASK)
 724		revents |= EPOLLHUP;
 725
 726	switch (sk->sk_state) {
 727	case TIPC_ESTABLISHED:
 728	case TIPC_CONNECTING:
 729		if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
 730			revents |= EPOLLOUT;
 731		/* fall thru' */
 732	case TIPC_LISTEN:
 733		if (!skb_queue_empty(&sk->sk_receive_queue))
 
 734			revents |= EPOLLIN | EPOLLRDNORM;
 735		break;
 736	case TIPC_OPEN:
 737		if (tsk->group_is_open && !tsk->cong_link_cnt)
 738			revents |= EPOLLOUT;
 739		if (!tipc_sk_type_connectionless(sk))
 740			break;
 741		if (skb_queue_empty(&sk->sk_receive_queue))
 742			break;
 743		revents |= EPOLLIN | EPOLLRDNORM;
 744		break;
 745	case TIPC_DISCONNECTING:
 746		revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
 747		break;
 748	}
 749	return revents;
 750}
 751
 752/**
 753 * tipc_sendmcast - send multicast message
 754 * @sock: socket structure
 755 * @seq: destination address
 756 * @msg: message to send
 757 * @dlen: length of data to send
 758 * @timeout: timeout to wait for wakeup
 759 *
 760 * Called from function tipc_sendmsg(), which has done all sanity checks
 761 * Returns the number of bytes sent on success, or errno
 762 */
 763static int tipc_sendmcast(struct  socket *sock, struct tipc_name_seq *seq,
 764			  struct msghdr *msg, size_t dlen, long timeout)
 765{
 766	struct sock *sk = sock->sk;
 767	struct tipc_sock *tsk = tipc_sk(sk);
 768	struct tipc_msg *hdr = &tsk->phdr;
 769	struct net *net = sock_net(sk);
 770	int mtu = tipc_bcast_get_mtu(net);
 771	struct tipc_mc_method *method = &tsk->mc_method;
 772	struct sk_buff_head pkts;
 773	struct tipc_nlist dsts;
 774	int rc;
 775
 776	if (tsk->group)
 777		return -EACCES;
 778
 779	/* Block or return if any destination link is congested */
 780	rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
 781	if (unlikely(rc))
 782		return rc;
 783
 784	/* Lookup destination nodes */
 785	tipc_nlist_init(&dsts, tipc_own_addr(net));
 786	tipc_nametbl_lookup_dst_nodes(net, seq->type, seq->lower,
 787				      seq->upper, &dsts);
 788	if (!dsts.local && !dsts.remote)
 789		return -EHOSTUNREACH;
 790
 791	/* Build message header */
 792	msg_set_type(hdr, TIPC_MCAST_MSG);
 793	msg_set_hdr_sz(hdr, MCAST_H_SIZE);
 794	msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
 795	msg_set_destport(hdr, 0);
 796	msg_set_destnode(hdr, 0);
 797	msg_set_nametype(hdr, seq->type);
 798	msg_set_namelower(hdr, seq->lower);
 799	msg_set_nameupper(hdr, seq->upper);
 800
 801	/* Build message as chain of buffers */
 802	skb_queue_head_init(&pkts);
 803	rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
 804
 805	/* Send message if build was successful */
 806	if (unlikely(rc == dlen))
 807		rc = tipc_mcast_xmit(net, &pkts, method, &dsts,
 
 
 808				     &tsk->cong_link_cnt);
 
 809
 810	tipc_nlist_purge(&dsts);
 811
 812	return rc ? rc : dlen;
 813}
 814
 815/**
 816 * tipc_send_group_msg - send a message to a member in the group
 817 * @net: network namespace
 
 818 * @m: message to send
 819 * @mb: group member
 820 * @dnode: destination node
 821 * @dport: destination port
 822 * @dlen: total length of message data
 823 */
 824static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
 825			       struct msghdr *m, struct tipc_member *mb,
 826			       u32 dnode, u32 dport, int dlen)
 827{
 828	u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
 829	struct tipc_mc_method *method = &tsk->mc_method;
 830	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
 831	struct tipc_msg *hdr = &tsk->phdr;
 832	struct sk_buff_head pkts;
 833	int mtu, rc;
 834
 835	/* Complete message header */
 836	msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
 837	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
 838	msg_set_destport(hdr, dport);
 839	msg_set_destnode(hdr, dnode);
 840	msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
 841
 842	/* Build message as chain of buffers */
 843	skb_queue_head_init(&pkts);
 844	mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
 845	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
 846	if (unlikely(rc != dlen))
 847		return rc;
 848
 849	/* Send message */
 850	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
 851	if (unlikely(rc == -ELINKCONG)) {
 852		tipc_dest_push(&tsk->cong_links, dnode, 0);
 853		tsk->cong_link_cnt++;
 854	}
 855
 856	/* Update send window */
 857	tipc_group_update_member(mb, blks);
 858
 859	/* A broadcast sent within next EXPIRE period must follow same path */
 860	method->rcast = true;
 861	method->mandatory = true;
 862	return dlen;
 863}
 864
 865/**
 866 * tipc_send_group_unicast - send message to a member in the group
 867 * @sock: socket structure
 868 * @m: message to send
 869 * @dlen: total length of message data
 870 * @timeout: timeout to wait for wakeup
 871 *
 872 * Called from function tipc_sendmsg(), which has done all sanity checks
 873 * Returns the number of bytes sent on success, or errno
 874 */
 875static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
 876				   int dlen, long timeout)
 877{
 878	struct sock *sk = sock->sk;
 879	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
 880	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
 881	struct tipc_sock *tsk = tipc_sk(sk);
 882	struct tipc_group *grp = tsk->group;
 883	struct net *net = sock_net(sk);
 884	struct tipc_member *mb = NULL;
 885	u32 node, port;
 886	int rc;
 887
 888	node = dest->addr.id.node;
 889	port = dest->addr.id.ref;
 890	if (!port && !node)
 891		return -EHOSTUNREACH;
 892
 893	/* Block or return if destination link or member is congested */
 894	rc = tipc_wait_for_cond(sock, &timeout,
 895				!tipc_dest_find(&tsk->cong_links, node, 0) &&
 896				!tipc_group_cong(grp, node, port, blks, &mb));
 
 
 897	if (unlikely(rc))
 898		return rc;
 899
 900	if (unlikely(!mb))
 901		return -EHOSTUNREACH;
 902
 903	rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
 904
 905	return rc ? rc : dlen;
 906}
 907
 908/**
 909 * tipc_send_group_anycast - send message to any member with given identity
 910 * @sock: socket structure
 911 * @m: message to send
 912 * @dlen: total length of message data
 913 * @timeout: timeout to wait for wakeup
 914 *
 915 * Called from function tipc_sendmsg(), which has done all sanity checks
 916 * Returns the number of bytes sent on success, or errno
 917 */
 918static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
 919				   int dlen, long timeout)
 920{
 921	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
 922	struct sock *sk = sock->sk;
 923	struct tipc_sock *tsk = tipc_sk(sk);
 924	struct list_head *cong_links = &tsk->cong_links;
 925	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
 926	struct tipc_group *grp = tsk->group;
 927	struct tipc_msg *hdr = &tsk->phdr;
 928	struct tipc_member *first = NULL;
 929	struct tipc_member *mbr = NULL;
 930	struct net *net = sock_net(sk);
 931	u32 node, port, exclude;
 932	struct list_head dsts;
 933	u32 type, inst, scope;
 934	int lookups = 0;
 935	int dstcnt, rc;
 936	bool cong;
 937
 938	INIT_LIST_HEAD(&dsts);
 
 939
 940	type = msg_nametype(hdr);
 941	inst = dest->addr.name.name.instance;
 942	scope = msg_lookup_scope(hdr);
 943	exclude = tipc_group_exclude(grp);
 944
 945	while (++lookups < 4) {
 946		first = NULL;
 947
 948		/* Look for a non-congested destination member, if any */
 949		while (1) {
 950			if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
 951						 &dstcnt, exclude, false))
 952				return -EHOSTUNREACH;
 953			tipc_dest_pop(&dsts, &node, &port);
 954			cong = tipc_group_cong(grp, node, port, blks, &mbr);
 
 955			if (!cong)
 956				break;
 957			if (mbr == first)
 958				break;
 959			if (!first)
 960				first = mbr;
 961		}
 962
 963		/* Start over if destination was not in member list */
 964		if (unlikely(!mbr))
 965			continue;
 966
 967		if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
 968			break;
 969
 970		/* Block or return if destination link or member is congested */
 971		rc = tipc_wait_for_cond(sock, &timeout,
 972					!tipc_dest_find(cong_links, node, 0) &&
 973					!tipc_group_cong(grp, node, port,
 
 974							 blks, &mbr));
 975		if (unlikely(rc))
 976			return rc;
 977
 978		/* Send, unless destination disappeared while waiting */
 979		if (likely(mbr))
 980			break;
 981	}
 982
 983	if (unlikely(lookups >= 4))
 984		return -EHOSTUNREACH;
 985
 986	rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
 987
 988	return rc ? rc : dlen;
 989}
 990
 991/**
 992 * tipc_send_group_bcast - send message to all members in communication group
 993 * @sk: 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 * Returns the number of bytes sent on success, or errno
1000 */
1001static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1002				 int dlen, long timeout)
1003{
1004	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1005	struct sock *sk = sock->sk;
1006	struct net *net = sock_net(sk);
1007	struct tipc_sock *tsk = tipc_sk(sk);
1008	struct tipc_group *grp = tsk->group;
1009	struct tipc_nlist *dsts = tipc_group_dests(grp);
1010	struct tipc_mc_method *method = &tsk->mc_method;
1011	bool ack = method->mandatory && method->rcast;
1012	int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1013	struct tipc_msg *hdr = &tsk->phdr;
1014	int mtu = tipc_bcast_get_mtu(net);
1015	struct sk_buff_head pkts;
1016	int rc = -EHOSTUNREACH;
1017
1018	if (!dsts->local && !dsts->remote)
1019		return -EHOSTUNREACH;
1020
1021	/* Block or return if any destination link or member is congested */
1022	rc = tipc_wait_for_cond(sock, &timeout,	!tsk->cong_link_cnt &&
1023				!tipc_group_bc_cong(grp, blks));
 
1024	if (unlikely(rc))
1025		return rc;
1026
 
 
 
 
1027	/* Complete message header */
1028	if (dest) {
1029		msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1030		msg_set_nameinst(hdr, dest->addr.name.name.instance);
1031	} else {
1032		msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1033		msg_set_nameinst(hdr, 0);
1034	}
1035	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1036	msg_set_destport(hdr, 0);
1037	msg_set_destnode(hdr, 0);
1038	msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(grp));
1039
1040	/* Avoid getting stuck with repeated forced replicasts */
1041	msg_set_grp_bc_ack_req(hdr, ack);
1042
1043	/* Build message as chain of buffers */
1044	skb_queue_head_init(&pkts);
1045	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1046	if (unlikely(rc != dlen))
1047		return rc;
1048
1049	/* Send message */
1050	rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1051	if (unlikely(rc))
1052		return rc;
1053
1054	/* Update broadcast sequence number and send windows */
1055	tipc_group_update_bc_members(tsk->group, blks, ack);
1056
1057	/* Broadcast link is now free to choose method for next broadcast */
1058	method->mandatory = false;
1059	method->expires = jiffies;
1060
1061	return dlen;
1062}
1063
1064/**
1065 * tipc_send_group_mcast - send message to all members with given identity
1066 * @sock: socket structure
1067 * @m: message to send
1068 * @dlen: total length of message data
1069 * @timeout: timeout to wait for wakeup
1070 *
1071 * Called from function tipc_sendmsg(), which has done all sanity checks
1072 * Returns the number of bytes sent on success, or errno
1073 */
1074static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1075				 int dlen, long timeout)
1076{
 
1077	struct sock *sk = sock->sk;
1078	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1079	struct tipc_sock *tsk = tipc_sk(sk);
1080	struct tipc_group *grp = tsk->group;
1081	struct tipc_msg *hdr = &tsk->phdr;
1082	struct net *net = sock_net(sk);
1083	u32 type, inst, scope, exclude;
1084	struct list_head dsts;
1085	u32 dstcnt;
1086
1087	INIT_LIST_HEAD(&dsts);
1088
1089	type = msg_nametype(hdr);
1090	inst = dest->addr.name.name.instance;
1091	scope = msg_lookup_scope(hdr);
1092	exclude = tipc_group_exclude(grp);
1093
1094	if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1095				 &dstcnt, exclude, true))
1096		return -EHOSTUNREACH;
1097
1098	if (dstcnt == 1) {
1099		tipc_dest_pop(&dsts, &dest->addr.id.node, &dest->addr.id.ref);
1100		return tipc_send_group_unicast(sock, m, dlen, timeout);
1101	}
1102
1103	tipc_dest_list_purge(&dsts);
1104	return tipc_send_group_bcast(sock, m, dlen, timeout);
1105}
1106
1107/**
1108 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
 
1109 * @arrvq: queue with arriving messages, to be cloned after destination lookup
1110 * @inputq: queue with cloned messages, delivered to socket after dest lookup
1111 *
1112 * Multi-threaded: parallel calls with reference to same queues may occur
1113 */
1114void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1115		       struct sk_buff_head *inputq)
1116{
1117	u32 self = tipc_own_addr(net);
1118	u32 type, lower, upper, scope;
1119	struct sk_buff *skb, *_skb;
1120	u32 portid, oport, onode;
1121	struct sk_buff_head tmpq;
1122	struct list_head dports;
1123	struct tipc_msg *hdr;
 
1124	int user, mtyp, hlen;
1125	bool exact;
1126
1127	__skb_queue_head_init(&tmpq);
1128	INIT_LIST_HEAD(&dports);
 
1129
 
1130	skb = tipc_skb_peek(arrvq, &inputq->lock);
1131	for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1132		hdr = buf_msg(skb);
1133		user = msg_user(hdr);
1134		mtyp = msg_type(hdr);
1135		hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1136		oport = msg_origport(hdr);
1137		onode = msg_orignode(hdr);
1138		type = msg_nametype(hdr);
 
 
 
 
 
 
1139
1140		if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1141			spin_lock_bh(&inputq->lock);
1142			if (skb_peek(arrvq) == skb) {
1143				__skb_dequeue(arrvq);
1144				__skb_queue_tail(inputq, skb);
1145			}
1146			kfree_skb(skb);
1147			spin_unlock_bh(&inputq->lock);
1148			continue;
1149		}
1150
1151		/* Group messages require exact scope match */
1152		if (msg_in_group(hdr)) {
1153			lower = 0;
1154			upper = ~0;
1155			scope = msg_lookup_scope(hdr);
1156			exact = true;
1157		} else {
1158			/* TIPC_NODE_SCOPE means "any scope" in this context */
1159			if (onode == self)
1160				scope = TIPC_NODE_SCOPE;
1161			else
1162				scope = TIPC_CLUSTER_SCOPE;
1163			exact = false;
1164			lower = msg_namelower(hdr);
1165			upper = msg_nameupper(hdr);
1166		}
1167
1168		/* Create destination port list: */
1169		tipc_nametbl_mc_lookup(net, type, lower, upper,
1170				       scope, exact, &dports);
1171
1172		/* Clone message per destination */
1173		while (tipc_dest_pop(&dports, NULL, &portid)) {
1174			_skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1175			if (_skb) {
1176				msg_set_destport(buf_msg(_skb), portid);
1177				__skb_queue_tail(&tmpq, _skb);
1178				continue;
1179			}
1180			pr_warn("Failed to clone mcast rcv buffer\n");
1181		}
1182		/* Append to inputq if not already done by other thread */
1183		spin_lock_bh(&inputq->lock);
1184		if (skb_peek(arrvq) == skb) {
1185			skb_queue_splice_tail_init(&tmpq, inputq);
 
1186			kfree_skb(__skb_dequeue(arrvq));
1187		}
1188		spin_unlock_bh(&inputq->lock);
1189		__skb_queue_purge(&tmpq);
1190		kfree_skb(skb);
1191	}
1192	tipc_sk_rcv(net, inputq);
1193}
1194
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1195/**
1196 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1197 * @tsk: receiving socket
1198 * @skb: pointer to message buffer.
 
 
1199 */
1200static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
 
1201				   struct sk_buff_head *xmitq)
1202{
1203	struct tipc_msg *hdr = buf_msg(skb);
1204	u32 onode = tsk_own_node(tsk);
1205	struct sock *sk = &tsk->sk;
1206	int mtyp = msg_type(hdr);
1207	bool conn_cong;
1208
1209	/* Ignore if connection cannot be validated: */
1210	if (!tsk_peer_msg(tsk, hdr))
 
1211		goto exit;
 
1212
1213	if (unlikely(msg_errcode(hdr))) {
1214		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1215		tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1216				      tsk_peer_port(tsk));
1217		sk->sk_state_change(sk);
1218		goto exit;
 
 
 
 
 
 
 
 
 
1219	}
1220
1221	tsk->probe_unacked = false;
1222
1223	if (mtyp == CONN_PROBE) {
1224		msg_set_type(hdr, CONN_PROBE_REPLY);
1225		if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1226			__skb_queue_tail(xmitq, skb);
1227		return;
1228	} else if (mtyp == CONN_ACK) {
1229		conn_cong = tsk_conn_cong(tsk);
 
1230		tsk->snt_unacked -= msg_conn_ack(hdr);
1231		if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1232			tsk->snd_win = msg_adv_win(hdr);
1233		if (conn_cong)
1234			sk->sk_write_space(sk);
1235	} else if (mtyp != CONN_PROBE_REPLY) {
1236		pr_warn("Received unknown CONN_PROTO msg\n");
1237	}
1238exit:
1239	kfree_skb(skb);
1240}
1241
1242/**
1243 * tipc_sendmsg - send message in connectionless manner
1244 * @sock: socket structure
1245 * @m: message to send
1246 * @dsz: amount of user data to be sent
1247 *
1248 * Message must have an destination specified explicitly.
1249 * Used for SOCK_RDM and SOCK_DGRAM messages,
1250 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1251 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1252 *
1253 * Returns the number of bytes sent on success, or errno otherwise
1254 */
1255static int tipc_sendmsg(struct socket *sock,
1256			struct msghdr *m, size_t dsz)
1257{
1258	struct sock *sk = sock->sk;
1259	int ret;
1260
1261	lock_sock(sk);
1262	ret = __tipc_sendmsg(sock, m, dsz);
1263	release_sock(sk);
1264
1265	return ret;
1266}
1267
1268static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1269{
1270	struct sock *sk = sock->sk;
1271	struct net *net = sock_net(sk);
1272	struct tipc_sock *tsk = tipc_sk(sk);
1273	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1274	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1275	struct list_head *clinks = &tsk->cong_links;
1276	bool syn = !tipc_sk_type_connectionless(sk);
1277	struct tipc_group *grp = tsk->group;
1278	struct tipc_msg *hdr = &tsk->phdr;
1279	struct tipc_name_seq *seq;
1280	struct sk_buff_head pkts;
1281	u32 dport, dnode = 0;
1282	u32 type, inst;
1283	int mtu, rc;
1284
1285	if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1286		return -EMSGSIZE;
1287
1288	if (likely(dest)) {
1289		if (unlikely(m->msg_namelen < sizeof(*dest)))
1290			return -EINVAL;
1291		if (unlikely(dest->family != AF_TIPC))
1292			return -EINVAL;
 
1293	}
1294
 
1295	if (grp) {
1296		if (!dest)
1297			return tipc_send_group_bcast(sock, m, dlen, timeout);
1298		if (dest->addrtype == TIPC_ADDR_NAME)
1299			return tipc_send_group_anycast(sock, m, dlen, timeout);
1300		if (dest->addrtype == TIPC_ADDR_ID)
1301			return tipc_send_group_unicast(sock, m, dlen, timeout);
1302		if (dest->addrtype == TIPC_ADDR_MCAST)
1303			return tipc_send_group_mcast(sock, m, dlen, timeout);
1304		return -EINVAL;
1305	}
1306
1307	if (unlikely(!dest)) {
1308		dest = &tsk->peer;
1309		if (!syn || dest->family != AF_TIPC)
1310			return -EDESTADDRREQ;
 
1311	}
1312
1313	if (unlikely(syn)) {
1314		if (sk->sk_state == TIPC_LISTEN)
1315			return -EPIPE;
1316		if (sk->sk_state != TIPC_OPEN)
1317			return -EISCONN;
1318		if (tsk->published)
1319			return -EOPNOTSUPP;
1320		if (dest->addrtype == TIPC_ADDR_NAME) {
1321			tsk->conn_type = dest->addr.name.name.type;
1322			tsk->conn_instance = dest->addr.name.name.instance;
1323		}
1324	}
1325
1326	seq = &dest->addr.nameseq;
1327	if (dest->addrtype == TIPC_ADDR_MCAST)
1328		return tipc_sendmcast(sock, seq, m, dlen, timeout);
1329
1330	if (dest->addrtype == TIPC_ADDR_NAME) {
1331		type = dest->addr.name.name.type;
1332		inst = dest->addr.name.name.instance;
1333		dnode = dest->addr.name.domain;
1334		msg_set_type(hdr, TIPC_NAMED_MSG);
1335		msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1336		msg_set_nametype(hdr, type);
1337		msg_set_nameinst(hdr, inst);
1338		msg_set_lookup_scope(hdr, tipc_node2scope(dnode));
1339		dport = tipc_nametbl_translate(net, type, inst, &dnode);
1340		msg_set_destnode(hdr, dnode);
1341		msg_set_destport(hdr, dport);
1342		if (unlikely(!dport && !dnode))
1343			return -EHOSTUNREACH;
1344	} else if (dest->addrtype == TIPC_ADDR_ID) {
1345		dnode = dest->addr.id.node;
1346		msg_set_type(hdr, TIPC_DIRECT_MSG);
1347		msg_set_lookup_scope(hdr, 0);
1348		msg_set_destnode(hdr, dnode);
1349		msg_set_destport(hdr, dest->addr.id.ref);
1350		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1351	} else {
1352		return -EINVAL;
1353	}
1354
1355	/* Block or return if destination link is congested */
1356	rc = tipc_wait_for_cond(sock, &timeout,
1357				!tipc_dest_find(clinks, dnode, 0));
1358	if (unlikely(rc))
1359		return rc;
1360
1361	skb_queue_head_init(&pkts);
1362	mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1363	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1364	if (unlikely(rc != dlen))
1365		return rc;
 
 
 
 
1366
1367	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
 
 
1368	if (unlikely(rc == -ELINKCONG)) {
1369		tipc_dest_push(clinks, dnode, 0);
1370		tsk->cong_link_cnt++;
1371		rc = 0;
1372	}
1373
1374	if (unlikely(syn && !rc))
1375		tipc_set_sk_state(sk, TIPC_CONNECTING);
 
 
 
 
 
1376
1377	return rc ? rc : dlen;
1378}
1379
1380/**
1381 * tipc_sendstream - send stream-oriented data
1382 * @sock: socket structure
1383 * @m: data to send
1384 * @dsz: total length of data to be transmitted
1385 *
1386 * Used for SOCK_STREAM data.
1387 *
1388 * Returns the number of bytes sent on success (or partial success),
1389 * or errno if no data sent
1390 */
1391static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1392{
1393	struct sock *sk = sock->sk;
1394	int ret;
1395
1396	lock_sock(sk);
1397	ret = __tipc_sendstream(sock, m, dsz);
1398	release_sock(sk);
1399
1400	return ret;
1401}
1402
1403static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1404{
1405	struct sock *sk = sock->sk;
1406	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1407	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
 
1408	struct tipc_sock *tsk = tipc_sk(sk);
1409	struct tipc_msg *hdr = &tsk->phdr;
1410	struct net *net = sock_net(sk);
1411	struct sk_buff_head pkts;
1412	u32 dnode = tsk_peer_node(tsk);
 
 
1413	int send, sent = 0;
1414	int rc = 0;
1415
1416	skb_queue_head_init(&pkts);
1417
1418	if (unlikely(dlen > INT_MAX))
1419		return -EMSGSIZE;
1420
1421	/* Handle implicit connection setup */
1422	if (unlikely(dest)) {
1423		rc = __tipc_sendmsg(sock, m, dlen);
1424		if (dlen && (dlen == rc))
 
1425			tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
 
1426		return rc;
1427	}
1428
1429	do {
1430		rc = tipc_wait_for_cond(sock, &timeout,
1431					(!tsk->cong_link_cnt &&
1432					 !tsk_conn_cong(tsk) &&
1433					 tipc_sk_connected(sk)));
1434		if (unlikely(rc))
1435			break;
1436
1437		send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1438		rc = tipc_msg_build(hdr, m, sent, send, tsk->max_pkt, &pkts);
1439		if (unlikely(rc != send))
1440			break;
1441
1442		rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1443		if (unlikely(rc == -ELINKCONG)) {
1444			tsk->cong_link_cnt = 1;
1445			rc = 0;
1446		}
1447		if (likely(!rc)) {
1448			tsk->snt_unacked += tsk_inc(tsk, send + MIN_H_SIZE);
 
1449			sent += send;
1450		}
1451	} while (sent < dlen && !rc);
1452
1453	return sent ? sent : rc;
1454}
1455
1456/**
1457 * tipc_send_packet - send a connection-oriented message
1458 * @sock: socket structure
1459 * @m: message to send
1460 * @dsz: length of data to be transmitted
1461 *
1462 * Used for SOCK_SEQPACKET messages.
1463 *
1464 * Returns the number of bytes sent on success, or errno otherwise
1465 */
1466static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1467{
1468	if (dsz > TIPC_MAX_USER_MSG_SIZE)
1469		return -EMSGSIZE;
1470
1471	return tipc_sendstream(sock, m, dsz);
1472}
1473
1474/* tipc_sk_finish_conn - complete the setup of a connection
1475 */
1476static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1477				u32 peer_node)
1478{
1479	struct sock *sk = &tsk->sk;
1480	struct net *net = sock_net(sk);
1481	struct tipc_msg *msg = &tsk->phdr;
1482
 
1483	msg_set_destnode(msg, peer_node);
1484	msg_set_destport(msg, peer_port);
1485	msg_set_type(msg, TIPC_CONN_MSG);
1486	msg_set_lookup_scope(msg, 0);
1487	msg_set_hdr_sz(msg, SHORT_H_SIZE);
1488
1489	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1490	tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1491	tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1492	tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1493	tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
 
 
1494	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1495		return;
1496
1497	/* Fall back to message based flow control */
1498	tsk->rcv_win = FLOWCTL_MSG_WIN;
1499	tsk->snd_win = FLOWCTL_MSG_WIN;
1500}
1501
1502/**
1503 * tipc_sk_set_orig_addr - capture sender's address for received message
1504 * @m: descriptor for message info
1505 * @hdr: received message header
1506 *
1507 * Note: Address is not captured if not requested by receiver.
1508 */
1509static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1510{
1511	DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1512	struct tipc_msg *hdr = buf_msg(skb);
1513
1514	if (!srcaddr)
1515		return;
1516
1517	srcaddr->sock.family = AF_TIPC;
1518	srcaddr->sock.addrtype = TIPC_ADDR_ID;
1519	srcaddr->sock.scope = 0;
1520	srcaddr->sock.addr.id.ref = msg_origport(hdr);
1521	srcaddr->sock.addr.id.node = msg_orignode(hdr);
1522	srcaddr->sock.addr.name.domain = 0;
1523	m->msg_namelen = sizeof(struct sockaddr_tipc);
1524
1525	if (!msg_in_group(hdr))
1526		return;
1527
1528	/* Group message users may also want to know sending member's id */
1529	srcaddr->member.family = AF_TIPC;
1530	srcaddr->member.addrtype = TIPC_ADDR_NAME;
1531	srcaddr->member.scope = 0;
1532	srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1533	srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1534	srcaddr->member.addr.name.domain = 0;
1535	m->msg_namelen = sizeof(*srcaddr);
1536}
1537
1538/**
1539 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1540 * @m: descriptor for message info
1541 * @msg: received message header
1542 * @tsk: TIPC port associated with message
1543 *
1544 * Note: Ancillary data is not captured if not requested by receiver.
1545 *
1546 * Returns 0 if successful, otherwise errno
1547 */
1548static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1549				 struct tipc_sock *tsk)
1550{
1551	u32 anc_data[3];
1552	u32 err;
1553	u32 dest_type;
1554	int has_name;
1555	int res;
1556
1557	if (likely(m->msg_controllen == 0))
1558		return 0;
1559
1560	/* Optionally capture errored message object(s) */
1561	err = msg ? msg_errcode(msg) : 0;
1562	if (unlikely(err)) {
1563		anc_data[0] = err;
1564		anc_data[1] = msg_data_sz(msg);
1565		res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1566		if (res)
1567			return res;
1568		if (anc_data[1]) {
1569			res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1570				       msg_data(msg));
1571			if (res)
1572				return res;
1573		}
 
 
1574	}
1575
1576	/* Optionally capture message destination object */
1577	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1578	switch (dest_type) {
1579	case TIPC_NAMED_MSG:
1580		has_name = 1;
1581		anc_data[0] = msg_nametype(msg);
1582		anc_data[1] = msg_namelower(msg);
1583		anc_data[2] = msg_namelower(msg);
1584		break;
1585	case TIPC_MCAST_MSG:
1586		has_name = 1;
1587		anc_data[0] = msg_nametype(msg);
1588		anc_data[1] = msg_namelower(msg);
1589		anc_data[2] = msg_nameupper(msg);
1590		break;
1591	case TIPC_CONN_MSG:
1592		has_name = (tsk->conn_type != 0);
1593		anc_data[0] = tsk->conn_type;
1594		anc_data[1] = tsk->conn_instance;
1595		anc_data[2] = tsk->conn_instance;
1596		break;
1597	default:
1598		has_name = 0;
1599	}
1600	if (has_name) {
1601		res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1602		if (res)
1603			return res;
1604	}
1605
1606	return 0;
1607}
1608
1609static void tipc_sk_send_ack(struct tipc_sock *tsk)
1610{
1611	struct sock *sk = &tsk->sk;
1612	struct net *net = sock_net(sk);
1613	struct sk_buff *skb = NULL;
1614	struct tipc_msg *msg;
1615	u32 peer_port = tsk_peer_port(tsk);
1616	u32 dnode = tsk_peer_node(tsk);
1617
1618	if (!tipc_sk_connected(sk))
1619		return;
1620	skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1621			      dnode, tsk_own_node(tsk), peer_port,
1622			      tsk->portid, TIPC_OK);
1623	if (!skb)
1624		return;
1625	msg = buf_msg(skb);
1626	msg_set_conn_ack(msg, tsk->rcv_unacked);
1627	tsk->rcv_unacked = 0;
1628
1629	/* Adjust to and advertize the correct window limit */
1630	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1631		tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1632		msg_set_adv_win(msg, tsk->rcv_win);
1633	}
1634	tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
 
 
 
 
 
 
 
 
 
 
 
 
1635}
1636
1637static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1638{
1639	struct sock *sk = sock->sk;
1640	DEFINE_WAIT(wait);
1641	long timeo = *timeop;
1642	int err = sock_error(sk);
1643
1644	if (err)
1645		return err;
1646
1647	for (;;) {
1648		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1649		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1650			if (sk->sk_shutdown & RCV_SHUTDOWN) {
1651				err = -ENOTCONN;
1652				break;
1653			}
 
1654			release_sock(sk);
1655			timeo = schedule_timeout(timeo);
 
1656			lock_sock(sk);
 
1657		}
1658		err = 0;
1659		if (!skb_queue_empty(&sk->sk_receive_queue))
1660			break;
1661		err = -EAGAIN;
1662		if (!timeo)
1663			break;
1664		err = sock_intr_errno(timeo);
1665		if (signal_pending(current))
1666			break;
1667
1668		err = sock_error(sk);
1669		if (err)
1670			break;
1671	}
1672	finish_wait(sk_sleep(sk), &wait);
1673	*timeop = timeo;
1674	return err;
1675}
1676
1677/**
1678 * tipc_recvmsg - receive packet-oriented message
 
1679 * @m: descriptor for message info
1680 * @buflen: length of user buffer area
1681 * @flags: receive flags
1682 *
1683 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1684 * If the complete message doesn't fit in user area, truncate it.
1685 *
1686 * Returns size of returned message data, errno otherwise
1687 */
1688static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1689			size_t buflen,	int flags)
1690{
1691	struct sock *sk = sock->sk;
1692	bool connected = !tipc_sk_type_connectionless(sk);
1693	struct tipc_sock *tsk = tipc_sk(sk);
1694	int rc, err, hlen, dlen, copy;
 
1695	struct sk_buff_head xmitq;
1696	struct tipc_msg *hdr;
1697	struct sk_buff *skb;
1698	bool grp_evt;
1699	long timeout;
1700
1701	/* Catch invalid receive requests */
1702	if (unlikely(!buflen))
1703		return -EINVAL;
1704
1705	lock_sock(sk);
1706	if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1707		rc = -ENOTCONN;
1708		goto exit;
1709	}
1710	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1711
1712	/* Step rcv queue to first msg with data or error; wait if necessary */
1713	do {
1714		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1715		if (unlikely(rc))
1716			goto exit;
1717		skb = skb_peek(&sk->sk_receive_queue);
 
1718		hdr = buf_msg(skb);
1719		dlen = msg_data_sz(hdr);
1720		hlen = msg_hdr_sz(hdr);
1721		err = msg_errcode(hdr);
1722		grp_evt = msg_is_grp_evt(hdr);
1723		if (likely(dlen || err))
1724			break;
1725		tsk_advance_rx_queue(sk);
1726	} while (1);
1727
1728	/* Collect msg meta data, including error code and rejected data */
1729	tipc_sk_set_orig_addr(m, skb);
1730	rc = tipc_sk_anc_data_recv(m, hdr, tsk);
1731	if (unlikely(rc))
1732		goto exit;
 
1733
1734	/* Capture data if non-error msg, otherwise just set return value */
1735	if (likely(!err)) {
1736		copy = min_t(int, dlen, buflen);
1737		if (unlikely(copy != dlen))
1738			m->msg_flags |= MSG_TRUNC;
1739		rc = skb_copy_datagram_msg(skb, hlen, m, copy);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1740	} else {
1741		copy = 0;
1742		rc = 0;
1743		if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1744			rc = -ECONNRESET;
 
 
1745	}
1746	if (unlikely(rc))
1747		goto exit;
1748
1749	/* Mark message as group event if applicable */
1750	if (unlikely(grp_evt)) {
1751		if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1752			m->msg_flags |= MSG_EOR;
1753		m->msg_flags |= MSG_OOB;
1754		copy = 0;
1755	}
1756
1757	/* Caption of data or error code/rejected data was successful */
1758	if (unlikely(flags & MSG_PEEK))
1759		goto exit;
1760
1761	/* Send group flow control advertisement when applicable */
1762	if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1763		skb_queue_head_init(&xmitq);
1764		tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1765					  msg_orignode(hdr), msg_origport(hdr),
1766					  &xmitq);
1767		tipc_node_distr_xmit(sock_net(sk), &xmitq);
1768	}
1769
 
 
 
1770	tsk_advance_rx_queue(sk);
1771
1772	if (likely(!connected))
1773		goto exit;
1774
1775	/* Send connection flow control advertisement when applicable */
1776	tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1777	if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1778		tipc_sk_send_ack(tsk);
1779exit:
1780	release_sock(sk);
1781	return rc ? rc : copy;
1782}
1783
1784/**
1785 * tipc_recvstream - receive stream-oriented data
 
1786 * @m: descriptor for message info
1787 * @buflen: total size of user buffer area
1788 * @flags: receive flags
1789 *
1790 * Used for SOCK_STREAM messages only.  If not enough data is available
1791 * will optionally wait for more; never truncates data.
1792 *
1793 * Returns size of returned message data, errno otherwise
1794 */
1795static int tipc_recvstream(struct socket *sock, struct msghdr *m,
1796			   size_t buflen, int flags)
1797{
1798	struct sock *sk = sock->sk;
1799	struct tipc_sock *tsk = tipc_sk(sk);
1800	struct sk_buff *skb;
1801	struct tipc_msg *hdr;
1802	struct tipc_skb_cb *skb_cb;
1803	bool peek = flags & MSG_PEEK;
1804	int offset, required, copy, copied = 0;
1805	int hlen, dlen, err, rc;
1806	long timeout;
1807
1808	/* Catch invalid receive attempts */
1809	if (unlikely(!buflen))
1810		return -EINVAL;
1811
1812	lock_sock(sk);
1813
1814	if (unlikely(sk->sk_state == TIPC_OPEN)) {
1815		rc = -ENOTCONN;
1816		goto exit;
1817	}
1818	required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
1819	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1820
1821	do {
1822		/* Look at first msg in receive queue; wait if necessary */
1823		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1824		if (unlikely(rc))
1825			break;
1826		skb = skb_peek(&sk->sk_receive_queue);
1827		skb_cb = TIPC_SKB_CB(skb);
1828		hdr = buf_msg(skb);
1829		dlen = msg_data_sz(hdr);
1830		hlen = msg_hdr_sz(hdr);
1831		err = msg_errcode(hdr);
1832
1833		/* Discard any empty non-errored (SYN-) message */
1834		if (unlikely(!dlen && !err)) {
1835			tsk_advance_rx_queue(sk);
1836			continue;
1837		}
1838
1839		/* Collect msg meta data, incl. error code and rejected data */
1840		if (!copied) {
1841			tipc_sk_set_orig_addr(m, skb);
1842			rc = tipc_sk_anc_data_recv(m, hdr, tsk);
1843			if (rc)
1844				break;
 
1845		}
1846
1847		/* Copy data if msg ok, otherwise return error/partial data */
1848		if (likely(!err)) {
1849			offset = skb_cb->bytes_read;
1850			copy = min_t(int, dlen - offset, buflen - copied);
1851			rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1852			if (unlikely(rc))
1853				break;
1854			copied += copy;
1855			offset += copy;
1856			if (unlikely(offset < dlen)) {
1857				if (!peek)
1858					skb_cb->bytes_read = offset;
1859				break;
1860			}
1861		} else {
1862			rc = 0;
1863			if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
1864				rc = -ECONNRESET;
1865			if (copied || rc)
1866				break;
1867		}
1868
1869		if (unlikely(peek))
1870			break;
1871
1872		tsk_advance_rx_queue(sk);
1873
1874		/* Send connection flow control advertisement when applicable */
1875		tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1876		if (unlikely(tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE))
1877			tipc_sk_send_ack(tsk);
1878
1879		/* Exit if all requested data or FIN/error received */
1880		if (copied == buflen || err)
1881			break;
1882
1883	} while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
1884exit:
1885	release_sock(sk);
1886	return copied ? copied : rc;
1887}
1888
1889/**
1890 * tipc_write_space - wake up thread if port congestion is released
1891 * @sk: socket
1892 */
1893static void tipc_write_space(struct sock *sk)
1894{
1895	struct socket_wq *wq;
1896
1897	rcu_read_lock();
1898	wq = rcu_dereference(sk->sk_wq);
1899	if (skwq_has_sleeper(wq))
1900		wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
1901						EPOLLWRNORM | EPOLLWRBAND);
1902	rcu_read_unlock();
1903}
1904
1905/**
1906 * tipc_data_ready - wake up threads to indicate messages have been received
1907 * @sk: socket
1908 * @len: the length of messages
1909 */
1910static void tipc_data_ready(struct sock *sk)
1911{
1912	struct socket_wq *wq;
1913
 
 
1914	rcu_read_lock();
1915	wq = rcu_dereference(sk->sk_wq);
1916	if (skwq_has_sleeper(wq))
1917		wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
1918						EPOLLRDNORM | EPOLLRDBAND);
1919	rcu_read_unlock();
1920}
1921
1922static void tipc_sock_destruct(struct sock *sk)
1923{
1924	__skb_queue_purge(&sk->sk_receive_queue);
1925}
1926
1927static void tipc_sk_proto_rcv(struct sock *sk,
1928			      struct sk_buff_head *inputq,
1929			      struct sk_buff_head *xmitq)
1930{
1931	struct sk_buff *skb = __skb_dequeue(inputq);
1932	struct tipc_sock *tsk = tipc_sk(sk);
1933	struct tipc_msg *hdr = buf_msg(skb);
1934	struct tipc_group *grp = tsk->group;
1935	bool wakeup = false;
1936
1937	switch (msg_user(hdr)) {
1938	case CONN_MANAGER:
1939		tipc_sk_conn_proto_rcv(tsk, skb, xmitq);
1940		return;
1941	case SOCK_WAKEUP:
1942		tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
 
 
1943		tsk->cong_link_cnt--;
1944		wakeup = true;
 
1945		break;
1946	case GROUP_PROTOCOL:
1947		tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
1948		break;
1949	case TOP_SRV:
1950		tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
1951				      hdr, inputq, xmitq);
1952		break;
1953	default:
1954		break;
1955	}
1956
1957	if (wakeup)
1958		sk->sk_write_space(sk);
1959
1960	kfree_skb(skb);
1961}
1962
1963/**
1964 * tipc_filter_connect - Handle incoming message for a connection-based socket
1965 * @tsk: TIPC socket
1966 * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1967 *
1968 * Returns true if everything ok, false otherwise
1969 */
1970static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
 
1971{
1972	struct sock *sk = &tsk->sk;
1973	struct net *net = sock_net(sk);
1974	struct tipc_msg *hdr = buf_msg(skb);
1975	u32 pport = msg_origport(hdr);
1976	u32 pnode = msg_orignode(hdr);
 
 
 
 
 
1977
1978	if (unlikely(msg_mcast(hdr)))
1979		return false;
 
1980
1981	switch (sk->sk_state) {
1982	case TIPC_CONNECTING:
1983		/* Accept only ACK or NACK message */
1984		if (unlikely(!msg_connected(hdr))) {
1985			if (pport != tsk_peer_port(tsk) ||
1986			    pnode != tsk_peer_node(tsk))
1987				return false;
1988
1989			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1990			sk->sk_err = ECONNREFUSED;
 
 
1991			sk->sk_state_change(sk);
1992			return true;
 
1993		}
 
 
 
1994
1995		if (unlikely(msg_errcode(hdr))) {
1996			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1997			sk->sk_err = ECONNREFUSED;
1998			sk->sk_state_change(sk);
1999			return true;
2000		}
2001
2002		if (unlikely(!msg_isdata(hdr))) {
2003			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2004			sk->sk_err = EINVAL;
2005			sk->sk_state_change(sk);
2006			return true;
2007		}
2008
2009		tipc_sk_finish_conn(tsk, msg_origport(hdr), msg_orignode(hdr));
2010		msg_set_importance(&tsk->phdr, msg_importance(hdr));
2011
2012		/* If 'ACK+' message, add to socket receive queue */
2013		if (msg_data_sz(hdr))
2014			return true;
2015
2016		/* If empty 'ACK-' message, wake up sleeping connect() */
2017		sk->sk_data_ready(sk);
2018
2019		/* 'ACK-' message is neither accepted nor rejected: */
2020		msg_set_dest_droppable(hdr, 1);
2021		return false;
2022
2023	case TIPC_OPEN:
2024	case TIPC_DISCONNECTING:
2025		break;
2026	case TIPC_LISTEN:
2027		/* Accept only SYN message */
2028		if (!msg_connected(hdr) && !(msg_errcode(hdr)))
 
 
 
2029			return true;
2030		break;
2031	case TIPC_ESTABLISHED:
 
 
2032		/* Accept only connection-based messages sent by peer */
2033		if (unlikely(!tsk_peer_msg(tsk, hdr)))
 
 
 
 
 
 
 
 
 
 
 
 
 
2034			return false;
2035
2036		if (unlikely(msg_errcode(hdr))) {
2037			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2038			/* Let timer expire on it's own */
2039			tipc_node_remove_conn(net, tsk_peer_node(tsk),
2040					      tsk->portid);
2041			sk->sk_state_change(sk);
2042		}
2043		return true;
2044	default:
2045		pr_err("Unknown sk_state %u\n", sk->sk_state);
2046	}
2047
2048	return false;
 
 
 
2049}
2050
2051/**
2052 * rcvbuf_limit - get proper overload limit of socket receive queue
2053 * @sk: socket
2054 * @skb: message
2055 *
2056 * For connection oriented messages, irrespective of importance,
2057 * default queue limit is 2 MB.
2058 *
2059 * For connectionless messages, queue limits are based on message
2060 * importance as follows:
2061 *
2062 * TIPC_LOW_IMPORTANCE       (2 MB)
2063 * TIPC_MEDIUM_IMPORTANCE    (4 MB)
2064 * TIPC_HIGH_IMPORTANCE      (8 MB)
2065 * TIPC_CRITICAL_IMPORTANCE  (16 MB)
2066 *
2067 * Returns overload limit according to corresponding message importance
2068 */
2069static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2070{
2071	struct tipc_sock *tsk = tipc_sk(sk);
2072	struct tipc_msg *hdr = buf_msg(skb);
2073
2074	if (unlikely(msg_in_group(hdr)))
2075		return sk->sk_rcvbuf;
2076
2077	if (unlikely(!msg_connected(hdr)))
2078		return sk->sk_rcvbuf << msg_importance(hdr);
2079
2080	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2081		return sk->sk_rcvbuf;
2082
2083	return FLOWCTL_MSG_LIM;
2084}
2085
2086/**
2087 * tipc_sk_filter_rcv - validate incoming message
2088 * @sk: socket
2089 * @skb: pointer to message.
 
2090 *
2091 * Enqueues message on receive queue if acceptable; optionally handles
2092 * disconnect indication for a connected socket.
2093 *
2094 * Called with socket lock already taken
2095 *
2096 */
2097static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2098			       struct sk_buff_head *xmitq)
2099{
2100	bool sk_conn = !tipc_sk_type_connectionless(sk);
2101	struct tipc_sock *tsk = tipc_sk(sk);
2102	struct tipc_group *grp = tsk->group;
2103	struct tipc_msg *hdr = buf_msg(skb);
2104	struct net *net = sock_net(sk);
2105	struct sk_buff_head inputq;
 
2106	int limit, err = TIPC_OK;
2107
 
2108	TIPC_SKB_CB(skb)->bytes_read = 0;
2109	__skb_queue_head_init(&inputq);
2110	__skb_queue_tail(&inputq, skb);
2111
2112	if (unlikely(!msg_isdata(hdr)))
2113		tipc_sk_proto_rcv(sk, &inputq, xmitq);
2114
2115	if (unlikely(grp))
2116		tipc_group_filter_msg(grp, &inputq, xmitq);
2117
 
 
 
2118	/* Validate and add to receive buffer if there is space */
2119	while ((skb = __skb_dequeue(&inputq))) {
2120		hdr = buf_msg(skb);
2121		limit = rcvbuf_limit(sk, skb);
2122		if ((sk_conn && !tipc_sk_filter_connect(tsk, skb)) ||
2123		    (!sk_conn && msg_connected(hdr)) ||
2124		    (!grp && msg_in_group(hdr)))
2125			err = TIPC_ERR_NO_PORT;
2126		else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
 
 
2127			atomic_inc(&sk->sk_drops);
2128			err = TIPC_ERR_OVERLOAD;
2129		}
2130
2131		if (unlikely(err)) {
2132			tipc_skb_reject(net, err, skb, xmitq);
 
 
 
 
2133			err = TIPC_OK;
2134			continue;
2135		}
2136		__skb_queue_tail(&sk->sk_receive_queue, skb);
2137		skb_set_owner_r(skb, sk);
 
 
2138		sk->sk_data_ready(sk);
2139	}
2140}
2141
2142/**
2143 * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2144 * @sk: socket
2145 * @skb: message
2146 *
2147 * Caller must hold socket lock
2148 */
2149static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2150{
2151	unsigned int before = sk_rmem_alloc_get(sk);
2152	struct sk_buff_head xmitq;
2153	unsigned int added;
2154
2155	__skb_queue_head_init(&xmitq);
2156
2157	tipc_sk_filter_rcv(sk, skb, &xmitq);
2158	added = sk_rmem_alloc_get(sk) - before;
2159	atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2160
2161	/* Send pending response/rejected messages, if any */
2162	tipc_node_distr_xmit(sock_net(sk), &xmitq);
2163	return 0;
2164}
2165
2166/**
2167 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2168 *                   inputq and try adding them to socket or backlog queue
2169 * @inputq: list of incoming buffers with potentially different destinations
2170 * @sk: socket where the buffers should be enqueued
2171 * @dport: port number for the socket
 
2172 *
2173 * Caller must hold socket lock
2174 */
2175static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2176			    u32 dport, struct sk_buff_head *xmitq)
2177{
2178	unsigned long time_limit = jiffies + 2;
2179	struct sk_buff *skb;
2180	unsigned int lim;
2181	atomic_t *dcnt;
2182	u32 onode;
2183
2184	while (skb_queue_len(inputq)) {
2185		if (unlikely(time_after_eq(jiffies, time_limit)))
2186			return;
2187
2188		skb = tipc_skb_dequeue(inputq, dport);
2189		if (unlikely(!skb))
2190			return;
2191
2192		/* Add message directly to receive queue if possible */
2193		if (!sock_owned_by_user(sk)) {
2194			tipc_sk_filter_rcv(sk, skb, xmitq);
2195			continue;
2196		}
2197
2198		/* Try backlog, compensating for double-counted bytes */
2199		dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2200		if (!sk->sk_backlog.len)
2201			atomic_set(dcnt, 0);
2202		lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2203		if (likely(!sk_add_backlog(sk, skb, lim)))
 
 
2204			continue;
 
2205
 
2206		/* Overload => reject message back to sender */
2207		onode = tipc_own_addr(sock_net(sk));
2208		atomic_inc(&sk->sk_drops);
2209		if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD))
 
 
2210			__skb_queue_tail(xmitq, skb);
 
2211		break;
2212	}
2213}
2214
2215/**
2216 * tipc_sk_rcv - handle a chain of incoming buffers
 
2217 * @inputq: buffer list containing the buffers
2218 * Consumes all buffers in list until inputq is empty
2219 * Note: may be called in multiple threads referring to the same queue
2220 */
2221void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2222{
2223	struct sk_buff_head xmitq;
2224	u32 dnode, dport = 0;
2225	int err;
2226	struct tipc_sock *tsk;
2227	struct sock *sk;
2228	struct sk_buff *skb;
2229
2230	__skb_queue_head_init(&xmitq);
2231	while (skb_queue_len(inputq)) {
2232		dport = tipc_skb_peek_port(inputq, dport);
2233		tsk = tipc_sk_lookup(net, dport);
2234
2235		if (likely(tsk)) {
2236			sk = &tsk->sk;
2237			if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2238				tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2239				spin_unlock_bh(&sk->sk_lock.slock);
2240			}
2241			/* Send pending response/rejected messages, if any */
2242			tipc_node_distr_xmit(sock_net(sk), &xmitq);
2243			sock_put(sk);
2244			continue;
2245		}
2246		/* No destination socket => dequeue skb if still there */
2247		skb = tipc_skb_dequeue(inputq, dport);
2248		if (!skb)
2249			return;
2250
2251		/* Try secondary lookup if unresolved named message */
2252		err = TIPC_ERR_NO_PORT;
2253		if (tipc_msg_lookup_dest(net, skb, &err))
2254			goto xmit;
2255
2256		/* Prepare for message rejection */
2257		if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2258			continue;
 
 
2259xmit:
2260		dnode = msg_destnode(buf_msg(skb));
2261		tipc_node_xmit_skb(net, skb, dnode, dport);
2262	}
2263}
2264
2265static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2266{
2267	DEFINE_WAIT_FUNC(wait, woken_wake_function);
2268	struct sock *sk = sock->sk;
2269	int done;
2270
2271	do {
2272		int err = sock_error(sk);
2273		if (err)
2274			return err;
2275		if (!*timeo_p)
2276			return -ETIMEDOUT;
2277		if (signal_pending(current))
2278			return sock_intr_errno(*timeo_p);
 
 
2279
2280		add_wait_queue(sk_sleep(sk), &wait);
2281		done = sk_wait_event(sk, timeo_p,
2282				     sk->sk_state != TIPC_CONNECTING, &wait);
2283		remove_wait_queue(sk_sleep(sk), &wait);
2284	} while (!done);
2285	return 0;
2286}
2287
 
 
 
 
 
 
 
 
 
 
2288/**
2289 * tipc_connect - establish a connection to another TIPC port
2290 * @sock: socket structure
2291 * @dest: socket address for destination port
2292 * @destlen: size of socket address data structure
2293 * @flags: file-related flags associated with socket
2294 *
2295 * Returns 0 on success, errno otherwise
2296 */
2297static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2298			int destlen, int flags)
2299{
2300	struct sock *sk = sock->sk;
2301	struct tipc_sock *tsk = tipc_sk(sk);
2302	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2303	struct msghdr m = {NULL,};
2304	long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2305	int previous;
2306	int res = 0;
2307
2308	if (destlen != sizeof(struct sockaddr_tipc))
2309		return -EINVAL;
2310
2311	lock_sock(sk);
2312
2313	if (tsk->group) {
2314		res = -EINVAL;
2315		goto exit;
2316	}
2317
2318	if (dst->family == AF_UNSPEC) {
2319		memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2320		if (!tipc_sk_type_connectionless(sk))
2321			res = -EINVAL;
2322		goto exit;
2323	} else if (dst->family != AF_TIPC) {
2324		res = -EINVAL;
2325	}
2326	if (dst->addrtype != TIPC_ADDR_ID && dst->addrtype != TIPC_ADDR_NAME)
2327		res = -EINVAL;
2328	if (res)
2329		goto exit;
2330
2331	/* DGRAM/RDM connect(), just save the destaddr */
2332	if (tipc_sk_type_connectionless(sk)) {
2333		memcpy(&tsk->peer, dest, destlen);
2334		goto exit;
 
 
 
2335	}
2336
2337	previous = sk->sk_state;
2338
2339	switch (sk->sk_state) {
2340	case TIPC_OPEN:
2341		/* Send a 'SYN-' to destination */
2342		m.msg_name = dest;
2343		m.msg_namelen = destlen;
 
2344
2345		/* If connect is in non-blocking case, set MSG_DONTWAIT to
2346		 * indicate send_msg() is never blocked.
2347		 */
2348		if (!timeout)
2349			m.msg_flags = MSG_DONTWAIT;
2350
2351		res = __tipc_sendmsg(sock, &m, 0);
2352		if ((res < 0) && (res != -EWOULDBLOCK))
2353			goto exit;
2354
2355		/* Just entered TIPC_CONNECTING state; the only
2356		 * difference is that return value in non-blocking
2357		 * case is EINPROGRESS, rather than EALREADY.
2358		 */
2359		res = -EINPROGRESS;
2360		/* fall thru' */
2361	case TIPC_CONNECTING:
2362		if (!timeout) {
2363			if (previous == TIPC_CONNECTING)
2364				res = -EALREADY;
2365			goto exit;
2366		}
2367		timeout = msecs_to_jiffies(timeout);
2368		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2369		res = tipc_wait_for_connect(sock, &timeout);
2370		break;
2371	case TIPC_ESTABLISHED:
2372		res = -EISCONN;
2373		break;
2374	default:
2375		res = -EINVAL;
2376	}
2377
2378exit:
2379	release_sock(sk);
2380	return res;
2381}
2382
2383/**
2384 * tipc_listen - allow socket to listen for incoming connections
2385 * @sock: socket structure
2386 * @len: (unused)
2387 *
2388 * Returns 0 on success, errno otherwise
2389 */
2390static int tipc_listen(struct socket *sock, int len)
2391{
2392	struct sock *sk = sock->sk;
2393	int res;
2394
2395	lock_sock(sk);
2396	res = tipc_set_sk_state(sk, TIPC_LISTEN);
2397	release_sock(sk);
2398
2399	return res;
2400}
2401
2402static int tipc_wait_for_accept(struct socket *sock, long timeo)
2403{
2404	struct sock *sk = sock->sk;
2405	DEFINE_WAIT(wait);
2406	int err;
2407
2408	/* True wake-one mechanism for incoming connections: only
2409	 * one process gets woken up, not the 'whole herd'.
2410	 * Since we do not 'race & poll' for established sockets
2411	 * anymore, the common case will execute the loop only once.
2412	*/
2413	for (;;) {
2414		prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2415					  TASK_INTERRUPTIBLE);
2416		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
 
2417			release_sock(sk);
2418			timeo = schedule_timeout(timeo);
2419			lock_sock(sk);
 
2420		}
2421		err = 0;
2422		if (!skb_queue_empty(&sk->sk_receive_queue))
2423			break;
2424		err = -EAGAIN;
2425		if (!timeo)
2426			break;
2427		err = sock_intr_errno(timeo);
2428		if (signal_pending(current))
2429			break;
2430	}
2431	finish_wait(sk_sleep(sk), &wait);
2432	return err;
2433}
2434
2435/**
2436 * tipc_accept - wait for connection request
2437 * @sock: listening socket
2438 * @newsock: new socket that is to be connected
2439 * @flags: file-related flags associated with socket
2440 *
2441 * Returns 0 on success, errno otherwise
2442 */
2443static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2444		       bool kern)
2445{
2446	struct sock *new_sk, *sk = sock->sk;
2447	struct sk_buff *buf;
2448	struct tipc_sock *new_tsock;
 
2449	struct tipc_msg *msg;
 
2450	long timeo;
2451	int res;
2452
2453	lock_sock(sk);
2454
2455	if (sk->sk_state != TIPC_LISTEN) {
2456		res = -EINVAL;
2457		goto exit;
2458	}
2459	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2460	res = tipc_wait_for_accept(sock, timeo);
2461	if (res)
2462		goto exit;
2463
2464	buf = skb_peek(&sk->sk_receive_queue);
2465
2466	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2467	if (res)
2468		goto exit;
2469	security_sk_clone(sock->sk, new_sock->sk);
2470
2471	new_sk = new_sock->sk;
2472	new_tsock = tipc_sk(new_sk);
2473	msg = buf_msg(buf);
2474
2475	/* we lock on new_sk; but lockdep sees the lock on sk */
2476	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2477
2478	/*
2479	 * Reject any stray messages received by new socket
2480	 * before the socket lock was taken (very, very unlikely)
2481	 */
2482	tsk_rej_rx_queue(new_sk);
2483
2484	/* Connect new socket to it's peer */
2485	tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2486
2487	tsk_set_importance(new_tsock, msg_importance(msg));
2488	if (msg_named(msg)) {
2489		new_tsock->conn_type = msg_nametype(msg);
2490		new_tsock->conn_instance = msg_nameinst(msg);
 
2491	}
2492
2493	/*
2494	 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2495	 * Respond to 'SYN+' by queuing it on new socket.
2496	 */
2497	if (!msg_data_sz(msg)) {
2498		struct msghdr m = {NULL,};
2499
2500		tsk_advance_rx_queue(sk);
2501		__tipc_sendstream(new_sock, &m, 0);
2502	} else {
2503		__skb_dequeue(&sk->sk_receive_queue);
2504		__skb_queue_head(&new_sk->sk_receive_queue, buf);
2505		skb_set_owner_r(buf, new_sk);
2506	}
 
 
2507	release_sock(new_sk);
2508exit:
2509	release_sock(sk);
2510	return res;
2511}
2512
2513/**
2514 * tipc_shutdown - shutdown socket connection
2515 * @sock: socket structure
2516 * @how: direction to close (must be SHUT_RDWR)
2517 *
2518 * Terminates connection (if necessary), then purges socket's receive queue.
2519 *
2520 * Returns 0 on success, errno otherwise
2521 */
2522static int tipc_shutdown(struct socket *sock, int how)
2523{
2524	struct sock *sk = sock->sk;
2525	int res;
2526
2527	if (how != SHUT_RDWR)
2528		return -EINVAL;
2529
2530	lock_sock(sk);
2531
 
2532	__tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2533	sk->sk_shutdown = SEND_SHUTDOWN;
2534
2535	if (sk->sk_state == TIPC_DISCONNECTING) {
2536		/* Discard any unreceived messages */
2537		__skb_queue_purge(&sk->sk_receive_queue);
2538
2539		/* Wake up anyone sleeping in poll */
2540		sk->sk_state_change(sk);
2541		res = 0;
2542	} else {
2543		res = -ENOTCONN;
2544	}
 
 
2545
2546	release_sock(sk);
2547	return res;
2548}
2549
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2550static void tipc_sk_timeout(struct timer_list *t)
2551{
2552	struct sock *sk = from_timer(sk, t, sk_timer);
2553	struct tipc_sock *tsk = tipc_sk(sk);
2554	u32 peer_port = tsk_peer_port(tsk);
2555	u32 peer_node = tsk_peer_node(tsk);
2556	u32 own_node = tsk_own_node(tsk);
2557	u32 own_port = tsk->portid;
2558	struct net *net = sock_net(sk);
2559	struct sk_buff *skb = NULL;
2560
 
2561	bh_lock_sock(sk);
2562	if (!tipc_sk_connected(sk))
2563		goto exit;
2564
2565	/* Try again later if socket is busy */
2566	if (sock_owned_by_user(sk)) {
2567		sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2568		goto exit;
 
 
2569	}
2570
2571	if (tsk->probe_unacked) {
2572		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2573		tipc_node_remove_conn(net, peer_node, peer_port);
2574		sk->sk_state_change(sk);
2575		goto exit;
 
 
 
 
 
 
 
 
 
2576	}
2577	/* Send new probe */
2578	skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2579			      peer_node, own_node, peer_port, own_port,
2580			      TIPC_OK);
2581	tsk->probe_unacked = true;
2582	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2583exit:
2584	bh_unlock_sock(sk);
2585	if (skb)
2586		tipc_node_xmit_skb(net, skb, peer_node, own_port);
2587	sock_put(sk);
2588}
2589
2590static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2591			   struct tipc_name_seq const *seq)
2592{
2593	struct sock *sk = &tsk->sk;
2594	struct net *net = sock_net(sk);
2595	struct publication *publ;
 
2596	u32 key;
2597
2598	if (scope != TIPC_NODE_SCOPE)
2599		scope = TIPC_CLUSTER_SCOPE;
2600
2601	if (tipc_sk_connected(sk))
2602		return -EINVAL;
2603	key = tsk->portid + tsk->pub_count + 1;
2604	if (key == tsk->portid)
2605		return -EADDRINUSE;
2606
2607	publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2608				    scope, tsk->portid, key);
2609	if (unlikely(!publ))
2610		return -EINVAL;
2611
2612	list_add(&publ->binding_sock, &tsk->publications);
2613	tsk->pub_count++;
2614	tsk->published = 1;
2615	return 0;
2616}
2617
2618static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2619			    struct tipc_name_seq const *seq)
2620{
2621	struct net *net = sock_net(&tsk->sk);
2622	struct publication *publ;
2623	struct publication *safe;
2624	int rc = -EINVAL;
2625
2626	if (scope != TIPC_NODE_SCOPE)
2627		scope = TIPC_CLUSTER_SCOPE;
2628
2629	list_for_each_entry_safe(publ, safe, &tsk->publications, binding_sock) {
2630		if (seq) {
2631			if (publ->scope != scope)
2632				continue;
2633			if (publ->type != seq->type)
2634				continue;
2635			if (publ->lower != seq->lower)
2636				continue;
2637			if (publ->upper != seq->upper)
2638				break;
2639			tipc_nametbl_withdraw(net, publ->type, publ->lower,
2640					      publ->upper, publ->key);
2641			rc = 0;
2642			break;
2643		}
2644		tipc_nametbl_withdraw(net, publ->type, publ->lower,
2645				      publ->upper, publ->key);
2646		rc = 0;
 
2647	}
2648	if (list_empty(&tsk->publications))
2649		tsk->published = 0;
 
 
2650	return rc;
2651}
2652
2653/* tipc_sk_reinit: set non-zero address in all existing sockets
2654 *                 when we go from standalone to network mode.
2655 */
2656void tipc_sk_reinit(struct net *net)
2657{
2658	struct tipc_net *tn = net_generic(net, tipc_net_id);
2659	struct rhashtable_iter iter;
2660	struct tipc_sock *tsk;
2661	struct tipc_msg *msg;
2662
2663	rhashtable_walk_enter(&tn->sk_rht, &iter);
2664
2665	do {
2666		rhashtable_walk_start(&iter);
2667
2668		while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2669			spin_lock_bh(&tsk->sk.sk_lock.slock);
 
 
2670			msg = &tsk->phdr;
2671			msg_set_prevnode(msg, tipc_own_addr(net));
2672			msg_set_orignode(msg, tipc_own_addr(net));
2673			spin_unlock_bh(&tsk->sk.sk_lock.slock);
 
 
2674		}
2675
2676		rhashtable_walk_stop(&iter);
2677	} while (tsk == ERR_PTR(-EAGAIN));
 
 
2678}
2679
2680static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2681{
2682	struct tipc_net *tn = net_generic(net, tipc_net_id);
2683	struct tipc_sock *tsk;
2684
2685	rcu_read_lock();
2686	tsk = rhashtable_lookup_fast(&tn->sk_rht, &portid, tsk_rht_params);
2687	if (tsk)
2688		sock_hold(&tsk->sk);
2689	rcu_read_unlock();
2690
2691	return tsk;
2692}
2693
2694static int tipc_sk_insert(struct tipc_sock *tsk)
2695{
2696	struct sock *sk = &tsk->sk;
2697	struct net *net = sock_net(sk);
2698	struct tipc_net *tn = net_generic(net, tipc_net_id);
2699	u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2700	u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2701
2702	while (remaining--) {
2703		portid++;
2704		if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2705			portid = TIPC_MIN_PORT;
2706		tsk->portid = portid;
2707		sock_hold(&tsk->sk);
2708		if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2709						   tsk_rht_params))
2710			return 0;
2711		sock_put(&tsk->sk);
2712	}
2713
2714	return -1;
2715}
2716
2717static void tipc_sk_remove(struct tipc_sock *tsk)
2718{
2719	struct sock *sk = &tsk->sk;
2720	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2721
2722	if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2723		WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
2724		__sock_put(sk);
2725	}
2726}
2727
2728static const struct rhashtable_params tsk_rht_params = {
2729	.nelem_hint = 192,
2730	.head_offset = offsetof(struct tipc_sock, node),
2731	.key_offset = offsetof(struct tipc_sock, portid),
2732	.key_len = sizeof(u32), /* portid */
2733	.max_size = 1048576,
2734	.min_size = 256,
2735	.automatic_shrinking = true,
2736};
2737
2738int tipc_sk_rht_init(struct net *net)
2739{
2740	struct tipc_net *tn = net_generic(net, tipc_net_id);
2741
2742	return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2743}
2744
2745void tipc_sk_rht_destroy(struct net *net)
2746{
2747	struct tipc_net *tn = net_generic(net, tipc_net_id);
2748
2749	/* Wait for socket readers to complete */
2750	synchronize_net();
2751
2752	rhashtable_destroy(&tn->sk_rht);
2753}
2754
2755static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
2756{
2757	struct net *net = sock_net(&tsk->sk);
2758	struct tipc_group *grp = tsk->group;
2759	struct tipc_msg *hdr = &tsk->phdr;
2760	struct tipc_name_seq seq;
2761	int rc;
2762
2763	if (mreq->type < TIPC_RESERVED_TYPES)
2764		return -EACCES;
2765	if (mreq->scope > TIPC_NODE_SCOPE)
2766		return -EINVAL;
 
 
2767	if (grp)
2768		return -EACCES;
2769	grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
2770	if (!grp)
2771		return -ENOMEM;
2772	tsk->group = grp;
2773	msg_set_lookup_scope(hdr, mreq->scope);
2774	msg_set_nametype(hdr, mreq->type);
2775	msg_set_dest_droppable(hdr, true);
2776	seq.type = mreq->type;
2777	seq.lower = mreq->instance;
2778	seq.upper = seq.lower;
2779	tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope);
2780	rc = tipc_sk_publish(tsk, mreq->scope, &seq);
2781	if (rc) {
2782		tipc_group_delete(net, grp);
2783		tsk->group = NULL;
2784		return rc;
2785	}
2786	/* Eliminate any risk that a broadcast overtakes sent JOINs */
2787	tsk->mc_method.rcast = true;
2788	tsk->mc_method.mandatory = true;
2789	tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
2790	return rc;
2791}
2792
2793static int tipc_sk_leave(struct tipc_sock *tsk)
2794{
2795	struct net *net = sock_net(&tsk->sk);
2796	struct tipc_group *grp = tsk->group;
2797	struct tipc_name_seq seq;
2798	int scope;
2799
2800	if (!grp)
2801		return -EINVAL;
2802	tipc_group_self(grp, &seq, &scope);
 
 
2803	tipc_group_delete(net, grp);
2804	tsk->group = NULL;
2805	tipc_sk_withdraw(tsk, scope, &seq);
2806	return 0;
2807}
2808
2809/**
2810 * tipc_setsockopt - set socket option
2811 * @sock: socket structure
2812 * @lvl: option level
2813 * @opt: option identifier
2814 * @ov: pointer to new option value
2815 * @ol: length of option value
2816 *
2817 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2818 * (to ease compatibility).
2819 *
2820 * Returns 0 on success, errno otherwise
2821 */
2822static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2823			   char __user *ov, unsigned int ol)
2824{
2825	struct sock *sk = sock->sk;
2826	struct tipc_sock *tsk = tipc_sk(sk);
2827	struct tipc_group_req mreq;
2828	u32 value = 0;
2829	int res = 0;
2830
2831	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2832		return 0;
2833	if (lvl != SOL_TIPC)
2834		return -ENOPROTOOPT;
2835
2836	switch (opt) {
2837	case TIPC_IMPORTANCE:
2838	case TIPC_SRC_DROPPABLE:
2839	case TIPC_DEST_DROPPABLE:
2840	case TIPC_CONN_TIMEOUT:
 
2841		if (ol < sizeof(value))
2842			return -EINVAL;
2843		if (get_user(value, (u32 __user *)ov))
2844			return -EFAULT;
2845		break;
2846	case TIPC_GROUP_JOIN:
2847		if (ol < sizeof(mreq))
2848			return -EINVAL;
2849		if (copy_from_user(&mreq, ov, sizeof(mreq)))
2850			return -EFAULT;
2851		break;
2852	default:
2853		if (ov || ol)
2854			return -EINVAL;
2855	}
2856
2857	lock_sock(sk);
2858
2859	switch (opt) {
2860	case TIPC_IMPORTANCE:
2861		res = tsk_set_importance(tsk, value);
2862		break;
2863	case TIPC_SRC_DROPPABLE:
2864		if (sock->type != SOCK_STREAM)
2865			tsk_set_unreliable(tsk, value);
2866		else
2867			res = -ENOPROTOOPT;
2868		break;
2869	case TIPC_DEST_DROPPABLE:
2870		tsk_set_unreturnable(tsk, value);
2871		break;
2872	case TIPC_CONN_TIMEOUT:
2873		tipc_sk(sk)->conn_timeout = value;
2874		break;
2875	case TIPC_MCAST_BROADCAST:
2876		tsk->mc_method.rcast = false;
2877		tsk->mc_method.mandatory = true;
2878		break;
2879	case TIPC_MCAST_REPLICAST:
2880		tsk->mc_method.rcast = true;
2881		tsk->mc_method.mandatory = true;
2882		break;
2883	case TIPC_GROUP_JOIN:
2884		res = tipc_sk_join(tsk, &mreq);
2885		break;
2886	case TIPC_GROUP_LEAVE:
2887		res = tipc_sk_leave(tsk);
2888		break;
 
 
 
 
2889	default:
2890		res = -EINVAL;
2891	}
2892
2893	release_sock(sk);
2894
2895	return res;
2896}
2897
2898/**
2899 * tipc_getsockopt - get socket option
2900 * @sock: socket structure
2901 * @lvl: option level
2902 * @opt: option identifier
2903 * @ov: receptacle for option value
2904 * @ol: receptacle for length of option value
2905 *
2906 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2907 * (to ease compatibility).
2908 *
2909 * Returns 0 on success, errno otherwise
2910 */
2911static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2912			   char __user *ov, int __user *ol)
2913{
2914	struct sock *sk = sock->sk;
2915	struct tipc_sock *tsk = tipc_sk(sk);
2916	struct tipc_name_seq seq;
2917	int len, scope;
2918	u32 value;
2919	int res;
2920
2921	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2922		return put_user(0, ol);
2923	if (lvl != SOL_TIPC)
2924		return -ENOPROTOOPT;
2925	res = get_user(len, ol);
2926	if (res)
2927		return res;
2928
2929	lock_sock(sk);
2930
2931	switch (opt) {
2932	case TIPC_IMPORTANCE:
2933		value = tsk_importance(tsk);
2934		break;
2935	case TIPC_SRC_DROPPABLE:
2936		value = tsk_unreliable(tsk);
2937		break;
2938	case TIPC_DEST_DROPPABLE:
2939		value = tsk_unreturnable(tsk);
2940		break;
2941	case TIPC_CONN_TIMEOUT:
2942		value = tsk->conn_timeout;
2943		/* no need to set "res", since already 0 at this point */
2944		break;
2945	case TIPC_NODE_RECVQ_DEPTH:
2946		value = 0; /* was tipc_queue_size, now obsolete */
2947		break;
2948	case TIPC_SOCK_RECVQ_DEPTH:
2949		value = skb_queue_len(&sk->sk_receive_queue);
2950		break;
 
 
 
2951	case TIPC_GROUP_JOIN:
2952		seq.type = 0;
2953		if (tsk->group)
2954			tipc_group_self(tsk->group, &seq, &scope);
2955		value = seq.type;
2956		break;
2957	default:
2958		res = -EINVAL;
2959	}
2960
2961	release_sock(sk);
2962
2963	if (res)
2964		return res;	/* "get" failed */
2965
2966	if (len < sizeof(value))
2967		return -EINVAL;
2968
2969	if (copy_to_user(ov, &value, sizeof(value)))
2970		return -EFAULT;
2971
2972	return put_user(sizeof(value), ol);
2973}
2974
2975static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2976{
2977	struct sock *sk = sock->sk;
 
2978	struct tipc_sioc_ln_req lnr;
2979	void __user *argp = (void __user *)arg;
2980
2981	switch (cmd) {
2982	case SIOCGETLINKNAME:
2983		if (copy_from_user(&lnr, argp, sizeof(lnr)))
2984			return -EFAULT;
2985		if (!tipc_node_get_linkname(sock_net(sk),
2986					    lnr.bearer_id & 0xffff, lnr.peer,
2987					    lnr.linkname, TIPC_MAX_LINK_NAME)) {
2988			if (copy_to_user(argp, &lnr, sizeof(lnr)))
2989				return -EFAULT;
2990			return 0;
2991		}
2992		return -EADDRNOTAVAIL;
 
 
 
 
 
 
 
 
2993	default:
2994		return -ENOIOCTLCMD;
2995	}
2996}
2997
2998static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
2999{
3000	struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3001	struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3002	u32 onode = tipc_own_addr(sock_net(sock1->sk));
3003
3004	tsk1->peer.family = AF_TIPC;
3005	tsk1->peer.addrtype = TIPC_ADDR_ID;
3006	tsk1->peer.scope = TIPC_NODE_SCOPE;
3007	tsk1->peer.addr.id.ref = tsk2->portid;
3008	tsk1->peer.addr.id.node = onode;
3009	tsk2->peer.family = AF_TIPC;
3010	tsk2->peer.addrtype = TIPC_ADDR_ID;
3011	tsk2->peer.scope = TIPC_NODE_SCOPE;
3012	tsk2->peer.addr.id.ref = tsk1->portid;
3013	tsk2->peer.addr.id.node = onode;
3014
3015	tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3016	tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3017	return 0;
3018}
3019
3020/* Protocol switches for the various types of TIPC sockets */
3021
3022static const struct proto_ops msg_ops = {
3023	.owner		= THIS_MODULE,
3024	.family		= AF_TIPC,
3025	.release	= tipc_release,
3026	.bind		= tipc_bind,
3027	.connect	= tipc_connect,
3028	.socketpair	= tipc_socketpair,
3029	.accept		= sock_no_accept,
3030	.getname	= tipc_getname,
3031	.poll		= tipc_poll,
3032	.ioctl		= tipc_ioctl,
3033	.listen		= sock_no_listen,
3034	.shutdown	= tipc_shutdown,
3035	.setsockopt	= tipc_setsockopt,
3036	.getsockopt	= tipc_getsockopt,
3037	.sendmsg	= tipc_sendmsg,
3038	.recvmsg	= tipc_recvmsg,
3039	.mmap		= sock_no_mmap,
3040	.sendpage	= sock_no_sendpage
3041};
3042
3043static const struct proto_ops packet_ops = {
3044	.owner		= THIS_MODULE,
3045	.family		= AF_TIPC,
3046	.release	= tipc_release,
3047	.bind		= tipc_bind,
3048	.connect	= tipc_connect,
3049	.socketpair	= tipc_socketpair,
3050	.accept		= tipc_accept,
3051	.getname	= tipc_getname,
3052	.poll		= tipc_poll,
3053	.ioctl		= tipc_ioctl,
3054	.listen		= tipc_listen,
3055	.shutdown	= tipc_shutdown,
3056	.setsockopt	= tipc_setsockopt,
3057	.getsockopt	= tipc_getsockopt,
3058	.sendmsg	= tipc_send_packet,
3059	.recvmsg	= tipc_recvmsg,
3060	.mmap		= sock_no_mmap,
3061	.sendpage	= sock_no_sendpage
3062};
3063
3064static const struct proto_ops stream_ops = {
3065	.owner		= THIS_MODULE,
3066	.family		= AF_TIPC,
3067	.release	= tipc_release,
3068	.bind		= tipc_bind,
3069	.connect	= tipc_connect,
3070	.socketpair	= tipc_socketpair,
3071	.accept		= tipc_accept,
3072	.getname	= tipc_getname,
3073	.poll		= tipc_poll,
3074	.ioctl		= tipc_ioctl,
3075	.listen		= tipc_listen,
3076	.shutdown	= tipc_shutdown,
3077	.setsockopt	= tipc_setsockopt,
3078	.getsockopt	= tipc_getsockopt,
3079	.sendmsg	= tipc_sendstream,
3080	.recvmsg	= tipc_recvstream,
3081	.mmap		= sock_no_mmap,
3082	.sendpage	= sock_no_sendpage
3083};
3084
3085static const struct net_proto_family tipc_family_ops = {
3086	.owner		= THIS_MODULE,
3087	.family		= AF_TIPC,
3088	.create		= tipc_sk_create
3089};
3090
3091static struct proto tipc_proto = {
3092	.name		= "TIPC",
3093	.owner		= THIS_MODULE,
3094	.obj_size	= sizeof(struct tipc_sock),
3095	.sysctl_rmem	= sysctl_tipc_rmem
3096};
3097
3098/**
3099 * tipc_socket_init - initialize TIPC socket interface
3100 *
3101 * Returns 0 on success, errno otherwise
3102 */
3103int tipc_socket_init(void)
3104{
3105	int res;
3106
3107	res = proto_register(&tipc_proto, 1);
3108	if (res) {
3109		pr_err("Failed to register TIPC protocol type\n");
3110		goto out;
3111	}
3112
3113	res = sock_register(&tipc_family_ops);
3114	if (res) {
3115		pr_err("Failed to register TIPC socket type\n");
3116		proto_unregister(&tipc_proto);
3117		goto out;
3118	}
3119 out:
3120	return res;
3121}
3122
3123/**
3124 * tipc_socket_stop - stop TIPC socket interface
3125 */
3126void tipc_socket_stop(void)
3127{
3128	sock_unregister(tipc_family_ops.family);
3129	proto_unregister(&tipc_proto);
3130}
3131
3132/* Caller should hold socket lock for the passed tipc socket. */
3133static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3134{
3135	u32 peer_node;
3136	u32 peer_port;
3137	struct nlattr *nest;
3138
3139	peer_node = tsk_peer_node(tsk);
3140	peer_port = tsk_peer_port(tsk);
3141
3142	nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
 
 
 
3143
3144	if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3145		goto msg_full;
3146	if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3147		goto msg_full;
3148
3149	if (tsk->conn_type != 0) {
3150		if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3151			goto msg_full;
3152		if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3153			goto msg_full;
3154		if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3155			goto msg_full;
3156	}
3157	nla_nest_end(skb, nest);
3158
3159	return 0;
3160
3161msg_full:
3162	nla_nest_cancel(skb, nest);
3163
3164	return -EMSGSIZE;
3165}
3166
3167static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3168			  *tsk)
3169{
3170	struct net *net = sock_net(skb->sk);
3171	struct sock *sk = &tsk->sk;
3172
3173	if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3174	    nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3175		return -EMSGSIZE;
3176
3177	if (tipc_sk_connected(sk)) {
3178		if (__tipc_nl_add_sk_con(skb, tsk))
3179			return -EMSGSIZE;
3180	} else if (!list_empty(&tsk->publications)) {
3181		if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3182			return -EMSGSIZE;
3183	}
3184	return 0;
3185}
3186
3187/* Caller should hold socket lock for the passed tipc socket. */
3188static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3189			    struct tipc_sock *tsk)
3190{
3191	struct nlattr *attrs;
3192	void *hdr;
3193
3194	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3195			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3196	if (!hdr)
3197		goto msg_cancel;
3198
3199	attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
3200	if (!attrs)
3201		goto genlmsg_cancel;
3202
3203	if (__tipc_nl_add_sk_info(skb, tsk))
3204		goto attr_msg_cancel;
3205
3206	nla_nest_end(skb, attrs);
3207	genlmsg_end(skb, hdr);
3208
3209	return 0;
3210
3211attr_msg_cancel:
3212	nla_nest_cancel(skb, attrs);
3213genlmsg_cancel:
3214	genlmsg_cancel(skb, hdr);
3215msg_cancel:
3216	return -EMSGSIZE;
3217}
3218
3219int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3220		    int (*skb_handler)(struct sk_buff *skb,
3221				       struct netlink_callback *cb,
3222				       struct tipc_sock *tsk))
3223{
3224	struct net *net = sock_net(skb->sk);
3225	struct tipc_net *tn = tipc_net(net);
3226	const struct bucket_table *tbl;
3227	u32 prev_portid = cb->args[1];
3228	u32 tbl_id = cb->args[0];
3229	struct rhash_head *pos;
3230	struct tipc_sock *tsk;
3231	int err;
3232
3233	rcu_read_lock();
3234	tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
3235	for (; tbl_id < tbl->size; tbl_id++) {
3236		rht_for_each_entry_rcu(tsk, pos, tbl, tbl_id, node) {
3237			spin_lock_bh(&tsk->sk.sk_lock.slock);
3238			if (prev_portid && prev_portid != tsk->portid) {
3239				spin_unlock_bh(&tsk->sk.sk_lock.slock);
3240				continue;
3241			}
 
3242
3243			err = skb_handler(skb, cb, tsk);
3244			if (err) {
3245				prev_portid = tsk->portid;
3246				spin_unlock_bh(&tsk->sk.sk_lock.slock);
3247				goto out;
3248			}
3249
3250			prev_portid = 0;
3251			spin_unlock_bh(&tsk->sk.sk_lock.slock);
3252		}
 
 
 
3253	}
 
3254out:
3255	rcu_read_unlock();
3256	cb->args[0] = tbl_id;
3257	cb->args[1] = prev_portid;
3258
3259	return skb->len;
3260}
3261EXPORT_SYMBOL(tipc_nl_sk_walk);
3262
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3263int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3264			   struct tipc_sock *tsk, u32 sk_filter_state,
3265			   u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3266{
3267	struct sock *sk = &tsk->sk;
3268	struct nlattr *attrs;
3269	struct nlattr *stat;
3270
3271	/*filter response w.r.t sk_state*/
3272	if (!(sk_filter_state & (1 << sk->sk_state)))
3273		return 0;
3274
3275	attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
3276	if (!attrs)
3277		goto msg_cancel;
3278
3279	if (__tipc_nl_add_sk_info(skb, tsk))
3280		goto attr_msg_cancel;
3281
3282	if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3283	    nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3284	    nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3285	    nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3286			from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3287					 sock_i_uid(sk))) ||
3288	    nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3289			      tipc_diag_gen_cookie(sk),
3290			      TIPC_NLA_SOCK_PAD))
3291		goto attr_msg_cancel;
3292
3293	stat = nla_nest_start(skb, TIPC_NLA_SOCK_STAT);
3294	if (!stat)
3295		goto attr_msg_cancel;
3296
3297	if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3298			skb_queue_len(&sk->sk_receive_queue)) ||
3299	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3300			skb_queue_len(&sk->sk_write_queue)) ||
3301	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3302			atomic_read(&sk->sk_drops)))
3303		goto stat_msg_cancel;
3304
3305	if (tsk->cong_link_cnt &&
3306	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3307		goto stat_msg_cancel;
3308
3309	if (tsk_conn_cong(tsk) &&
3310	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3311		goto stat_msg_cancel;
3312
3313	nla_nest_end(skb, stat);
 
 
 
 
 
3314	nla_nest_end(skb, attrs);
3315
3316	return 0;
3317
3318stat_msg_cancel:
3319	nla_nest_cancel(skb, stat);
3320attr_msg_cancel:
3321	nla_nest_cancel(skb, attrs);
3322msg_cancel:
3323	return -EMSGSIZE;
3324}
3325EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3326
3327int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3328{
3329	return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3330}
3331
3332/* Caller should hold socket lock for the passed tipc socket. */
3333static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3334				 struct netlink_callback *cb,
3335				 struct publication *publ)
3336{
3337	void *hdr;
3338	struct nlattr *attrs;
3339
3340	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3341			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3342	if (!hdr)
3343		goto msg_cancel;
3344
3345	attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
3346	if (!attrs)
3347		goto genlmsg_cancel;
3348
3349	if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3350		goto attr_msg_cancel;
3351	if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3352		goto attr_msg_cancel;
3353	if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3354		goto attr_msg_cancel;
3355	if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3356		goto attr_msg_cancel;
3357
3358	nla_nest_end(skb, attrs);
3359	genlmsg_end(skb, hdr);
3360
3361	return 0;
3362
3363attr_msg_cancel:
3364	nla_nest_cancel(skb, attrs);
3365genlmsg_cancel:
3366	genlmsg_cancel(skb, hdr);
3367msg_cancel:
3368	return -EMSGSIZE;
3369}
3370
3371/* Caller should hold socket lock for the passed tipc socket. */
3372static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3373				  struct netlink_callback *cb,
3374				  struct tipc_sock *tsk, u32 *last_publ)
3375{
3376	int err;
3377	struct publication *p;
3378
3379	if (*last_publ) {
3380		list_for_each_entry(p, &tsk->publications, binding_sock) {
3381			if (p->key == *last_publ)
3382				break;
3383		}
3384		if (p->key != *last_publ) {
3385			/* We never set seq or call nl_dump_check_consistent()
3386			 * this means that setting prev_seq here will cause the
3387			 * consistence check to fail in the netlink callback
3388			 * handler. Resulting in the last NLMSG_DONE message
3389			 * having the NLM_F_DUMP_INTR flag set.
3390			 */
3391			cb->prev_seq = 1;
3392			*last_publ = 0;
3393			return -EPIPE;
3394		}
3395	} else {
3396		p = list_first_entry(&tsk->publications, struct publication,
3397				     binding_sock);
3398	}
3399
3400	list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3401		err = __tipc_nl_add_sk_publ(skb, cb, p);
3402		if (err) {
3403			*last_publ = p->key;
3404			return err;
3405		}
3406	}
3407	*last_publ = 0;
3408
3409	return 0;
3410}
3411
3412int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3413{
3414	int err;
3415	u32 tsk_portid = cb->args[0];
3416	u32 last_publ = cb->args[1];
3417	u32 done = cb->args[2];
3418	struct net *net = sock_net(skb->sk);
3419	struct tipc_sock *tsk;
3420
3421	if (!tsk_portid) {
3422		struct nlattr **attrs;
3423		struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3424
3425		err = tipc_nlmsg_parse(cb->nlh, &attrs);
3426		if (err)
3427			return err;
3428
3429		if (!attrs[TIPC_NLA_SOCK])
3430			return -EINVAL;
3431
3432		err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
3433				       attrs[TIPC_NLA_SOCK],
3434				       tipc_nl_sock_policy, NULL);
3435		if (err)
3436			return err;
3437
3438		if (!sock[TIPC_NLA_SOCK_REF])
3439			return -EINVAL;
3440
3441		tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3442	}
3443
3444	if (done)
3445		return 0;
3446
3447	tsk = tipc_sk_lookup(net, tsk_portid);
3448	if (!tsk)
3449		return -EINVAL;
3450
3451	lock_sock(&tsk->sk);
3452	err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3453	if (!err)
3454		done = 1;
3455	release_sock(&tsk->sk);
3456	sock_put(&tsk->sk);
3457
3458	cb->args[0] = tsk_portid;
3459	cb->args[1] = last_publ;
3460	cb->args[2] = done;
3461
3462	return skb->len;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3463}