name_table.c

   1/*
   2 * net/tipc/name_table.c: TIPC name table code
   3 *
   4 * Copyright (c) 2000-2006, 2014-2018, Ericsson AB
   5 * Copyright (c) 2004-2008, 2010-2014, 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 <net/sock.h>
  39#include <linux/list_sort.h>
  40#include <linux/rbtree_augmented.h>
  41#include "core.h"
  42#include "netlink.h"
  43#include "name_table.h"
  44#include "name_distr.h"
  45#include "subscr.h"
  46#include "bcast.h"
  47#include "addr.h"
  48#include "node.h"
  49#include "group.h"
  50
  51/**
  52 * struct service_range - container for all bindings of a service range
  53 * @lower: service range lower bound
  54 * @upper: service range upper bound
  55 * @tree_node: member of service range RB tree
  56 * @max: largest 'upper' in this node subtree
  57 * @local_publ: list of identical publications made from this node
  58 *   Used by closest_first lookup and multicast lookup algorithm
  59 * @all_publ: all publications identical to this one, whatever node and scope
  60 *   Used by round-robin lookup algorithm
  61 */
  62struct service_range {
  63	u32 lower;
  64	u32 upper;
  65	struct rb_node tree_node;
  66	u32 max;
  67	struct list_head local_publ;
  68	struct list_head all_publ;
  69};
  70
  71/**
  72 * struct tipc_service - container for all published instances of a service type
  73 * @type: 32 bit 'type' value for service
  74 * @publ_cnt: increasing counter for publications in this service
  75 * @ranges: rb tree containing all service ranges for this service
  76 * @service_list: links to adjacent name ranges in hash chain
  77 * @subscriptions: list of subscriptions for this service type
  78 * @lock: spinlock controlling access to pertaining service ranges/publications
  79 * @rcu: RCU callback head used for deferred freeing
  80 */
  81struct tipc_service {
  82	u32 type;
  83	u32 publ_cnt;
  84	struct rb_root ranges;
  85	struct hlist_node service_list;
  86	struct list_head subscriptions;
  87	spinlock_t lock; /* Covers service range list */
  88	struct rcu_head rcu;
  89};
  90
  91#define service_range_upper(sr) ((sr)->upper)
  92RB_DECLARE_CALLBACKS_MAX(static, sr_callbacks,
  93			 struct service_range, tree_node, u32, max,
  94			 service_range_upper)
  95
  96#define service_range_entry(rbtree_node)				\
  97	(container_of(rbtree_node, struct service_range, tree_node))
  98
  99#define service_range_overlap(sr, start, end)				\
 100	((sr)->lower <= (end) && (sr)->upper >= (start))
 101
 102/**
 103 * service_range_foreach_match - iterate over tipc service rbtree for each
 104 *                               range match
 105 * @sr: the service range pointer as a loop cursor
 106 * @sc: the pointer to tipc service which holds the service range rbtree
 107 * @start: beginning of the search range (end >= start) for matching
 108 * @end: end of the search range (end >= start) for matching
 109 */
 110#define service_range_foreach_match(sr, sc, start, end)			\
 111	for (sr = service_range_match_first((sc)->ranges.rb_node,	\
 112					    start,			\
 113					    end);			\
 114	     sr;							\
 115	     sr = service_range_match_next(&(sr)->tree_node,		\
 116					   start,			\
 117					   end))
 118
 119/**
 120 * service_range_match_first - find first service range matching a range
 121 * @n: the root node of service range rbtree for searching
 122 * @start: beginning of the search range (end >= start) for matching
 123 * @end: end of the search range (end >= start) for matching
 124 *
 125 * Return: the leftmost service range node in the rbtree that overlaps the
 126 * specific range if any. Otherwise, returns NULL.
 127 */
 128static struct service_range *service_range_match_first(struct rb_node *n,
 129						       u32 start, u32 end)
 130{
 131	struct service_range *sr;
 132	struct rb_node *l, *r;
 133
 134	/* Non overlaps in tree at all? */
 135	if (!n || service_range_entry(n)->max < start)
 136		return NULL;
 137
 138	while (n) {
 139		l = n->rb_left;
 140		if (l && service_range_entry(l)->max >= start) {
 141			/* A leftmost overlap range node must be one in the left
 142			 * subtree. If not, it has lower > end, then nodes on
 143			 * the right side cannot satisfy the condition either.
 144			 */
 145			n = l;
 146			continue;
 147		}
 148
 149		/* No one in the left subtree can match, return if this node is
 150		 * an overlap i.e. leftmost.
 151		 */
 152		sr = service_range_entry(n);
 153		if (service_range_overlap(sr, start, end))
 154			return sr;
 155
 156		/* Ok, try to lookup on the right side */
 157		r = n->rb_right;
 158		if (sr->lower <= end &&
 159		    r && service_range_entry(r)->max >= start) {
 160			n = r;
 161			continue;
 162		}
 163		break;
 164	}
 165
 166	return NULL;
 167}
 168
 169/**
 170 * service_range_match_next - find next service range matching a range
 171 * @n: a node in service range rbtree from which the searching starts
 172 * @start: beginning of the search range (end >= start) for matching
 173 * @end: end of the search range (end >= start) for matching
 174 *
 175 * Return: the next service range node to the given node in the rbtree that
 176 * overlaps the specific range if any. Otherwise, returns NULL.
 177 */
 178static struct service_range *service_range_match_next(struct rb_node *n,
 179						      u32 start, u32 end)
 180{
 181	struct service_range *sr;
 182	struct rb_node *p, *r;
 183
 184	while (n) {
 185		r = n->rb_right;
 186		if (r && service_range_entry(r)->max >= start)
 187			/* A next overlap range node must be one in the right
 188			 * subtree. If not, it has lower > end, then any next
 189			 * successor (- an ancestor) of this node cannot
 190			 * satisfy the condition either.
 191			 */
 192			return service_range_match_first(r, start, end);
 193
 194		/* No one in the right subtree can match, go up to find an
 195		 * ancestor of this node which is parent of a left-hand child.
 196		 */
 197		while ((p = rb_parent(n)) && n == p->rb_right)
 198			n = p;
 199		if (!p)
 200			break;
 201
 202		/* Return if this ancestor is an overlap */
 203		sr = service_range_entry(p);
 204		if (service_range_overlap(sr, start, end))
 205			return sr;
 206
 207		/* Ok, try to lookup more from this ancestor */
 208		if (sr->lower <= end) {
 209			n = p;
 210			continue;
 211		}
 212		break;
 213	}
 214
 215	return NULL;
 216}
 217
 218static int hash(int x)
 219{
 220	return x & (TIPC_NAMETBL_SIZE - 1);
 221}
 222
 223/**
 224 * tipc_publ_create - create a publication structure
 225 * @ua: the service range the user is binding to
 226 * @sk: the address of the socket that is bound
 227 * @key: publication key
 228 */
 229static struct publication *tipc_publ_create(struct tipc_uaddr *ua,
 230					    struct tipc_socket_addr *sk,
 231					    u32 key)
 232{
 233	struct publication *p = kzalloc(sizeof(*p), GFP_ATOMIC);
 234
 235	if (!p)
 236		return NULL;
 237
 238	p->sr = ua->sr;
 239	p->sk = *sk;
 240	p->scope = ua->scope;
 241	p->key = key;
 242	INIT_LIST_HEAD(&p->binding_sock);
 243	INIT_LIST_HEAD(&p->binding_node);
 244	INIT_LIST_HEAD(&p->local_publ);
 245	INIT_LIST_HEAD(&p->all_publ);
 246	INIT_LIST_HEAD(&p->list);
 247	return p;
 248}
 249
 250/**
 251 * tipc_service_create - create a service structure for the specified 'type'
 252 * @net: network namespace
 253 * @ua: address representing the service to be bound
 254 *
 255 * Allocates a single range structure and sets it to all 0's.
 256 */
 257static struct tipc_service *tipc_service_create(struct net *net,
 258						struct tipc_uaddr *ua)
 259{
 260	struct name_table *nt = tipc_name_table(net);
 261	struct tipc_service *service;
 262	struct hlist_head *hd;
 263
 264	service = kzalloc(sizeof(*service), GFP_ATOMIC);
 265	if (!service) {
 266		pr_warn("Service creation failed, no memory\n");
 267		return NULL;
 268	}
 269
 270	spin_lock_init(&service->lock);
 271	service->type = ua->sr.type;
 272	service->ranges = RB_ROOT;
 273	INIT_HLIST_NODE(&service->service_list);
 274	INIT_LIST_HEAD(&service->subscriptions);
 275	hd = &nt->services[hash(ua->sr.type)];
 276	hlist_add_head_rcu(&service->service_list, hd);
 277	return service;
 278}
 279
 280/*  tipc_service_find_range - find service range matching publication parameters
 281 */
 282static struct service_range *tipc_service_find_range(struct tipc_service *sc,
 283						     struct tipc_uaddr *ua)
 284{
 285	struct service_range *sr;
 286
 287	service_range_foreach_match(sr, sc, ua->sr.lower, ua->sr.upper) {
 288		/* Look for exact match */
 289		if (sr->lower == ua->sr.lower && sr->upper == ua->sr.upper)
 290			return sr;
 291	}
 292
 293	return NULL;
 294}
 295
 296static struct service_range *tipc_service_create_range(struct tipc_service *sc,
 297						       struct publication *p)
 298{
 299	struct rb_node **n, *parent = NULL;
 300	struct service_range *sr;
 301	u32 lower = p->sr.lower;
 302	u32 upper = p->sr.upper;
 303
 304	n = &sc->ranges.rb_node;
 305	while (*n) {
 306		parent = *n;
 307		sr = service_range_entry(parent);
 308		if (lower == sr->lower && upper == sr->upper)
 309			return sr;
 310		if (sr->max < upper)
 311			sr->max = upper;
 312		if (lower <= sr->lower)
 313			n = &parent->rb_left;
 314		else
 315			n = &parent->rb_right;
 316	}
 317	sr = kzalloc(sizeof(*sr), GFP_ATOMIC);
 318	if (!sr)
 319		return NULL;
 320	sr->lower = lower;
 321	sr->upper = upper;
 322	sr->max = upper;
 323	INIT_LIST_HEAD(&sr->local_publ);
 324	INIT_LIST_HEAD(&sr->all_publ);
 325	rb_link_node(&sr->tree_node, parent, n);
 326	rb_insert_augmented(&sr->tree_node, &sc->ranges, &sr_callbacks);
 327	return sr;
 328}
 329
 330static bool tipc_service_insert_publ(struct net *net,
 331				     struct tipc_service *sc,
 332				     struct publication *p)
 333{
 334	struct tipc_subscription *sub, *tmp;
 335	struct service_range *sr;
 336	struct publication *_p;
 337	u32 node = p->sk.node;
 338	bool first = false;
 339	bool res = false;
 340	u32 key = p->key;
 341
 342	spin_lock_bh(&sc->lock);
 343	sr = tipc_service_create_range(sc, p);
 344	if (!sr)
 345		goto  exit;
 346
 347	first = list_empty(&sr->all_publ);
 348
 349	/* Return if the publication already exists */
 350	list_for_each_entry(_p, &sr->all_publ, all_publ) {
 351		if (_p->key == key && _p->sk.ref == p->sk.ref &&
 352		    (!_p->sk.node || _p->sk.node == node)) {
 353			pr_debug("Failed to bind duplicate %u,%u,%u/%u:%u/%u\n",
 354				 p->sr.type, p->sr.lower, p->sr.upper,
 355				 node, p->sk.ref, key);
 356			goto exit;
 357		}
 358	}
 359
 360	if (in_own_node(net, p->sk.node))
 361		list_add(&p->local_publ, &sr->local_publ);
 362	list_add(&p->all_publ, &sr->all_publ);
 363	p->id = sc->publ_cnt++;
 364
 365	/* Any subscriptions waiting for notification?  */
 366	list_for_each_entry_safe(sub, tmp, &sc->subscriptions, service_list) {
 367		tipc_sub_report_overlap(sub, p, TIPC_PUBLISHED, first);
 368	}
 369	res = true;
 370exit:
 371	if (!res)
 372		pr_warn("Failed to bind to %u,%u,%u\n",
 373			p->sr.type, p->sr.lower, p->sr.upper);
 374	spin_unlock_bh(&sc->lock);
 375	return res;
 376}
 377
 378/**
 379 * tipc_service_remove_publ - remove a publication from a service
 380 * @r: service_range to remove publication from
 381 * @sk: address publishing socket
 382 * @key: target publication key
 383 */
 384static struct publication *tipc_service_remove_publ(struct service_range *r,
 385						    struct tipc_socket_addr *sk,
 386						    u32 key)
 387{
 388	struct publication *p;
 389	u32 node = sk->node;
 390
 391	list_for_each_entry(p, &r->all_publ, all_publ) {
 392		if (p->key != key || p->sk.ref != sk->ref ||
 393		    (node && node != p->sk.node))
 394			continue;
 395		list_del(&p->all_publ);
 396		list_del(&p->local_publ);
 397		return p;
 398	}
 399	return NULL;
 400}
 401
 402/*
 403 * Code reused: time_after32() for the same purpose
 404 */
 405#define publication_after(pa, pb) time_after32((pa)->id, (pb)->id)
 406static int tipc_publ_sort(void *priv, const struct list_head *a,
 407			  const struct list_head *b)
 408{
 409	struct publication *pa, *pb;
 410
 411	pa = container_of(a, struct publication, list);
 412	pb = container_of(b, struct publication, list);
 413	return publication_after(pa, pb);
 414}
 415
 416/**
 417 * tipc_service_subscribe - attach a subscription, and optionally
 418 * issue the prescribed number of events if there is any service
 419 * range overlapping with the requested range
 420 * @service: the tipc_service to attach the @sub to
 421 * @sub: the subscription to attach
 422 */
 423static void tipc_service_subscribe(struct tipc_service *service,
 424				   struct tipc_subscription *sub)
 425{
 426	struct publication *p, *first, *tmp;
 427	struct list_head publ_list;
 428	struct service_range *sr;
 429	u32 filter, lower, upper;
 430
 431	filter = sub->s.filter;
 432	lower = sub->s.seq.lower;
 433	upper = sub->s.seq.upper;
 434
 435	tipc_sub_get(sub);
 436	list_add(&sub->service_list, &service->subscriptions);
 437
 438	if (filter & TIPC_SUB_NO_STATUS)
 439		return;
 440
 441	INIT_LIST_HEAD(&publ_list);
 442	service_range_foreach_match(sr, service, lower, upper) {
 443		first = NULL;
 444		list_for_each_entry(p, &sr->all_publ, all_publ) {
 445			if (filter & TIPC_SUB_PORTS)
 446				list_add_tail(&p->list, &publ_list);
 447			else if (!first || publication_after(first, p))
 448				/* Pick this range's *first* publication */
 449				first = p;
 450		}
 451		if (first)
 452			list_add_tail(&first->list, &publ_list);
 453	}
 454
 455	/* Sort the publications before reporting */
 456	list_sort(NULL, &publ_list, tipc_publ_sort);
 457	list_for_each_entry_safe(p, tmp, &publ_list, list) {
 458		tipc_sub_report_overlap(sub, p, TIPC_PUBLISHED, true);
 459		list_del_init(&p->list);
 460	}
 461}
 462
 463static struct tipc_service *tipc_service_find(struct net *net,
 464					      struct tipc_uaddr *ua)
 465{
 466	struct name_table *nt = tipc_name_table(net);
 467	struct hlist_head *service_head;
 468	struct tipc_service *service;
 469
 470	service_head = &nt->services[hash(ua->sr.type)];
 471	hlist_for_each_entry_rcu(service, service_head, service_list) {
 472		if (service->type == ua->sr.type)
 473			return service;
 474	}
 475	return NULL;
 476};
 477
 478struct publication *tipc_nametbl_insert_publ(struct net *net,
 479					     struct tipc_uaddr *ua,
 480					     struct tipc_socket_addr *sk,
 481					     u32 key)
 482{
 483	struct tipc_service *sc;
 484	struct publication *p;
 485
 486	p = tipc_publ_create(ua, sk, key);
 487	if (!p)
 488		return NULL;
 489
 490	sc = tipc_service_find(net, ua);
 491	if (!sc)
 492		sc = tipc_service_create(net, ua);
 493	if (sc && tipc_service_insert_publ(net, sc, p))
 494		return p;
 495	kfree(p);
 496	return NULL;
 497}
 498
 499struct publication *tipc_nametbl_remove_publ(struct net *net,
 500					     struct tipc_uaddr *ua,
 501					     struct tipc_socket_addr *sk,
 502					     u32 key)
 503{
 504	struct tipc_subscription *sub, *tmp;
 505	struct publication *p = NULL;
 506	struct service_range *sr;
 507	struct tipc_service *sc;
 508	bool last;
 509
 510	sc = tipc_service_find(net, ua);
 511	if (!sc)
 512		goto exit;
 513
 514	spin_lock_bh(&sc->lock);
 515	sr = tipc_service_find_range(sc, ua);
 516	if (!sr)
 517		goto unlock;
 518	p = tipc_service_remove_publ(sr, sk, key);
 519	if (!p)
 520		goto unlock;
 521
 522	/* Notify any waiting subscriptions */
 523	last = list_empty(&sr->all_publ);
 524	list_for_each_entry_safe(sub, tmp, &sc->subscriptions, service_list) {
 525		tipc_sub_report_overlap(sub, p, TIPC_WITHDRAWN, last);
 526	}
 527
 528	/* Remove service range item if this was its last publication */
 529	if (list_empty(&sr->all_publ)) {
 530		rb_erase_augmented(&sr->tree_node, &sc->ranges, &sr_callbacks);
 531		kfree(sr);
 532	}
 533
 534	/* Delete service item if no more publications and subscriptions */
 535	if (RB_EMPTY_ROOT(&sc->ranges) && list_empty(&sc->subscriptions)) {
 536		hlist_del_init_rcu(&sc->service_list);
 537		kfree_rcu(sc, rcu);
 538	}
 539unlock:
 540	spin_unlock_bh(&sc->lock);
 541exit:
 542	if (!p) {
 543		pr_err("Failed to remove unknown binding: %u,%u,%u/%u:%u/%u\n",
 544		       ua->sr.type, ua->sr.lower, ua->sr.upper,
 545		       sk->node, sk->ref, key);
 546	}
 547	return p;
 548}
 549
 550/**
 551 * tipc_nametbl_lookup_anycast - perform service instance to socket translation
 552 * @net: network namespace
 553 * @ua: service address to look up
 554 * @sk: address to socket we want to find
 555 *
 556 * On entry, a non-zero 'sk->node' indicates the node where we want lookup to be
 557 * performed, which may not be this one.
 558 *
 559 * On exit:
 560 *
 561 * - If lookup is deferred to another node, leave 'sk->node' unchanged and
 562 *   return 'true'.
 563 * - If lookup is successful, set the 'sk->node' and 'sk->ref' (== portid) which
 564 *   represent the bound socket and return 'true'.
 565 * - If lookup fails, return 'false'
 566 *
 567 * Note that for legacy users (node configured with Z.C.N address format) the
 568 * 'closest-first' lookup algorithm must be maintained, i.e., if sk.node is 0
 569 * we must look in the local binding list first
 570 */
 571bool tipc_nametbl_lookup_anycast(struct net *net,
 572				 struct tipc_uaddr *ua,
 573				 struct tipc_socket_addr *sk)
 574{
 575	struct tipc_net *tn = tipc_net(net);
 576	bool legacy = tn->legacy_addr_format;
 577	u32 self = tipc_own_addr(net);
 578	u32 inst = ua->sa.instance;
 579	struct service_range *r;
 580	struct tipc_service *sc;
 581	struct publication *p;
 582	struct list_head *l;
 583	bool res = false;
 584
 585	if (!tipc_in_scope(legacy, sk->node, self))
 586		return true;
 587
 588	rcu_read_lock();
 589	sc = tipc_service_find(net, ua);
 590	if (unlikely(!sc))
 591		goto exit;
 592
 593	spin_lock_bh(&sc->lock);
 594	service_range_foreach_match(r, sc, inst, inst) {
 595		/* Select lookup algo: local, closest-first or round-robin */
 596		if (sk->node == self) {
 597			l = &r->local_publ;
 598			if (list_empty(l))
 599				continue;
 600			p = list_first_entry(l, struct publication, local_publ);
 601			list_move_tail(&p->local_publ, &r->local_publ);
 602		} else if (legacy && !sk->node && !list_empty(&r->local_publ)) {
 603			l = &r->local_publ;
 604			p = list_first_entry(l, struct publication, local_publ);
 605			list_move_tail(&p->local_publ, &r->local_publ);
 606		} else {
 607			l = &r->all_publ;
 608			p = list_first_entry(l, struct publication, all_publ);
 609			list_move_tail(&p->all_publ, &r->all_publ);
 610		}
 611		*sk = p->sk;
 612		res = true;
 613		/* Todo: as for legacy, pick the first matching range only, a
 614		 * "true" round-robin will be performed as needed.
 615		 */
 616		break;
 617	}
 618	spin_unlock_bh(&sc->lock);
 619
 620exit:
 621	rcu_read_unlock();
 622	return res;
 623}
 624
 625/* tipc_nametbl_lookup_group(): lookup destinaton(s) in a communication group
 626 * Returns a list of one (== group anycast) or more (== group multicast)
 627 * destination socket/node pairs matching the given address.
 628 * The requester may or may not want to exclude himself from the list.
 629 */
 630bool tipc_nametbl_lookup_group(struct net *net, struct tipc_uaddr *ua,
 631			       struct list_head *dsts, int *dstcnt,
 632			       u32 exclude, bool mcast)
 633{
 634	u32 self = tipc_own_addr(net);
 635	u32 inst = ua->sa.instance;
 636	struct service_range *sr;
 637	struct tipc_service *sc;
 638	struct publication *p;
 639
 640	*dstcnt = 0;
 641	rcu_read_lock();
 642	sc = tipc_service_find(net, ua);
 643	if (unlikely(!sc))
 644		goto exit;
 645
 646	spin_lock_bh(&sc->lock);
 647
 648	/* Todo: a full search i.e. service_range_foreach_match() instead? */
 649	sr = service_range_match_first(sc->ranges.rb_node, inst, inst);
 650	if (!sr)
 651		goto no_match;
 652
 653	list_for_each_entry(p, &sr->all_publ, all_publ) {
 654		if (p->scope != ua->scope)
 655			continue;
 656		if (p->sk.ref == exclude && p->sk.node == self)
 657			continue;
 658		tipc_dest_push(dsts, p->sk.node, p->sk.ref);
 659		(*dstcnt)++;
 660		if (mcast)
 661			continue;
 662		list_move_tail(&p->all_publ, &sr->all_publ);
 663		break;
 664	}
 665no_match:
 666	spin_unlock_bh(&sc->lock);
 667exit:
 668	rcu_read_unlock();
 669	return !list_empty(dsts);
 670}
 671
 672/* tipc_nametbl_lookup_mcast_sockets(): look up node local destinaton sockets
 673 *                                      matching the given address
 674 * Used on nodes which have received a multicast/broadcast message
 675 * Returns a list of local sockets
 676 */
 677void tipc_nametbl_lookup_mcast_sockets(struct net *net, struct tipc_uaddr *ua,
 678				       struct list_head *dports)
 679{
 680	struct service_range *sr;
 681	struct tipc_service *sc;
 682	struct publication *p;
 683	u8 scope = ua->scope;
 684
 685	rcu_read_lock();
 686	sc = tipc_service_find(net, ua);
 687	if (!sc)
 688		goto exit;
 689
 690	spin_lock_bh(&sc->lock);
 691	service_range_foreach_match(sr, sc, ua->sr.lower, ua->sr.upper) {
 692		list_for_each_entry(p, &sr->local_publ, local_publ) {
 693			if (scope == p->scope || scope == TIPC_ANY_SCOPE)
 694				tipc_dest_push(dports, 0, p->sk.ref);
 695		}
 696	}
 697	spin_unlock_bh(&sc->lock);
 698exit:
 699	rcu_read_unlock();
 700}
 701
 702/* tipc_nametbl_lookup_mcast_nodes(): look up all destination nodes matching
 703 *                                    the given address. Used in sending node.
 704 * Used on nodes which are sending out a multicast/broadcast message
 705 * Returns a list of nodes, including own node if applicable
 706 */
 707void tipc_nametbl_lookup_mcast_nodes(struct net *net, struct tipc_uaddr *ua,
 708				     struct tipc_nlist *nodes)
 709{
 710	struct service_range *sr;
 711	struct tipc_service *sc;
 712	struct publication *p;
 713
 714	rcu_read_lock();
 715	sc = tipc_service_find(net, ua);
 716	if (!sc)
 717		goto exit;
 718
 719	spin_lock_bh(&sc->lock);
 720	service_range_foreach_match(sr, sc, ua->sr.lower, ua->sr.upper) {
 721		list_for_each_entry(p, &sr->all_publ, all_publ) {
 722			tipc_nlist_add(nodes, p->sk.node);
 723		}
 724	}
 725	spin_unlock_bh(&sc->lock);
 726exit:
 727	rcu_read_unlock();
 728}
 729
 730/* tipc_nametbl_build_group - build list of communication group members
 731 */
 732void tipc_nametbl_build_group(struct net *net, struct tipc_group *grp,
 733			      struct tipc_uaddr *ua)
 734{
 735	struct service_range *sr;
 736	struct tipc_service *sc;
 737	struct publication *p;
 738	struct rb_node *n;
 739
 740	rcu_read_lock();
 741	sc = tipc_service_find(net, ua);
 742	if (!sc)
 743		goto exit;
 744
 745	spin_lock_bh(&sc->lock);
 746	for (n = rb_first(&sc->ranges); n; n = rb_next(n)) {
 747		sr = container_of(n, struct service_range, tree_node);
 748		list_for_each_entry(p, &sr->all_publ, all_publ) {
 749			if (p->scope != ua->scope)
 750				continue;
 751			tipc_group_add_member(grp, p->sk.node, p->sk.ref,
 752					      p->sr.lower);
 753		}
 754	}
 755	spin_unlock_bh(&sc->lock);
 756exit:
 757	rcu_read_unlock();
 758}
 759
 760/* tipc_nametbl_publish - add service binding to name table
 761 */
 762struct publication *tipc_nametbl_publish(struct net *net, struct tipc_uaddr *ua,
 763					 struct tipc_socket_addr *sk, u32 key)
 764{
 765	struct name_table *nt = tipc_name_table(net);
 766	struct tipc_net *tn = tipc_net(net);
 767	struct publication *p = NULL;
 768	struct sk_buff *skb = NULL;
 769	u32 rc_dests;
 770
 771	spin_lock_bh(&tn->nametbl_lock);
 772
 773	if (nt->local_publ_count >= TIPC_MAX_PUBL) {
 774		pr_warn("Bind failed, max limit %u reached\n", TIPC_MAX_PUBL);
 775		goto exit;
 776	}
 777
 778	p = tipc_nametbl_insert_publ(net, ua, sk, key);
 779	if (p) {
 780		nt->local_publ_count++;
 781		skb = tipc_named_publish(net, p);
 782	}
 783	rc_dests = nt->rc_dests;
 784exit:
 785	spin_unlock_bh(&tn->nametbl_lock);
 786
 787	if (skb)
 788		tipc_node_broadcast(net, skb, rc_dests);
 789	return p;
 790
 791}
 792
 793/**
 794 * tipc_nametbl_withdraw - withdraw a service binding
 795 * @net: network namespace
 796 * @ua: service address/range being unbound
 797 * @sk: address of the socket being unbound from
 798 * @key: target publication key
 799 */
 800void tipc_nametbl_withdraw(struct net *net, struct tipc_uaddr *ua,
 801			   struct tipc_socket_addr *sk, u32 key)
 802{
 803	struct name_table *nt = tipc_name_table(net);
 804	struct tipc_net *tn = tipc_net(net);
 805	struct sk_buff *skb = NULL;
 806	struct publication *p;
 807	u32 rc_dests;
 808
 809	spin_lock_bh(&tn->nametbl_lock);
 810
 811	p = tipc_nametbl_remove_publ(net, ua, sk, key);
 812	if (p) {
 813		nt->local_publ_count--;
 814		skb = tipc_named_withdraw(net, p);
 815		list_del_init(&p->binding_sock);
 816		kfree_rcu(p, rcu);
 817	}
 818	rc_dests = nt->rc_dests;
 819	spin_unlock_bh(&tn->nametbl_lock);
 820
 821	if (skb)
 822		tipc_node_broadcast(net, skb, rc_dests);
 823}
 824
 825/**
 826 * tipc_nametbl_subscribe - add a subscription object to the name table
 827 * @sub: subscription to add
 828 */
 829bool tipc_nametbl_subscribe(struct tipc_subscription *sub)
 830{
 831	struct tipc_net *tn = tipc_net(sub->net);
 832	u32 type = sub->s.seq.type;
 833	struct tipc_service *sc;
 834	struct tipc_uaddr ua;
 835	bool res = true;
 836
 837	tipc_uaddr(&ua, TIPC_SERVICE_RANGE, TIPC_NODE_SCOPE, type,
 838		   sub->s.seq.lower, sub->s.seq.upper);
 839	spin_lock_bh(&tn->nametbl_lock);
 840	sc = tipc_service_find(sub->net, &ua);
 841	if (!sc)
 842		sc = tipc_service_create(sub->net, &ua);
 843	if (sc) {
 844		spin_lock_bh(&sc->lock);
 845		tipc_service_subscribe(sc, sub);
 846		spin_unlock_bh(&sc->lock);
 847	} else {
 848		pr_warn("Failed to subscribe for {%u,%u,%u}\n",
 849			type, sub->s.seq.lower, sub->s.seq.upper);
 850		res = false;
 851	}
 852	spin_unlock_bh(&tn->nametbl_lock);
 853	return res;
 854}
 855
 856/**
 857 * tipc_nametbl_unsubscribe - remove a subscription object from name table
 858 * @sub: subscription to remove
 859 */
 860void tipc_nametbl_unsubscribe(struct tipc_subscription *sub)
 861{
 862	struct tipc_net *tn = tipc_net(sub->net);
 863	struct tipc_service *sc;
 864	struct tipc_uaddr ua;
 865
 866	tipc_uaddr(&ua, TIPC_SERVICE_RANGE, TIPC_NODE_SCOPE,
 867		   sub->s.seq.type, sub->s.seq.lower, sub->s.seq.upper);
 868	spin_lock_bh(&tn->nametbl_lock);
 869	sc = tipc_service_find(sub->net, &ua);
 870	if (!sc)
 871		goto exit;
 872
 873	spin_lock_bh(&sc->lock);
 874	list_del_init(&sub->service_list);
 875	tipc_sub_put(sub);
 876
 877	/* Delete service item if no more publications and subscriptions */
 878	if (RB_EMPTY_ROOT(&sc->ranges) && list_empty(&sc->subscriptions)) {
 879		hlist_del_init_rcu(&sc->service_list);
 880		kfree_rcu(sc, rcu);
 881	}
 882	spin_unlock_bh(&sc->lock);
 883exit:
 884	spin_unlock_bh(&tn->nametbl_lock);
 885}
 886
 887int tipc_nametbl_init(struct net *net)
 888{
 889	struct tipc_net *tn = tipc_net(net);
 890	struct name_table *nt;
 891	int i;
 892
 893	nt = kzalloc(sizeof(*nt), GFP_KERNEL);
 894	if (!nt)
 895		return -ENOMEM;
 896
 897	for (i = 0; i < TIPC_NAMETBL_SIZE; i++)
 898		INIT_HLIST_HEAD(&nt->services[i]);
 899
 900	INIT_LIST_HEAD(&nt->node_scope);
 901	INIT_LIST_HEAD(&nt->cluster_scope);
 902	rwlock_init(&nt->cluster_scope_lock);
 903	tn->nametbl = nt;
 904	spin_lock_init(&tn->nametbl_lock);
 905	return 0;
 906}
 907
 908/**
 909 * tipc_service_delete - purge all publications for a service and delete it
 910 * @net: the associated network namespace
 911 * @sc: tipc_service to delete
 912 */
 913static void tipc_service_delete(struct net *net, struct tipc_service *sc)
 914{
 915	struct service_range *sr, *tmpr;
 916	struct publication *p, *tmp;
 917
 918	spin_lock_bh(&sc->lock);
 919	rbtree_postorder_for_each_entry_safe(sr, tmpr, &sc->ranges, tree_node) {
 920		list_for_each_entry_safe(p, tmp, &sr->all_publ, all_publ) {
 921			tipc_service_remove_publ(sr, &p->sk, p->key);
 922			kfree_rcu(p, rcu);
 923		}
 924		rb_erase_augmented(&sr->tree_node, &sc->ranges, &sr_callbacks);
 925		kfree(sr);
 926	}
 927	hlist_del_init_rcu(&sc->service_list);
 928	spin_unlock_bh(&sc->lock);
 929	kfree_rcu(sc, rcu);
 930}
 931
 932void tipc_nametbl_stop(struct net *net)
 933{
 934	struct name_table *nt = tipc_name_table(net);
 935	struct tipc_net *tn = tipc_net(net);
 936	struct hlist_head *service_head;
 937	struct tipc_service *service;
 938	u32 i;
 939
 940	/* Verify name table is empty and purge any lingering
 941	 * publications, then release the name table
 942	 */
 943	spin_lock_bh(&tn->nametbl_lock);
 944	for (i = 0; i < TIPC_NAMETBL_SIZE; i++) {
 945		if (hlist_empty(&nt->services[i]))
 946			continue;
 947		service_head = &nt->services[i];
 948		hlist_for_each_entry_rcu(service, service_head, service_list) {
 949			tipc_service_delete(net, service);
 950		}
 951	}
 952	spin_unlock_bh(&tn->nametbl_lock);
 953
 954	/* TODO: clear tn->nametbl, implement proper RCU rules ? */
 955	kfree_rcu(nt, rcu);
 956}
 957
 958static int __tipc_nl_add_nametable_publ(struct tipc_nl_msg *msg,
 959					struct tipc_service *service,
 960					struct service_range *sr,
 961					u32 *last_key)
 962{
 963	struct publication *p;
 964	struct nlattr *attrs;
 965	struct nlattr *b;
 966	void *hdr;
 967
 968	if (*last_key) {
 969		list_for_each_entry(p, &sr->all_publ, all_publ)
 970			if (p->key == *last_key)
 971				break;
 972		if (list_entry_is_head(p, &sr->all_publ, all_publ))
 973			return -EPIPE;
 974	} else {
 975		p = list_first_entry(&sr->all_publ,
 976				     struct publication,
 977				     all_publ);
 978	}
 979
 980	list_for_each_entry_from(p, &sr->all_publ, all_publ) {
 981		*last_key = p->key;
 982
 983		hdr = genlmsg_put(msg->skb, msg->portid, msg->seq,
 984				  &tipc_genl_family, NLM_F_MULTI,
 985				  TIPC_NL_NAME_TABLE_GET);
 986		if (!hdr)
 987			return -EMSGSIZE;
 988
 989		attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_NAME_TABLE);
 990		if (!attrs)
 991			goto msg_full;
 992
 993		b = nla_nest_start_noflag(msg->skb, TIPC_NLA_NAME_TABLE_PUBL);
 994		if (!b)
 995			goto attr_msg_full;
 996
 997		if (nla_put_u32(msg->skb, TIPC_NLA_PUBL_TYPE, service->type))
 998			goto publ_msg_full;
 999		if (nla_put_u32(msg->skb, TIPC_NLA_PUBL_LOWER, sr->lower))
1000			goto publ_msg_full;
1001		if (nla_put_u32(msg->skb, TIPC_NLA_PUBL_UPPER, sr->upper))
1002			goto publ_msg_full;
1003		if (nla_put_u32(msg->skb, TIPC_NLA_PUBL_SCOPE, p->scope))
1004			goto publ_msg_full;
1005		if (nla_put_u32(msg->skb, TIPC_NLA_PUBL_NODE, p->sk.node))
1006			goto publ_msg_full;
1007		if (nla_put_u32(msg->skb, TIPC_NLA_PUBL_REF, p->sk.ref))
1008			goto publ_msg_full;
1009		if (nla_put_u32(msg->skb, TIPC_NLA_PUBL_KEY, p->key))
1010			goto publ_msg_full;
1011
1012		nla_nest_end(msg->skb, b);
1013		nla_nest_end(msg->skb, attrs);
1014		genlmsg_end(msg->skb, hdr);
1015	}
1016	*last_key = 0;
1017
1018	return 0;
1019
1020publ_msg_full:
1021	nla_nest_cancel(msg->skb, b);
1022attr_msg_full:
1023	nla_nest_cancel(msg->skb, attrs);
1024msg_full:
1025	genlmsg_cancel(msg->skb, hdr);
1026
1027	return -EMSGSIZE;
1028}
1029
1030static int __tipc_nl_service_range_list(struct tipc_nl_msg *msg,
1031					struct tipc_service *sc,
1032					u32 *last_lower, u32 *last_key)
1033{
1034	struct service_range *sr;
1035	struct rb_node *n;
1036	int err;
1037
1038	for (n = rb_first(&sc->ranges); n; n = rb_next(n)) {
1039		sr = container_of(n, struct service_range, tree_node);
1040		if (sr->lower < *last_lower)
1041			continue;
1042		err = __tipc_nl_add_nametable_publ(msg, sc, sr, last_key);
1043		if (err) {
1044			*last_lower = sr->lower;
1045			return err;
1046		}
1047	}
1048	*last_lower = 0;
1049	return 0;
1050}
1051
1052static int tipc_nl_service_list(struct net *net, struct tipc_nl_msg *msg,
1053				u32 *last_type, u32 *last_lower, u32 *last_key)
1054{
1055	struct tipc_net *tn = tipc_net(net);
1056	struct tipc_service *service = NULL;
1057	struct hlist_head *head;
1058	struct tipc_uaddr ua;
1059	int err;
1060	int i;
1061
1062	if (*last_type)
1063		i = hash(*last_type);
1064	else
1065		i = 0;
1066
1067	for (; i < TIPC_NAMETBL_SIZE; i++) {
1068		head = &tn->nametbl->services[i];
1069
1070		if (*last_type ||
1071		    (!i && *last_key && (*last_lower == *last_key))) {
1072			tipc_uaddr(&ua, TIPC_SERVICE_RANGE, TIPC_NODE_SCOPE,
1073				   *last_type, *last_lower, *last_lower);
1074			service = tipc_service_find(net, &ua);
1075			if (!service)
1076				return -EPIPE;
1077		} else {
1078			hlist_for_each_entry_rcu(service, head, service_list)
1079				break;
1080			if (!service)
1081				continue;
1082		}
1083
1084		hlist_for_each_entry_from_rcu(service, service_list) {
1085			spin_lock_bh(&service->lock);
1086			err = __tipc_nl_service_range_list(msg, service,
1087							   last_lower,
1088							   last_key);
1089
1090			if (err) {
1091				*last_type = service->type;
1092				spin_unlock_bh(&service->lock);
1093				return err;
1094			}
1095			spin_unlock_bh(&service->lock);
1096		}
1097		*last_type = 0;
1098	}
1099	return 0;
1100}
1101
1102int tipc_nl_name_table_dump(struct sk_buff *skb, struct netlink_callback *cb)
1103{
1104	struct net *net = sock_net(skb->sk);
1105	u32 last_type = cb->args[0];
1106	u32 last_lower = cb->args[1];
1107	u32 last_key = cb->args[2];
1108	int done = cb->args[3];
1109	struct tipc_nl_msg msg;
1110	int err;
1111
1112	if (done)
1113		return 0;
1114
1115	msg.skb = skb;
1116	msg.portid = NETLINK_CB(cb->skb).portid;
1117	msg.seq = cb->nlh->nlmsg_seq;
1118
1119	rcu_read_lock();
1120	err = tipc_nl_service_list(net, &msg, &last_type,
1121				   &last_lower, &last_key);
1122	if (!err) {
1123		done = 1;
1124	} else if (err != -EMSGSIZE) {
1125		/* We never set seq or call nl_dump_check_consistent() this
1126		 * means that setting prev_seq here will cause the consistence
1127		 * check to fail in the netlink callback handler. Resulting in
1128		 * the NLMSG_DONE message having the NLM_F_DUMP_INTR flag set if
1129		 * we got an error.
1130		 */
1131		cb->prev_seq = 1;
1132	}
1133	rcu_read_unlock();
1134
1135	cb->args[0] = last_type;
1136	cb->args[1] = last_lower;
1137	cb->args[2] = last_key;
1138	cb->args[3] = done;
1139
1140	return skb->len;
1141}
1142
1143struct tipc_dest *tipc_dest_find(struct list_head *l, u32 node, u32 port)
1144{
1145	struct tipc_dest *dst;
1146
1147	list_for_each_entry(dst, l, list) {
1148		if (dst->node == node && dst->port == port)
1149			return dst;
1150	}
1151	return NULL;
1152}
1153
1154bool tipc_dest_push(struct list_head *l, u32 node, u32 port)
1155{
1156	struct tipc_dest *dst;
1157
1158	if (tipc_dest_find(l, node, port))
1159		return false;
1160
1161	dst = kmalloc(sizeof(*dst), GFP_ATOMIC);
1162	if (unlikely(!dst))
1163		return false;
1164	dst->node = node;
1165	dst->port = port;
1166	list_add(&dst->list, l);
1167	return true;
1168}
1169
1170bool tipc_dest_pop(struct list_head *l, u32 *node, u32 *port)
1171{
1172	struct tipc_dest *dst;
1173
1174	if (list_empty(l))
1175		return false;
1176	dst = list_first_entry(l, typeof(*dst), list);
1177	if (port)
1178		*port = dst->port;
1179	if (node)
1180		*node = dst->node;
1181	list_del(&dst->list);
1182	kfree(dst);
1183	return true;
1184}
1185
1186bool tipc_dest_del(struct list_head *l, u32 node, u32 port)
1187{
1188	struct tipc_dest *dst;
1189
1190	dst = tipc_dest_find(l, node, port);
1191	if (!dst)
1192		return false;
1193	list_del(&dst->list);
1194	kfree(dst);
1195	return true;
1196}
1197
1198void tipc_dest_list_purge(struct list_head *l)
1199{
1200	struct tipc_dest *dst, *tmp;
1201
1202	list_for_each_entry_safe(dst, tmp, l, list) {
1203		list_del(&dst->list);
1204		kfree(dst);
1205	}
1206}