xref: /openbmc/linux/net/tipc/name_table.c (revision 66db239c)
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  */
62 struct 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  */
81 struct 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)
92 RB_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  */
128 static 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  */
178 static 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 
218 static 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  */
229 static 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  * @type: service type
253  * @hd: name_table services list
254  *
255  * Allocates a single range structure and sets it to all 0's.
256  */
257 static struct tipc_service *tipc_service_create(u32 type, struct hlist_head *hd)
258 {
259 	struct tipc_service *service = kzalloc(sizeof(*service), GFP_ATOMIC);
260 
261 	if (!service) {
262 		pr_warn("Service creation failed, no memory\n");
263 		return NULL;
264 	}
265 
266 	spin_lock_init(&service->lock);
267 	service->type = type;
268 	service->ranges = RB_ROOT;
269 	INIT_HLIST_NODE(&service->service_list);
270 	INIT_LIST_HEAD(&service->subscriptions);
271 	hlist_add_head_rcu(&service->service_list, hd);
272 	return service;
273 }
274 
275 /*  tipc_service_find_range - find service range matching publication parameters
276  */
277 static struct service_range *tipc_service_find_range(struct tipc_service *sc,
278 						     u32 lower, u32 upper)
279 {
280 	struct service_range *sr;
281 
282 	service_range_foreach_match(sr, sc, lower, upper) {
283 		/* Look for exact match */
284 		if (sr->lower == lower && sr->upper == upper)
285 			return sr;
286 	}
287 
288 	return NULL;
289 }
290 
291 static struct service_range *tipc_service_create_range(struct tipc_service *sc,
292 						       u32 lower, u32 upper)
293 {
294 	struct rb_node **n, *parent = NULL;
295 	struct service_range *sr;
296 
297 	n = &sc->ranges.rb_node;
298 	while (*n) {
299 		parent = *n;
300 		sr = service_range_entry(parent);
301 		if (lower == sr->lower && upper == sr->upper)
302 			return sr;
303 		if (sr->max < upper)
304 			sr->max = upper;
305 		if (lower <= sr->lower)
306 			n = &parent->rb_left;
307 		else
308 			n = &parent->rb_right;
309 	}
310 	sr = kzalloc(sizeof(*sr), GFP_ATOMIC);
311 	if (!sr)
312 		return NULL;
313 	sr->lower = lower;
314 	sr->upper = upper;
315 	sr->max = upper;
316 	INIT_LIST_HEAD(&sr->local_publ);
317 	INIT_LIST_HEAD(&sr->all_publ);
318 	rb_link_node(&sr->tree_node, parent, n);
319 	rb_insert_augmented(&sr->tree_node, &sc->ranges, &sr_callbacks);
320 	return sr;
321 }
322 
323 static bool tipc_service_insert_publ(struct net *net,
324 				     struct tipc_service *sc,
325 				     struct publication *p)
326 {
327 	struct tipc_subscription *sub, *tmp;
328 	struct service_range *sr;
329 	struct publication *_p;
330 	u32 node = p->sk.node;
331 	bool first = false;
332 	bool res = false;
333 
334 	spin_lock_bh(&sc->lock);
335 	sr = tipc_service_create_range(sc, p->sr.lower, p->sr.upper);
336 	if (!sr)
337 		goto  exit;
338 
339 	first = list_empty(&sr->all_publ);
340 
341 	/* Return if the publication already exists */
342 	list_for_each_entry(_p, &sr->all_publ, all_publ) {
343 		if (_p->key == p->key && (!_p->sk.node || _p->sk.node == node))
344 			goto exit;
345 	}
346 
347 	if (in_own_node(net, p->sk.node))
348 		list_add(&p->local_publ, &sr->local_publ);
349 	list_add(&p->all_publ, &sr->all_publ);
350 	p->id = sc->publ_cnt++;
351 
352 	/* Any subscriptions waiting for notification?  */
353 	list_for_each_entry_safe(sub, tmp, &sc->subscriptions, service_list) {
354 		tipc_sub_report_overlap(sub, p->sr.lower, p->sr.upper,
355 					TIPC_PUBLISHED, p->sk.ref, p->sk.node,
356 					p->scope, first);
357 	}
358 	res = true;
359 exit:
360 	if (!res)
361 		pr_warn("Failed to bind to %u,%u,%u\n",
362 			p->sr.type, p->sr.lower, p->sr.upper);
363 	spin_unlock_bh(&sc->lock);
364 	return res;
365 }
366 
367 /**
368  * tipc_service_remove_publ - remove a publication from a service
369  * @r: service_range to remove publication from
370  * @sk: address publishing socket
371  * @key: target publication key
372  */
373 static struct publication *tipc_service_remove_publ(struct service_range *r,
374 						    struct tipc_socket_addr *sk,
375 						    u32 key)
376 {
377 	struct publication *p;
378 	u32 node = sk->node;
379 
380 	list_for_each_entry(p, &r->all_publ, all_publ) {
381 		if (p->key != key || (node && node != p->sk.node))
382 			continue;
383 		list_del(&p->all_publ);
384 		list_del(&p->local_publ);
385 		return p;
386 	}
387 	return NULL;
388 }
389 
390 /*
391  * Code reused: time_after32() for the same purpose
392  */
393 #define publication_after(pa, pb) time_after32((pa)->id, (pb)->id)
394 static int tipc_publ_sort(void *priv, struct list_head *a,
395 			  struct list_head *b)
396 {
397 	struct publication *pa, *pb;
398 
399 	pa = container_of(a, struct publication, list);
400 	pb = container_of(b, struct publication, list);
401 	return publication_after(pa, pb);
402 }
403 
404 /**
405  * tipc_service_subscribe - attach a subscription, and optionally
406  * issue the prescribed number of events if there is any service
407  * range overlapping with the requested range
408  * @service: the tipc_service to attach the @sub to
409  * @sub: the subscription to attach
410  */
411 static void tipc_service_subscribe(struct tipc_service *service,
412 				   struct tipc_subscription *sub)
413 {
414 	struct tipc_subscr *sb = &sub->evt.s;
415 	struct publication *p, *first, *tmp;
416 	struct list_head publ_list;
417 	struct service_range *sr;
418 	struct tipc_service_range r;
419 	u32 filter;
420 
421 	r.type = tipc_sub_read(sb, seq.type);
422 	r.lower = tipc_sub_read(sb, seq.lower);
423 	r.upper = tipc_sub_read(sb, seq.upper);
424 	filter = tipc_sub_read(sb, filter);
425 
426 	tipc_sub_get(sub);
427 	list_add(&sub->service_list, &service->subscriptions);
428 
429 	if (filter & TIPC_SUB_NO_STATUS)
430 		return;
431 
432 	INIT_LIST_HEAD(&publ_list);
433 	service_range_foreach_match(sr, service, r.lower, r.upper) {
434 		first = NULL;
435 		list_for_each_entry(p, &sr->all_publ, all_publ) {
436 			if (filter & TIPC_SUB_PORTS)
437 				list_add_tail(&p->list, &publ_list);
438 			else if (!first || publication_after(first, p))
439 				/* Pick this range's *first* publication */
440 				first = p;
441 		}
442 		if (first)
443 			list_add_tail(&first->list, &publ_list);
444 	}
445 
446 	/* Sort the publications before reporting */
447 	list_sort(NULL, &publ_list, tipc_publ_sort);
448 	list_for_each_entry_safe(p, tmp, &publ_list, list) {
449 		tipc_sub_report_overlap(sub, p->sr.lower, p->sr.upper,
450 					TIPC_PUBLISHED, p->sk.ref, p->sk.node,
451 					p->scope, true);
452 		list_del_init(&p->list);
453 	}
454 }
455 
456 static struct tipc_service *tipc_service_find(struct net *net, u32 type)
457 {
458 	struct name_table *nt = tipc_name_table(net);
459 	struct hlist_head *service_head;
460 	struct tipc_service *service;
461 
462 	service_head = &nt->services[hash(type)];
463 	hlist_for_each_entry_rcu(service, service_head, service_list) {
464 		if (service->type == type)
465 			return service;
466 	}
467 	return NULL;
468 };
469 
470 struct publication *tipc_nametbl_insert_publ(struct net *net,
471 					     struct tipc_uaddr *ua,
472 					     struct tipc_socket_addr *sk,
473 					     u32 key)
474 {
475 	struct name_table *nt = tipc_name_table(net);
476 	struct tipc_service *sc;
477 	struct publication *p;
478 	u32 type = ua->sr.type;
479 
480 	p = tipc_publ_create(ua, sk, key);
481 	if (!p)
482 		return NULL;
483 
484 	if (ua->sr.lower > ua->sr.upper) {
485 		pr_debug("Failed to bind illegal {%u,%u,%u} from node %u\n",
486 			 type, ua->sr.lower, ua->sr.upper, sk->node);
487 		return NULL;
488 	}
489 	sc = tipc_service_find(net, type);
490 	if (!sc)
491 		sc = tipc_service_create(type, &nt->services[hash(type)]);
492 	if (sc && tipc_service_insert_publ(net, sc, p))
493 		return p;
494 	kfree(p);
495 	return NULL;
496 }
497 
498 struct publication *tipc_nametbl_remove_publ(struct net *net,
499 					     struct tipc_uaddr *ua,
500 					     struct tipc_socket_addr *sk,
501 					     u32 key)
502 {
503 	struct tipc_subscription *sub, *tmp;
504 	struct publication *p = NULL;
505 	struct service_range *sr;
506 	struct tipc_service *sc;
507 	u32 upper = ua->sr.upper;
508 	u32 lower = ua->sr.lower;
509 	bool last;
510 
511 	sc = tipc_service_find(net, ua->sr.type);
512 	if (!sc)
513 		return NULL;
514 
515 	spin_lock_bh(&sc->lock);
516 	sr = tipc_service_find_range(sc, lower, upper);
517 	if (!sr)
518 		goto exit;
519 	p = tipc_service_remove_publ(sr, sk, key);
520 	if (!p)
521 		goto exit;
522 
523 	/* Notify any waiting subscriptions */
524 	last = list_empty(&sr->all_publ);
525 	list_for_each_entry_safe(sub, tmp, &sc->subscriptions, service_list) {
526 		tipc_sub_report_overlap(sub, lower, upper, TIPC_WITHDRAWN,
527 					sk->ref, sk->node, ua->scope, last);
528 	}
529 
530 	/* Remove service range item if this was its last publication */
531 	if (list_empty(&sr->all_publ)) {
532 		rb_erase_augmented(&sr->tree_node, &sc->ranges, &sr_callbacks);
533 		kfree(sr);
534 	}
535 
536 	/* Delete service item if this no more publications and subscriptions */
537 	if (RB_EMPTY_ROOT(&sc->ranges) && list_empty(&sc->subscriptions)) {
538 		hlist_del_init_rcu(&sc->service_list);
539 		kfree_rcu(sc, rcu);
540 	}
541 exit:
542 	spin_unlock_bh(&sc->lock);
543 	return p;
544 }
545 
546 /**
547  * tipc_nametbl_lookup_anycast - perform service instance to socket translation
548  * @net: network namespace
549  * @type: message type
550  * @instance: message instance
551  * @dnode: the search domain used during translation
552  *
553  * On entry, 'dnode' is the search domain used during the lookup
554  *
555  * On exit:
556  * - if lookup is deferred to another node, leave 'dnode' unchanged and return 0
557  * - if lookup is attempted and succeeds, set 'dnode' to the publishing node and
558  *   return the published (non-zero) port number
559  * - if lookup is attempted and fails, set 'dnode' to 0 and return 0
560  *
561  * Note that for legacy users (node configured with Z.C.N address format) the
562  * 'closest-first' lookup algorithm must be maintained, i.e., if dnode is 0
563  * we must look in the local binding list first
564  */
565 u32 tipc_nametbl_lookup_anycast(struct net *net, u32 type,
566 				u32 instance, u32 *dnode)
567 {
568 	struct tipc_net *tn = tipc_net(net);
569 	bool legacy = tn->legacy_addr_format;
570 	u32 self = tipc_own_addr(net);
571 	struct service_range *sr;
572 	struct tipc_service *sc;
573 	struct list_head *list;
574 	struct publication *p;
575 	u32 port = 0;
576 	u32 node = 0;
577 
578 	if (!tipc_in_scope(legacy, *dnode, self))
579 		return 0;
580 
581 	rcu_read_lock();
582 	sc = tipc_service_find(net, type);
583 	if (unlikely(!sc))
584 		goto exit;
585 
586 	spin_lock_bh(&sc->lock);
587 	service_range_foreach_match(sr, sc, instance, instance) {
588 		/* Select lookup algo: local, closest-first or round-robin */
589 		if (*dnode == self) {
590 			list = &sr->local_publ;
591 			if (list_empty(list))
592 				continue;
593 			p = list_first_entry(list, struct publication,
594 					     local_publ);
595 			list_move_tail(&p->local_publ, &sr->local_publ);
596 		} else if (legacy && !*dnode && !list_empty(&sr->local_publ)) {
597 			list = &sr->local_publ;
598 			p = list_first_entry(list, struct publication,
599 					     local_publ);
600 			list_move_tail(&p->local_publ, &sr->local_publ);
601 		} else {
602 			list = &sr->all_publ;
603 			p = list_first_entry(list, struct publication,
604 					     all_publ);
605 			list_move_tail(&p->all_publ, &sr->all_publ);
606 		}
607 		port = p->sk.ref;
608 		node = p->sk.node;
609 		/* Todo: as for legacy, pick the first matching range only, a
610 		 * "true" round-robin will be performed as needed.
611 		 */
612 		break;
613 	}
614 	spin_unlock_bh(&sc->lock);
615 
616 exit:
617 	rcu_read_unlock();
618 	*dnode = node;
619 	return port;
620 }
621 
622 /* tipc_nametbl_lookup_group(): lookup destinaton(s) in a communication group
623  * Returns a list of one (== group anycast) or more (== group multicast)
624  * destination socket/node pairs matching the given address.
625  * The requester may or may not want to exclude himself from the list.
626  */
627 bool tipc_nametbl_lookup_group(struct net *net, u32 type, u32 instance,
628 			       u32 scope, struct list_head *dsts,
629 			       int *dstcnt, u32 exclude,
630 			       bool mcast)
631 {
632 	u32 self = tipc_own_addr(net);
633 	struct service_range *sr;
634 	struct tipc_service *sc;
635 	struct publication *p;
636 
637 	*dstcnt = 0;
638 	rcu_read_lock();
639 	sc = tipc_service_find(net, type);
640 	if (unlikely(!sc))
641 		goto exit;
642 
643 	spin_lock_bh(&sc->lock);
644 
645 	/* Todo: a full search i.e. service_range_foreach_match() instead? */
646 	sr = service_range_match_first(sc->ranges.rb_node, instance, instance);
647 	if (!sr)
648 		goto no_match;
649 
650 	list_for_each_entry(p, &sr->all_publ, all_publ) {
651 		if (p->scope != scope)
652 			continue;
653 		if (p->sk.ref == exclude && p->sk.node == self)
654 			continue;
655 		tipc_dest_push(dsts, p->sk.node, p->sk.ref);
656 		(*dstcnt)++;
657 		if (mcast)
658 			continue;
659 		list_move_tail(&p->all_publ, &sr->all_publ);
660 		break;
661 	}
662 no_match:
663 	spin_unlock_bh(&sc->lock);
664 exit:
665 	rcu_read_unlock();
666 	return !list_empty(dsts);
667 }
668 
669 /* tipc_nametbl_lookup_mcast_sockets(): look up node local destinaton sockets
670  *                                      matching the given address
671  * Used on nodes which have received a multicast/broadcast message
672  * Returns a list of local sockets
673  */
674 void tipc_nametbl_lookup_mcast_sockets(struct net *net, u32 type, u32 lower,
675 				       u32 upper, u32 scope, bool exact,
676 				       struct list_head *dports)
677 {
678 	struct service_range *sr;
679 	struct tipc_service *sc;
680 	struct publication *p;
681 
682 	rcu_read_lock();
683 	sc = tipc_service_find(net, type);
684 	if (!sc)
685 		goto exit;
686 
687 	spin_lock_bh(&sc->lock);
688 	service_range_foreach_match(sr, sc, lower, upper) {
689 		list_for_each_entry(p, &sr->local_publ, local_publ) {
690 			if (p->scope == scope || (!exact && p->scope < scope))
691 				tipc_dest_push(dports, 0, p->sk.ref);
692 		}
693 	}
694 	spin_unlock_bh(&sc->lock);
695 exit:
696 	rcu_read_unlock();
697 }
698 
699 /* tipc_nametbl_lookup_mcast_nodes(): look up all destination nodes matching
700  *                                    the given address. Used in sending node.
701  * Used on nodes which are sending out a multicast/broadcast message
702  * Returns a list of nodes, including own node if applicable
703  */
704 void tipc_nametbl_lookup_mcast_nodes(struct net *net, u32 type, u32 lower,
705 				     u32 upper, struct tipc_nlist *nodes)
706 {
707 	struct service_range *sr;
708 	struct tipc_service *sc;
709 	struct publication *p;
710 
711 	rcu_read_lock();
712 	sc = tipc_service_find(net, type);
713 	if (!sc)
714 		goto exit;
715 
716 	spin_lock_bh(&sc->lock);
717 	service_range_foreach_match(sr, sc, lower, upper) {
718 		list_for_each_entry(p, &sr->all_publ, all_publ) {
719 			tipc_nlist_add(nodes, p->sk.node);
720 		}
721 	}
722 	spin_unlock_bh(&sc->lock);
723 exit:
724 	rcu_read_unlock();
725 }
726 
727 /* tipc_nametbl_build_group - build list of communication group members
728  */
729 void tipc_nametbl_build_group(struct net *net, struct tipc_group *grp,
730 			      u32 type, u32 scope)
731 {
732 	struct service_range *sr;
733 	struct tipc_service *sc;
734 	struct publication *p;
735 	struct rb_node *n;
736 
737 	rcu_read_lock();
738 	sc = tipc_service_find(net, type);
739 	if (!sc)
740 		goto exit;
741 
742 	spin_lock_bh(&sc->lock);
743 	for (n = rb_first(&sc->ranges); n; n = rb_next(n)) {
744 		sr = container_of(n, struct service_range, tree_node);
745 		list_for_each_entry(p, &sr->all_publ, all_publ) {
746 			if (p->scope != scope)
747 				continue;
748 			tipc_group_add_member(grp, p->sk.node, p->sk.ref, p->sr.lower);
749 		}
750 	}
751 	spin_unlock_bh(&sc->lock);
752 exit:
753 	rcu_read_unlock();
754 }
755 
756 /* tipc_nametbl_publish - add service binding to name table
757  */
758 struct publication *tipc_nametbl_publish(struct net *net, struct tipc_uaddr *ua,
759 					 struct tipc_socket_addr *sk, u32 key)
760 {
761 	struct name_table *nt = tipc_name_table(net);
762 	struct tipc_net *tn = tipc_net(net);
763 	struct publication *p = NULL;
764 	struct sk_buff *skb = NULL;
765 	u32 rc_dests;
766 
767 	spin_lock_bh(&tn->nametbl_lock);
768 
769 	if (nt->local_publ_count >= TIPC_MAX_PUBL) {
770 		pr_warn("Bind failed, max limit %u reached\n", TIPC_MAX_PUBL);
771 		goto exit;
772 	}
773 
774 	p = tipc_nametbl_insert_publ(net, ua, sk, key);
775 	if (p) {
776 		nt->local_publ_count++;
777 		skb = tipc_named_publish(net, p);
778 	}
779 	rc_dests = nt->rc_dests;
780 exit:
781 	spin_unlock_bh(&tn->nametbl_lock);
782 
783 	if (skb)
784 		tipc_node_broadcast(net, skb, rc_dests);
785 	return p;
786 
787 }
788 
789 /**
790  * tipc_nametbl_withdraw - withdraw a service binding
791  * @net: network namespace
792  * @ua: service address/range being unbound
793  * @sk: address of the socket being unbound from
794  * @key: target publication key
795  */
796 void tipc_nametbl_withdraw(struct net *net, struct tipc_uaddr *ua,
797 			   struct tipc_socket_addr *sk, u32 key)
798 {
799 	struct name_table *nt = tipc_name_table(net);
800 	struct tipc_net *tn = tipc_net(net);
801 	struct sk_buff *skb = NULL;
802 	struct publication *p;
803 	u32 rc_dests;
804 
805 	spin_lock_bh(&tn->nametbl_lock);
806 
807 	p = tipc_nametbl_remove_publ(net, ua, sk, key);
808 	if (p) {
809 		nt->local_publ_count--;
810 		skb = tipc_named_withdraw(net, p);
811 		list_del_init(&p->binding_sock);
812 		kfree_rcu(p, rcu);
813 	} else {
814 		pr_err("Failed to remove local publication {%u,%u,%u}/%u\n",
815 		       ua->sr.type, ua->sr.lower, ua->sr.upper, key);
816 	}
817 	rc_dests = nt->rc_dests;
818 	spin_unlock_bh(&tn->nametbl_lock);
819 
820 	if (skb)
821 		tipc_node_broadcast(net, skb, rc_dests);
822 }
823 
824 /**
825  * tipc_nametbl_subscribe - add a subscription object to the name table
826  * @sub: subscription to add
827  */
828 bool tipc_nametbl_subscribe(struct tipc_subscription *sub)
829 {
830 	struct name_table *nt = tipc_name_table(sub->net);
831 	struct tipc_net *tn = tipc_net(sub->net);
832 	struct tipc_subscr *s = &sub->evt.s;
833 	u32 type = tipc_sub_read(s, seq.type);
834 	struct tipc_service *sc;
835 	bool res = true;
836 
837 	spin_lock_bh(&tn->nametbl_lock);
838 	sc = tipc_service_find(sub->net, type);
839 	if (!sc)
840 		sc = tipc_service_create(type, &nt->services[hash(type)]);
841 	if (sc) {
842 		spin_lock_bh(&sc->lock);
843 		tipc_service_subscribe(sc, sub);
844 		spin_unlock_bh(&sc->lock);
845 	} else {
846 		pr_warn("Failed to subscribe for {%u,%u,%u}\n", type,
847 			tipc_sub_read(s, seq.lower),
848 			tipc_sub_read(s, seq.upper));
849 		res = false;
850 	}
851 	spin_unlock_bh(&tn->nametbl_lock);
852 	return res;
853 }
854 
855 /**
856  * tipc_nametbl_unsubscribe - remove a subscription object from name table
857  * @sub: subscription to remove
858  */
859 void tipc_nametbl_unsubscribe(struct tipc_subscription *sub)
860 {
861 	struct tipc_net *tn = tipc_net(sub->net);
862 	struct tipc_subscr *s = &sub->evt.s;
863 	u32 type = tipc_sub_read(s, seq.type);
864 	struct tipc_service *sc;
865 
866 	spin_lock_bh(&tn->nametbl_lock);
867 	sc = tipc_service_find(sub->net, type);
868 	if (!sc)
869 		goto exit;
870 
871 	spin_lock_bh(&sc->lock);
872 	list_del_init(&sub->service_list);
873 	tipc_sub_put(sub);
874 
875 	/* Delete service item if no more publications and subscriptions */
876 	if (RB_EMPTY_ROOT(&sc->ranges) && list_empty(&sc->subscriptions)) {
877 		hlist_del_init_rcu(&sc->service_list);
878 		kfree_rcu(sc, rcu);
879 	}
880 	spin_unlock_bh(&sc->lock);
881 exit:
882 	spin_unlock_bh(&tn->nametbl_lock);
883 }
884 
885 int tipc_nametbl_init(struct net *net)
886 {
887 	struct tipc_net *tn = tipc_net(net);
888 	struct name_table *nt;
889 	int i;
890 
891 	nt = kzalloc(sizeof(*nt), GFP_KERNEL);
892 	if (!nt)
893 		return -ENOMEM;
894 
895 	for (i = 0; i < TIPC_NAMETBL_SIZE; i++)
896 		INIT_HLIST_HEAD(&nt->services[i]);
897 
898 	INIT_LIST_HEAD(&nt->node_scope);
899 	INIT_LIST_HEAD(&nt->cluster_scope);
900 	rwlock_init(&nt->cluster_scope_lock);
901 	tn->nametbl = nt;
902 	spin_lock_init(&tn->nametbl_lock);
903 	return 0;
904 }
905 
906 /**
907  * tipc_service_delete - purge all publications for a service and delete it
908  * @net: the associated network namespace
909  * @sc: tipc_service to delete
910  */
911 static void tipc_service_delete(struct net *net, struct tipc_service *sc)
912 {
913 	struct service_range *sr, *tmpr;
914 	struct publication *p, *tmp;
915 
916 	spin_lock_bh(&sc->lock);
917 	rbtree_postorder_for_each_entry_safe(sr, tmpr, &sc->ranges, tree_node) {
918 		list_for_each_entry_safe(p, tmp, &sr->all_publ, all_publ) {
919 			tipc_service_remove_publ(sr, &p->sk, p->key);
920 			kfree_rcu(p, rcu);
921 		}
922 		rb_erase_augmented(&sr->tree_node, &sc->ranges, &sr_callbacks);
923 		kfree(sr);
924 	}
925 	hlist_del_init_rcu(&sc->service_list);
926 	spin_unlock_bh(&sc->lock);
927 	kfree_rcu(sc, rcu);
928 }
929 
930 void tipc_nametbl_stop(struct net *net)
931 {
932 	struct name_table *nt = tipc_name_table(net);
933 	struct tipc_net *tn = tipc_net(net);
934 	struct hlist_head *service_head;
935 	struct tipc_service *service;
936 	u32 i;
937 
938 	/* Verify name table is empty and purge any lingering
939 	 * publications, then release the name table
940 	 */
941 	spin_lock_bh(&tn->nametbl_lock);
942 	for (i = 0; i < TIPC_NAMETBL_SIZE; i++) {
943 		if (hlist_empty(&nt->services[i]))
944 			continue;
945 		service_head = &nt->services[i];
946 		hlist_for_each_entry_rcu(service, service_head, service_list) {
947 			tipc_service_delete(net, service);
948 		}
949 	}
950 	spin_unlock_bh(&tn->nametbl_lock);
951 
952 	synchronize_net();
953 	kfree(nt);
954 }
955 
956 static int __tipc_nl_add_nametable_publ(struct tipc_nl_msg *msg,
957 					struct tipc_service *service,
958 					struct service_range *sr,
959 					u32 *last_key)
960 {
961 	struct publication *p;
962 	struct nlattr *attrs;
963 	struct nlattr *b;
964 	void *hdr;
965 
966 	if (*last_key) {
967 		list_for_each_entry(p, &sr->all_publ, all_publ)
968 			if (p->key == *last_key)
969 				break;
970 		if (p->key != *last_key)
971 			return -EPIPE;
972 	} else {
973 		p = list_first_entry(&sr->all_publ,
974 				     struct publication,
975 				     all_publ);
976 	}
977 
978 	list_for_each_entry_from(p, &sr->all_publ, all_publ) {
979 		*last_key = p->key;
980 
981 		hdr = genlmsg_put(msg->skb, msg->portid, msg->seq,
982 				  &tipc_genl_family, NLM_F_MULTI,
983 				  TIPC_NL_NAME_TABLE_GET);
984 		if (!hdr)
985 			return -EMSGSIZE;
986 
987 		attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_NAME_TABLE);
988 		if (!attrs)
989 			goto msg_full;
990 
991 		b = nla_nest_start_noflag(msg->skb, TIPC_NLA_NAME_TABLE_PUBL);
992 		if (!b)
993 			goto attr_msg_full;
994 
995 		if (nla_put_u32(msg->skb, TIPC_NLA_PUBL_TYPE, service->type))
996 			goto publ_msg_full;
997 		if (nla_put_u32(msg->skb, TIPC_NLA_PUBL_LOWER, sr->lower))
998 			goto publ_msg_full;
999 		if (nla_put_u32(msg->skb, TIPC_NLA_PUBL_UPPER, sr->upper))
1000 			goto publ_msg_full;
1001 		if (nla_put_u32(msg->skb, TIPC_NLA_PUBL_SCOPE, p->scope))
1002 			goto publ_msg_full;
1003 		if (nla_put_u32(msg->skb, TIPC_NLA_PUBL_NODE, p->sk.node))
1004 			goto publ_msg_full;
1005 		if (nla_put_u32(msg->skb, TIPC_NLA_PUBL_REF, p->sk.ref))
1006 			goto publ_msg_full;
1007 		if (nla_put_u32(msg->skb, TIPC_NLA_PUBL_KEY, p->key))
1008 			goto publ_msg_full;
1009 
1010 		nla_nest_end(msg->skb, b);
1011 		nla_nest_end(msg->skb, attrs);
1012 		genlmsg_end(msg->skb, hdr);
1013 	}
1014 	*last_key = 0;
1015 
1016 	return 0;
1017 
1018 publ_msg_full:
1019 	nla_nest_cancel(msg->skb, b);
1020 attr_msg_full:
1021 	nla_nest_cancel(msg->skb, attrs);
1022 msg_full:
1023 	genlmsg_cancel(msg->skb, hdr);
1024 
1025 	return -EMSGSIZE;
1026 }
1027 
1028 static int __tipc_nl_service_range_list(struct tipc_nl_msg *msg,
1029 					struct tipc_service *sc,
1030 					u32 *last_lower, u32 *last_key)
1031 {
1032 	struct service_range *sr;
1033 	struct rb_node *n;
1034 	int err;
1035 
1036 	for (n = rb_first(&sc->ranges); n; n = rb_next(n)) {
1037 		sr = container_of(n, struct service_range, tree_node);
1038 		if (sr->lower < *last_lower)
1039 			continue;
1040 		err = __tipc_nl_add_nametable_publ(msg, sc, sr, last_key);
1041 		if (err) {
1042 			*last_lower = sr->lower;
1043 			return err;
1044 		}
1045 	}
1046 	*last_lower = 0;
1047 	return 0;
1048 }
1049 
1050 static int tipc_nl_service_list(struct net *net, struct tipc_nl_msg *msg,
1051 				u32 *last_type, u32 *last_lower, u32 *last_key)
1052 {
1053 	struct tipc_net *tn = tipc_net(net);
1054 	struct tipc_service *service = NULL;
1055 	struct hlist_head *head;
1056 	int err;
1057 	int i;
1058 
1059 	if (*last_type)
1060 		i = hash(*last_type);
1061 	else
1062 		i = 0;
1063 
1064 	for (; i < TIPC_NAMETBL_SIZE; i++) {
1065 		head = &tn->nametbl->services[i];
1066 
1067 		if (*last_type ||
1068 		    (!i && *last_key && (*last_lower == *last_key))) {
1069 			service = tipc_service_find(net, *last_type);
1070 			if (!service)
1071 				return -EPIPE;
1072 		} else {
1073 			hlist_for_each_entry_rcu(service, head, service_list)
1074 				break;
1075 			if (!service)
1076 				continue;
1077 		}
1078 
1079 		hlist_for_each_entry_from_rcu(service, service_list) {
1080 			spin_lock_bh(&service->lock);
1081 			err = __tipc_nl_service_range_list(msg, service,
1082 							   last_lower,
1083 							   last_key);
1084 
1085 			if (err) {
1086 				*last_type = service->type;
1087 				spin_unlock_bh(&service->lock);
1088 				return err;
1089 			}
1090 			spin_unlock_bh(&service->lock);
1091 		}
1092 		*last_type = 0;
1093 	}
1094 	return 0;
1095 }
1096 
1097 int tipc_nl_name_table_dump(struct sk_buff *skb, struct netlink_callback *cb)
1098 {
1099 	struct net *net = sock_net(skb->sk);
1100 	u32 last_type = cb->args[0];
1101 	u32 last_lower = cb->args[1];
1102 	u32 last_key = cb->args[2];
1103 	int done = cb->args[3];
1104 	struct tipc_nl_msg msg;
1105 	int err;
1106 
1107 	if (done)
1108 		return 0;
1109 
1110 	msg.skb = skb;
1111 	msg.portid = NETLINK_CB(cb->skb).portid;
1112 	msg.seq = cb->nlh->nlmsg_seq;
1113 
1114 	rcu_read_lock();
1115 	err = tipc_nl_service_list(net, &msg, &last_type,
1116 				   &last_lower, &last_key);
1117 	if (!err) {
1118 		done = 1;
1119 	} else if (err != -EMSGSIZE) {
1120 		/* We never set seq or call nl_dump_check_consistent() this
1121 		 * means that setting prev_seq here will cause the consistence
1122 		 * check to fail in the netlink callback handler. Resulting in
1123 		 * the NLMSG_DONE message having the NLM_F_DUMP_INTR flag set if
1124 		 * we got an error.
1125 		 */
1126 		cb->prev_seq = 1;
1127 	}
1128 	rcu_read_unlock();
1129 
1130 	cb->args[0] = last_type;
1131 	cb->args[1] = last_lower;
1132 	cb->args[2] = last_key;
1133 	cb->args[3] = done;
1134 
1135 	return skb->len;
1136 }
1137 
1138 struct tipc_dest *tipc_dest_find(struct list_head *l, u32 node, u32 port)
1139 {
1140 	struct tipc_dest *dst;
1141 
1142 	list_for_each_entry(dst, l, list) {
1143 		if (dst->node == node && dst->port == port)
1144 			return dst;
1145 	}
1146 	return NULL;
1147 }
1148 
1149 bool tipc_dest_push(struct list_head *l, u32 node, u32 port)
1150 {
1151 	struct tipc_dest *dst;
1152 
1153 	if (tipc_dest_find(l, node, port))
1154 		return false;
1155 
1156 	dst = kmalloc(sizeof(*dst), GFP_ATOMIC);
1157 	if (unlikely(!dst))
1158 		return false;
1159 	dst->node = node;
1160 	dst->port = port;
1161 	list_add(&dst->list, l);
1162 	return true;
1163 }
1164 
1165 bool tipc_dest_pop(struct list_head *l, u32 *node, u32 *port)
1166 {
1167 	struct tipc_dest *dst;
1168 
1169 	if (list_empty(l))
1170 		return false;
1171 	dst = list_first_entry(l, typeof(*dst), list);
1172 	if (port)
1173 		*port = dst->port;
1174 	if (node)
1175 		*node = dst->node;
1176 	list_del(&dst->list);
1177 	kfree(dst);
1178 	return true;
1179 }
1180 
1181 bool tipc_dest_del(struct list_head *l, u32 node, u32 port)
1182 {
1183 	struct tipc_dest *dst;
1184 
1185 	dst = tipc_dest_find(l, node, port);
1186 	if (!dst)
1187 		return false;
1188 	list_del(&dst->list);
1189 	kfree(dst);
1190 	return true;
1191 }
1192 
1193 void tipc_dest_list_purge(struct list_head *l)
1194 {
1195 	struct tipc_dest *dst, *tmp;
1196 
1197 	list_for_each_entry_safe(dst, tmp, l, list) {
1198 		list_del(&dst->list);
1199 		kfree(dst);
1200 	}
1201 }
1202 
1203 int tipc_dest_list_len(struct list_head *l)
1204 {
1205 	struct tipc_dest *dst;
1206 	int i = 0;
1207 
1208 	list_for_each_entry(dst, l, list) {
1209 		i++;
1210 	}
1211 	return i;
1212 }
1213