node.c revision a635b46bd884efc1fc98819cb5a200da255d575c
1/*
2 * net/tipc/node.c: TIPC node management routines
3 *
4 * Copyright (c) 2000-2006, Ericsson AB
5 * Copyright (c) 2005-2006, 2010-2011, 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 "core.h"
38#include "config.h"
39#include "node.h"
40#include "name_distr.h"
41
42#define NODE_HTABLE_SIZE 512
43
44static void node_lost_contact(struct tipc_node *n_ptr);
45static void node_established_contact(struct tipc_node *n_ptr);
46
47static DEFINE_SPINLOCK(node_create_lock);
48
49static struct hlist_head node_htable[NODE_HTABLE_SIZE];
50LIST_HEAD(tipc_node_list);
51static u32 tipc_num_nodes;
52
53static atomic_t tipc_num_links = ATOMIC_INIT(0);
54
55/*
56 * A trivial power-of-two bitmask technique is used for speed, since this
57 * operation is done for every incoming TIPC packet. The number of hash table
58 * entries has been chosen so that no hash chain exceeds 8 nodes and will
59 * usually be much smaller (typically only a single node).
60 */
61static inline unsigned int tipc_hashfn(u32 addr)
62{
63	return addr & (NODE_HTABLE_SIZE - 1);
64}
65
66/*
67 * tipc_node_find - locate specified node object, if it exists
68 */
69
70struct tipc_node *tipc_node_find(u32 addr)
71{
72	struct tipc_node *node;
73	struct hlist_node *pos;
74
75	if (unlikely(!in_own_cluster(addr)))
76		return NULL;
77
78	hlist_for_each_entry(node, pos, &node_htable[tipc_hashfn(addr)], hash) {
79		if (node->addr == addr)
80			return node;
81	}
82	return NULL;
83}
84
85/**
86 * tipc_node_create - create neighboring node
87 *
88 * Currently, this routine is called by neighbor discovery code, which holds
89 * net_lock for reading only.  We must take node_create_lock to ensure a node
90 * isn't created twice if two different bearers discover the node at the same
91 * time.  (It would be preferable to switch to holding net_lock in write mode,
92 * but this is a non-trivial change.)
93 */
94
95struct tipc_node *tipc_node_create(u32 addr)
96{
97	struct tipc_node *n_ptr, *temp_node;
98
99	spin_lock_bh(&node_create_lock);
100
101	n_ptr = tipc_node_find(addr);
102	if (n_ptr) {
103		spin_unlock_bh(&node_create_lock);
104		return n_ptr;
105	}
106
107	n_ptr = kzalloc(sizeof(*n_ptr), GFP_ATOMIC);
108	if (!n_ptr) {
109		spin_unlock_bh(&node_create_lock);
110		warn("Node creation failed, no memory\n");
111		return NULL;
112	}
113
114	n_ptr->addr = addr;
115	spin_lock_init(&n_ptr->lock);
116	INIT_HLIST_NODE(&n_ptr->hash);
117	INIT_LIST_HEAD(&n_ptr->list);
118	INIT_LIST_HEAD(&n_ptr->nsub);
119
120	hlist_add_head(&n_ptr->hash, &node_htable[tipc_hashfn(addr)]);
121
122	list_for_each_entry(temp_node, &tipc_node_list, list) {
123		if (n_ptr->addr < temp_node->addr)
124			break;
125	}
126	list_add_tail(&n_ptr->list, &temp_node->list);
127	n_ptr->block_setup = WAIT_PEER_DOWN;
128	n_ptr->signature = INVALID_NODE_SIG;
129
130	tipc_num_nodes++;
131
132	spin_unlock_bh(&node_create_lock);
133	return n_ptr;
134}
135
136void tipc_node_delete(struct tipc_node *n_ptr)
137{
138	list_del(&n_ptr->list);
139	hlist_del(&n_ptr->hash);
140	kfree(n_ptr);
141
142	tipc_num_nodes--;
143}
144
145
146/**
147 * tipc_node_link_up - handle addition of link
148 *
149 * Link becomes active (alone or shared) or standby, depending on its priority.
150 */
151
152void tipc_node_link_up(struct tipc_node *n_ptr, struct tipc_link *l_ptr)
153{
154	struct tipc_link **active = &n_ptr->active_links[0];
155
156	n_ptr->working_links++;
157
158	info("Established link <%s> on network plane %c\n",
159	     l_ptr->name, l_ptr->b_ptr->net_plane);
160
161	if (!active[0]) {
162		active[0] = active[1] = l_ptr;
163		node_established_contact(n_ptr);
164		return;
165	}
166	if (l_ptr->priority < active[0]->priority) {
167		info("New link <%s> becomes standby\n", l_ptr->name);
168		return;
169	}
170	tipc_link_send_duplicate(active[0], l_ptr);
171	if (l_ptr->priority == active[0]->priority) {
172		active[0] = l_ptr;
173		return;
174	}
175	info("Old link <%s> becomes standby\n", active[0]->name);
176	if (active[1] != active[0])
177		info("Old link <%s> becomes standby\n", active[1]->name);
178	active[0] = active[1] = l_ptr;
179}
180
181/**
182 * node_select_active_links - select active link
183 */
184
185static void node_select_active_links(struct tipc_node *n_ptr)
186{
187	struct tipc_link **active = &n_ptr->active_links[0];
188	u32 i;
189	u32 highest_prio = 0;
190
191	active[0] = active[1] = NULL;
192
193	for (i = 0; i < MAX_BEARERS; i++) {
194		struct tipc_link *l_ptr = n_ptr->links[i];
195
196		if (!l_ptr || !tipc_link_is_up(l_ptr) ||
197		    (l_ptr->priority < highest_prio))
198			continue;
199
200		if (l_ptr->priority > highest_prio) {
201			highest_prio = l_ptr->priority;
202			active[0] = active[1] = l_ptr;
203		} else {
204			active[1] = l_ptr;
205		}
206	}
207}
208
209/**
210 * tipc_node_link_down - handle loss of link
211 */
212
213void tipc_node_link_down(struct tipc_node *n_ptr, struct tipc_link *l_ptr)
214{
215	struct tipc_link **active;
216
217	n_ptr->working_links--;
218
219	if (!tipc_link_is_active(l_ptr)) {
220		info("Lost standby link <%s> on network plane %c\n",
221		     l_ptr->name, l_ptr->b_ptr->net_plane);
222		return;
223	}
224	info("Lost link <%s> on network plane %c\n",
225		l_ptr->name, l_ptr->b_ptr->net_plane);
226
227	active = &n_ptr->active_links[0];
228	if (active[0] == l_ptr)
229		active[0] = active[1];
230	if (active[1] == l_ptr)
231		active[1] = active[0];
232	if (active[0] == l_ptr)
233		node_select_active_links(n_ptr);
234	if (tipc_node_is_up(n_ptr))
235		tipc_link_changeover(l_ptr);
236	else
237		node_lost_contact(n_ptr);
238}
239
240int tipc_node_active_links(struct tipc_node *n_ptr)
241{
242	return n_ptr->active_links[0] != NULL;
243}
244
245int tipc_node_redundant_links(struct tipc_node *n_ptr)
246{
247	return n_ptr->working_links > 1;
248}
249
250int tipc_node_is_up(struct tipc_node *n_ptr)
251{
252	return tipc_node_active_links(n_ptr);
253}
254
255void tipc_node_attach_link(struct tipc_node *n_ptr, struct tipc_link *l_ptr)
256{
257	n_ptr->links[l_ptr->b_ptr->identity] = l_ptr;
258	atomic_inc(&tipc_num_links);
259	n_ptr->link_cnt++;
260}
261
262void tipc_node_detach_link(struct tipc_node *n_ptr, struct tipc_link *l_ptr)
263{
264	n_ptr->links[l_ptr->b_ptr->identity] = NULL;
265	atomic_dec(&tipc_num_links);
266	n_ptr->link_cnt--;
267}
268
269/*
270 * Routing table management - five cases to handle:
271 *
272 * 1: A link towards a zone/cluster external node comes up.
273 *    => Send a multicast message updating routing tables of all
274 *    system nodes within own cluster that the new destination
275 *    can be reached via this node.
276 *    (node.establishedContact()=>cluster.multicastNewRoute())
277 *
278 * 2: A link towards a slave node comes up.
279 *    => Send a multicast message updating routing tables of all
280 *    system nodes within own cluster that the new destination
281 *    can be reached via this node.
282 *    (node.establishedContact()=>cluster.multicastNewRoute())
283 *    => Send a  message to the slave node about existence
284 *    of all system nodes within cluster:
285 *    (node.establishedContact()=>cluster.sendLocalRoutes())
286 *
287 * 3: A new cluster local system node becomes available.
288 *    => Send message(s) to this particular node containing
289 *    information about all cluster external and slave
290 *     nodes which can be reached via this node.
291 *    (node.establishedContact()==>network.sendExternalRoutes())
292 *    (node.establishedContact()==>network.sendSlaveRoutes())
293 *    => Send messages to all directly connected slave nodes
294 *    containing information about the existence of the new node
295 *    (node.establishedContact()=>cluster.multicastNewRoute())
296 *
297 * 4: The link towards a zone/cluster external node or slave
298 *    node goes down.
299 *    => Send a multcast message updating routing tables of all
300 *    nodes within cluster that the new destination can not any
301 *    longer be reached via this node.
302 *    (node.lostAllLinks()=>cluster.bcastLostRoute())
303 *
304 * 5: A cluster local system node becomes unavailable.
305 *    => Remove all references to this node from the local
306 *    routing tables. Note: This is a completely node
307 *    local operation.
308 *    (node.lostAllLinks()=>network.removeAsRouter())
309 *    => Send messages to all directly connected slave nodes
310 *    containing information about loss of the node
311 *    (node.establishedContact()=>cluster.multicastLostRoute())
312 *
313 */
314
315static void node_established_contact(struct tipc_node *n_ptr)
316{
317	tipc_k_signal((Handler)tipc_named_node_up, n_ptr->addr);
318
319	/* Syncronize broadcast acks */
320	n_ptr->bclink.acked = tipc_bclink_get_last_sent();
321
322	if (n_ptr->bclink.supportable) {
323		tipc_bclink_add_node(n_ptr->addr);
324		n_ptr->bclink.supported = 1;
325	}
326}
327
328static void node_name_purge_complete(unsigned long node_addr)
329{
330	struct tipc_node *n_ptr;
331
332	read_lock_bh(&tipc_net_lock);
333	n_ptr = tipc_node_find(node_addr);
334	if (n_ptr) {
335		tipc_node_lock(n_ptr);
336		n_ptr->block_setup &= ~WAIT_NAMES_GONE;
337		tipc_node_unlock(n_ptr);
338	}
339	read_unlock_bh(&tipc_net_lock);
340}
341
342static void node_lost_contact(struct tipc_node *n_ptr)
343{
344	char addr_string[16];
345	u32 i;
346
347	info("Lost contact with %s\n",
348	     tipc_addr_string_fill(addr_string, n_ptr->addr));
349
350	/* Flush broadcast link info associated with lost node */
351
352	if (n_ptr->bclink.supported) {
353		while (n_ptr->bclink.deferred_head) {
354			struct sk_buff *buf = n_ptr->bclink.deferred_head;
355			n_ptr->bclink.deferred_head = buf->next;
356			buf_discard(buf);
357		}
358		n_ptr->bclink.deferred_size = 0;
359
360		if (n_ptr->bclink.defragm) {
361			buf_discard(n_ptr->bclink.defragm);
362			n_ptr->bclink.defragm = NULL;
363		}
364
365		tipc_bclink_remove_node(n_ptr->addr);
366		tipc_bclink_acknowledge(n_ptr, INVALID_LINK_SEQ);
367
368		n_ptr->bclink.supported = 0;
369	}
370
371	/* Abort link changeover */
372	for (i = 0; i < MAX_BEARERS; i++) {
373		struct tipc_link *l_ptr = n_ptr->links[i];
374		if (!l_ptr)
375			continue;
376		l_ptr->reset_checkpoint = l_ptr->next_in_no;
377		l_ptr->exp_msg_count = 0;
378		tipc_link_reset_fragments(l_ptr);
379	}
380
381	/* Notify subscribers */
382	tipc_nodesub_notify(n_ptr);
383
384	/* Prevent re-contact with node until cleanup is done */
385
386	n_ptr->block_setup = WAIT_PEER_DOWN | WAIT_NAMES_GONE;
387	tipc_k_signal((Handler)node_name_purge_complete, n_ptr->addr);
388}
389
390struct sk_buff *tipc_node_get_nodes(const void *req_tlv_area, int req_tlv_space)
391{
392	u32 domain;
393	struct sk_buff *buf;
394	struct tipc_node *n_ptr;
395	struct tipc_node_info node_info;
396	u32 payload_size;
397
398	if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_NET_ADDR))
399		return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
400
401	domain = ntohl(*(__be32 *)TLV_DATA(req_tlv_area));
402	if (!tipc_addr_domain_valid(domain))
403		return tipc_cfg_reply_error_string(TIPC_CFG_INVALID_VALUE
404						   " (network address)");
405
406	read_lock_bh(&tipc_net_lock);
407	if (!tipc_num_nodes) {
408		read_unlock_bh(&tipc_net_lock);
409		return tipc_cfg_reply_none();
410	}
411
412	/* For now, get space for all other nodes */
413
414	payload_size = TLV_SPACE(sizeof(node_info)) * tipc_num_nodes;
415	if (payload_size > 32768u) {
416		read_unlock_bh(&tipc_net_lock);
417		return tipc_cfg_reply_error_string(TIPC_CFG_NOT_SUPPORTED
418						   " (too many nodes)");
419	}
420	buf = tipc_cfg_reply_alloc(payload_size);
421	if (!buf) {
422		read_unlock_bh(&tipc_net_lock);
423		return NULL;
424	}
425
426	/* Add TLVs for all nodes in scope */
427
428	list_for_each_entry(n_ptr, &tipc_node_list, list) {
429		if (!tipc_in_scope(domain, n_ptr->addr))
430			continue;
431		node_info.addr = htonl(n_ptr->addr);
432		node_info.up = htonl(tipc_node_is_up(n_ptr));
433		tipc_cfg_append_tlv(buf, TIPC_TLV_NODE_INFO,
434				    &node_info, sizeof(node_info));
435	}
436
437	read_unlock_bh(&tipc_net_lock);
438	return buf;
439}
440
441struct sk_buff *tipc_node_get_links(const void *req_tlv_area, int req_tlv_space)
442{
443	u32 domain;
444	struct sk_buff *buf;
445	struct tipc_node *n_ptr;
446	struct tipc_link_info link_info;
447	u32 payload_size;
448
449	if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_NET_ADDR))
450		return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
451
452	domain = ntohl(*(__be32 *)TLV_DATA(req_tlv_area));
453	if (!tipc_addr_domain_valid(domain))
454		return tipc_cfg_reply_error_string(TIPC_CFG_INVALID_VALUE
455						   " (network address)");
456
457	if (tipc_mode != TIPC_NET_MODE)
458		return tipc_cfg_reply_none();
459
460	read_lock_bh(&tipc_net_lock);
461
462	/* Get space for all unicast links + broadcast link */
463
464	payload_size = TLV_SPACE(sizeof(link_info)) *
465		(atomic_read(&tipc_num_links) + 1);
466	if (payload_size > 32768u) {
467		read_unlock_bh(&tipc_net_lock);
468		return tipc_cfg_reply_error_string(TIPC_CFG_NOT_SUPPORTED
469						   " (too many links)");
470	}
471	buf = tipc_cfg_reply_alloc(payload_size);
472	if (!buf) {
473		read_unlock_bh(&tipc_net_lock);
474		return NULL;
475	}
476
477	/* Add TLV for broadcast link */
478
479	link_info.dest = htonl(tipc_cluster_mask(tipc_own_addr));
480	link_info.up = htonl(1);
481	strlcpy(link_info.str, tipc_bclink_name, TIPC_MAX_LINK_NAME);
482	tipc_cfg_append_tlv(buf, TIPC_TLV_LINK_INFO, &link_info, sizeof(link_info));
483
484	/* Add TLVs for any other links in scope */
485
486	list_for_each_entry(n_ptr, &tipc_node_list, list) {
487		u32 i;
488
489		if (!tipc_in_scope(domain, n_ptr->addr))
490			continue;
491		tipc_node_lock(n_ptr);
492		for (i = 0; i < MAX_BEARERS; i++) {
493			if (!n_ptr->links[i])
494				continue;
495			link_info.dest = htonl(n_ptr->addr);
496			link_info.up = htonl(tipc_link_is_up(n_ptr->links[i]));
497			strcpy(link_info.str, n_ptr->links[i]->name);
498			tipc_cfg_append_tlv(buf, TIPC_TLV_LINK_INFO,
499					    &link_info, sizeof(link_info));
500		}
501		tipc_node_unlock(n_ptr);
502	}
503
504	read_unlock_bh(&tipc_net_lock);
505	return buf;
506}
507