xref: /linux/fs/ocfs2/cluster/nodemanager.c (revision e58e871becec2d3b04ed91c0c16fe8deac9c9dfa)
1 /* -*- mode: c; c-basic-offset: 8; -*-
2  * vim: noexpandtab sw=8 ts=8 sts=0:
3  *
4  * Copyright (C) 2004, 2005 Oracle.  All rights reserved.
5  *
6  * This program is free software; you can redistribute it and/or
7  * modify it under the terms of the GNU General Public
8  * License as published by the Free Software Foundation; either
9  * version 2 of the License, or (at your option) any later version.
10  *
11  * This program is distributed in the hope that it will be useful,
12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
14  * General Public License for more details.
15  *
16  * You should have received a copy of the GNU General Public
17  * License along with this program; if not, write to the
18  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
19  * Boston, MA 021110-1307, USA.
20  */
21 
22 #include <linux/slab.h>
23 #include <linux/kernel.h>
24 #include <linux/module.h>
25 #include <linux/configfs.h>
26 
27 #include "tcp.h"
28 #include "nodemanager.h"
29 #include "heartbeat.h"
30 #include "masklog.h"
31 #include "sys.h"
32 
33 /* for now we operate under the assertion that there can be only one
34  * cluster active at a time.  Changing this will require trickling
35  * cluster references throughout where nodes are looked up */
36 struct o2nm_cluster *o2nm_single_cluster = NULL;
37 
38 char *o2nm_fence_method_desc[O2NM_FENCE_METHODS] = {
39 		"reset",	/* O2NM_FENCE_RESET */
40 		"panic",	/* O2NM_FENCE_PANIC */
41 };
42 
43 struct o2nm_node *o2nm_get_node_by_num(u8 node_num)
44 {
45 	struct o2nm_node *node = NULL;
46 
47 	if (node_num >= O2NM_MAX_NODES || o2nm_single_cluster == NULL)
48 		goto out;
49 
50 	read_lock(&o2nm_single_cluster->cl_nodes_lock);
51 	node = o2nm_single_cluster->cl_nodes[node_num];
52 	if (node)
53 		config_item_get(&node->nd_item);
54 	read_unlock(&o2nm_single_cluster->cl_nodes_lock);
55 out:
56 	return node;
57 }
58 EXPORT_SYMBOL_GPL(o2nm_get_node_by_num);
59 
60 int o2nm_configured_node_map(unsigned long *map, unsigned bytes)
61 {
62 	struct o2nm_cluster *cluster = o2nm_single_cluster;
63 
64 	BUG_ON(bytes < (sizeof(cluster->cl_nodes_bitmap)));
65 
66 	if (cluster == NULL)
67 		return -EINVAL;
68 
69 	read_lock(&cluster->cl_nodes_lock);
70 	memcpy(map, cluster->cl_nodes_bitmap, sizeof(cluster->cl_nodes_bitmap));
71 	read_unlock(&cluster->cl_nodes_lock);
72 
73 	return 0;
74 }
75 EXPORT_SYMBOL_GPL(o2nm_configured_node_map);
76 
77 static struct o2nm_node *o2nm_node_ip_tree_lookup(struct o2nm_cluster *cluster,
78 						  __be32 ip_needle,
79 						  struct rb_node ***ret_p,
80 						  struct rb_node **ret_parent)
81 {
82 	struct rb_node **p = &cluster->cl_node_ip_tree.rb_node;
83 	struct rb_node *parent = NULL;
84 	struct o2nm_node *node, *ret = NULL;
85 
86 	while (*p) {
87 		int cmp;
88 
89 		parent = *p;
90 		node = rb_entry(parent, struct o2nm_node, nd_ip_node);
91 
92 		cmp = memcmp(&ip_needle, &node->nd_ipv4_address,
93 				sizeof(ip_needle));
94 		if (cmp < 0)
95 			p = &(*p)->rb_left;
96 		else if (cmp > 0)
97 			p = &(*p)->rb_right;
98 		else {
99 			ret = node;
100 			break;
101 		}
102 	}
103 
104 	if (ret_p != NULL)
105 		*ret_p = p;
106 	if (ret_parent != NULL)
107 		*ret_parent = parent;
108 
109 	return ret;
110 }
111 
112 struct o2nm_node *o2nm_get_node_by_ip(__be32 addr)
113 {
114 	struct o2nm_node *node = NULL;
115 	struct o2nm_cluster *cluster = o2nm_single_cluster;
116 
117 	if (cluster == NULL)
118 		goto out;
119 
120 	read_lock(&cluster->cl_nodes_lock);
121 	node = o2nm_node_ip_tree_lookup(cluster, addr, NULL, NULL);
122 	if (node)
123 		config_item_get(&node->nd_item);
124 	read_unlock(&cluster->cl_nodes_lock);
125 
126 out:
127 	return node;
128 }
129 EXPORT_SYMBOL_GPL(o2nm_get_node_by_ip);
130 
131 void o2nm_node_put(struct o2nm_node *node)
132 {
133 	config_item_put(&node->nd_item);
134 }
135 EXPORT_SYMBOL_GPL(o2nm_node_put);
136 
137 void o2nm_node_get(struct o2nm_node *node)
138 {
139 	config_item_get(&node->nd_item);
140 }
141 EXPORT_SYMBOL_GPL(o2nm_node_get);
142 
143 u8 o2nm_this_node(void)
144 {
145 	u8 node_num = O2NM_MAX_NODES;
146 
147 	if (o2nm_single_cluster && o2nm_single_cluster->cl_has_local)
148 		node_num = o2nm_single_cluster->cl_local_node;
149 
150 	return node_num;
151 }
152 EXPORT_SYMBOL_GPL(o2nm_this_node);
153 
154 /* node configfs bits */
155 
156 static struct o2nm_cluster *to_o2nm_cluster(struct config_item *item)
157 {
158 	return item ?
159 		container_of(to_config_group(item), struct o2nm_cluster,
160 			     cl_group)
161 		: NULL;
162 }
163 
164 static struct o2nm_node *to_o2nm_node(struct config_item *item)
165 {
166 	return item ? container_of(item, struct o2nm_node, nd_item) : NULL;
167 }
168 
169 static void o2nm_node_release(struct config_item *item)
170 {
171 	struct o2nm_node *node = to_o2nm_node(item);
172 	kfree(node);
173 }
174 
175 static ssize_t o2nm_node_num_show(struct config_item *item, char *page)
176 {
177 	return sprintf(page, "%d\n", to_o2nm_node(item)->nd_num);
178 }
179 
180 static struct o2nm_cluster *to_o2nm_cluster_from_node(struct o2nm_node *node)
181 {
182 	/* through the first node_set .parent
183 	 * mycluster/nodes/mynode == o2nm_cluster->o2nm_node_group->o2nm_node */
184 	return to_o2nm_cluster(node->nd_item.ci_parent->ci_parent);
185 }
186 
187 enum {
188 	O2NM_NODE_ATTR_NUM = 0,
189 	O2NM_NODE_ATTR_PORT,
190 	O2NM_NODE_ATTR_ADDRESS,
191 };
192 
193 static ssize_t o2nm_node_num_store(struct config_item *item, const char *page,
194 				   size_t count)
195 {
196 	struct o2nm_node *node = to_o2nm_node(item);
197 	struct o2nm_cluster *cluster = to_o2nm_cluster_from_node(node);
198 	unsigned long tmp;
199 	char *p = (char *)page;
200 	int ret = 0;
201 
202 	tmp = simple_strtoul(p, &p, 0);
203 	if (!p || (*p && (*p != '\n')))
204 		return -EINVAL;
205 
206 	if (tmp >= O2NM_MAX_NODES)
207 		return -ERANGE;
208 
209 	/* once we're in the cl_nodes tree networking can look us up by
210 	 * node number and try to use our address and port attributes
211 	 * to connect to this node.. make sure that they've been set
212 	 * before writing the node attribute? */
213 	if (!test_bit(O2NM_NODE_ATTR_ADDRESS, &node->nd_set_attributes) ||
214 	    !test_bit(O2NM_NODE_ATTR_PORT, &node->nd_set_attributes))
215 		return -EINVAL; /* XXX */
216 
217 	write_lock(&cluster->cl_nodes_lock);
218 	if (cluster->cl_nodes[tmp])
219 		ret = -EEXIST;
220 	else if (test_and_set_bit(O2NM_NODE_ATTR_NUM,
221 			&node->nd_set_attributes))
222 		ret = -EBUSY;
223 	else  {
224 		cluster->cl_nodes[tmp] = node;
225 		node->nd_num = tmp;
226 		set_bit(tmp, cluster->cl_nodes_bitmap);
227 	}
228 	write_unlock(&cluster->cl_nodes_lock);
229 	if (ret)
230 		return ret;
231 
232 	return count;
233 }
234 static ssize_t o2nm_node_ipv4_port_show(struct config_item *item, char *page)
235 {
236 	return sprintf(page, "%u\n", ntohs(to_o2nm_node(item)->nd_ipv4_port));
237 }
238 
239 static ssize_t o2nm_node_ipv4_port_store(struct config_item *item,
240 					 const char *page, size_t count)
241 {
242 	struct o2nm_node *node = to_o2nm_node(item);
243 	unsigned long tmp;
244 	char *p = (char *)page;
245 
246 	tmp = simple_strtoul(p, &p, 0);
247 	if (!p || (*p && (*p != '\n')))
248 		return -EINVAL;
249 
250 	if (tmp == 0)
251 		return -EINVAL;
252 	if (tmp >= (u16)-1)
253 		return -ERANGE;
254 
255 	if (test_and_set_bit(O2NM_NODE_ATTR_PORT, &node->nd_set_attributes))
256 		return -EBUSY;
257 	node->nd_ipv4_port = htons(tmp);
258 
259 	return count;
260 }
261 
262 static ssize_t o2nm_node_ipv4_address_show(struct config_item *item, char *page)
263 {
264 	return sprintf(page, "%pI4\n", &to_o2nm_node(item)->nd_ipv4_address);
265 }
266 
267 static ssize_t o2nm_node_ipv4_address_store(struct config_item *item,
268 					    const char *page,
269 					    size_t count)
270 {
271 	struct o2nm_node *node = to_o2nm_node(item);
272 	struct o2nm_cluster *cluster = to_o2nm_cluster_from_node(node);
273 	int ret, i;
274 	struct rb_node **p, *parent;
275 	unsigned int octets[4];
276 	__be32 ipv4_addr = 0;
277 
278 	ret = sscanf(page, "%3u.%3u.%3u.%3u", &octets[3], &octets[2],
279 		     &octets[1], &octets[0]);
280 	if (ret != 4)
281 		return -EINVAL;
282 
283 	for (i = 0; i < ARRAY_SIZE(octets); i++) {
284 		if (octets[i] > 255)
285 			return -ERANGE;
286 		be32_add_cpu(&ipv4_addr, octets[i] << (i * 8));
287 	}
288 
289 	ret = 0;
290 	write_lock(&cluster->cl_nodes_lock);
291 	if (o2nm_node_ip_tree_lookup(cluster, ipv4_addr, &p, &parent))
292 		ret = -EEXIST;
293 	else if (test_and_set_bit(O2NM_NODE_ATTR_ADDRESS,
294 			&node->nd_set_attributes))
295 		ret = -EBUSY;
296 	else {
297 		rb_link_node(&node->nd_ip_node, parent, p);
298 		rb_insert_color(&node->nd_ip_node, &cluster->cl_node_ip_tree);
299 	}
300 	write_unlock(&cluster->cl_nodes_lock);
301 	if (ret)
302 		return ret;
303 
304 	memcpy(&node->nd_ipv4_address, &ipv4_addr, sizeof(ipv4_addr));
305 
306 	return count;
307 }
308 
309 static ssize_t o2nm_node_local_show(struct config_item *item, char *page)
310 {
311 	return sprintf(page, "%d\n", to_o2nm_node(item)->nd_local);
312 }
313 
314 static ssize_t o2nm_node_local_store(struct config_item *item, const char *page,
315 				     size_t count)
316 {
317 	struct o2nm_node *node = to_o2nm_node(item);
318 	struct o2nm_cluster *cluster = to_o2nm_cluster_from_node(node);
319 	unsigned long tmp;
320 	char *p = (char *)page;
321 	ssize_t ret;
322 
323 	tmp = simple_strtoul(p, &p, 0);
324 	if (!p || (*p && (*p != '\n')))
325 		return -EINVAL;
326 
327 	tmp = !!tmp; /* boolean of whether this node wants to be local */
328 
329 	/* setting local turns on networking rx for now so we require having
330 	 * set everything else first */
331 	if (!test_bit(O2NM_NODE_ATTR_ADDRESS, &node->nd_set_attributes) ||
332 	    !test_bit(O2NM_NODE_ATTR_NUM, &node->nd_set_attributes) ||
333 	    !test_bit(O2NM_NODE_ATTR_PORT, &node->nd_set_attributes))
334 		return -EINVAL; /* XXX */
335 
336 	/* the only failure case is trying to set a new local node
337 	 * when a different one is already set */
338 	if (tmp && tmp == cluster->cl_has_local &&
339 	    cluster->cl_local_node != node->nd_num)
340 		return -EBUSY;
341 
342 	/* bring up the rx thread if we're setting the new local node. */
343 	if (tmp && !cluster->cl_has_local) {
344 		ret = o2net_start_listening(node);
345 		if (ret)
346 			return ret;
347 	}
348 
349 	if (!tmp && cluster->cl_has_local &&
350 	    cluster->cl_local_node == node->nd_num) {
351 		o2net_stop_listening(node);
352 		cluster->cl_local_node = O2NM_INVALID_NODE_NUM;
353 	}
354 
355 	node->nd_local = tmp;
356 	if (node->nd_local) {
357 		cluster->cl_has_local = tmp;
358 		cluster->cl_local_node = node->nd_num;
359 	}
360 
361 	return count;
362 }
363 
364 CONFIGFS_ATTR(o2nm_node_, num);
365 CONFIGFS_ATTR(o2nm_node_, ipv4_port);
366 CONFIGFS_ATTR(o2nm_node_, ipv4_address);
367 CONFIGFS_ATTR(o2nm_node_, local);
368 
369 static struct configfs_attribute *o2nm_node_attrs[] = {
370 	&o2nm_node_attr_num,
371 	&o2nm_node_attr_ipv4_port,
372 	&o2nm_node_attr_ipv4_address,
373 	&o2nm_node_attr_local,
374 	NULL,
375 };
376 
377 static struct configfs_item_operations o2nm_node_item_ops = {
378 	.release		= o2nm_node_release,
379 };
380 
381 static struct config_item_type o2nm_node_type = {
382 	.ct_item_ops	= &o2nm_node_item_ops,
383 	.ct_attrs	= o2nm_node_attrs,
384 	.ct_owner	= THIS_MODULE,
385 };
386 
387 /* node set */
388 
389 struct o2nm_node_group {
390 	struct config_group ns_group;
391 	/* some stuff? */
392 };
393 
394 #if 0
395 static struct o2nm_node_group *to_o2nm_node_group(struct config_group *group)
396 {
397 	return group ?
398 		container_of(group, struct o2nm_node_group, ns_group)
399 		: NULL;
400 }
401 #endif
402 
403 static ssize_t o2nm_cluster_attr_write(const char *page, ssize_t count,
404                                        unsigned int *val)
405 {
406 	unsigned long tmp;
407 	char *p = (char *)page;
408 
409 	tmp = simple_strtoul(p, &p, 0);
410 	if (!p || (*p && (*p != '\n')))
411 		return -EINVAL;
412 
413 	if (tmp == 0)
414 		return -EINVAL;
415 	if (tmp >= (u32)-1)
416 		return -ERANGE;
417 
418 	*val = tmp;
419 
420 	return count;
421 }
422 
423 static ssize_t o2nm_cluster_idle_timeout_ms_show(struct config_item *item,
424 	char *page)
425 {
426 	return sprintf(page, "%u\n", to_o2nm_cluster(item)->cl_idle_timeout_ms);
427 }
428 
429 static ssize_t o2nm_cluster_idle_timeout_ms_store(struct config_item *item,
430 	const char *page, size_t count)
431 {
432 	struct o2nm_cluster *cluster = to_o2nm_cluster(item);
433 	ssize_t ret;
434 	unsigned int val;
435 
436 	ret =  o2nm_cluster_attr_write(page, count, &val);
437 
438 	if (ret > 0) {
439 		if (cluster->cl_idle_timeout_ms != val
440 			&& o2net_num_connected_peers()) {
441 			mlog(ML_NOTICE,
442 			     "o2net: cannot change idle timeout after "
443 			     "the first peer has agreed to it."
444 			     "  %d connected peers\n",
445 			     o2net_num_connected_peers());
446 			ret = -EINVAL;
447 		} else if (val <= cluster->cl_keepalive_delay_ms) {
448 			mlog(ML_NOTICE, "o2net: idle timeout must be larger "
449 			     "than keepalive delay\n");
450 			ret = -EINVAL;
451 		} else {
452 			cluster->cl_idle_timeout_ms = val;
453 		}
454 	}
455 
456 	return ret;
457 }
458 
459 static ssize_t o2nm_cluster_keepalive_delay_ms_show(
460 	struct config_item *item, char *page)
461 {
462 	return sprintf(page, "%u\n",
463 			to_o2nm_cluster(item)->cl_keepalive_delay_ms);
464 }
465 
466 static ssize_t o2nm_cluster_keepalive_delay_ms_store(
467 	struct config_item *item, const char *page, size_t count)
468 {
469 	struct o2nm_cluster *cluster = to_o2nm_cluster(item);
470 	ssize_t ret;
471 	unsigned int val;
472 
473 	ret =  o2nm_cluster_attr_write(page, count, &val);
474 
475 	if (ret > 0) {
476 		if (cluster->cl_keepalive_delay_ms != val
477 		    && o2net_num_connected_peers()) {
478 			mlog(ML_NOTICE,
479 			     "o2net: cannot change keepalive delay after"
480 			     " the first peer has agreed to it."
481 			     "  %d connected peers\n",
482 			     o2net_num_connected_peers());
483 			ret = -EINVAL;
484 		} else if (val >= cluster->cl_idle_timeout_ms) {
485 			mlog(ML_NOTICE, "o2net: keepalive delay must be "
486 			     "smaller than idle timeout\n");
487 			ret = -EINVAL;
488 		} else {
489 			cluster->cl_keepalive_delay_ms = val;
490 		}
491 	}
492 
493 	return ret;
494 }
495 
496 static ssize_t o2nm_cluster_reconnect_delay_ms_show(
497 	struct config_item *item, char *page)
498 {
499 	return sprintf(page, "%u\n",
500 			to_o2nm_cluster(item)->cl_reconnect_delay_ms);
501 }
502 
503 static ssize_t o2nm_cluster_reconnect_delay_ms_store(
504 	struct config_item *item, const char *page, size_t count)
505 {
506 	return o2nm_cluster_attr_write(page, count,
507                                &to_o2nm_cluster(item)->cl_reconnect_delay_ms);
508 }
509 
510 static ssize_t o2nm_cluster_fence_method_show(
511 	struct config_item *item, char *page)
512 {
513 	struct o2nm_cluster *cluster = to_o2nm_cluster(item);
514 	ssize_t ret = 0;
515 
516 	if (cluster)
517 		ret = sprintf(page, "%s\n",
518 			      o2nm_fence_method_desc[cluster->cl_fence_method]);
519 	return ret;
520 }
521 
522 static ssize_t o2nm_cluster_fence_method_store(
523 	struct config_item *item, const char *page, size_t count)
524 {
525 	unsigned int i;
526 
527 	if (page[count - 1] != '\n')
528 		goto bail;
529 
530 	for (i = 0; i < O2NM_FENCE_METHODS; ++i) {
531 		if (count != strlen(o2nm_fence_method_desc[i]) + 1)
532 			continue;
533 		if (strncasecmp(page, o2nm_fence_method_desc[i], count - 1))
534 			continue;
535 		if (to_o2nm_cluster(item)->cl_fence_method != i) {
536 			printk(KERN_INFO "ocfs2: Changing fence method to %s\n",
537 			       o2nm_fence_method_desc[i]);
538 			to_o2nm_cluster(item)->cl_fence_method = i;
539 		}
540 		return count;
541 	}
542 
543 bail:
544 	return -EINVAL;
545 }
546 
547 CONFIGFS_ATTR(o2nm_cluster_, idle_timeout_ms);
548 CONFIGFS_ATTR(o2nm_cluster_, keepalive_delay_ms);
549 CONFIGFS_ATTR(o2nm_cluster_, reconnect_delay_ms);
550 CONFIGFS_ATTR(o2nm_cluster_, fence_method);
551 
552 static struct configfs_attribute *o2nm_cluster_attrs[] = {
553 	&o2nm_cluster_attr_idle_timeout_ms,
554 	&o2nm_cluster_attr_keepalive_delay_ms,
555 	&o2nm_cluster_attr_reconnect_delay_ms,
556 	&o2nm_cluster_attr_fence_method,
557 	NULL,
558 };
559 
560 static struct config_item *o2nm_node_group_make_item(struct config_group *group,
561 						     const char *name)
562 {
563 	struct o2nm_node *node = NULL;
564 
565 	if (strlen(name) > O2NM_MAX_NAME_LEN)
566 		return ERR_PTR(-ENAMETOOLONG);
567 
568 	node = kzalloc(sizeof(struct o2nm_node), GFP_KERNEL);
569 	if (node == NULL)
570 		return ERR_PTR(-ENOMEM);
571 
572 	strcpy(node->nd_name, name); /* use item.ci_namebuf instead? */
573 	config_item_init_type_name(&node->nd_item, name, &o2nm_node_type);
574 	spin_lock_init(&node->nd_lock);
575 
576 	mlog(ML_CLUSTER, "o2nm: Registering node %s\n", name);
577 
578 	return &node->nd_item;
579 }
580 
581 static void o2nm_node_group_drop_item(struct config_group *group,
582 				      struct config_item *item)
583 {
584 	struct o2nm_node *node = to_o2nm_node(item);
585 	struct o2nm_cluster *cluster = to_o2nm_cluster(group->cg_item.ci_parent);
586 
587 	o2net_disconnect_node(node);
588 
589 	if (cluster->cl_has_local &&
590 	    (cluster->cl_local_node == node->nd_num)) {
591 		cluster->cl_has_local = 0;
592 		cluster->cl_local_node = O2NM_INVALID_NODE_NUM;
593 		o2net_stop_listening(node);
594 	}
595 
596 	/* XXX call into net to stop this node from trading messages */
597 
598 	write_lock(&cluster->cl_nodes_lock);
599 
600 	/* XXX sloppy */
601 	if (node->nd_ipv4_address)
602 		rb_erase(&node->nd_ip_node, &cluster->cl_node_ip_tree);
603 
604 	/* nd_num might be 0 if the node number hasn't been set.. */
605 	if (cluster->cl_nodes[node->nd_num] == node) {
606 		cluster->cl_nodes[node->nd_num] = NULL;
607 		clear_bit(node->nd_num, cluster->cl_nodes_bitmap);
608 	}
609 	write_unlock(&cluster->cl_nodes_lock);
610 
611 	mlog(ML_CLUSTER, "o2nm: Unregistered node %s\n",
612 	     config_item_name(&node->nd_item));
613 
614 	config_item_put(item);
615 }
616 
617 static struct configfs_group_operations o2nm_node_group_group_ops = {
618 	.make_item	= o2nm_node_group_make_item,
619 	.drop_item	= o2nm_node_group_drop_item,
620 };
621 
622 static struct config_item_type o2nm_node_group_type = {
623 	.ct_group_ops	= &o2nm_node_group_group_ops,
624 	.ct_owner	= THIS_MODULE,
625 };
626 
627 /* cluster */
628 
629 static void o2nm_cluster_release(struct config_item *item)
630 {
631 	struct o2nm_cluster *cluster = to_o2nm_cluster(item);
632 
633 	kfree(cluster);
634 }
635 
636 static struct configfs_item_operations o2nm_cluster_item_ops = {
637 	.release	= o2nm_cluster_release,
638 };
639 
640 static struct config_item_type o2nm_cluster_type = {
641 	.ct_item_ops	= &o2nm_cluster_item_ops,
642 	.ct_attrs	= o2nm_cluster_attrs,
643 	.ct_owner	= THIS_MODULE,
644 };
645 
646 /* cluster set */
647 
648 struct o2nm_cluster_group {
649 	struct configfs_subsystem cs_subsys;
650 	/* some stuff? */
651 };
652 
653 #if 0
654 static struct o2nm_cluster_group *to_o2nm_cluster_group(struct config_group *group)
655 {
656 	return group ?
657 		container_of(to_configfs_subsystem(group), struct o2nm_cluster_group, cs_subsys)
658 	       : NULL;
659 }
660 #endif
661 
662 static struct config_group *o2nm_cluster_group_make_group(struct config_group *group,
663 							  const char *name)
664 {
665 	struct o2nm_cluster *cluster = NULL;
666 	struct o2nm_node_group *ns = NULL;
667 	struct config_group *o2hb_group = NULL, *ret = NULL;
668 
669 	/* this runs under the parent dir's i_mutex; there can be only
670 	 * one caller in here at a time */
671 	if (o2nm_single_cluster)
672 		return ERR_PTR(-ENOSPC);
673 
674 	cluster = kzalloc(sizeof(struct o2nm_cluster), GFP_KERNEL);
675 	ns = kzalloc(sizeof(struct o2nm_node_group), GFP_KERNEL);
676 	o2hb_group = o2hb_alloc_hb_set();
677 	if (cluster == NULL || ns == NULL || o2hb_group == NULL)
678 		goto out;
679 
680 	config_group_init_type_name(&cluster->cl_group, name,
681 				    &o2nm_cluster_type);
682 	configfs_add_default_group(&ns->ns_group, &cluster->cl_group);
683 
684 	config_group_init_type_name(&ns->ns_group, "node",
685 				    &o2nm_node_group_type);
686 	configfs_add_default_group(o2hb_group, &cluster->cl_group);
687 
688 	rwlock_init(&cluster->cl_nodes_lock);
689 	cluster->cl_node_ip_tree = RB_ROOT;
690 	cluster->cl_reconnect_delay_ms = O2NET_RECONNECT_DELAY_MS_DEFAULT;
691 	cluster->cl_idle_timeout_ms    = O2NET_IDLE_TIMEOUT_MS_DEFAULT;
692 	cluster->cl_keepalive_delay_ms = O2NET_KEEPALIVE_DELAY_MS_DEFAULT;
693 	cluster->cl_fence_method       = O2NM_FENCE_RESET;
694 
695 	ret = &cluster->cl_group;
696 	o2nm_single_cluster = cluster;
697 
698 out:
699 	if (ret == NULL) {
700 		kfree(cluster);
701 		kfree(ns);
702 		o2hb_free_hb_set(o2hb_group);
703 		ret = ERR_PTR(-ENOMEM);
704 	}
705 
706 	return ret;
707 }
708 
709 static void o2nm_cluster_group_drop_item(struct config_group *group, struct config_item *item)
710 {
711 	struct o2nm_cluster *cluster = to_o2nm_cluster(item);
712 
713 	BUG_ON(o2nm_single_cluster != cluster);
714 	o2nm_single_cluster = NULL;
715 
716 	configfs_remove_default_groups(&cluster->cl_group);
717 	config_item_put(item);
718 }
719 
720 static struct configfs_group_operations o2nm_cluster_group_group_ops = {
721 	.make_group	= o2nm_cluster_group_make_group,
722 	.drop_item	= o2nm_cluster_group_drop_item,
723 };
724 
725 static struct config_item_type o2nm_cluster_group_type = {
726 	.ct_group_ops	= &o2nm_cluster_group_group_ops,
727 	.ct_owner	= THIS_MODULE,
728 };
729 
730 static struct o2nm_cluster_group o2nm_cluster_group = {
731 	.cs_subsys = {
732 		.su_group = {
733 			.cg_item = {
734 				.ci_namebuf = "cluster",
735 				.ci_type = &o2nm_cluster_group_type,
736 			},
737 		},
738 	},
739 };
740 
741 int o2nm_depend_item(struct config_item *item)
742 {
743 	return configfs_depend_item(&o2nm_cluster_group.cs_subsys, item);
744 }
745 
746 void o2nm_undepend_item(struct config_item *item)
747 {
748 	configfs_undepend_item(item);
749 }
750 
751 int o2nm_depend_this_node(void)
752 {
753 	int ret = 0;
754 	struct o2nm_node *local_node;
755 
756 	local_node = o2nm_get_node_by_num(o2nm_this_node());
757 	if (!local_node) {
758 		ret = -EINVAL;
759 		goto out;
760 	}
761 
762 	ret = o2nm_depend_item(&local_node->nd_item);
763 	o2nm_node_put(local_node);
764 
765 out:
766 	return ret;
767 }
768 
769 void o2nm_undepend_this_node(void)
770 {
771 	struct o2nm_node *local_node;
772 
773 	local_node = o2nm_get_node_by_num(o2nm_this_node());
774 	BUG_ON(!local_node);
775 
776 	o2nm_undepend_item(&local_node->nd_item);
777 	o2nm_node_put(local_node);
778 }
779 
780 
781 static void __exit exit_o2nm(void)
782 {
783 	/* XXX sync with hb callbacks and shut down hb? */
784 	o2net_unregister_hb_callbacks();
785 	configfs_unregister_subsystem(&o2nm_cluster_group.cs_subsys);
786 	o2cb_sys_shutdown();
787 
788 	o2net_exit();
789 	o2hb_exit();
790 }
791 
792 static int __init init_o2nm(void)
793 {
794 	int ret = -1;
795 
796 	ret = o2hb_init();
797 	if (ret)
798 		goto out;
799 
800 	ret = o2net_init();
801 	if (ret)
802 		goto out_o2hb;
803 
804 	ret = o2net_register_hb_callbacks();
805 	if (ret)
806 		goto out_o2net;
807 
808 	config_group_init(&o2nm_cluster_group.cs_subsys.su_group);
809 	mutex_init(&o2nm_cluster_group.cs_subsys.su_mutex);
810 	ret = configfs_register_subsystem(&o2nm_cluster_group.cs_subsys);
811 	if (ret) {
812 		printk(KERN_ERR "nodemanager: Registration returned %d\n", ret);
813 		goto out_callbacks;
814 	}
815 
816 	ret = o2cb_sys_init();
817 	if (!ret)
818 		goto out;
819 
820 	configfs_unregister_subsystem(&o2nm_cluster_group.cs_subsys);
821 out_callbacks:
822 	o2net_unregister_hb_callbacks();
823 out_o2net:
824 	o2net_exit();
825 out_o2hb:
826 	o2hb_exit();
827 out:
828 	return ret;
829 }
830 
831 MODULE_AUTHOR("Oracle");
832 MODULE_LICENSE("GPL");
833 MODULE_DESCRIPTION("OCFS2 cluster management");
834 
835 module_init(init_o2nm)
836 module_exit(exit_o2nm)
837