xref: /openbmc/linux/fs/ocfs2/cluster/nodemanager.c (revision 79acbb3ff2d8095b692e1502b9eb2ccec348de26)
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/kernel.h>
23 #include <linux/module.h>
24 #include <linux/sysctl.h>
25 #include <linux/configfs.h>
26 
27 #include "endian.h"
28 #include "tcp.h"
29 #include "nodemanager.h"
30 #include "heartbeat.h"
31 #include "masklog.h"
32 #include "sys.h"
33 #include "ver.h"
34 
35 /* for now we operate under the assertion that there can be only one
36  * cluster active at a time.  Changing this will require trickling
37  * cluster references throughout where nodes are looked up */
38 static struct o2nm_cluster *o2nm_single_cluster = NULL;
39 
40 #define OCFS2_MAX_HB_CTL_PATH 256
41 static char ocfs2_hb_ctl_path[OCFS2_MAX_HB_CTL_PATH] = "/sbin/ocfs2_hb_ctl";
42 
43 static ctl_table ocfs2_nm_table[] = {
44 	{
45 		.ctl_name	= 1,
46 		.procname	= "hb_ctl_path",
47 		.data		= ocfs2_hb_ctl_path,
48 		.maxlen		= OCFS2_MAX_HB_CTL_PATH,
49 		.mode		= 0644,
50 		.proc_handler	= &proc_dostring,
51 		.strategy	= &sysctl_string,
52 	},
53 	{ .ctl_name = 0 }
54 };
55 
56 static ctl_table ocfs2_mod_table[] = {
57 	{
58 		.ctl_name	= KERN_OCFS2_NM,
59 		.procname	= "nm",
60 		.data		= NULL,
61 		.maxlen		= 0,
62 		.mode		= 0555,
63 		.child		= ocfs2_nm_table
64 	},
65 	{ .ctl_name = 0}
66 };
67 
68 static ctl_table ocfs2_kern_table[] = {
69 	{
70 		.ctl_name	= KERN_OCFS2,
71 		.procname	= "ocfs2",
72 		.data		= NULL,
73 		.maxlen		= 0,
74 		.mode		= 0555,
75 		.child		= ocfs2_mod_table
76 	},
77 	{ .ctl_name = 0}
78 };
79 
80 static ctl_table ocfs2_root_table[] = {
81 	{
82 		.ctl_name	= CTL_FS,
83 		.procname	= "fs",
84 		.data		= NULL,
85 		.maxlen		= 0,
86 		.mode		= 0555,
87 		.child		= ocfs2_kern_table
88 	},
89 	{ .ctl_name = 0 }
90 };
91 
92 static struct ctl_table_header *ocfs2_table_header = NULL;
93 
94 const char *o2nm_get_hb_ctl_path(void)
95 {
96 	return ocfs2_hb_ctl_path;
97 }
98 EXPORT_SYMBOL_GPL(o2nm_get_hb_ctl_path);
99 
100 struct o2nm_cluster {
101 	struct config_group	cl_group;
102 	unsigned		cl_has_local:1;
103 	u8			cl_local_node;
104 	rwlock_t		cl_nodes_lock;
105 	struct o2nm_node  	*cl_nodes[O2NM_MAX_NODES];
106 	struct rb_root		cl_node_ip_tree;
107 	/* this bitmap is part of a hack for disk bitmap.. will go eventually. - zab */
108 	unsigned long	cl_nodes_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
109 };
110 
111 struct o2nm_node *o2nm_get_node_by_num(u8 node_num)
112 {
113 	struct o2nm_node *node = NULL;
114 
115 	if (node_num >= O2NM_MAX_NODES || o2nm_single_cluster == NULL)
116 		goto out;
117 
118 	read_lock(&o2nm_single_cluster->cl_nodes_lock);
119 	node = o2nm_single_cluster->cl_nodes[node_num];
120 	if (node)
121 		config_item_get(&node->nd_item);
122 	read_unlock(&o2nm_single_cluster->cl_nodes_lock);
123 out:
124 	return node;
125 }
126 EXPORT_SYMBOL_GPL(o2nm_get_node_by_num);
127 
128 int o2nm_configured_node_map(unsigned long *map, unsigned bytes)
129 {
130 	struct o2nm_cluster *cluster = o2nm_single_cluster;
131 
132 	BUG_ON(bytes < (sizeof(cluster->cl_nodes_bitmap)));
133 
134 	if (cluster == NULL)
135 		return -EINVAL;
136 
137 	read_lock(&cluster->cl_nodes_lock);
138 	memcpy(map, cluster->cl_nodes_bitmap, sizeof(cluster->cl_nodes_bitmap));
139 	read_unlock(&cluster->cl_nodes_lock);
140 
141 	return 0;
142 }
143 EXPORT_SYMBOL_GPL(o2nm_configured_node_map);
144 
145 static struct o2nm_node *o2nm_node_ip_tree_lookup(struct o2nm_cluster *cluster,
146 						  __be32 ip_needle,
147 						  struct rb_node ***ret_p,
148 						  struct rb_node **ret_parent)
149 {
150 	struct rb_node **p = &cluster->cl_node_ip_tree.rb_node;
151 	struct rb_node *parent = NULL;
152 	struct o2nm_node *node, *ret = NULL;
153 
154 	while (*p) {
155 		int cmp;
156 
157 		parent = *p;
158 		node = rb_entry(parent, struct o2nm_node, nd_ip_node);
159 
160 		cmp = memcmp(&ip_needle, &node->nd_ipv4_address,
161 				sizeof(ip_needle));
162 		if (cmp < 0)
163 			p = &(*p)->rb_left;
164 		else if (cmp > 0)
165 			p = &(*p)->rb_right;
166 		else {
167 			ret = node;
168 			break;
169 		}
170 	}
171 
172 	if (ret_p != NULL)
173 		*ret_p = p;
174 	if (ret_parent != NULL)
175 		*ret_parent = parent;
176 
177 	return ret;
178 }
179 
180 struct o2nm_node *o2nm_get_node_by_ip(__be32 addr)
181 {
182 	struct o2nm_node *node = NULL;
183 	struct o2nm_cluster *cluster = o2nm_single_cluster;
184 
185 	if (cluster == NULL)
186 		goto out;
187 
188 	read_lock(&cluster->cl_nodes_lock);
189 	node = o2nm_node_ip_tree_lookup(cluster, addr, NULL, NULL);
190 	if (node)
191 		config_item_get(&node->nd_item);
192 	read_unlock(&cluster->cl_nodes_lock);
193 
194 out:
195 	return node;
196 }
197 EXPORT_SYMBOL_GPL(o2nm_get_node_by_ip);
198 
199 void o2nm_node_put(struct o2nm_node *node)
200 {
201 	config_item_put(&node->nd_item);
202 }
203 EXPORT_SYMBOL_GPL(o2nm_node_put);
204 
205 void o2nm_node_get(struct o2nm_node *node)
206 {
207 	config_item_get(&node->nd_item);
208 }
209 EXPORT_SYMBOL_GPL(o2nm_node_get);
210 
211 u8 o2nm_this_node(void)
212 {
213 	u8 node_num = O2NM_MAX_NODES;
214 
215 	if (o2nm_single_cluster && o2nm_single_cluster->cl_has_local)
216 		node_num = o2nm_single_cluster->cl_local_node;
217 
218 	return node_num;
219 }
220 EXPORT_SYMBOL_GPL(o2nm_this_node);
221 
222 /* node configfs bits */
223 
224 static struct o2nm_cluster *to_o2nm_cluster(struct config_item *item)
225 {
226 	return item ?
227 		container_of(to_config_group(item), struct o2nm_cluster,
228 			     cl_group)
229 		: NULL;
230 }
231 
232 static struct o2nm_node *to_o2nm_node(struct config_item *item)
233 {
234 	return item ? container_of(item, struct o2nm_node, nd_item) : NULL;
235 }
236 
237 static void o2nm_node_release(struct config_item *item)
238 {
239 	struct o2nm_node *node = to_o2nm_node(item);
240 	kfree(node);
241 }
242 
243 static ssize_t o2nm_node_num_read(struct o2nm_node *node, char *page)
244 {
245 	return sprintf(page, "%d\n", node->nd_num);
246 }
247 
248 static struct o2nm_cluster *to_o2nm_cluster_from_node(struct o2nm_node *node)
249 {
250 	/* through the first node_set .parent
251 	 * mycluster/nodes/mynode == o2nm_cluster->o2nm_node_group->o2nm_node */
252 	return to_o2nm_cluster(node->nd_item.ci_parent->ci_parent);
253 }
254 
255 enum {
256 	O2NM_NODE_ATTR_NUM = 0,
257 	O2NM_NODE_ATTR_PORT,
258 	O2NM_NODE_ATTR_ADDRESS,
259 	O2NM_NODE_ATTR_LOCAL,
260 };
261 
262 static ssize_t o2nm_node_num_write(struct o2nm_node *node, const char *page,
263 				   size_t count)
264 {
265 	struct o2nm_cluster *cluster = to_o2nm_cluster_from_node(node);
266 	unsigned long tmp;
267 	char *p = (char *)page;
268 
269 	tmp = simple_strtoul(p, &p, 0);
270 	if (!p || (*p && (*p != '\n')))
271 		return -EINVAL;
272 
273 	if (tmp >= O2NM_MAX_NODES)
274 		return -ERANGE;
275 
276 	/* once we're in the cl_nodes tree networking can look us up by
277 	 * node number and try to use our address and port attributes
278 	 * to connect to this node.. make sure that they've been set
279 	 * before writing the node attribute? */
280 	if (!test_bit(O2NM_NODE_ATTR_ADDRESS, &node->nd_set_attributes) ||
281 	    !test_bit(O2NM_NODE_ATTR_PORT, &node->nd_set_attributes))
282 		return -EINVAL; /* XXX */
283 
284 	write_lock(&cluster->cl_nodes_lock);
285 	if (cluster->cl_nodes[tmp])
286 		p = NULL;
287 	else  {
288 		cluster->cl_nodes[tmp] = node;
289 		node->nd_num = tmp;
290 		set_bit(tmp, cluster->cl_nodes_bitmap);
291 	}
292 	write_unlock(&cluster->cl_nodes_lock);
293 	if (p == NULL)
294 		return -EEXIST;
295 
296 	return count;
297 }
298 static ssize_t o2nm_node_ipv4_port_read(struct o2nm_node *node, char *page)
299 {
300 	return sprintf(page, "%u\n", ntohs(node->nd_ipv4_port));
301 }
302 
303 static ssize_t o2nm_node_ipv4_port_write(struct o2nm_node *node,
304 					 const char *page, size_t count)
305 {
306 	unsigned long tmp;
307 	char *p = (char *)page;
308 
309 	tmp = simple_strtoul(p, &p, 0);
310 	if (!p || (*p && (*p != '\n')))
311 		return -EINVAL;
312 
313 	if (tmp == 0)
314 		return -EINVAL;
315 	if (tmp >= (u16)-1)
316 		return -ERANGE;
317 
318 	node->nd_ipv4_port = htons(tmp);
319 
320 	return count;
321 }
322 
323 static ssize_t o2nm_node_ipv4_address_read(struct o2nm_node *node, char *page)
324 {
325 	return sprintf(page, "%u.%u.%u.%u\n", NIPQUAD(node->nd_ipv4_address));
326 }
327 
328 static ssize_t o2nm_node_ipv4_address_write(struct o2nm_node *node,
329 					    const char *page,
330 					    size_t count)
331 {
332 	struct o2nm_cluster *cluster = to_o2nm_cluster_from_node(node);
333 	int ret, i;
334 	struct rb_node **p, *parent;
335 	unsigned int octets[4];
336 	__be32 ipv4_addr = 0;
337 
338 	ret = sscanf(page, "%3u.%3u.%3u.%3u", &octets[3], &octets[2],
339 		     &octets[1], &octets[0]);
340 	if (ret != 4)
341 		return -EINVAL;
342 
343 	for (i = 0; i < ARRAY_SIZE(octets); i++) {
344 		if (octets[i] > 255)
345 			return -ERANGE;
346 		be32_add_cpu(&ipv4_addr, octets[i] << (i * 8));
347 	}
348 
349 	ret = 0;
350 	write_lock(&cluster->cl_nodes_lock);
351 	if (o2nm_node_ip_tree_lookup(cluster, ipv4_addr, &p, &parent))
352 		ret = -EEXIST;
353 	else {
354 		rb_link_node(&node->nd_ip_node, parent, p);
355 		rb_insert_color(&node->nd_ip_node, &cluster->cl_node_ip_tree);
356 	}
357 	write_unlock(&cluster->cl_nodes_lock);
358 	if (ret)
359 		return ret;
360 
361 	memcpy(&node->nd_ipv4_address, &ipv4_addr, sizeof(ipv4_addr));
362 
363 	return count;
364 }
365 
366 static ssize_t o2nm_node_local_read(struct o2nm_node *node, char *page)
367 {
368 	return sprintf(page, "%d\n", node->nd_local);
369 }
370 
371 static ssize_t o2nm_node_local_write(struct o2nm_node *node, const char *page,
372 				     size_t count)
373 {
374 	struct o2nm_cluster *cluster = to_o2nm_cluster_from_node(node);
375 	unsigned long tmp;
376 	char *p = (char *)page;
377 	ssize_t ret;
378 
379 	tmp = simple_strtoul(p, &p, 0);
380 	if (!p || (*p && (*p != '\n')))
381 		return -EINVAL;
382 
383 	tmp = !!tmp; /* boolean of whether this node wants to be local */
384 
385 	/* setting local turns on networking rx for now so we require having
386 	 * set everything else first */
387 	if (!test_bit(O2NM_NODE_ATTR_ADDRESS, &node->nd_set_attributes) ||
388 	    !test_bit(O2NM_NODE_ATTR_NUM, &node->nd_set_attributes) ||
389 	    !test_bit(O2NM_NODE_ATTR_PORT, &node->nd_set_attributes))
390 		return -EINVAL; /* XXX */
391 
392 	/* the only failure case is trying to set a new local node
393 	 * when a different one is already set */
394 	if (tmp && tmp == cluster->cl_has_local &&
395 	    cluster->cl_local_node != node->nd_num)
396 		return -EBUSY;
397 
398 	/* bring up the rx thread if we're setting the new local node. */
399 	if (tmp && !cluster->cl_has_local) {
400 		ret = o2net_start_listening(node);
401 		if (ret)
402 			return ret;
403 	}
404 
405 	if (!tmp && cluster->cl_has_local &&
406 	    cluster->cl_local_node == node->nd_num) {
407 		o2net_stop_listening(node);
408 		cluster->cl_local_node = O2NM_INVALID_NODE_NUM;
409 	}
410 
411 	node->nd_local = tmp;
412 	if (node->nd_local) {
413 		cluster->cl_has_local = tmp;
414 		cluster->cl_local_node = node->nd_num;
415 	}
416 
417 	return count;
418 }
419 
420 struct o2nm_node_attribute {
421 	struct configfs_attribute attr;
422 	ssize_t (*show)(struct o2nm_node *, char *);
423 	ssize_t (*store)(struct o2nm_node *, const char *, size_t);
424 };
425 
426 static struct o2nm_node_attribute o2nm_node_attr_num = {
427 	.attr	= { .ca_owner = THIS_MODULE,
428 		    .ca_name = "num",
429 		    .ca_mode = S_IRUGO | S_IWUSR },
430 	.show	= o2nm_node_num_read,
431 	.store	= o2nm_node_num_write,
432 };
433 
434 static struct o2nm_node_attribute o2nm_node_attr_ipv4_port = {
435 	.attr	= { .ca_owner = THIS_MODULE,
436 		    .ca_name = "ipv4_port",
437 		    .ca_mode = S_IRUGO | S_IWUSR },
438 	.show	= o2nm_node_ipv4_port_read,
439 	.store	= o2nm_node_ipv4_port_write,
440 };
441 
442 static struct o2nm_node_attribute o2nm_node_attr_ipv4_address = {
443 	.attr	= { .ca_owner = THIS_MODULE,
444 		    .ca_name = "ipv4_address",
445 		    .ca_mode = S_IRUGO | S_IWUSR },
446 	.show	= o2nm_node_ipv4_address_read,
447 	.store	= o2nm_node_ipv4_address_write,
448 };
449 
450 static struct o2nm_node_attribute o2nm_node_attr_local = {
451 	.attr	= { .ca_owner = THIS_MODULE,
452 		    .ca_name = "local",
453 		    .ca_mode = S_IRUGO | S_IWUSR },
454 	.show	= o2nm_node_local_read,
455 	.store	= o2nm_node_local_write,
456 };
457 
458 static struct configfs_attribute *o2nm_node_attrs[] = {
459 	[O2NM_NODE_ATTR_NUM] = &o2nm_node_attr_num.attr,
460 	[O2NM_NODE_ATTR_PORT] = &o2nm_node_attr_ipv4_port.attr,
461 	[O2NM_NODE_ATTR_ADDRESS] = &o2nm_node_attr_ipv4_address.attr,
462 	[O2NM_NODE_ATTR_LOCAL] = &o2nm_node_attr_local.attr,
463 	NULL,
464 };
465 
466 static int o2nm_attr_index(struct configfs_attribute *attr)
467 {
468 	int i;
469 	for (i = 0; i < ARRAY_SIZE(o2nm_node_attrs); i++) {
470 		if (attr == o2nm_node_attrs[i])
471 			return i;
472 	}
473 	BUG();
474 	return 0;
475 }
476 
477 static ssize_t o2nm_node_show(struct config_item *item,
478 			      struct configfs_attribute *attr,
479 			      char *page)
480 {
481 	struct o2nm_node *node = to_o2nm_node(item);
482 	struct o2nm_node_attribute *o2nm_node_attr =
483 		container_of(attr, struct o2nm_node_attribute, attr);
484 	ssize_t ret = 0;
485 
486 	if (o2nm_node_attr->show)
487 		ret = o2nm_node_attr->show(node, page);
488 	return ret;
489 }
490 
491 static ssize_t o2nm_node_store(struct config_item *item,
492 			       struct configfs_attribute *attr,
493 			       const char *page, size_t count)
494 {
495 	struct o2nm_node *node = to_o2nm_node(item);
496 	struct o2nm_node_attribute *o2nm_node_attr =
497 		container_of(attr, struct o2nm_node_attribute, attr);
498 	ssize_t ret;
499 	int attr_index = o2nm_attr_index(attr);
500 
501 	if (o2nm_node_attr->store == NULL) {
502 		ret = -EINVAL;
503 		goto out;
504 	}
505 
506 	if (test_bit(attr_index, &node->nd_set_attributes))
507 		return -EBUSY;
508 
509 	ret = o2nm_node_attr->store(node, page, count);
510 	if (ret < count)
511 		goto out;
512 
513 	set_bit(attr_index, &node->nd_set_attributes);
514 out:
515 	return ret;
516 }
517 
518 static struct configfs_item_operations o2nm_node_item_ops = {
519 	.release		= o2nm_node_release,
520 	.show_attribute		= o2nm_node_show,
521 	.store_attribute	= o2nm_node_store,
522 };
523 
524 static struct config_item_type o2nm_node_type = {
525 	.ct_item_ops	= &o2nm_node_item_ops,
526 	.ct_attrs	= o2nm_node_attrs,
527 	.ct_owner	= THIS_MODULE,
528 };
529 
530 /* node set */
531 
532 struct o2nm_node_group {
533 	struct config_group ns_group;
534 	/* some stuff? */
535 };
536 
537 #if 0
538 static struct o2nm_node_group *to_o2nm_node_group(struct config_group *group)
539 {
540 	return group ?
541 		container_of(group, struct o2nm_node_group, ns_group)
542 		: NULL;
543 }
544 #endif
545 
546 static struct config_item *o2nm_node_group_make_item(struct config_group *group,
547 						     const char *name)
548 {
549 	struct o2nm_node *node = NULL;
550 	struct config_item *ret = NULL;
551 
552 	if (strlen(name) > O2NM_MAX_NAME_LEN)
553 		goto out; /* ENAMETOOLONG */
554 
555 	node = kcalloc(1, sizeof(struct o2nm_node), GFP_KERNEL);
556 	if (node == NULL)
557 		goto out; /* ENOMEM */
558 
559 	strcpy(node->nd_name, name); /* use item.ci_namebuf instead? */
560 	config_item_init_type_name(&node->nd_item, name, &o2nm_node_type);
561 	spin_lock_init(&node->nd_lock);
562 
563 	ret = &node->nd_item;
564 
565 out:
566 	if (ret == NULL)
567 		kfree(node);
568 
569 	return ret;
570 }
571 
572 static void o2nm_node_group_drop_item(struct config_group *group,
573 				      struct config_item *item)
574 {
575 	struct o2nm_node *node = to_o2nm_node(item);
576 	struct o2nm_cluster *cluster = to_o2nm_cluster(group->cg_item.ci_parent);
577 
578 	o2net_disconnect_node(node);
579 
580 	if (cluster->cl_has_local &&
581 	    (cluster->cl_local_node == node->nd_num)) {
582 		cluster->cl_has_local = 0;
583 		cluster->cl_local_node = O2NM_INVALID_NODE_NUM;
584 		o2net_stop_listening(node);
585 	}
586 
587 	/* XXX call into net to stop this node from trading messages */
588 
589 	write_lock(&cluster->cl_nodes_lock);
590 
591 	/* XXX sloppy */
592 	if (node->nd_ipv4_address)
593 		rb_erase(&node->nd_ip_node, &cluster->cl_node_ip_tree);
594 
595 	/* nd_num might be 0 if the node number hasn't been set.. */
596 	if (cluster->cl_nodes[node->nd_num] == node) {
597 		cluster->cl_nodes[node->nd_num] = NULL;
598 		clear_bit(node->nd_num, cluster->cl_nodes_bitmap);
599 	}
600 	write_unlock(&cluster->cl_nodes_lock);
601 
602 	config_item_put(item);
603 }
604 
605 static struct configfs_group_operations o2nm_node_group_group_ops = {
606 	.make_item	= o2nm_node_group_make_item,
607 	.drop_item	= o2nm_node_group_drop_item,
608 };
609 
610 static struct config_item_type o2nm_node_group_type = {
611 	.ct_group_ops	= &o2nm_node_group_group_ops,
612 	.ct_owner	= THIS_MODULE,
613 };
614 
615 /* cluster */
616 
617 static void o2nm_cluster_release(struct config_item *item)
618 {
619 	struct o2nm_cluster *cluster = to_o2nm_cluster(item);
620 
621 	kfree(cluster->cl_group.default_groups);
622 	kfree(cluster);
623 }
624 
625 static struct configfs_item_operations o2nm_cluster_item_ops = {
626 	.release	= o2nm_cluster_release,
627 };
628 
629 static struct config_item_type o2nm_cluster_type = {
630 	.ct_item_ops	= &o2nm_cluster_item_ops,
631 	.ct_owner	= THIS_MODULE,
632 };
633 
634 /* cluster set */
635 
636 struct o2nm_cluster_group {
637 	struct configfs_subsystem cs_subsys;
638 	/* some stuff? */
639 };
640 
641 #if 0
642 static struct o2nm_cluster_group *to_o2nm_cluster_group(struct config_group *group)
643 {
644 	return group ?
645 		container_of(to_configfs_subsystem(group), struct o2nm_cluster_group, cs_subsys)
646 	       : NULL;
647 }
648 #endif
649 
650 static struct config_group *o2nm_cluster_group_make_group(struct config_group *group,
651 							  const char *name)
652 {
653 	struct o2nm_cluster *cluster = NULL;
654 	struct o2nm_node_group *ns = NULL;
655 	struct config_group *o2hb_group = NULL, *ret = NULL;
656 	void *defs = NULL;
657 
658 	/* this runs under the parent dir's i_mutex; there can be only
659 	 * one caller in here at a time */
660 	if (o2nm_single_cluster)
661 		goto out; /* ENOSPC */
662 
663 	cluster = kcalloc(1, sizeof(struct o2nm_cluster), GFP_KERNEL);
664 	ns = kcalloc(1, sizeof(struct o2nm_node_group), GFP_KERNEL);
665 	defs = kcalloc(3, sizeof(struct config_group *), GFP_KERNEL);
666 	o2hb_group = o2hb_alloc_hb_set();
667 	if (cluster == NULL || ns == NULL || o2hb_group == NULL || defs == NULL)
668 		goto out;
669 
670 	config_group_init_type_name(&cluster->cl_group, name,
671 				    &o2nm_cluster_type);
672 	config_group_init_type_name(&ns->ns_group, "node",
673 				    &o2nm_node_group_type);
674 
675 	cluster->cl_group.default_groups = defs;
676 	cluster->cl_group.default_groups[0] = &ns->ns_group;
677 	cluster->cl_group.default_groups[1] = o2hb_group;
678 	cluster->cl_group.default_groups[2] = NULL;
679 	rwlock_init(&cluster->cl_nodes_lock);
680 	cluster->cl_node_ip_tree = RB_ROOT;
681 
682 	ret = &cluster->cl_group;
683 	o2nm_single_cluster = cluster;
684 
685 out:
686 	if (ret == NULL) {
687 		kfree(cluster);
688 		kfree(ns);
689 		o2hb_free_hb_set(o2hb_group);
690 		kfree(defs);
691 	}
692 
693 	return ret;
694 }
695 
696 static void o2nm_cluster_group_drop_item(struct config_group *group, struct config_item *item)
697 {
698 	struct o2nm_cluster *cluster = to_o2nm_cluster(item);
699 	int i;
700 	struct config_item *killme;
701 
702 	BUG_ON(o2nm_single_cluster != cluster);
703 	o2nm_single_cluster = NULL;
704 
705 	for (i = 0; cluster->cl_group.default_groups[i]; i++) {
706 		killme = &cluster->cl_group.default_groups[i]->cg_item;
707 		cluster->cl_group.default_groups[i] = NULL;
708 		config_item_put(killme);
709 	}
710 
711 	config_item_put(item);
712 }
713 
714 static struct configfs_group_operations o2nm_cluster_group_group_ops = {
715 	.make_group	= o2nm_cluster_group_make_group,
716 	.drop_item	= o2nm_cluster_group_drop_item,
717 };
718 
719 static struct config_item_type o2nm_cluster_group_type = {
720 	.ct_group_ops	= &o2nm_cluster_group_group_ops,
721 	.ct_owner	= THIS_MODULE,
722 };
723 
724 static struct o2nm_cluster_group o2nm_cluster_group = {
725 	.cs_subsys = {
726 		.su_group = {
727 			.cg_item = {
728 				.ci_namebuf = "cluster",
729 				.ci_type = &o2nm_cluster_group_type,
730 			},
731 		},
732 	},
733 };
734 
735 static void __exit exit_o2nm(void)
736 {
737 	if (ocfs2_table_header)
738 		unregister_sysctl_table(ocfs2_table_header);
739 
740 	/* XXX sync with hb callbacks and shut down hb? */
741 	o2net_unregister_hb_callbacks();
742 	configfs_unregister_subsystem(&o2nm_cluster_group.cs_subsys);
743 	o2cb_sys_shutdown();
744 
745 	o2net_exit();
746 }
747 
748 static int __init init_o2nm(void)
749 {
750 	int ret = -1;
751 
752 	cluster_print_version();
753 
754 	o2hb_init();
755 	o2net_init();
756 
757 	ocfs2_table_header = register_sysctl_table(ocfs2_root_table, 0);
758 	if (!ocfs2_table_header) {
759 		printk(KERN_ERR "nodemanager: unable to register sysctl\n");
760 		ret = -ENOMEM; /* or something. */
761 		goto out_o2net;
762 	}
763 
764 	ret = o2net_register_hb_callbacks();
765 	if (ret)
766 		goto out_sysctl;
767 
768 	config_group_init(&o2nm_cluster_group.cs_subsys.su_group);
769 	init_MUTEX(&o2nm_cluster_group.cs_subsys.su_sem);
770 	ret = configfs_register_subsystem(&o2nm_cluster_group.cs_subsys);
771 	if (ret) {
772 		printk(KERN_ERR "nodemanager: Registration returned %d\n", ret);
773 		goto out_callbacks;
774 	}
775 
776 	ret = o2cb_sys_init();
777 	if (!ret)
778 		goto out;
779 
780 	configfs_unregister_subsystem(&o2nm_cluster_group.cs_subsys);
781 out_callbacks:
782 	o2net_unregister_hb_callbacks();
783 out_sysctl:
784 	unregister_sysctl_table(ocfs2_table_header);
785 out_o2net:
786 	o2net_exit();
787 out:
788 	return ret;
789 }
790 
791 MODULE_AUTHOR("Oracle");
792 MODULE_LICENSE("GPL");
793 
794 module_init(init_o2nm)
795 module_exit(exit_o2nm)
796