xref: /linux/fs/ocfs2/cluster/nodemanager.c (revision 765532c8aaac624b5f8687af6d319c6a1138a257)
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 #include "ver.h"
33 
34 /* for now we operate under the assertion that there can be only one
35  * cluster active at a time.  Changing this will require trickling
36  * cluster references throughout where nodes are looked up */
37 struct o2nm_cluster *o2nm_single_cluster = NULL;
38 
39 char *o2nm_fence_method_desc[O2NM_FENCE_METHODS] = {
40 		"reset",	/* O2NM_FENCE_RESET */
41 		"panic",	/* O2NM_FENCE_PANIC */
42 };
43 
44 struct o2nm_node *o2nm_get_node_by_num(u8 node_num)
45 {
46 	struct o2nm_node *node = NULL;
47 
48 	if (node_num >= O2NM_MAX_NODES || o2nm_single_cluster == NULL)
49 		goto out;
50 
51 	read_lock(&o2nm_single_cluster->cl_nodes_lock);
52 	node = o2nm_single_cluster->cl_nodes[node_num];
53 	if (node)
54 		config_item_get(&node->nd_item);
55 	read_unlock(&o2nm_single_cluster->cl_nodes_lock);
56 out:
57 	return node;
58 }
59 EXPORT_SYMBOL_GPL(o2nm_get_node_by_num);
60 
61 int o2nm_configured_node_map(unsigned long *map, unsigned bytes)
62 {
63 	struct o2nm_cluster *cluster = o2nm_single_cluster;
64 
65 	BUG_ON(bytes < (sizeof(cluster->cl_nodes_bitmap)));
66 
67 	if (cluster == NULL)
68 		return -EINVAL;
69 
70 	read_lock(&cluster->cl_nodes_lock);
71 	memcpy(map, cluster->cl_nodes_bitmap, sizeof(cluster->cl_nodes_bitmap));
72 	read_unlock(&cluster->cl_nodes_lock);
73 
74 	return 0;
75 }
76 EXPORT_SYMBOL_GPL(o2nm_configured_node_map);
77 
78 static struct o2nm_node *o2nm_node_ip_tree_lookup(struct o2nm_cluster *cluster,
79 						  __be32 ip_needle,
80 						  struct rb_node ***ret_p,
81 						  struct rb_node **ret_parent)
82 {
83 	struct rb_node **p = &cluster->cl_node_ip_tree.rb_node;
84 	struct rb_node *parent = NULL;
85 	struct o2nm_node *node, *ret = NULL;
86 
87 	while (*p) {
88 		int cmp;
89 
90 		parent = *p;
91 		node = rb_entry(parent, struct o2nm_node, nd_ip_node);
92 
93 		cmp = memcmp(&ip_needle, &node->nd_ipv4_address,
94 				sizeof(ip_needle));
95 		if (cmp < 0)
96 			p = &(*p)->rb_left;
97 		else if (cmp > 0)
98 			p = &(*p)->rb_right;
99 		else {
100 			ret = node;
101 			break;
102 		}
103 	}
104 
105 	if (ret_p != NULL)
106 		*ret_p = p;
107 	if (ret_parent != NULL)
108 		*ret_parent = parent;
109 
110 	return ret;
111 }
112 
113 struct o2nm_node *o2nm_get_node_by_ip(__be32 addr)
114 {
115 	struct o2nm_node *node = NULL;
116 	struct o2nm_cluster *cluster = o2nm_single_cluster;
117 
118 	if (cluster == NULL)
119 		goto out;
120 
121 	read_lock(&cluster->cl_nodes_lock);
122 	node = o2nm_node_ip_tree_lookup(cluster, addr, NULL, NULL);
123 	if (node)
124 		config_item_get(&node->nd_item);
125 	read_unlock(&cluster->cl_nodes_lock);
126 
127 out:
128 	return node;
129 }
130 EXPORT_SYMBOL_GPL(o2nm_get_node_by_ip);
131 
132 void o2nm_node_put(struct o2nm_node *node)
133 {
134 	config_item_put(&node->nd_item);
135 }
136 EXPORT_SYMBOL_GPL(o2nm_node_put);
137 
138 void o2nm_node_get(struct o2nm_node *node)
139 {
140 	config_item_get(&node->nd_item);
141 }
142 EXPORT_SYMBOL_GPL(o2nm_node_get);
143 
144 u8 o2nm_this_node(void)
145 {
146 	u8 node_num = O2NM_MAX_NODES;
147 
148 	if (o2nm_single_cluster && o2nm_single_cluster->cl_has_local)
149 		node_num = o2nm_single_cluster->cl_local_node;
150 
151 	return node_num;
152 }
153 EXPORT_SYMBOL_GPL(o2nm_this_node);
154 
155 /* node configfs bits */
156 
157 static struct o2nm_cluster *to_o2nm_cluster(struct config_item *item)
158 {
159 	return item ?
160 		container_of(to_config_group(item), struct o2nm_cluster,
161 			     cl_group)
162 		: NULL;
163 }
164 
165 static struct o2nm_node *to_o2nm_node(struct config_item *item)
166 {
167 	return item ? container_of(item, struct o2nm_node, nd_item) : NULL;
168 }
169 
170 static void o2nm_node_release(struct config_item *item)
171 {
172 	struct o2nm_node *node = to_o2nm_node(item);
173 	kfree(node);
174 }
175 
176 static ssize_t o2nm_node_num_read(struct o2nm_node *node, char *page)
177 {
178 	return sprintf(page, "%d\n", node->nd_num);
179 }
180 
181 static struct o2nm_cluster *to_o2nm_cluster_from_node(struct o2nm_node *node)
182 {
183 	/* through the first node_set .parent
184 	 * mycluster/nodes/mynode == o2nm_cluster->o2nm_node_group->o2nm_node */
185 	return to_o2nm_cluster(node->nd_item.ci_parent->ci_parent);
186 }
187 
188 enum {
189 	O2NM_NODE_ATTR_NUM = 0,
190 	O2NM_NODE_ATTR_PORT,
191 	O2NM_NODE_ATTR_ADDRESS,
192 	O2NM_NODE_ATTR_LOCAL,
193 };
194 
195 static ssize_t o2nm_node_num_write(struct o2nm_node *node, const char *page,
196 				   size_t count)
197 {
198 	struct o2nm_cluster *cluster = to_o2nm_cluster_from_node(node);
199 	unsigned long tmp;
200 	char *p = (char *)page;
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 		p = NULL;
220 	else  {
221 		cluster->cl_nodes[tmp] = node;
222 		node->nd_num = tmp;
223 		set_bit(tmp, cluster->cl_nodes_bitmap);
224 	}
225 	write_unlock(&cluster->cl_nodes_lock);
226 	if (p == NULL)
227 		return -EEXIST;
228 
229 	return count;
230 }
231 static ssize_t o2nm_node_ipv4_port_read(struct o2nm_node *node, char *page)
232 {
233 	return sprintf(page, "%u\n", ntohs(node->nd_ipv4_port));
234 }
235 
236 static ssize_t o2nm_node_ipv4_port_write(struct o2nm_node *node,
237 					 const char *page, size_t count)
238 {
239 	unsigned long tmp;
240 	char *p = (char *)page;
241 
242 	tmp = simple_strtoul(p, &p, 0);
243 	if (!p || (*p && (*p != '\n')))
244 		return -EINVAL;
245 
246 	if (tmp == 0)
247 		return -EINVAL;
248 	if (tmp >= (u16)-1)
249 		return -ERANGE;
250 
251 	node->nd_ipv4_port = htons(tmp);
252 
253 	return count;
254 }
255 
256 static ssize_t o2nm_node_ipv4_address_read(struct o2nm_node *node, char *page)
257 {
258 	return sprintf(page, "%pI4\n", &node->nd_ipv4_address);
259 }
260 
261 static ssize_t o2nm_node_ipv4_address_write(struct o2nm_node *node,
262 					    const char *page,
263 					    size_t count)
264 {
265 	struct o2nm_cluster *cluster = to_o2nm_cluster_from_node(node);
266 	int ret, i;
267 	struct rb_node **p, *parent;
268 	unsigned int octets[4];
269 	__be32 ipv4_addr = 0;
270 
271 	ret = sscanf(page, "%3u.%3u.%3u.%3u", &octets[3], &octets[2],
272 		     &octets[1], &octets[0]);
273 	if (ret != 4)
274 		return -EINVAL;
275 
276 	for (i = 0; i < ARRAY_SIZE(octets); i++) {
277 		if (octets[i] > 255)
278 			return -ERANGE;
279 		be32_add_cpu(&ipv4_addr, octets[i] << (i * 8));
280 	}
281 
282 	ret = 0;
283 	write_lock(&cluster->cl_nodes_lock);
284 	if (o2nm_node_ip_tree_lookup(cluster, ipv4_addr, &p, &parent))
285 		ret = -EEXIST;
286 	else {
287 		rb_link_node(&node->nd_ip_node, parent, p);
288 		rb_insert_color(&node->nd_ip_node, &cluster->cl_node_ip_tree);
289 	}
290 	write_unlock(&cluster->cl_nodes_lock);
291 	if (ret)
292 		return ret;
293 
294 	memcpy(&node->nd_ipv4_address, &ipv4_addr, sizeof(ipv4_addr));
295 
296 	return count;
297 }
298 
299 static ssize_t o2nm_node_local_read(struct o2nm_node *node, char *page)
300 {
301 	return sprintf(page, "%d\n", node->nd_local);
302 }
303 
304 static ssize_t o2nm_node_local_write(struct o2nm_node *node, const char *page,
305 				     size_t count)
306 {
307 	struct o2nm_cluster *cluster = to_o2nm_cluster_from_node(node);
308 	unsigned long tmp;
309 	char *p = (char *)page;
310 	ssize_t ret;
311 
312 	tmp = simple_strtoul(p, &p, 0);
313 	if (!p || (*p && (*p != '\n')))
314 		return -EINVAL;
315 
316 	tmp = !!tmp; /* boolean of whether this node wants to be local */
317 
318 	/* setting local turns on networking rx for now so we require having
319 	 * set everything else first */
320 	if (!test_bit(O2NM_NODE_ATTR_ADDRESS, &node->nd_set_attributes) ||
321 	    !test_bit(O2NM_NODE_ATTR_NUM, &node->nd_set_attributes) ||
322 	    !test_bit(O2NM_NODE_ATTR_PORT, &node->nd_set_attributes))
323 		return -EINVAL; /* XXX */
324 
325 	/* the only failure case is trying to set a new local node
326 	 * when a different one is already set */
327 	if (tmp && tmp == cluster->cl_has_local &&
328 	    cluster->cl_local_node != node->nd_num)
329 		return -EBUSY;
330 
331 	/* bring up the rx thread if we're setting the new local node. */
332 	if (tmp && !cluster->cl_has_local) {
333 		ret = o2net_start_listening(node);
334 		if (ret)
335 			return ret;
336 	}
337 
338 	if (!tmp && cluster->cl_has_local &&
339 	    cluster->cl_local_node == node->nd_num) {
340 		o2net_stop_listening(node);
341 		cluster->cl_local_node = O2NM_INVALID_NODE_NUM;
342 	}
343 
344 	node->nd_local = tmp;
345 	if (node->nd_local) {
346 		cluster->cl_has_local = tmp;
347 		cluster->cl_local_node = node->nd_num;
348 	}
349 
350 	return count;
351 }
352 
353 struct o2nm_node_attribute {
354 	struct configfs_attribute attr;
355 	ssize_t (*show)(struct o2nm_node *, char *);
356 	ssize_t (*store)(struct o2nm_node *, const char *, size_t);
357 };
358 
359 static struct o2nm_node_attribute o2nm_node_attr_num = {
360 	.attr	= { .ca_owner = THIS_MODULE,
361 		    .ca_name = "num",
362 		    .ca_mode = S_IRUGO | S_IWUSR },
363 	.show	= o2nm_node_num_read,
364 	.store	= o2nm_node_num_write,
365 };
366 
367 static struct o2nm_node_attribute o2nm_node_attr_ipv4_port = {
368 	.attr	= { .ca_owner = THIS_MODULE,
369 		    .ca_name = "ipv4_port",
370 		    .ca_mode = S_IRUGO | S_IWUSR },
371 	.show	= o2nm_node_ipv4_port_read,
372 	.store	= o2nm_node_ipv4_port_write,
373 };
374 
375 static struct o2nm_node_attribute o2nm_node_attr_ipv4_address = {
376 	.attr	= { .ca_owner = THIS_MODULE,
377 		    .ca_name = "ipv4_address",
378 		    .ca_mode = S_IRUGO | S_IWUSR },
379 	.show	= o2nm_node_ipv4_address_read,
380 	.store	= o2nm_node_ipv4_address_write,
381 };
382 
383 static struct o2nm_node_attribute o2nm_node_attr_local = {
384 	.attr	= { .ca_owner = THIS_MODULE,
385 		    .ca_name = "local",
386 		    .ca_mode = S_IRUGO | S_IWUSR },
387 	.show	= o2nm_node_local_read,
388 	.store	= o2nm_node_local_write,
389 };
390 
391 static struct configfs_attribute *o2nm_node_attrs[] = {
392 	[O2NM_NODE_ATTR_NUM] = &o2nm_node_attr_num.attr,
393 	[O2NM_NODE_ATTR_PORT] = &o2nm_node_attr_ipv4_port.attr,
394 	[O2NM_NODE_ATTR_ADDRESS] = &o2nm_node_attr_ipv4_address.attr,
395 	[O2NM_NODE_ATTR_LOCAL] = &o2nm_node_attr_local.attr,
396 	NULL,
397 };
398 
399 static int o2nm_attr_index(struct configfs_attribute *attr)
400 {
401 	int i;
402 	for (i = 0; i < ARRAY_SIZE(o2nm_node_attrs); i++) {
403 		if (attr == o2nm_node_attrs[i])
404 			return i;
405 	}
406 	BUG();
407 	return 0;
408 }
409 
410 static ssize_t o2nm_node_show(struct config_item *item,
411 			      struct configfs_attribute *attr,
412 			      char *page)
413 {
414 	struct o2nm_node *node = to_o2nm_node(item);
415 	struct o2nm_node_attribute *o2nm_node_attr =
416 		container_of(attr, struct o2nm_node_attribute, attr);
417 	ssize_t ret = 0;
418 
419 	if (o2nm_node_attr->show)
420 		ret = o2nm_node_attr->show(node, page);
421 	return ret;
422 }
423 
424 static ssize_t o2nm_node_store(struct config_item *item,
425 			       struct configfs_attribute *attr,
426 			       const char *page, size_t count)
427 {
428 	struct o2nm_node *node = to_o2nm_node(item);
429 	struct o2nm_node_attribute *o2nm_node_attr =
430 		container_of(attr, struct o2nm_node_attribute, attr);
431 	ssize_t ret;
432 	int attr_index = o2nm_attr_index(attr);
433 
434 	if (o2nm_node_attr->store == NULL) {
435 		ret = -EINVAL;
436 		goto out;
437 	}
438 
439 	if (test_bit(attr_index, &node->nd_set_attributes))
440 		return -EBUSY;
441 
442 	ret = o2nm_node_attr->store(node, page, count);
443 	if (ret < count)
444 		goto out;
445 
446 	set_bit(attr_index, &node->nd_set_attributes);
447 out:
448 	return ret;
449 }
450 
451 static struct configfs_item_operations o2nm_node_item_ops = {
452 	.release		= o2nm_node_release,
453 	.show_attribute		= o2nm_node_show,
454 	.store_attribute	= o2nm_node_store,
455 };
456 
457 static struct config_item_type o2nm_node_type = {
458 	.ct_item_ops	= &o2nm_node_item_ops,
459 	.ct_attrs	= o2nm_node_attrs,
460 	.ct_owner	= THIS_MODULE,
461 };
462 
463 /* node set */
464 
465 struct o2nm_node_group {
466 	struct config_group ns_group;
467 	/* some stuff? */
468 };
469 
470 #if 0
471 static struct o2nm_node_group *to_o2nm_node_group(struct config_group *group)
472 {
473 	return group ?
474 		container_of(group, struct o2nm_node_group, ns_group)
475 		: NULL;
476 }
477 #endif
478 
479 struct o2nm_cluster_attribute {
480 	struct configfs_attribute attr;
481 	ssize_t (*show)(struct o2nm_cluster *, char *);
482 	ssize_t (*store)(struct o2nm_cluster *, const char *, size_t);
483 };
484 
485 static ssize_t o2nm_cluster_attr_write(const char *page, ssize_t count,
486                                        unsigned int *val)
487 {
488 	unsigned long tmp;
489 	char *p = (char *)page;
490 
491 	tmp = simple_strtoul(p, &p, 0);
492 	if (!p || (*p && (*p != '\n')))
493 		return -EINVAL;
494 
495 	if (tmp == 0)
496 		return -EINVAL;
497 	if (tmp >= (u32)-1)
498 		return -ERANGE;
499 
500 	*val = tmp;
501 
502 	return count;
503 }
504 
505 static ssize_t o2nm_cluster_attr_idle_timeout_ms_read(
506 	struct o2nm_cluster *cluster, char *page)
507 {
508 	return sprintf(page, "%u\n", cluster->cl_idle_timeout_ms);
509 }
510 
511 static ssize_t o2nm_cluster_attr_idle_timeout_ms_write(
512 	struct o2nm_cluster *cluster, const char *page, size_t count)
513 {
514 	ssize_t ret;
515 	unsigned int val;
516 
517 	ret =  o2nm_cluster_attr_write(page, count, &val);
518 
519 	if (ret > 0) {
520 		if (cluster->cl_idle_timeout_ms != val
521 			&& o2net_num_connected_peers()) {
522 			mlog(ML_NOTICE,
523 			     "o2net: cannot change idle timeout after "
524 			     "the first peer has agreed to it."
525 			     "  %d connected peers\n",
526 			     o2net_num_connected_peers());
527 			ret = -EINVAL;
528 		} else if (val <= cluster->cl_keepalive_delay_ms) {
529 			mlog(ML_NOTICE, "o2net: idle timeout must be larger "
530 			     "than keepalive delay\n");
531 			ret = -EINVAL;
532 		} else {
533 			cluster->cl_idle_timeout_ms = val;
534 		}
535 	}
536 
537 	return ret;
538 }
539 
540 static ssize_t o2nm_cluster_attr_keepalive_delay_ms_read(
541 	struct o2nm_cluster *cluster, char *page)
542 {
543 	return sprintf(page, "%u\n", cluster->cl_keepalive_delay_ms);
544 }
545 
546 static ssize_t o2nm_cluster_attr_keepalive_delay_ms_write(
547 	struct o2nm_cluster *cluster, const char *page, size_t count)
548 {
549 	ssize_t ret;
550 	unsigned int val;
551 
552 	ret =  o2nm_cluster_attr_write(page, count, &val);
553 
554 	if (ret > 0) {
555 		if (cluster->cl_keepalive_delay_ms != val
556 		    && o2net_num_connected_peers()) {
557 			mlog(ML_NOTICE,
558 			     "o2net: cannot change keepalive delay after"
559 			     " the first peer has agreed to it."
560 			     "  %d connected peers\n",
561 			     o2net_num_connected_peers());
562 			ret = -EINVAL;
563 		} else if (val >= cluster->cl_idle_timeout_ms) {
564 			mlog(ML_NOTICE, "o2net: keepalive delay must be "
565 			     "smaller than idle timeout\n");
566 			ret = -EINVAL;
567 		} else {
568 			cluster->cl_keepalive_delay_ms = val;
569 		}
570 	}
571 
572 	return ret;
573 }
574 
575 static ssize_t o2nm_cluster_attr_reconnect_delay_ms_read(
576 	struct o2nm_cluster *cluster, char *page)
577 {
578 	return sprintf(page, "%u\n", cluster->cl_reconnect_delay_ms);
579 }
580 
581 static ssize_t o2nm_cluster_attr_reconnect_delay_ms_write(
582 	struct o2nm_cluster *cluster, const char *page, size_t count)
583 {
584 	return o2nm_cluster_attr_write(page, count,
585 	                               &cluster->cl_reconnect_delay_ms);
586 }
587 
588 static ssize_t o2nm_cluster_attr_fence_method_read(
589 	struct o2nm_cluster *cluster, char *page)
590 {
591 	ssize_t ret = 0;
592 
593 	if (cluster)
594 		ret = sprintf(page, "%s\n",
595 			      o2nm_fence_method_desc[cluster->cl_fence_method]);
596 	return ret;
597 }
598 
599 static ssize_t o2nm_cluster_attr_fence_method_write(
600 	struct o2nm_cluster *cluster, const char *page, size_t count)
601 {
602 	unsigned int i;
603 
604 	if (page[count - 1] != '\n')
605 		goto bail;
606 
607 	for (i = 0; i < O2NM_FENCE_METHODS; ++i) {
608 		if (count != strlen(o2nm_fence_method_desc[i]) + 1)
609 			continue;
610 		if (strncasecmp(page, o2nm_fence_method_desc[i], count - 1))
611 			continue;
612 		if (cluster->cl_fence_method != i) {
613 			printk(KERN_INFO "ocfs2: Changing fence method to %s\n",
614 			       o2nm_fence_method_desc[i]);
615 			cluster->cl_fence_method = i;
616 		}
617 		return count;
618 	}
619 
620 bail:
621 	return -EINVAL;
622 }
623 
624 static struct o2nm_cluster_attribute o2nm_cluster_attr_idle_timeout_ms = {
625 	.attr	= { .ca_owner = THIS_MODULE,
626 		    .ca_name = "idle_timeout_ms",
627 		    .ca_mode = S_IRUGO | S_IWUSR },
628 	.show	= o2nm_cluster_attr_idle_timeout_ms_read,
629 	.store	= o2nm_cluster_attr_idle_timeout_ms_write,
630 };
631 
632 static struct o2nm_cluster_attribute o2nm_cluster_attr_keepalive_delay_ms = {
633 	.attr	= { .ca_owner = THIS_MODULE,
634 		    .ca_name = "keepalive_delay_ms",
635 		    .ca_mode = S_IRUGO | S_IWUSR },
636 	.show	= o2nm_cluster_attr_keepalive_delay_ms_read,
637 	.store	= o2nm_cluster_attr_keepalive_delay_ms_write,
638 };
639 
640 static struct o2nm_cluster_attribute o2nm_cluster_attr_reconnect_delay_ms = {
641 	.attr	= { .ca_owner = THIS_MODULE,
642 		    .ca_name = "reconnect_delay_ms",
643 		    .ca_mode = S_IRUGO | S_IWUSR },
644 	.show	= o2nm_cluster_attr_reconnect_delay_ms_read,
645 	.store	= o2nm_cluster_attr_reconnect_delay_ms_write,
646 };
647 
648 static struct o2nm_cluster_attribute o2nm_cluster_attr_fence_method = {
649 	.attr	= { .ca_owner = THIS_MODULE,
650 		    .ca_name = "fence_method",
651 		    .ca_mode = S_IRUGO | S_IWUSR },
652 	.show	= o2nm_cluster_attr_fence_method_read,
653 	.store	= o2nm_cluster_attr_fence_method_write,
654 };
655 
656 static struct configfs_attribute *o2nm_cluster_attrs[] = {
657 	&o2nm_cluster_attr_idle_timeout_ms.attr,
658 	&o2nm_cluster_attr_keepalive_delay_ms.attr,
659 	&o2nm_cluster_attr_reconnect_delay_ms.attr,
660 	&o2nm_cluster_attr_fence_method.attr,
661 	NULL,
662 };
663 static ssize_t o2nm_cluster_show(struct config_item *item,
664                                  struct configfs_attribute *attr,
665                                  char *page)
666 {
667 	struct o2nm_cluster *cluster = to_o2nm_cluster(item);
668 	struct o2nm_cluster_attribute *o2nm_cluster_attr =
669 		container_of(attr, struct o2nm_cluster_attribute, attr);
670 	ssize_t ret = 0;
671 
672 	if (o2nm_cluster_attr->show)
673 		ret = o2nm_cluster_attr->show(cluster, page);
674 	return ret;
675 }
676 
677 static ssize_t o2nm_cluster_store(struct config_item *item,
678                                   struct configfs_attribute *attr,
679                                   const char *page, size_t count)
680 {
681 	struct o2nm_cluster *cluster = to_o2nm_cluster(item);
682 	struct o2nm_cluster_attribute *o2nm_cluster_attr =
683 		container_of(attr, struct o2nm_cluster_attribute, attr);
684 	ssize_t ret;
685 
686 	if (o2nm_cluster_attr->store == NULL) {
687 		ret = -EINVAL;
688 		goto out;
689 	}
690 
691 	ret = o2nm_cluster_attr->store(cluster, page, count);
692 	if (ret < count)
693 		goto out;
694 out:
695 	return ret;
696 }
697 
698 static struct config_item *o2nm_node_group_make_item(struct config_group *group,
699 						     const char *name)
700 {
701 	struct o2nm_node *node = NULL;
702 
703 	if (strlen(name) > O2NM_MAX_NAME_LEN)
704 		return ERR_PTR(-ENAMETOOLONG);
705 
706 	node = kzalloc(sizeof(struct o2nm_node), GFP_KERNEL);
707 	if (node == NULL)
708 		return ERR_PTR(-ENOMEM);
709 
710 	strcpy(node->nd_name, name); /* use item.ci_namebuf instead? */
711 	config_item_init_type_name(&node->nd_item, name, &o2nm_node_type);
712 	spin_lock_init(&node->nd_lock);
713 
714 	mlog(ML_CLUSTER, "o2nm: Registering node %s\n", name);
715 
716 	return &node->nd_item;
717 }
718 
719 static void o2nm_node_group_drop_item(struct config_group *group,
720 				      struct config_item *item)
721 {
722 	struct o2nm_node *node = to_o2nm_node(item);
723 	struct o2nm_cluster *cluster = to_o2nm_cluster(group->cg_item.ci_parent);
724 
725 	o2net_disconnect_node(node);
726 
727 	if (cluster->cl_has_local &&
728 	    (cluster->cl_local_node == node->nd_num)) {
729 		cluster->cl_has_local = 0;
730 		cluster->cl_local_node = O2NM_INVALID_NODE_NUM;
731 		o2net_stop_listening(node);
732 	}
733 
734 	/* XXX call into net to stop this node from trading messages */
735 
736 	write_lock(&cluster->cl_nodes_lock);
737 
738 	/* XXX sloppy */
739 	if (node->nd_ipv4_address)
740 		rb_erase(&node->nd_ip_node, &cluster->cl_node_ip_tree);
741 
742 	/* nd_num might be 0 if the node number hasn't been set.. */
743 	if (cluster->cl_nodes[node->nd_num] == node) {
744 		cluster->cl_nodes[node->nd_num] = NULL;
745 		clear_bit(node->nd_num, cluster->cl_nodes_bitmap);
746 	}
747 	write_unlock(&cluster->cl_nodes_lock);
748 
749 	mlog(ML_CLUSTER, "o2nm: Unregistered node %s\n",
750 	     config_item_name(&node->nd_item));
751 
752 	config_item_put(item);
753 }
754 
755 static struct configfs_group_operations o2nm_node_group_group_ops = {
756 	.make_item	= o2nm_node_group_make_item,
757 	.drop_item	= o2nm_node_group_drop_item,
758 };
759 
760 static struct config_item_type o2nm_node_group_type = {
761 	.ct_group_ops	= &o2nm_node_group_group_ops,
762 	.ct_owner	= THIS_MODULE,
763 };
764 
765 /* cluster */
766 
767 static void o2nm_cluster_release(struct config_item *item)
768 {
769 	struct o2nm_cluster *cluster = to_o2nm_cluster(item);
770 
771 	kfree(cluster->cl_group.default_groups);
772 	kfree(cluster);
773 }
774 
775 static struct configfs_item_operations o2nm_cluster_item_ops = {
776 	.release	= o2nm_cluster_release,
777 	.show_attribute		= o2nm_cluster_show,
778 	.store_attribute	= o2nm_cluster_store,
779 };
780 
781 static struct config_item_type o2nm_cluster_type = {
782 	.ct_item_ops	= &o2nm_cluster_item_ops,
783 	.ct_attrs	= o2nm_cluster_attrs,
784 	.ct_owner	= THIS_MODULE,
785 };
786 
787 /* cluster set */
788 
789 struct o2nm_cluster_group {
790 	struct configfs_subsystem cs_subsys;
791 	/* some stuff? */
792 };
793 
794 #if 0
795 static struct o2nm_cluster_group *to_o2nm_cluster_group(struct config_group *group)
796 {
797 	return group ?
798 		container_of(to_configfs_subsystem(group), struct o2nm_cluster_group, cs_subsys)
799 	       : NULL;
800 }
801 #endif
802 
803 static struct config_group *o2nm_cluster_group_make_group(struct config_group *group,
804 							  const char *name)
805 {
806 	struct o2nm_cluster *cluster = NULL;
807 	struct o2nm_node_group *ns = NULL;
808 	struct config_group *o2hb_group = NULL, *ret = NULL;
809 	void *defs = NULL;
810 
811 	/* this runs under the parent dir's i_mutex; there can be only
812 	 * one caller in here at a time */
813 	if (o2nm_single_cluster)
814 		return ERR_PTR(-ENOSPC);
815 
816 	cluster = kzalloc(sizeof(struct o2nm_cluster), GFP_KERNEL);
817 	ns = kzalloc(sizeof(struct o2nm_node_group), GFP_KERNEL);
818 	defs = kcalloc(3, sizeof(struct config_group *), GFP_KERNEL);
819 	o2hb_group = o2hb_alloc_hb_set();
820 	if (cluster == NULL || ns == NULL || o2hb_group == NULL || defs == NULL)
821 		goto out;
822 
823 	config_group_init_type_name(&cluster->cl_group, name,
824 				    &o2nm_cluster_type);
825 	config_group_init_type_name(&ns->ns_group, "node",
826 				    &o2nm_node_group_type);
827 
828 	cluster->cl_group.default_groups = defs;
829 	cluster->cl_group.default_groups[0] = &ns->ns_group;
830 	cluster->cl_group.default_groups[1] = o2hb_group;
831 	cluster->cl_group.default_groups[2] = NULL;
832 	rwlock_init(&cluster->cl_nodes_lock);
833 	cluster->cl_node_ip_tree = RB_ROOT;
834 	cluster->cl_reconnect_delay_ms = O2NET_RECONNECT_DELAY_MS_DEFAULT;
835 	cluster->cl_idle_timeout_ms    = O2NET_IDLE_TIMEOUT_MS_DEFAULT;
836 	cluster->cl_keepalive_delay_ms = O2NET_KEEPALIVE_DELAY_MS_DEFAULT;
837 	cluster->cl_fence_method       = O2NM_FENCE_RESET;
838 
839 	ret = &cluster->cl_group;
840 	o2nm_single_cluster = cluster;
841 
842 out:
843 	if (ret == NULL) {
844 		kfree(cluster);
845 		kfree(ns);
846 		o2hb_free_hb_set(o2hb_group);
847 		kfree(defs);
848 		ret = ERR_PTR(-ENOMEM);
849 	}
850 
851 	return ret;
852 }
853 
854 static void o2nm_cluster_group_drop_item(struct config_group *group, struct config_item *item)
855 {
856 	struct o2nm_cluster *cluster = to_o2nm_cluster(item);
857 	int i;
858 	struct config_item *killme;
859 
860 	BUG_ON(o2nm_single_cluster != cluster);
861 	o2nm_single_cluster = NULL;
862 
863 	for (i = 0; cluster->cl_group.default_groups[i]; i++) {
864 		killme = &cluster->cl_group.default_groups[i]->cg_item;
865 		cluster->cl_group.default_groups[i] = NULL;
866 		config_item_put(killme);
867 	}
868 
869 	config_item_put(item);
870 }
871 
872 static struct configfs_group_operations o2nm_cluster_group_group_ops = {
873 	.make_group	= o2nm_cluster_group_make_group,
874 	.drop_item	= o2nm_cluster_group_drop_item,
875 };
876 
877 static struct config_item_type o2nm_cluster_group_type = {
878 	.ct_group_ops	= &o2nm_cluster_group_group_ops,
879 	.ct_owner	= THIS_MODULE,
880 };
881 
882 static struct o2nm_cluster_group o2nm_cluster_group = {
883 	.cs_subsys = {
884 		.su_group = {
885 			.cg_item = {
886 				.ci_namebuf = "cluster",
887 				.ci_type = &o2nm_cluster_group_type,
888 			},
889 		},
890 	},
891 };
892 
893 int o2nm_depend_item(struct config_item *item)
894 {
895 	return configfs_depend_item(&o2nm_cluster_group.cs_subsys, item);
896 }
897 
898 void o2nm_undepend_item(struct config_item *item)
899 {
900 	configfs_undepend_item(&o2nm_cluster_group.cs_subsys, item);
901 }
902 
903 int o2nm_depend_this_node(void)
904 {
905 	int ret = 0;
906 	struct o2nm_node *local_node;
907 
908 	local_node = o2nm_get_node_by_num(o2nm_this_node());
909 	if (!local_node) {
910 		ret = -EINVAL;
911 		goto out;
912 	}
913 
914 	ret = o2nm_depend_item(&local_node->nd_item);
915 	o2nm_node_put(local_node);
916 
917 out:
918 	return ret;
919 }
920 
921 void o2nm_undepend_this_node(void)
922 {
923 	struct o2nm_node *local_node;
924 
925 	local_node = o2nm_get_node_by_num(o2nm_this_node());
926 	BUG_ON(!local_node);
927 
928 	o2nm_undepend_item(&local_node->nd_item);
929 	o2nm_node_put(local_node);
930 }
931 
932 
933 static void __exit exit_o2nm(void)
934 {
935 	/* XXX sync with hb callbacks and shut down hb? */
936 	o2net_unregister_hb_callbacks();
937 	configfs_unregister_subsystem(&o2nm_cluster_group.cs_subsys);
938 	o2cb_sys_shutdown();
939 
940 	o2net_exit();
941 	o2hb_exit();
942 }
943 
944 static int __init init_o2nm(void)
945 {
946 	int ret = -1;
947 
948 	cluster_print_version();
949 
950 	ret = o2hb_init();
951 	if (ret)
952 		goto out;
953 
954 	ret = o2net_init();
955 	if (ret)
956 		goto out_o2hb;
957 
958 	ret = o2net_register_hb_callbacks();
959 	if (ret)
960 		goto out_o2net;
961 
962 	config_group_init(&o2nm_cluster_group.cs_subsys.su_group);
963 	mutex_init(&o2nm_cluster_group.cs_subsys.su_mutex);
964 	ret = configfs_register_subsystem(&o2nm_cluster_group.cs_subsys);
965 	if (ret) {
966 		printk(KERN_ERR "nodemanager: Registration returned %d\n", ret);
967 		goto out_callbacks;
968 	}
969 
970 	ret = o2cb_sys_init();
971 	if (!ret)
972 		goto out;
973 
974 	configfs_unregister_subsystem(&o2nm_cluster_group.cs_subsys);
975 out_callbacks:
976 	o2net_unregister_hb_callbacks();
977 out_o2net:
978 	o2net_exit();
979 out_o2hb:
980 	o2hb_exit();
981 out:
982 	return ret;
983 }
984 
985 MODULE_AUTHOR("Oracle");
986 MODULE_LICENSE("GPL");
987 
988 module_init(init_o2nm)
989 module_exit(exit_o2nm)
990