1 /* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
6 * Register ourselves with the heartbaet service, keep our node maps
7 * up to date, and fire off recovery when needed.
9 * Copyright (C) 2002, 2004 Oracle. All rights reserved.
11 * This program is free software; you can redistribute it and/or
12 * modify it under the terms of the GNU General Public
13 * License as published by the Free Software Foundation; either
14 * version 2 of the License, or (at your option) any later version.
16 * This program is distributed in the hope that it will be useful,
17 * but WITHOUT ANY WARRANTY; without even the implied warranty of
18 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
19 * General Public License for more details.
21 * You should have received a copy of the GNU General Public
22 * License along with this program; if not, write to the
23 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
24 * Boston, MA 021110-1307, USA.
28 #include <linux/types.h>
29 #include <linux/slab.h>
30 #include <linux/highmem.h>
31 #include <linux/kmod.h>
33 #include <cluster/heartbeat.h>
34 #include <cluster/nodemanager.h>
36 #include <dlm/dlmapi.h>
38 #define MLOG_MASK_PREFIX ML_SUPER
39 #include <cluster/masklog.h>
44 #include "heartbeat.h"
49 #include "buffer_head_io.h"
51 #define OCFS2_HB_NODE_DOWN_PRI (0x0000002)
52 #define OCFS2_HB_NODE_UP_PRI OCFS2_HB_NODE_DOWN_PRI
54 static inline void __ocfs2_node_map_set_bit(struct ocfs2_node_map *map,
56 static inline void __ocfs2_node_map_clear_bit(struct ocfs2_node_map *map,
58 static inline int __ocfs2_node_map_is_empty(struct ocfs2_node_map *map);
59 static void __ocfs2_node_map_dup(struct ocfs2_node_map *target,
60 struct ocfs2_node_map *from);
61 static void __ocfs2_node_map_set(struct ocfs2_node_map *target,
62 struct ocfs2_node_map *from);
64 void ocfs2_init_node_maps(struct ocfs2_super *osb)
66 spin_lock_init(&osb->node_map_lock);
67 ocfs2_node_map_init(&osb->mounted_map);
68 ocfs2_node_map_init(&osb->recovery_map);
69 ocfs2_node_map_init(&osb->umount_map);
72 static void ocfs2_do_node_down(int node_num,
73 struct ocfs2_super *osb)
75 BUG_ON(osb->node_num == node_num);
77 mlog(0, "ocfs2: node down event for %d\n", node_num);
81 * No DLM means we're not even ready to participate yet.
82 * We check the slots after the DLM comes up, so we will
83 * notice the node death then. We can safely ignore it
89 if (ocfs2_node_map_test_bit(osb, &osb->umount_map, node_num)) {
90 /* If a node is in the umount map, then we've been
91 * expecting him to go down and we know ahead of time
92 * that recovery is not necessary. */
93 ocfs2_node_map_clear_bit(osb, &osb->umount_map, node_num);
97 ocfs2_recovery_thread(osb, node_num);
99 ocfs2_remove_node_from_vote_queues(osb, node_num);
102 static void ocfs2_hb_node_down_cb(struct o2nm_node *node,
106 ocfs2_do_node_down(node_num, (struct ocfs2_super *) data);
109 /* Called from the dlm when it's about to evict a node. We may also
110 * get a heartbeat callback later. */
111 static void ocfs2_dlm_eviction_cb(int node_num,
114 struct ocfs2_super *osb = (struct ocfs2_super *) data;
115 struct super_block *sb = osb->sb;
117 mlog(ML_NOTICE, "device (%u,%u): dlm has evicted node %d\n",
118 MAJOR(sb->s_dev), MINOR(sb->s_dev), node_num);
120 ocfs2_do_node_down(node_num, osb);
123 static void ocfs2_hb_node_up_cb(struct o2nm_node *node,
127 struct ocfs2_super *osb = data;
129 BUG_ON(osb->node_num == node_num);
131 mlog(0, "node up event for %d\n", node_num);
132 ocfs2_node_map_clear_bit(osb, &osb->umount_map, node_num);
135 void ocfs2_setup_hb_callbacks(struct ocfs2_super *osb)
137 o2hb_setup_callback(&osb->osb_hb_down, O2HB_NODE_DOWN_CB,
138 ocfs2_hb_node_down_cb, osb,
139 OCFS2_HB_NODE_DOWN_PRI);
141 o2hb_setup_callback(&osb->osb_hb_up, O2HB_NODE_UP_CB,
142 ocfs2_hb_node_up_cb, osb, OCFS2_HB_NODE_UP_PRI);
144 /* Not exactly a heartbeat callback, but leads to essentially
145 * the same path so we set it up here. */
146 dlm_setup_eviction_cb(&osb->osb_eviction_cb,
147 ocfs2_dlm_eviction_cb,
151 /* Most functions here are just stubs for now... */
152 int ocfs2_register_hb_callbacks(struct ocfs2_super *osb)
156 status = o2hb_register_callback(&osb->osb_hb_down);
162 status = o2hb_register_callback(&osb->osb_hb_up);
170 void ocfs2_clear_hb_callbacks(struct ocfs2_super *osb)
174 status = o2hb_unregister_callback(&osb->osb_hb_down);
178 status = o2hb_unregister_callback(&osb->osb_hb_up);
183 void ocfs2_stop_heartbeat(struct ocfs2_super *osb)
186 char *argv[5], *envp[3];
188 if (!osb->uuid_str) {
189 /* This can happen if we don't get far enough in mount... */
190 mlog(0, "No UUID with which to stop heartbeat!\n\n");
194 argv[0] = (char *)o2nm_get_hb_ctl_path();
197 argv[3] = osb->uuid_str;
200 mlog(0, "Run: %s %s %s %s\n", argv[0], argv[1], argv[2], argv[3]);
202 /* minimal command environment taken from cpu_run_sbin_hotplug */
204 envp[1] = "PATH=/sbin:/bin:/usr/sbin:/usr/bin";
207 ret = call_usermodehelper(argv[0], argv, envp, 1);
212 /* special case -1 for now
213 * TODO: should *really* make sure the calling func never passes -1!! */
214 void ocfs2_node_map_init(struct ocfs2_node_map *map)
216 map->num_nodes = OCFS2_NODE_MAP_MAX_NODES;
217 memset(map->map, 0, BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES) *
218 sizeof(unsigned long));
221 static inline void __ocfs2_node_map_set_bit(struct ocfs2_node_map *map,
224 set_bit(bit, map->map);
227 void ocfs2_node_map_set_bit(struct ocfs2_super *osb,
228 struct ocfs2_node_map *map,
233 BUG_ON(bit >= map->num_nodes);
234 spin_lock(&osb->node_map_lock);
235 __ocfs2_node_map_set_bit(map, bit);
236 spin_unlock(&osb->node_map_lock);
239 static inline void __ocfs2_node_map_clear_bit(struct ocfs2_node_map *map,
242 clear_bit(bit, map->map);
245 void ocfs2_node_map_clear_bit(struct ocfs2_super *osb,
246 struct ocfs2_node_map *map,
251 BUG_ON(bit >= map->num_nodes);
252 spin_lock(&osb->node_map_lock);
253 __ocfs2_node_map_clear_bit(map, bit);
254 spin_unlock(&osb->node_map_lock);
257 int ocfs2_node_map_test_bit(struct ocfs2_super *osb,
258 struct ocfs2_node_map *map,
262 if (bit >= map->num_nodes) {
263 mlog(ML_ERROR, "bit=%d map->num_nodes=%d\n", bit, map->num_nodes);
266 spin_lock(&osb->node_map_lock);
267 ret = test_bit(bit, map->map);
268 spin_unlock(&osb->node_map_lock);
272 static inline int __ocfs2_node_map_is_empty(struct ocfs2_node_map *map)
275 bit = find_next_bit(map->map, map->num_nodes, 0);
276 if (bit < map->num_nodes)
281 int ocfs2_node_map_is_empty(struct ocfs2_super *osb,
282 struct ocfs2_node_map *map)
285 BUG_ON(map->num_nodes == 0);
286 spin_lock(&osb->node_map_lock);
287 ret = __ocfs2_node_map_is_empty(map);
288 spin_unlock(&osb->node_map_lock);
292 static void __ocfs2_node_map_dup(struct ocfs2_node_map *target,
293 struct ocfs2_node_map *from)
295 BUG_ON(from->num_nodes == 0);
296 ocfs2_node_map_init(target);
297 __ocfs2_node_map_set(target, from);
300 /* returns 1 if bit is the only bit set in target, 0 otherwise */
301 int ocfs2_node_map_is_only(struct ocfs2_super *osb,
302 struct ocfs2_node_map *target,
305 struct ocfs2_node_map temp;
308 spin_lock(&osb->node_map_lock);
309 __ocfs2_node_map_dup(&temp, target);
310 __ocfs2_node_map_clear_bit(&temp, bit);
311 ret = __ocfs2_node_map_is_empty(&temp);
312 spin_unlock(&osb->node_map_lock);
317 static void __ocfs2_node_map_set(struct ocfs2_node_map *target,
318 struct ocfs2_node_map *from)
322 BUG_ON(target->num_nodes != from->num_nodes);
323 BUG_ON(target->num_nodes == 0);
325 num_longs = BITS_TO_LONGS(target->num_nodes);
326 for (i = 0; i < num_longs; i++)
327 target->map[i] = from->map[i];
330 /* Returns whether the recovery bit was actually set - it may not be
331 * if a node is still marked as needing recovery */
332 int ocfs2_recovery_map_set(struct ocfs2_super *osb,
337 spin_lock(&osb->node_map_lock);
339 __ocfs2_node_map_clear_bit(&osb->mounted_map, num);
341 if (!test_bit(num, osb->recovery_map.map)) {
342 __ocfs2_node_map_set_bit(&osb->recovery_map, num);
346 spin_unlock(&osb->node_map_lock);
351 void ocfs2_recovery_map_clear(struct ocfs2_super *osb,
354 ocfs2_node_map_clear_bit(osb, &osb->recovery_map, num);
357 int ocfs2_node_map_iterate(struct ocfs2_super *osb,
358 struct ocfs2_node_map *map,
363 idx = O2NM_INVALID_NODE_NUM;
364 spin_lock(&osb->node_map_lock);
365 if ((i != O2NM_INVALID_NODE_NUM) &&
367 (i < map->num_nodes)) {
368 while(i < map->num_nodes) {
369 if (test_bit(i, map->map)) {
376 spin_unlock(&osb->node_map_lock);