aboutsummaryrefslogtreecommitdiffstats
path: root/fs/ocfs2/heartbeat.c
diff options
context:
space:
mode:
authorMark Fasheh <mark.fasheh@oracle.com>2005-12-15 17:31:24 -0500
committerJoel Becker <joel.becker@oracle.com>2006-01-03 14:45:47 -0500
commitccd979bdbce9fba8412beb3f1de68a9d0171b12c (patch)
treec50ed941849ce06ccadd4ce27599b3ef9fdbe2ae /fs/ocfs2/heartbeat.c
parent8df08c89c668e1bd922a053fdb5ba1fadbecbb38 (diff)
[PATCH] OCFS2: The Second Oracle Cluster Filesystem
The OCFS2 file system module. Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com> Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Diffstat (limited to 'fs/ocfs2/heartbeat.c')
-rw-r--r--fs/ocfs2/heartbeat.c378
1 files changed, 378 insertions, 0 deletions
diff --git a/fs/ocfs2/heartbeat.c b/fs/ocfs2/heartbeat.c
new file mode 100644
index 000000000000..0bbd22f46c80
--- /dev/null
+++ b/fs/ocfs2/heartbeat.c
@@ -0,0 +1,378 @@
1/* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
3 *
4 * heartbeat.c
5 *
6 * Register ourselves with the heartbaet service, keep our node maps
7 * up to date, and fire off recovery when needed.
8 *
9 * Copyright (C) 2002, 2004 Oracle. All rights reserved.
10 *
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.
15 *
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.
20 *
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.
25 */
26
27#include <linux/fs.h>
28#include <linux/types.h>
29#include <linux/slab.h>
30#include <linux/highmem.h>
31#include <linux/kmod.h>
32
33#include <cluster/heartbeat.h>
34#include <cluster/nodemanager.h>
35
36#include <dlm/dlmapi.h>
37
38#define MLOG_MASK_PREFIX ML_SUPER
39#include <cluster/masklog.h>
40
41#include "ocfs2.h"
42
43#include "alloc.h"
44#include "heartbeat.h"
45#include "inode.h"
46#include "journal.h"
47#include "vote.h"
48
49#include "buffer_head_io.h"
50
51#define OCFS2_HB_NODE_DOWN_PRI (0x0000002)
52#define OCFS2_HB_NODE_UP_PRI OCFS2_HB_NODE_DOWN_PRI
53
54static inline void __ocfs2_node_map_set_bit(struct ocfs2_node_map *map,
55 int bit);
56static inline void __ocfs2_node_map_clear_bit(struct ocfs2_node_map *map,
57 int bit);
58static inline int __ocfs2_node_map_is_empty(struct ocfs2_node_map *map);
59static void __ocfs2_node_map_dup(struct ocfs2_node_map *target,
60 struct ocfs2_node_map *from);
61static void __ocfs2_node_map_set(struct ocfs2_node_map *target,
62 struct ocfs2_node_map *from);
63
64void ocfs2_init_node_maps(struct ocfs2_super *osb)
65{
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);
70}
71
72static void ocfs2_do_node_down(int node_num,
73 struct ocfs2_super *osb)
74{
75 BUG_ON(osb->node_num == node_num);
76
77 mlog(0, "ocfs2: node down event for %d\n", node_num);
78
79 if (!osb->dlm) {
80 /*
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
84 * here.
85 */
86 return;
87 }
88
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);
94 return;
95 }
96
97 ocfs2_recovery_thread(osb, node_num);
98
99 ocfs2_remove_node_from_vote_queues(osb, node_num);
100}
101
102static void ocfs2_hb_node_down_cb(struct o2nm_node *node,
103 int node_num,
104 void *data)
105{
106 ocfs2_do_node_down(node_num, (struct ocfs2_super *) data);
107}
108
109/* Called from the dlm when it's about to evict a node. We may also
110 * get a heartbeat callback later. */
111static void ocfs2_dlm_eviction_cb(int node_num,
112 void *data)
113{
114 struct ocfs2_super *osb = (struct ocfs2_super *) data;
115 struct super_block *sb = osb->sb;
116
117 mlog(ML_NOTICE, "device (%u,%u): dlm has evicted node %d\n",
118 MAJOR(sb->s_dev), MINOR(sb->s_dev), node_num);
119
120 ocfs2_do_node_down(node_num, osb);
121}
122
123static void ocfs2_hb_node_up_cb(struct o2nm_node *node,
124 int node_num,
125 void *data)
126{
127 struct ocfs2_super *osb = data;
128
129 BUG_ON(osb->node_num == node_num);
130
131 mlog(0, "node up event for %d\n", node_num);
132 ocfs2_node_map_clear_bit(osb, &osb->umount_map, node_num);
133}
134
135void ocfs2_setup_hb_callbacks(struct ocfs2_super *osb)
136{
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);
140
141 o2hb_setup_callback(&osb->osb_hb_up, O2HB_NODE_UP_CB,
142 ocfs2_hb_node_up_cb, osb, OCFS2_HB_NODE_UP_PRI);
143
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,
148 osb);
149}
150
151/* Most functions here are just stubs for now... */
152int ocfs2_register_hb_callbacks(struct ocfs2_super *osb)
153{
154 int status;
155
156 status = o2hb_register_callback(&osb->osb_hb_down);
157 if (status < 0) {
158 mlog_errno(status);
159 goto bail;
160 }
161
162 status = o2hb_register_callback(&osb->osb_hb_up);
163 if (status < 0)
164 mlog_errno(status);
165
166bail:
167 return status;
168}
169
170void ocfs2_clear_hb_callbacks(struct ocfs2_super *osb)
171{
172 int status;
173
174 status = o2hb_unregister_callback(&osb->osb_hb_down);
175 if (status < 0)
176 mlog_errno(status);
177
178 status = o2hb_unregister_callback(&osb->osb_hb_up);
179 if (status < 0)
180 mlog_errno(status);
181}
182
183void ocfs2_stop_heartbeat(struct ocfs2_super *osb)
184{
185 int ret;
186 char *argv[5], *envp[3];
187
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");
191 return;
192 }
193
194 argv[0] = (char *)o2nm_get_hb_ctl_path();
195 argv[1] = "-K";
196 argv[2] = "-u";
197 argv[3] = osb->uuid_str;
198 argv[4] = NULL;
199
200 mlog(0, "Run: %s %s %s %s\n", argv[0], argv[1], argv[2], argv[3]);
201
202 /* minimal command environment taken from cpu_run_sbin_hotplug */
203 envp[0] = "HOME=/";
204 envp[1] = "PATH=/sbin:/bin:/usr/sbin:/usr/bin";
205 envp[2] = NULL;
206
207 ret = call_usermodehelper(argv[0], argv, envp, 1);
208 if (ret < 0)
209 mlog_errno(ret);
210}
211
212/* special case -1 for now
213 * TODO: should *really* make sure the calling func never passes -1!! */
214void ocfs2_node_map_init(struct ocfs2_node_map *map)
215{
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));
219}
220
221static inline void __ocfs2_node_map_set_bit(struct ocfs2_node_map *map,
222 int bit)
223{
224 set_bit(bit, map->map);
225}
226
227void ocfs2_node_map_set_bit(struct ocfs2_super *osb,
228 struct ocfs2_node_map *map,
229 int bit)
230{
231 if (bit==-1)
232 return;
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);
237}
238
239static inline void __ocfs2_node_map_clear_bit(struct ocfs2_node_map *map,
240 int bit)
241{
242 clear_bit(bit, map->map);
243}
244
245void ocfs2_node_map_clear_bit(struct ocfs2_super *osb,
246 struct ocfs2_node_map *map,
247 int bit)
248{
249 if (bit==-1)
250 return;
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);
255}
256
257int ocfs2_node_map_test_bit(struct ocfs2_super *osb,
258 struct ocfs2_node_map *map,
259 int bit)
260{
261 int ret;
262 if (bit >= map->num_nodes) {
263 mlog(ML_ERROR, "bit=%d map->num_nodes=%d\n", bit, map->num_nodes);
264 BUG();
265 }
266 spin_lock(&osb->node_map_lock);
267 ret = test_bit(bit, map->map);
268 spin_unlock(&osb->node_map_lock);
269 return ret;
270}
271
272static inline int __ocfs2_node_map_is_empty(struct ocfs2_node_map *map)
273{
274 int bit;
275 bit = find_next_bit(map->map, map->num_nodes, 0);
276 if (bit < map->num_nodes)
277 return 0;
278 return 1;
279}
280
281int ocfs2_node_map_is_empty(struct ocfs2_super *osb,
282 struct ocfs2_node_map *map)
283{
284 int ret;
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);
289 return ret;
290}
291
292static void __ocfs2_node_map_dup(struct ocfs2_node_map *target,
293 struct ocfs2_node_map *from)
294{
295 BUG_ON(from->num_nodes == 0);
296 ocfs2_node_map_init(target);
297 __ocfs2_node_map_set(target, from);
298}
299
300/* returns 1 if bit is the only bit set in target, 0 otherwise */
301int ocfs2_node_map_is_only(struct ocfs2_super *osb,
302 struct ocfs2_node_map *target,
303 int bit)
304{
305 struct ocfs2_node_map temp;
306 int ret;
307
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);
313
314 return ret;
315}
316
317static void __ocfs2_node_map_set(struct ocfs2_node_map *target,
318 struct ocfs2_node_map *from)
319{
320 int num_longs, i;
321
322 BUG_ON(target->num_nodes != from->num_nodes);
323 BUG_ON(target->num_nodes == 0);
324
325 num_longs = BITS_TO_LONGS(target->num_nodes);
326 for (i = 0; i < num_longs; i++)
327 target->map[i] = from->map[i];
328}
329
330/* Returns whether the recovery bit was actually set - it may not be
331 * if a node is still marked as needing recovery */
332int ocfs2_recovery_map_set(struct ocfs2_super *osb,
333 int num)
334{
335 int set = 0;
336
337 spin_lock(&osb->node_map_lock);
338
339 __ocfs2_node_map_clear_bit(&osb->mounted_map, num);
340
341 if (!test_bit(num, osb->recovery_map.map)) {
342 __ocfs2_node_map_set_bit(&osb->recovery_map, num);
343 set = 1;
344 }
345
346 spin_unlock(&osb->node_map_lock);
347
348 return set;
349}
350
351void ocfs2_recovery_map_clear(struct ocfs2_super *osb,
352 int num)
353{
354 ocfs2_node_map_clear_bit(osb, &osb->recovery_map, num);
355}
356
357int ocfs2_node_map_iterate(struct ocfs2_super *osb,
358 struct ocfs2_node_map *map,
359 int idx)
360{
361 int i = idx;
362
363 idx = O2NM_INVALID_NODE_NUM;
364 spin_lock(&osb->node_map_lock);
365 if ((i != O2NM_INVALID_NODE_NUM) &&
366 (i >= 0) &&
367 (i < map->num_nodes)) {
368 while(i < map->num_nodes) {
369 if (test_bit(i, map->map)) {
370 idx = i;
371 break;
372 }
373 i++;
374 }
375 }
376 spin_unlock(&osb->node_map_lock);
377 return idx;
378}