blob: c0efd9489fe8168f88d28c1abe6acdb2cedb3385 [file] [log] [blame]
Mark Fashehccd979b2005-12-15 14:31:24 -08001/* -*- 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
Mark Fashehccd979b2005-12-15 14:31:24 -080033#include <dlm/dlmapi.h>
34
35#define MLOG_MASK_PREFIX ML_SUPER
36#include <cluster/masklog.h>
37
38#include "ocfs2.h"
39
40#include "alloc.h"
41#include "heartbeat.h"
42#include "inode.h"
43#include "journal.h"
Mark Fashehccd979b2005-12-15 14:31:24 -080044
45#include "buffer_head_io.h"
46
Mark Fashehccd979b2005-12-15 14:31:24 -080047static inline void __ocfs2_node_map_set_bit(struct ocfs2_node_map *map,
48 int bit);
49static inline void __ocfs2_node_map_clear_bit(struct ocfs2_node_map *map,
50 int bit);
51static inline int __ocfs2_node_map_is_empty(struct ocfs2_node_map *map);
52static void __ocfs2_node_map_dup(struct ocfs2_node_map *target,
53 struct ocfs2_node_map *from);
54static void __ocfs2_node_map_set(struct ocfs2_node_map *target,
55 struct ocfs2_node_map *from);
56
57void ocfs2_init_node_maps(struct ocfs2_super *osb)
58{
59 spin_lock_init(&osb->node_map_lock);
Mark Fashehccd979b2005-12-15 14:31:24 -080060 ocfs2_node_map_init(&osb->recovery_map);
Mark Fashehb4df6ed2006-02-22 17:35:08 -080061 ocfs2_node_map_init(&osb->osb_recovering_orphan_dirs);
Mark Fashehccd979b2005-12-15 14:31:24 -080062}
63
64static void ocfs2_do_node_down(int node_num,
65 struct ocfs2_super *osb)
66{
67 BUG_ON(osb->node_num == node_num);
68
69 mlog(0, "ocfs2: node down event for %d\n", node_num);
70
71 if (!osb->dlm) {
72 /*
73 * No DLM means we're not even ready to participate yet.
74 * We check the slots after the DLM comes up, so we will
75 * notice the node death then. We can safely ignore it
76 * here.
77 */
78 return;
79 }
80
Mark Fashehccd979b2005-12-15 14:31:24 -080081 ocfs2_recovery_thread(osb, node_num);
Mark Fashehccd979b2005-12-15 14:31:24 -080082}
83
Mark Fashehccd979b2005-12-15 14:31:24 -080084/* Called from the dlm when it's about to evict a node. We may also
85 * get a heartbeat callback later. */
86static void ocfs2_dlm_eviction_cb(int node_num,
87 void *data)
88{
89 struct ocfs2_super *osb = (struct ocfs2_super *) data;
90 struct super_block *sb = osb->sb;
91
92 mlog(ML_NOTICE, "device (%u,%u): dlm has evicted node %d\n",
93 MAJOR(sb->s_dev), MINOR(sb->s_dev), node_num);
94
95 ocfs2_do_node_down(node_num, osb);
96}
97
Mark Fashehccd979b2005-12-15 14:31:24 -080098void ocfs2_setup_hb_callbacks(struct ocfs2_super *osb)
99{
Mark Fashehccd979b2005-12-15 14:31:24 -0800100 /* Not exactly a heartbeat callback, but leads to essentially
101 * the same path so we set it up here. */
102 dlm_setup_eviction_cb(&osb->osb_eviction_cb,
103 ocfs2_dlm_eviction_cb,
104 osb);
105}
106
Mark Fashehccd979b2005-12-15 14:31:24 -0800107void ocfs2_stop_heartbeat(struct ocfs2_super *osb)
108{
109 int ret;
110 char *argv[5], *envp[3];
111
Sunil Mushranc271c5c2006-12-05 17:56:35 -0800112 if (ocfs2_mount_local(osb))
113 return;
114
Mark Fashehccd979b2005-12-15 14:31:24 -0800115 if (!osb->uuid_str) {
116 /* This can happen if we don't get far enough in mount... */
117 mlog(0, "No UUID with which to stop heartbeat!\n\n");
118 return;
119 }
120
121 argv[0] = (char *)o2nm_get_hb_ctl_path();
122 argv[1] = "-K";
123 argv[2] = "-u";
124 argv[3] = osb->uuid_str;
125 argv[4] = NULL;
126
127 mlog(0, "Run: %s %s %s %s\n", argv[0], argv[1], argv[2], argv[3]);
128
129 /* minimal command environment taken from cpu_run_sbin_hotplug */
130 envp[0] = "HOME=/";
131 envp[1] = "PATH=/sbin:/bin:/usr/sbin:/usr/bin";
132 envp[2] = NULL;
133
Jeremy Fitzhardinge86313c42007-07-17 18:37:03 -0700134 ret = call_usermodehelper(argv[0], argv, envp, UMH_WAIT_PROC);
Mark Fashehccd979b2005-12-15 14:31:24 -0800135 if (ret < 0)
136 mlog_errno(ret);
137}
138
139/* special case -1 for now
140 * TODO: should *really* make sure the calling func never passes -1!! */
141void ocfs2_node_map_init(struct ocfs2_node_map *map)
142{
143 map->num_nodes = OCFS2_NODE_MAP_MAX_NODES;
144 memset(map->map, 0, BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES) *
145 sizeof(unsigned long));
146}
147
148static inline void __ocfs2_node_map_set_bit(struct ocfs2_node_map *map,
149 int bit)
150{
151 set_bit(bit, map->map);
152}
153
154void ocfs2_node_map_set_bit(struct ocfs2_super *osb,
155 struct ocfs2_node_map *map,
156 int bit)
157{
158 if (bit==-1)
159 return;
160 BUG_ON(bit >= map->num_nodes);
161 spin_lock(&osb->node_map_lock);
162 __ocfs2_node_map_set_bit(map, bit);
163 spin_unlock(&osb->node_map_lock);
164}
165
166static inline void __ocfs2_node_map_clear_bit(struct ocfs2_node_map *map,
167 int bit)
168{
169 clear_bit(bit, map->map);
170}
171
172void ocfs2_node_map_clear_bit(struct ocfs2_super *osb,
173 struct ocfs2_node_map *map,
174 int bit)
175{
176 if (bit==-1)
177 return;
178 BUG_ON(bit >= map->num_nodes);
179 spin_lock(&osb->node_map_lock);
180 __ocfs2_node_map_clear_bit(map, bit);
181 spin_unlock(&osb->node_map_lock);
182}
183
184int ocfs2_node_map_test_bit(struct ocfs2_super *osb,
185 struct ocfs2_node_map *map,
186 int bit)
187{
188 int ret;
189 if (bit >= map->num_nodes) {
190 mlog(ML_ERROR, "bit=%d map->num_nodes=%d\n", bit, map->num_nodes);
191 BUG();
192 }
193 spin_lock(&osb->node_map_lock);
194 ret = test_bit(bit, map->map);
195 spin_unlock(&osb->node_map_lock);
196 return ret;
197}
198
199static inline int __ocfs2_node_map_is_empty(struct ocfs2_node_map *map)
200{
201 int bit;
202 bit = find_next_bit(map->map, map->num_nodes, 0);
203 if (bit < map->num_nodes)
204 return 0;
205 return 1;
206}
207
208int ocfs2_node_map_is_empty(struct ocfs2_super *osb,
209 struct ocfs2_node_map *map)
210{
211 int ret;
212 BUG_ON(map->num_nodes == 0);
213 spin_lock(&osb->node_map_lock);
214 ret = __ocfs2_node_map_is_empty(map);
215 spin_unlock(&osb->node_map_lock);
216 return ret;
217}
218
219static void __ocfs2_node_map_dup(struct ocfs2_node_map *target,
220 struct ocfs2_node_map *from)
221{
222 BUG_ON(from->num_nodes == 0);
223 ocfs2_node_map_init(target);
224 __ocfs2_node_map_set(target, from);
225}
226
227/* returns 1 if bit is the only bit set in target, 0 otherwise */
228int ocfs2_node_map_is_only(struct ocfs2_super *osb,
229 struct ocfs2_node_map *target,
230 int bit)
231{
232 struct ocfs2_node_map temp;
233 int ret;
234
235 spin_lock(&osb->node_map_lock);
236 __ocfs2_node_map_dup(&temp, target);
237 __ocfs2_node_map_clear_bit(&temp, bit);
238 ret = __ocfs2_node_map_is_empty(&temp);
239 spin_unlock(&osb->node_map_lock);
240
241 return ret;
242}
243
244static void __ocfs2_node_map_set(struct ocfs2_node_map *target,
245 struct ocfs2_node_map *from)
246{
247 int num_longs, i;
248
249 BUG_ON(target->num_nodes != from->num_nodes);
250 BUG_ON(target->num_nodes == 0);
251
252 num_longs = BITS_TO_LONGS(target->num_nodes);
253 for (i = 0; i < num_longs; i++)
254 target->map[i] = from->map[i];
255}
256
257/* Returns whether the recovery bit was actually set - it may not be
258 * if a node is still marked as needing recovery */
259int ocfs2_recovery_map_set(struct ocfs2_super *osb,
260 int num)
261{
262 int set = 0;
263
264 spin_lock(&osb->node_map_lock);
265
Mark Fashehccd979b2005-12-15 14:31:24 -0800266 if (!test_bit(num, osb->recovery_map.map)) {
267 __ocfs2_node_map_set_bit(&osb->recovery_map, num);
268 set = 1;
269 }
270
271 spin_unlock(&osb->node_map_lock);
272
273 return set;
274}
275
276void ocfs2_recovery_map_clear(struct ocfs2_super *osb,
277 int num)
278{
279 ocfs2_node_map_clear_bit(osb, &osb->recovery_map, num);
280}
281
282int ocfs2_node_map_iterate(struct ocfs2_super *osb,
283 struct ocfs2_node_map *map,
284 int idx)
285{
286 int i = idx;
287
288 idx = O2NM_INVALID_NODE_NUM;
289 spin_lock(&osb->node_map_lock);
290 if ((i != O2NM_INVALID_NODE_NUM) &&
291 (i >= 0) &&
292 (i < map->num_nodes)) {
293 while(i < map->num_nodes) {
294 if (test_bit(i, map->map)) {
295 idx = i;
296 break;
297 }
298 i++;
299 }
300 }
301 spin_unlock(&osb->node_map_lock);
302 return idx;
303}