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 <dlm/dlmapi.h>
35 #define MLOG_MASK_PREFIX ML_SUPER
36 #include <cluster/masklog.h>
41 #include "heartbeat.h"
45 #include "buffer_head_io.h"
47 static inline void __ocfs2_node_map_set_bit(struct ocfs2_node_map
*map
,
49 static inline void __ocfs2_node_map_clear_bit(struct ocfs2_node_map
*map
,
51 static inline int __ocfs2_node_map_is_empty(struct ocfs2_node_map
*map
);
53 /* special case -1 for now
54 * TODO: should *really* make sure the calling func never passes -1!! */
55 static void ocfs2_node_map_init(struct ocfs2_node_map
*map
)
57 map
->num_nodes
= OCFS2_NODE_MAP_MAX_NODES
;
58 memset(map
->map
, 0, BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES
) *
59 sizeof(unsigned long));
62 void ocfs2_init_node_maps(struct ocfs2_super
*osb
)
64 spin_lock_init(&osb
->node_map_lock
);
65 ocfs2_node_map_init(&osb
->recovery_map
);
66 ocfs2_node_map_init(&osb
->osb_recovering_orphan_dirs
);
69 static void ocfs2_do_node_down(int node_num
,
70 struct ocfs2_super
*osb
)
72 BUG_ON(osb
->node_num
== node_num
);
74 mlog(0, "ocfs2: node down event for %d\n", node_num
);
78 * No DLM means we're not even ready to participate yet.
79 * We check the slots after the DLM comes up, so we will
80 * notice the node death then. We can safely ignore it
86 ocfs2_recovery_thread(osb
, node_num
);
89 /* Called from the dlm when it's about to evict a node. We may also
90 * get a heartbeat callback later. */
91 static void ocfs2_dlm_eviction_cb(int node_num
,
94 struct ocfs2_super
*osb
= (struct ocfs2_super
*) data
;
95 struct super_block
*sb
= osb
->sb
;
97 mlog(ML_NOTICE
, "device (%u,%u): dlm has evicted node %d\n",
98 MAJOR(sb
->s_dev
), MINOR(sb
->s_dev
), node_num
);
100 ocfs2_do_node_down(node_num
, osb
);
103 void ocfs2_setup_hb_callbacks(struct ocfs2_super
*osb
)
105 /* Not exactly a heartbeat callback, but leads to essentially
106 * the same path so we set it up here. */
107 dlm_setup_eviction_cb(&osb
->osb_eviction_cb
,
108 ocfs2_dlm_eviction_cb
,
112 void ocfs2_stop_heartbeat(struct ocfs2_super
*osb
)
115 char *argv
[5], *envp
[3];
117 if (ocfs2_mount_local(osb
))
120 if (!osb
->uuid_str
) {
121 /* This can happen if we don't get far enough in mount... */
122 mlog(0, "No UUID with which to stop heartbeat!\n\n");
126 argv
[0] = (char *)o2nm_get_hb_ctl_path();
129 argv
[3] = osb
->uuid_str
;
132 mlog(0, "Run: %s %s %s %s\n", argv
[0], argv
[1], argv
[2], argv
[3]);
134 /* minimal command environment taken from cpu_run_sbin_hotplug */
136 envp
[1] = "PATH=/sbin:/bin:/usr/sbin:/usr/bin";
139 ret
= call_usermodehelper(argv
[0], argv
, envp
, UMH_WAIT_PROC
);
144 static inline void __ocfs2_node_map_set_bit(struct ocfs2_node_map
*map
,
147 set_bit(bit
, map
->map
);
150 void ocfs2_node_map_set_bit(struct ocfs2_super
*osb
,
151 struct ocfs2_node_map
*map
,
156 BUG_ON(bit
>= map
->num_nodes
);
157 spin_lock(&osb
->node_map_lock
);
158 __ocfs2_node_map_set_bit(map
, bit
);
159 spin_unlock(&osb
->node_map_lock
);
162 static inline void __ocfs2_node_map_clear_bit(struct ocfs2_node_map
*map
,
165 clear_bit(bit
, map
->map
);
168 void ocfs2_node_map_clear_bit(struct ocfs2_super
*osb
,
169 struct ocfs2_node_map
*map
,
174 BUG_ON(bit
>= map
->num_nodes
);
175 spin_lock(&osb
->node_map_lock
);
176 __ocfs2_node_map_clear_bit(map
, bit
);
177 spin_unlock(&osb
->node_map_lock
);
180 int ocfs2_node_map_test_bit(struct ocfs2_super
*osb
,
181 struct ocfs2_node_map
*map
,
185 if (bit
>= map
->num_nodes
) {
186 mlog(ML_ERROR
, "bit=%d map->num_nodes=%d\n", bit
, map
->num_nodes
);
189 spin_lock(&osb
->node_map_lock
);
190 ret
= test_bit(bit
, map
->map
);
191 spin_unlock(&osb
->node_map_lock
);
195 static inline int __ocfs2_node_map_is_empty(struct ocfs2_node_map
*map
)
198 bit
= find_next_bit(map
->map
, map
->num_nodes
, 0);
199 if (bit
< map
->num_nodes
)
204 int ocfs2_node_map_is_empty(struct ocfs2_super
*osb
,
205 struct ocfs2_node_map
*map
)
208 BUG_ON(map
->num_nodes
== 0);
209 spin_lock(&osb
->node_map_lock
);
210 ret
= __ocfs2_node_map_is_empty(map
);
211 spin_unlock(&osb
->node_map_lock
);
217 static void __ocfs2_node_map_dup(struct ocfs2_node_map
*target
,
218 struct ocfs2_node_map
*from
)
220 BUG_ON(from
->num_nodes
== 0);
221 ocfs2_node_map_init(target
);
222 __ocfs2_node_map_set(target
, from
);
225 /* returns 1 if bit is the only bit set in target, 0 otherwise */
226 int ocfs2_node_map_is_only(struct ocfs2_super
*osb
,
227 struct ocfs2_node_map
*target
,
230 struct ocfs2_node_map temp
;
233 spin_lock(&osb
->node_map_lock
);
234 __ocfs2_node_map_dup(&temp
, target
);
235 __ocfs2_node_map_clear_bit(&temp
, bit
);
236 ret
= __ocfs2_node_map_is_empty(&temp
);
237 spin_unlock(&osb
->node_map_lock
);
242 static void __ocfs2_node_map_set(struct ocfs2_node_map
*target
,
243 struct ocfs2_node_map
*from
)
247 BUG_ON(target
->num_nodes
!= from
->num_nodes
);
248 BUG_ON(target
->num_nodes
== 0);
250 num_longs
= BITS_TO_LONGS(target
->num_nodes
);
251 for (i
= 0; i
< num_longs
; i
++)
252 target
->map
[i
] = from
->map
[i
];
257 /* Returns whether the recovery bit was actually set - it may not be
258 * if a node is still marked as needing recovery */
259 int ocfs2_recovery_map_set(struct ocfs2_super
*osb
,
264 spin_lock(&osb
->node_map_lock
);
266 if (!test_bit(num
, osb
->recovery_map
.map
)) {
267 __ocfs2_node_map_set_bit(&osb
->recovery_map
, num
);
271 spin_unlock(&osb
->node_map_lock
);
276 void ocfs2_recovery_map_clear(struct ocfs2_super
*osb
,
279 ocfs2_node_map_clear_bit(osb
, &osb
->recovery_map
, num
);
282 int ocfs2_node_map_iterate(struct ocfs2_super
*osb
,
283 struct ocfs2_node_map
*map
,
288 idx
= O2NM_INVALID_NODE_NUM
;
289 spin_lock(&osb
->node_map_lock
);
290 if ((i
!= O2NM_INVALID_NODE_NUM
) &&
292 (i
< map
->num_nodes
)) {
293 while(i
< map
->num_nodes
) {
294 if (test_bit(i
, map
->map
)) {
301 spin_unlock(&osb
->node_map_lock
);