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
);
70 ocfs2_node_map_init(&osb
->osb_recovering_orphan_dirs
);
73 static void ocfs2_do_node_down(int node_num
,
74 struct ocfs2_super
*osb
)
76 BUG_ON(osb
->node_num
== node_num
);
78 mlog(0, "ocfs2: node down event for %d\n", node_num
);
82 * No DLM means we're not even ready to participate yet.
83 * We check the slots after the DLM comes up, so we will
84 * notice the node death then. We can safely ignore it
90 if (ocfs2_node_map_test_bit(osb
, &osb
->umount_map
, node_num
)) {
91 /* If a node is in the umount map, then we've been
92 * expecting him to go down and we know ahead of time
93 * that recovery is not necessary. */
94 ocfs2_node_map_clear_bit(osb
, &osb
->umount_map
, node_num
);
98 ocfs2_recovery_thread(osb
, node_num
);
100 ocfs2_remove_node_from_vote_queues(osb
, node_num
);
103 static void ocfs2_hb_node_down_cb(struct o2nm_node
*node
,
107 ocfs2_do_node_down(node_num
, (struct ocfs2_super
*) data
);
110 /* Called from the dlm when it's about to evict a node. We may also
111 * get a heartbeat callback later. */
112 static void ocfs2_dlm_eviction_cb(int node_num
,
115 struct ocfs2_super
*osb
= (struct ocfs2_super
*) data
;
116 struct super_block
*sb
= osb
->sb
;
118 mlog(ML_NOTICE
, "device (%u,%u): dlm has evicted node %d\n",
119 MAJOR(sb
->s_dev
), MINOR(sb
->s_dev
), node_num
);
121 ocfs2_do_node_down(node_num
, osb
);
124 static void ocfs2_hb_node_up_cb(struct o2nm_node
*node
,
128 struct ocfs2_super
*osb
= data
;
130 BUG_ON(osb
->node_num
== node_num
);
132 mlog(0, "node up event for %d\n", node_num
);
133 ocfs2_node_map_clear_bit(osb
, &osb
->umount_map
, node_num
);
136 void ocfs2_setup_hb_callbacks(struct ocfs2_super
*osb
)
138 o2hb_setup_callback(&osb
->osb_hb_down
, O2HB_NODE_DOWN_CB
,
139 ocfs2_hb_node_down_cb
, osb
,
140 OCFS2_HB_NODE_DOWN_PRI
);
142 o2hb_setup_callback(&osb
->osb_hb_up
, O2HB_NODE_UP_CB
,
143 ocfs2_hb_node_up_cb
, osb
, OCFS2_HB_NODE_UP_PRI
);
145 /* Not exactly a heartbeat callback, but leads to essentially
146 * the same path so we set it up here. */
147 dlm_setup_eviction_cb(&osb
->osb_eviction_cb
,
148 ocfs2_dlm_eviction_cb
,
152 /* Most functions here are just stubs for now... */
153 int ocfs2_register_hb_callbacks(struct ocfs2_super
*osb
)
157 if (ocfs2_mount_local(osb
))
160 status
= o2hb_register_callback(osb
->uuid_str
, &osb
->osb_hb_down
);
166 status
= o2hb_register_callback(osb
->uuid_str
, &osb
->osb_hb_up
);
169 o2hb_unregister_callback(osb
->uuid_str
, &osb
->osb_hb_down
);
176 void ocfs2_clear_hb_callbacks(struct ocfs2_super
*osb
)
178 if (ocfs2_mount_local(osb
))
181 o2hb_unregister_callback(osb
->uuid_str
, &osb
->osb_hb_down
);
182 o2hb_unregister_callback(osb
->uuid_str
, &osb
->osb_hb_up
);
185 void ocfs2_stop_heartbeat(struct ocfs2_super
*osb
)
188 char *argv
[5], *envp
[3];
190 if (ocfs2_mount_local(osb
))
193 if (!osb
->uuid_str
) {
194 /* This can happen if we don't get far enough in mount... */
195 mlog(0, "No UUID with which to stop heartbeat!\n\n");
199 argv
[0] = (char *)o2nm_get_hb_ctl_path();
202 argv
[3] = osb
->uuid_str
;
205 mlog(0, "Run: %s %s %s %s\n", argv
[0], argv
[1], argv
[2], argv
[3]);
207 /* minimal command environment taken from cpu_run_sbin_hotplug */
209 envp
[1] = "PATH=/sbin:/bin:/usr/sbin:/usr/bin";
212 ret
= call_usermodehelper(argv
[0], argv
, envp
, UMH_WAIT_PROC
);
217 /* special case -1 for now
218 * TODO: should *really* make sure the calling func never passes -1!! */
219 void ocfs2_node_map_init(struct ocfs2_node_map
*map
)
221 map
->num_nodes
= OCFS2_NODE_MAP_MAX_NODES
;
222 memset(map
->map
, 0, BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES
) *
223 sizeof(unsigned long));
226 static inline void __ocfs2_node_map_set_bit(struct ocfs2_node_map
*map
,
229 set_bit(bit
, map
->map
);
232 void ocfs2_node_map_set_bit(struct ocfs2_super
*osb
,
233 struct ocfs2_node_map
*map
,
238 BUG_ON(bit
>= map
->num_nodes
);
239 spin_lock(&osb
->node_map_lock
);
240 __ocfs2_node_map_set_bit(map
, bit
);
241 spin_unlock(&osb
->node_map_lock
);
244 static inline void __ocfs2_node_map_clear_bit(struct ocfs2_node_map
*map
,
247 clear_bit(bit
, map
->map
);
250 void ocfs2_node_map_clear_bit(struct ocfs2_super
*osb
,
251 struct ocfs2_node_map
*map
,
256 BUG_ON(bit
>= map
->num_nodes
);
257 spin_lock(&osb
->node_map_lock
);
258 __ocfs2_node_map_clear_bit(map
, bit
);
259 spin_unlock(&osb
->node_map_lock
);
262 int ocfs2_node_map_test_bit(struct ocfs2_super
*osb
,
263 struct ocfs2_node_map
*map
,
267 if (bit
>= map
->num_nodes
) {
268 mlog(ML_ERROR
, "bit=%d map->num_nodes=%d\n", bit
, map
->num_nodes
);
271 spin_lock(&osb
->node_map_lock
);
272 ret
= test_bit(bit
, map
->map
);
273 spin_unlock(&osb
->node_map_lock
);
277 static inline int __ocfs2_node_map_is_empty(struct ocfs2_node_map
*map
)
280 bit
= find_next_bit(map
->map
, map
->num_nodes
, 0);
281 if (bit
< map
->num_nodes
)
286 int ocfs2_node_map_is_empty(struct ocfs2_super
*osb
,
287 struct ocfs2_node_map
*map
)
290 BUG_ON(map
->num_nodes
== 0);
291 spin_lock(&osb
->node_map_lock
);
292 ret
= __ocfs2_node_map_is_empty(map
);
293 spin_unlock(&osb
->node_map_lock
);
297 static void __ocfs2_node_map_dup(struct ocfs2_node_map
*target
,
298 struct ocfs2_node_map
*from
)
300 BUG_ON(from
->num_nodes
== 0);
301 ocfs2_node_map_init(target
);
302 __ocfs2_node_map_set(target
, from
);
305 /* returns 1 if bit is the only bit set in target, 0 otherwise */
306 int ocfs2_node_map_is_only(struct ocfs2_super
*osb
,
307 struct ocfs2_node_map
*target
,
310 struct ocfs2_node_map temp
;
313 spin_lock(&osb
->node_map_lock
);
314 __ocfs2_node_map_dup(&temp
, target
);
315 __ocfs2_node_map_clear_bit(&temp
, bit
);
316 ret
= __ocfs2_node_map_is_empty(&temp
);
317 spin_unlock(&osb
->node_map_lock
);
322 static void __ocfs2_node_map_set(struct ocfs2_node_map
*target
,
323 struct ocfs2_node_map
*from
)
327 BUG_ON(target
->num_nodes
!= from
->num_nodes
);
328 BUG_ON(target
->num_nodes
== 0);
330 num_longs
= BITS_TO_LONGS(target
->num_nodes
);
331 for (i
= 0; i
< num_longs
; i
++)
332 target
->map
[i
] = from
->map
[i
];
335 /* Returns whether the recovery bit was actually set - it may not be
336 * if a node is still marked as needing recovery */
337 int ocfs2_recovery_map_set(struct ocfs2_super
*osb
,
342 spin_lock(&osb
->node_map_lock
);
344 __ocfs2_node_map_clear_bit(&osb
->mounted_map
, num
);
346 if (!test_bit(num
, osb
->recovery_map
.map
)) {
347 __ocfs2_node_map_set_bit(&osb
->recovery_map
, num
);
351 spin_unlock(&osb
->node_map_lock
);
356 void ocfs2_recovery_map_clear(struct ocfs2_super
*osb
,
359 ocfs2_node_map_clear_bit(osb
, &osb
->recovery_map
, num
);
362 int ocfs2_node_map_iterate(struct ocfs2_super
*osb
,
363 struct ocfs2_node_map
*map
,
368 idx
= O2NM_INVALID_NODE_NUM
;
369 spin_lock(&osb
->node_map_lock
);
370 if ((i
!= O2NM_INVALID_NODE_NUM
) &&
372 (i
< map
->num_nodes
)) {
373 while(i
< map
->num_nodes
) {
374 if (test_bit(i
, map
->map
)) {
381 spin_unlock(&osb
->node_map_lock
);