Commit | Line | Data |
---|---|---|
ccd979bd MF |
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 | ||
54 | static inline void __ocfs2_node_map_set_bit(struct ocfs2_node_map *map, | |
55 | int bit); | |
56 | static inline void __ocfs2_node_map_clear_bit(struct ocfs2_node_map *map, | |
57 | int bit); | |
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); | |
63 | ||
64 | void 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); | |
b4df6ed8 | 70 | ocfs2_node_map_init(&osb->osb_recovering_orphan_dirs); |
ccd979bd MF |
71 | } |
72 | ||
73 | static void ocfs2_do_node_down(int node_num, | |
74 | struct ocfs2_super *osb) | |
75 | { | |
76 | BUG_ON(osb->node_num == node_num); | |
77 | ||
78 | mlog(0, "ocfs2: node down event for %d\n", node_num); | |
79 | ||
80 | if (!osb->dlm) { | |
81 | /* | |
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 | |
85 | * here. | |
86 | */ | |
87 | return; | |
88 | } | |
89 | ||
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); | |
95 | return; | |
96 | } | |
97 | ||
98 | ocfs2_recovery_thread(osb, node_num); | |
99 | ||
100 | ocfs2_remove_node_from_vote_queues(osb, node_num); | |
101 | } | |
102 | ||
103 | static void ocfs2_hb_node_down_cb(struct o2nm_node *node, | |
104 | int node_num, | |
105 | void *data) | |
106 | { | |
107 | ocfs2_do_node_down(node_num, (struct ocfs2_super *) data); | |
108 | } | |
109 | ||
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, | |
113 | void *data) | |
114 | { | |
115 | struct ocfs2_super *osb = (struct ocfs2_super *) data; | |
116 | struct super_block *sb = osb->sb; | |
117 | ||
118 | mlog(ML_NOTICE, "device (%u,%u): dlm has evicted node %d\n", | |
119 | MAJOR(sb->s_dev), MINOR(sb->s_dev), node_num); | |
120 | ||
121 | ocfs2_do_node_down(node_num, osb); | |
122 | } | |
123 | ||
124 | static void ocfs2_hb_node_up_cb(struct o2nm_node *node, | |
125 | int node_num, | |
126 | void *data) | |
127 | { | |
128 | struct ocfs2_super *osb = data; | |
129 | ||
130 | BUG_ON(osb->node_num == node_num); | |
131 | ||
132 | mlog(0, "node up event for %d\n", node_num); | |
133 | ocfs2_node_map_clear_bit(osb, &osb->umount_map, node_num); | |
134 | } | |
135 | ||
136 | void ocfs2_setup_hb_callbacks(struct ocfs2_super *osb) | |
137 | { | |
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); | |
141 | ||
142 | o2hb_setup_callback(&osb->osb_hb_up, O2HB_NODE_UP_CB, | |
143 | ocfs2_hb_node_up_cb, osb, OCFS2_HB_NODE_UP_PRI); | |
144 | ||
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, | |
149 | osb); | |
150 | } | |
151 | ||
152 | /* Most functions here are just stubs for now... */ | |
153 | int ocfs2_register_hb_callbacks(struct ocfs2_super *osb) | |
154 | { | |
155 | int status; | |
156 | ||
c271c5c2 SM |
157 | if (ocfs2_mount_local(osb)) |
158 | return 0; | |
159 | ||
14829422 | 160 | status = o2hb_register_callback(osb->uuid_str, &osb->osb_hb_down); |
ccd979bd MF |
161 | if (status < 0) { |
162 | mlog_errno(status); | |
163 | goto bail; | |
164 | } | |
165 | ||
14829422 | 166 | status = o2hb_register_callback(osb->uuid_str, &osb->osb_hb_up); |
c24f72cc | 167 | if (status < 0) { |
ccd979bd | 168 | mlog_errno(status); |
14829422 | 169 | o2hb_unregister_callback(osb->uuid_str, &osb->osb_hb_down); |
c24f72cc | 170 | } |
ccd979bd MF |
171 | |
172 | bail: | |
173 | return status; | |
174 | } | |
175 | ||
176 | void ocfs2_clear_hb_callbacks(struct ocfs2_super *osb) | |
177 | { | |
c271c5c2 SM |
178 | if (ocfs2_mount_local(osb)) |
179 | return; | |
180 | ||
14829422 JB |
181 | o2hb_unregister_callback(osb->uuid_str, &osb->osb_hb_down); |
182 | o2hb_unregister_callback(osb->uuid_str, &osb->osb_hb_up); | |
ccd979bd MF |
183 | } |
184 | ||
185 | void ocfs2_stop_heartbeat(struct ocfs2_super *osb) | |
186 | { | |
187 | int ret; | |
188 | char *argv[5], *envp[3]; | |
189 | ||
c271c5c2 SM |
190 | if (ocfs2_mount_local(osb)) |
191 | return; | |
192 | ||
ccd979bd MF |
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"); | |
196 | return; | |
197 | } | |
198 | ||
199 | argv[0] = (char *)o2nm_get_hb_ctl_path(); | |
200 | argv[1] = "-K"; | |
201 | argv[2] = "-u"; | |
202 | argv[3] = osb->uuid_str; | |
203 | argv[4] = NULL; | |
204 | ||
205 | mlog(0, "Run: %s %s %s %s\n", argv[0], argv[1], argv[2], argv[3]); | |
206 | ||
207 | /* minimal command environment taken from cpu_run_sbin_hotplug */ | |
208 | envp[0] = "HOME=/"; | |
209 | envp[1] = "PATH=/sbin:/bin:/usr/sbin:/usr/bin"; | |
210 | envp[2] = NULL; | |
211 | ||
86313c48 | 212 | ret = call_usermodehelper(argv[0], argv, envp, UMH_WAIT_PROC); |
ccd979bd MF |
213 | if (ret < 0) |
214 | mlog_errno(ret); | |
215 | } | |
216 | ||
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) | |
220 | { | |
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)); | |
224 | } | |
225 | ||
226 | static inline void __ocfs2_node_map_set_bit(struct ocfs2_node_map *map, | |
227 | int bit) | |
228 | { | |
229 | set_bit(bit, map->map); | |
230 | } | |
231 | ||
232 | void ocfs2_node_map_set_bit(struct ocfs2_super *osb, | |
233 | struct ocfs2_node_map *map, | |
234 | int bit) | |
235 | { | |
236 | if (bit==-1) | |
237 | return; | |
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); | |
242 | } | |
243 | ||
244 | static inline void __ocfs2_node_map_clear_bit(struct ocfs2_node_map *map, | |
245 | int bit) | |
246 | { | |
247 | clear_bit(bit, map->map); | |
248 | } | |
249 | ||
250 | void ocfs2_node_map_clear_bit(struct ocfs2_super *osb, | |
251 | struct ocfs2_node_map *map, | |
252 | int bit) | |
253 | { | |
254 | if (bit==-1) | |
255 | return; | |
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); | |
260 | } | |
261 | ||
262 | int ocfs2_node_map_test_bit(struct ocfs2_super *osb, | |
263 | struct ocfs2_node_map *map, | |
264 | int bit) | |
265 | { | |
266 | int ret; | |
267 | if (bit >= map->num_nodes) { | |
268 | mlog(ML_ERROR, "bit=%d map->num_nodes=%d\n", bit, map->num_nodes); | |
269 | BUG(); | |
270 | } | |
271 | spin_lock(&osb->node_map_lock); | |
272 | ret = test_bit(bit, map->map); | |
273 | spin_unlock(&osb->node_map_lock); | |
274 | return ret; | |
275 | } | |
276 | ||
277 | static inline int __ocfs2_node_map_is_empty(struct ocfs2_node_map *map) | |
278 | { | |
279 | int bit; | |
280 | bit = find_next_bit(map->map, map->num_nodes, 0); | |
281 | if (bit < map->num_nodes) | |
282 | return 0; | |
283 | return 1; | |
284 | } | |
285 | ||
286 | int ocfs2_node_map_is_empty(struct ocfs2_super *osb, | |
287 | struct ocfs2_node_map *map) | |
288 | { | |
289 | int ret; | |
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); | |
294 | return ret; | |
295 | } | |
296 | ||
297 | static void __ocfs2_node_map_dup(struct ocfs2_node_map *target, | |
298 | struct ocfs2_node_map *from) | |
299 | { | |
300 | BUG_ON(from->num_nodes == 0); | |
301 | ocfs2_node_map_init(target); | |
302 | __ocfs2_node_map_set(target, from); | |
303 | } | |
304 | ||
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, | |
308 | int bit) | |
309 | { | |
310 | struct ocfs2_node_map temp; | |
311 | int ret; | |
312 | ||
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); | |
318 | ||
319 | return ret; | |
320 | } | |
321 | ||
322 | static void __ocfs2_node_map_set(struct ocfs2_node_map *target, | |
323 | struct ocfs2_node_map *from) | |
324 | { | |
325 | int num_longs, i; | |
326 | ||
327 | BUG_ON(target->num_nodes != from->num_nodes); | |
328 | BUG_ON(target->num_nodes == 0); | |
329 | ||
330 | num_longs = BITS_TO_LONGS(target->num_nodes); | |
331 | for (i = 0; i < num_longs; i++) | |
332 | target->map[i] = from->map[i]; | |
333 | } | |
334 | ||
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, | |
338 | int num) | |
339 | { | |
340 | int set = 0; | |
341 | ||
342 | spin_lock(&osb->node_map_lock); | |
343 | ||
344 | __ocfs2_node_map_clear_bit(&osb->mounted_map, num); | |
345 | ||
346 | if (!test_bit(num, osb->recovery_map.map)) { | |
347 | __ocfs2_node_map_set_bit(&osb->recovery_map, num); | |
348 | set = 1; | |
349 | } | |
350 | ||
351 | spin_unlock(&osb->node_map_lock); | |
352 | ||
353 | return set; | |
354 | } | |
355 | ||
356 | void ocfs2_recovery_map_clear(struct ocfs2_super *osb, | |
357 | int num) | |
358 | { | |
359 | ocfs2_node_map_clear_bit(osb, &osb->recovery_map, num); | |
360 | } | |
361 | ||
362 | int ocfs2_node_map_iterate(struct ocfs2_super *osb, | |
363 | struct ocfs2_node_map *map, | |
364 | int idx) | |
365 | { | |
366 | int i = idx; | |
367 | ||
368 | idx = O2NM_INVALID_NODE_NUM; | |
369 | spin_lock(&osb->node_map_lock); | |
370 | if ((i != O2NM_INVALID_NODE_NUM) && | |
371 | (i >= 0) && | |
372 | (i < map->num_nodes)) { | |
373 | while(i < map->num_nodes) { | |
374 | if (test_bit(i, map->map)) { | |
375 | idx = i; | |
376 | break; | |
377 | } | |
378 | i++; | |
379 | } | |
380 | } | |
381 | spin_unlock(&osb->node_map_lock); | |
382 | return idx; | |
383 | } |