Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 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); |
Mark Fasheh | b4df6ed | 2006-02-22 17:35:08 -0800 | [diff] [blame] | 70 | ocfs2_node_map_init(&osb->osb_recovering_orphan_dirs); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 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 | |
| 157 | status = o2hb_register_callback(&osb->osb_hb_down); |
| 158 | if (status < 0) { |
| 159 | mlog_errno(status); |
| 160 | goto bail; |
| 161 | } |
| 162 | |
| 163 | status = o2hb_register_callback(&osb->osb_hb_up); |
| 164 | if (status < 0) |
| 165 | mlog_errno(status); |
| 166 | |
| 167 | bail: |
| 168 | return status; |
| 169 | } |
| 170 | |
| 171 | void ocfs2_clear_hb_callbacks(struct ocfs2_super *osb) |
| 172 | { |
| 173 | int status; |
| 174 | |
| 175 | status = o2hb_unregister_callback(&osb->osb_hb_down); |
| 176 | if (status < 0) |
| 177 | mlog_errno(status); |
| 178 | |
| 179 | status = o2hb_unregister_callback(&osb->osb_hb_up); |
| 180 | if (status < 0) |
| 181 | mlog_errno(status); |
| 182 | } |
| 183 | |
| 184 | void ocfs2_stop_heartbeat(struct ocfs2_super *osb) |
| 185 | { |
| 186 | int ret; |
| 187 | char *argv[5], *envp[3]; |
| 188 | |
| 189 | if (!osb->uuid_str) { |
| 190 | /* This can happen if we don't get far enough in mount... */ |
| 191 | mlog(0, "No UUID with which to stop heartbeat!\n\n"); |
| 192 | return; |
| 193 | } |
| 194 | |
| 195 | argv[0] = (char *)o2nm_get_hb_ctl_path(); |
| 196 | argv[1] = "-K"; |
| 197 | argv[2] = "-u"; |
| 198 | argv[3] = osb->uuid_str; |
| 199 | argv[4] = NULL; |
| 200 | |
| 201 | mlog(0, "Run: %s %s %s %s\n", argv[0], argv[1], argv[2], argv[3]); |
| 202 | |
| 203 | /* minimal command environment taken from cpu_run_sbin_hotplug */ |
| 204 | envp[0] = "HOME=/"; |
| 205 | envp[1] = "PATH=/sbin:/bin:/usr/sbin:/usr/bin"; |
| 206 | envp[2] = NULL; |
| 207 | |
| 208 | ret = call_usermodehelper(argv[0], argv, envp, 1); |
| 209 | if (ret < 0) |
| 210 | mlog_errno(ret); |
| 211 | } |
| 212 | |
| 213 | /* special case -1 for now |
| 214 | * TODO: should *really* make sure the calling func never passes -1!! */ |
| 215 | void ocfs2_node_map_init(struct ocfs2_node_map *map) |
| 216 | { |
| 217 | map->num_nodes = OCFS2_NODE_MAP_MAX_NODES; |
| 218 | memset(map->map, 0, BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES) * |
| 219 | sizeof(unsigned long)); |
| 220 | } |
| 221 | |
| 222 | static inline void __ocfs2_node_map_set_bit(struct ocfs2_node_map *map, |
| 223 | int bit) |
| 224 | { |
| 225 | set_bit(bit, map->map); |
| 226 | } |
| 227 | |
| 228 | void ocfs2_node_map_set_bit(struct ocfs2_super *osb, |
| 229 | struct ocfs2_node_map *map, |
| 230 | int bit) |
| 231 | { |
| 232 | if (bit==-1) |
| 233 | return; |
| 234 | BUG_ON(bit >= map->num_nodes); |
| 235 | spin_lock(&osb->node_map_lock); |
| 236 | __ocfs2_node_map_set_bit(map, bit); |
| 237 | spin_unlock(&osb->node_map_lock); |
| 238 | } |
| 239 | |
| 240 | static inline void __ocfs2_node_map_clear_bit(struct ocfs2_node_map *map, |
| 241 | int bit) |
| 242 | { |
| 243 | clear_bit(bit, map->map); |
| 244 | } |
| 245 | |
| 246 | void ocfs2_node_map_clear_bit(struct ocfs2_super *osb, |
| 247 | struct ocfs2_node_map *map, |
| 248 | int bit) |
| 249 | { |
| 250 | if (bit==-1) |
| 251 | return; |
| 252 | BUG_ON(bit >= map->num_nodes); |
| 253 | spin_lock(&osb->node_map_lock); |
| 254 | __ocfs2_node_map_clear_bit(map, bit); |
| 255 | spin_unlock(&osb->node_map_lock); |
| 256 | } |
| 257 | |
| 258 | int ocfs2_node_map_test_bit(struct ocfs2_super *osb, |
| 259 | struct ocfs2_node_map *map, |
| 260 | int bit) |
| 261 | { |
| 262 | int ret; |
| 263 | if (bit >= map->num_nodes) { |
| 264 | mlog(ML_ERROR, "bit=%d map->num_nodes=%d\n", bit, map->num_nodes); |
| 265 | BUG(); |
| 266 | } |
| 267 | spin_lock(&osb->node_map_lock); |
| 268 | ret = test_bit(bit, map->map); |
| 269 | spin_unlock(&osb->node_map_lock); |
| 270 | return ret; |
| 271 | } |
| 272 | |
| 273 | static inline int __ocfs2_node_map_is_empty(struct ocfs2_node_map *map) |
| 274 | { |
| 275 | int bit; |
| 276 | bit = find_next_bit(map->map, map->num_nodes, 0); |
| 277 | if (bit < map->num_nodes) |
| 278 | return 0; |
| 279 | return 1; |
| 280 | } |
| 281 | |
| 282 | int ocfs2_node_map_is_empty(struct ocfs2_super *osb, |
| 283 | struct ocfs2_node_map *map) |
| 284 | { |
| 285 | int ret; |
| 286 | BUG_ON(map->num_nodes == 0); |
| 287 | spin_lock(&osb->node_map_lock); |
| 288 | ret = __ocfs2_node_map_is_empty(map); |
| 289 | spin_unlock(&osb->node_map_lock); |
| 290 | return ret; |
| 291 | } |
| 292 | |
| 293 | static void __ocfs2_node_map_dup(struct ocfs2_node_map *target, |
| 294 | struct ocfs2_node_map *from) |
| 295 | { |
| 296 | BUG_ON(from->num_nodes == 0); |
| 297 | ocfs2_node_map_init(target); |
| 298 | __ocfs2_node_map_set(target, from); |
| 299 | } |
| 300 | |
| 301 | /* returns 1 if bit is the only bit set in target, 0 otherwise */ |
| 302 | int ocfs2_node_map_is_only(struct ocfs2_super *osb, |
| 303 | struct ocfs2_node_map *target, |
| 304 | int bit) |
| 305 | { |
| 306 | struct ocfs2_node_map temp; |
| 307 | int ret; |
| 308 | |
| 309 | spin_lock(&osb->node_map_lock); |
| 310 | __ocfs2_node_map_dup(&temp, target); |
| 311 | __ocfs2_node_map_clear_bit(&temp, bit); |
| 312 | ret = __ocfs2_node_map_is_empty(&temp); |
| 313 | spin_unlock(&osb->node_map_lock); |
| 314 | |
| 315 | return ret; |
| 316 | } |
| 317 | |
| 318 | static void __ocfs2_node_map_set(struct ocfs2_node_map *target, |
| 319 | struct ocfs2_node_map *from) |
| 320 | { |
| 321 | int num_longs, i; |
| 322 | |
| 323 | BUG_ON(target->num_nodes != from->num_nodes); |
| 324 | BUG_ON(target->num_nodes == 0); |
| 325 | |
| 326 | num_longs = BITS_TO_LONGS(target->num_nodes); |
| 327 | for (i = 0; i < num_longs; i++) |
| 328 | target->map[i] = from->map[i]; |
| 329 | } |
| 330 | |
| 331 | /* Returns whether the recovery bit was actually set - it may not be |
| 332 | * if a node is still marked as needing recovery */ |
| 333 | int ocfs2_recovery_map_set(struct ocfs2_super *osb, |
| 334 | int num) |
| 335 | { |
| 336 | int set = 0; |
| 337 | |
| 338 | spin_lock(&osb->node_map_lock); |
| 339 | |
| 340 | __ocfs2_node_map_clear_bit(&osb->mounted_map, num); |
| 341 | |
| 342 | if (!test_bit(num, osb->recovery_map.map)) { |
| 343 | __ocfs2_node_map_set_bit(&osb->recovery_map, num); |
| 344 | set = 1; |
| 345 | } |
| 346 | |
| 347 | spin_unlock(&osb->node_map_lock); |
| 348 | |
| 349 | return set; |
| 350 | } |
| 351 | |
| 352 | void ocfs2_recovery_map_clear(struct ocfs2_super *osb, |
| 353 | int num) |
| 354 | { |
| 355 | ocfs2_node_map_clear_bit(osb, &osb->recovery_map, num); |
| 356 | } |
| 357 | |
| 358 | int ocfs2_node_map_iterate(struct ocfs2_super *osb, |
| 359 | struct ocfs2_node_map *map, |
| 360 | int idx) |
| 361 | { |
| 362 | int i = idx; |
| 363 | |
| 364 | idx = O2NM_INVALID_NODE_NUM; |
| 365 | spin_lock(&osb->node_map_lock); |
| 366 | if ((i != O2NM_INVALID_NODE_NUM) && |
| 367 | (i >= 0) && |
| 368 | (i < map->num_nodes)) { |
| 369 | while(i < map->num_nodes) { |
| 370 | if (test_bit(i, map->map)) { |
| 371 | idx = i; |
| 372 | break; |
| 373 | } |
| 374 | i++; |
| 375 | } |
| 376 | } |
| 377 | spin_unlock(&osb->node_map_lock); |
| 378 | return idx; |
| 379 | } |