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 <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" 44 45 #include "buffer_head_io.h" 46 47 static inline void __ocfs2_node_map_set_bit(struct ocfs2_node_map *map, 48 int bit); 49 static inline void __ocfs2_node_map_clear_bit(struct ocfs2_node_map *map, 50 int bit); 51 static inline int __ocfs2_node_map_is_empty(struct ocfs2_node_map *map); 52 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) 56 { 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)); 60 } 61 62 void ocfs2_init_node_maps(struct ocfs2_super *osb) 63 { 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); 67 } 68 69 static void ocfs2_do_node_down(int node_num, 70 struct ocfs2_super *osb) 71 { 72 BUG_ON(osb->node_num == node_num); 73 74 mlog(0, "ocfs2: node down event for %d\n", node_num); 75 76 if (!osb->dlm) { 77 /* 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 81 * here. 82 */ 83 return; 84 } 85 86 ocfs2_recovery_thread(osb, node_num); 87 } 88 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, 92 void *data) 93 { 94 struct ocfs2_super *osb = (struct ocfs2_super *) data; 95 struct super_block *sb = osb->sb; 96 97 mlog(ML_NOTICE, "device (%u,%u): dlm has evicted node %d\n", 98 MAJOR(sb->s_dev), MINOR(sb->s_dev), node_num); 99 100 ocfs2_do_node_down(node_num, osb); 101 } 102 103 void ocfs2_setup_hb_callbacks(struct ocfs2_super *osb) 104 { 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, 109 osb); 110 } 111 112 void ocfs2_stop_heartbeat(struct ocfs2_super *osb) 113 { 114 int ret; 115 char *argv[5], *envp[3]; 116 117 if (ocfs2_mount_local(osb)) 118 return; 119 120 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"); 123 return; 124 } 125 126 argv[0] = (char *)o2nm_get_hb_ctl_path(); 127 argv[1] = "-K"; 128 argv[2] = "-u"; 129 argv[3] = osb->uuid_str; 130 argv[4] = NULL; 131 132 mlog(0, "Run: %s %s %s %s\n", argv[0], argv[1], argv[2], argv[3]); 133 134 /* minimal command environment taken from cpu_run_sbin_hotplug */ 135 envp[0] = "HOME=/"; 136 envp[1] = "PATH=/sbin:/bin:/usr/sbin:/usr/bin"; 137 envp[2] = NULL; 138 139 ret = call_usermodehelper(argv[0], argv, envp, UMH_WAIT_PROC); 140 if (ret < 0) 141 mlog_errno(ret); 142 } 143 144 static inline void __ocfs2_node_map_set_bit(struct ocfs2_node_map *map, 145 int bit) 146 { 147 set_bit(bit, map->map); 148 } 149 150 void ocfs2_node_map_set_bit(struct ocfs2_super *osb, 151 struct ocfs2_node_map *map, 152 int bit) 153 { 154 if (bit==-1) 155 return; 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); 160 } 161 162 static inline void __ocfs2_node_map_clear_bit(struct ocfs2_node_map *map, 163 int bit) 164 { 165 clear_bit(bit, map->map); 166 } 167 168 void ocfs2_node_map_clear_bit(struct ocfs2_super *osb, 169 struct ocfs2_node_map *map, 170 int bit) 171 { 172 if (bit==-1) 173 return; 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); 178 } 179 180 int ocfs2_node_map_test_bit(struct ocfs2_super *osb, 181 struct ocfs2_node_map *map, 182 int bit) 183 { 184 int ret; 185 if (bit >= map->num_nodes) { 186 mlog(ML_ERROR, "bit=%d map->num_nodes=%d\n", bit, map->num_nodes); 187 BUG(); 188 } 189 spin_lock(&osb->node_map_lock); 190 ret = test_bit(bit, map->map); 191 spin_unlock(&osb->node_map_lock); 192 return ret; 193 } 194 195 static inline int __ocfs2_node_map_is_empty(struct ocfs2_node_map *map) 196 { 197 int bit; 198 bit = find_next_bit(map->map, map->num_nodes, 0); 199 if (bit < map->num_nodes) 200 return 0; 201 return 1; 202 } 203 204 int ocfs2_node_map_is_empty(struct ocfs2_super *osb, 205 struct ocfs2_node_map *map) 206 { 207 int ret; 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); 212 return ret; 213 } 214 215 #if 0 216 217 static void __ocfs2_node_map_dup(struct ocfs2_node_map *target, 218 struct ocfs2_node_map *from) 219 { 220 BUG_ON(from->num_nodes == 0); 221 ocfs2_node_map_init(target); 222 __ocfs2_node_map_set(target, from); 223 } 224 225 /* 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, 228 int bit) 229 { 230 struct ocfs2_node_map temp; 231 int ret; 232 233 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); 238 239 return ret; 240 } 241 242 static void __ocfs2_node_map_set(struct ocfs2_node_map *target, 243 struct ocfs2_node_map *from) 244 { 245 int num_longs, i; 246 247 BUG_ON(target->num_nodes != from->num_nodes); 248 BUG_ON(target->num_nodes == 0); 249 250 num_longs = BITS_TO_LONGS(target->num_nodes); 251 for (i = 0; i < num_longs; i++) 252 target->map[i] = from->map[i]; 253 } 254 255 #endif /* 0 */ 256 257 /* 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, 260 int num) 261 { 262 int set = 0; 263 264 spin_lock(&osb->node_map_lock); 265 266 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 } 275 276 void ocfs2_recovery_map_clear(struct ocfs2_super *osb, 277 int num) 278 { 279 ocfs2_node_map_clear_bit(osb, &osb->recovery_map, num); 280 } 281 282 int 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 } 304