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 | * ocfs2.h |
| 5 | * |
| 6 | * Defines macros and structures used in OCFS2 |
| 7 | * |
| 8 | * Copyright (C) 2002, 2004 Oracle. All rights reserved. |
| 9 | * |
| 10 | * This program is free software; you can redistribute it and/or |
| 11 | * modify it under the terms of the GNU General Public |
| 12 | * License as published by the Free Software Foundation; either |
| 13 | * version 2 of the License, or (at your option) any later version. |
| 14 | * |
| 15 | * This program is distributed in the hope that it will be useful, |
| 16 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 17 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 18 | * General Public License for more details. |
| 19 | * |
| 20 | * You should have received a copy of the GNU General Public |
| 21 | * License along with this program; if not, write to the |
| 22 | * Free Software Foundation, Inc., 59 Temple Place - Suite 330, |
| 23 | * Boston, MA 021110-1307, USA. |
| 24 | */ |
| 25 | |
| 26 | #ifndef OCFS2_H |
| 27 | #define OCFS2_H |
| 28 | |
| 29 | #include <linux/spinlock.h> |
| 30 | #include <linux/sched.h> |
| 31 | #include <linux/wait.h> |
| 32 | #include <linux/list.h> |
| 33 | #include <linux/rbtree.h> |
| 34 | #include <linux/workqueue.h> |
| 35 | #include <linux/kref.h> |
Arjan van de Ven | c74ec2f | 2006-01-13 21:54:23 -0800 | [diff] [blame] | 36 | #include <linux/mutex.h> |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 37 | |
| 38 | #include "cluster/nodemanager.h" |
| 39 | #include "cluster/heartbeat.h" |
| 40 | #include "cluster/tcp.h" |
| 41 | |
| 42 | #include "dlm/dlmapi.h" |
| 43 | |
| 44 | #include "ocfs2_fs.h" |
| 45 | #include "endian.h" |
| 46 | #include "ocfs2_lockid.h" |
| 47 | |
| 48 | struct ocfs2_extent_map { |
| 49 | u32 em_clusters; |
| 50 | struct rb_root em_extents; |
| 51 | }; |
| 52 | |
| 53 | /* Most user visible OCFS2 inodes will have very few pieces of |
| 54 | * metadata, but larger files (including bitmaps, etc) must be taken |
| 55 | * into account when designing an access scheme. We allow a small |
| 56 | * amount of inlined blocks to be stored on an array and grow the |
| 57 | * structure into a rb tree when necessary. */ |
| 58 | #define OCFS2_INODE_MAX_CACHE_ARRAY 2 |
| 59 | |
| 60 | struct ocfs2_caching_info { |
| 61 | unsigned int ci_num_cached; |
| 62 | union { |
| 63 | sector_t ci_array[OCFS2_INODE_MAX_CACHE_ARRAY]; |
| 64 | struct rb_root ci_tree; |
| 65 | } ci_cache; |
| 66 | }; |
| 67 | |
| 68 | /* this limits us to 256 nodes |
| 69 | * if we need more, we can do a kmalloc for the map */ |
| 70 | #define OCFS2_NODE_MAP_MAX_NODES 256 |
| 71 | struct ocfs2_node_map { |
| 72 | u16 num_nodes; |
| 73 | unsigned long map[BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES)]; |
| 74 | }; |
| 75 | |
| 76 | enum ocfs2_ast_action { |
| 77 | OCFS2_AST_INVALID = 0, |
| 78 | OCFS2_AST_ATTACH, |
| 79 | OCFS2_AST_CONVERT, |
| 80 | OCFS2_AST_DOWNCONVERT, |
| 81 | }; |
| 82 | |
| 83 | /* actions for an unlockast function to take. */ |
| 84 | enum ocfs2_unlock_action { |
| 85 | OCFS2_UNLOCK_INVALID = 0, |
| 86 | OCFS2_UNLOCK_CANCEL_CONVERT, |
| 87 | OCFS2_UNLOCK_DROP_LOCK, |
| 88 | }; |
| 89 | |
| 90 | /* ocfs2_lock_res->l_flags flags. */ |
| 91 | #define OCFS2_LOCK_ATTACHED (0x00000001) /* have we initialized |
| 92 | * the lvb */ |
| 93 | #define OCFS2_LOCK_BUSY (0x00000002) /* we are currently in |
| 94 | * dlm_lock */ |
| 95 | #define OCFS2_LOCK_BLOCKED (0x00000004) /* blocked waiting to |
| 96 | * downconvert*/ |
| 97 | #define OCFS2_LOCK_LOCAL (0x00000008) /* newly created inode */ |
| 98 | #define OCFS2_LOCK_NEEDS_REFRESH (0x00000010) |
| 99 | #define OCFS2_LOCK_REFRESHING (0x00000020) |
| 100 | #define OCFS2_LOCK_INITIALIZED (0x00000040) /* track initialization |
| 101 | * for shutdown paths */ |
| 102 | #define OCFS2_LOCK_FREEING (0x00000080) /* help dlmglue track |
| 103 | * when to skip queueing |
| 104 | * a lock because it's |
| 105 | * about to be |
| 106 | * dropped. */ |
| 107 | #define OCFS2_LOCK_QUEUED (0x00000100) /* queued for downconvert */ |
| 108 | |
| 109 | struct ocfs2_lock_res_ops; |
| 110 | |
| 111 | typedef void (*ocfs2_lock_callback)(int status, unsigned long data); |
| 112 | |
| 113 | struct ocfs2_lock_res { |
| 114 | void *l_priv; |
| 115 | struct ocfs2_lock_res_ops *l_ops; |
| 116 | spinlock_t l_lock; |
| 117 | |
| 118 | struct list_head l_blocked_list; |
| 119 | struct list_head l_mask_waiters; |
| 120 | |
| 121 | enum ocfs2_lock_type l_type; |
| 122 | unsigned long l_flags; |
| 123 | char l_name[OCFS2_LOCK_ID_MAX_LEN]; |
| 124 | int l_level; |
| 125 | unsigned int l_ro_holders; |
| 126 | unsigned int l_ex_holders; |
| 127 | struct dlm_lockstatus l_lksb; |
| 128 | |
| 129 | /* used from AST/BAST funcs. */ |
| 130 | enum ocfs2_ast_action l_action; |
| 131 | enum ocfs2_unlock_action l_unlock_action; |
| 132 | int l_requested; |
| 133 | int l_blocking; |
| 134 | |
| 135 | wait_queue_head_t l_event; |
| 136 | |
| 137 | struct list_head l_debug_list; |
| 138 | }; |
| 139 | |
| 140 | struct ocfs2_dlm_debug { |
| 141 | struct kref d_refcnt; |
| 142 | struct dentry *d_locking_state; |
| 143 | struct list_head d_lockres_tracking; |
| 144 | }; |
| 145 | |
| 146 | enum ocfs2_vol_state |
| 147 | { |
| 148 | VOLUME_INIT = 0, |
| 149 | VOLUME_MOUNTED, |
| 150 | VOLUME_DISMOUNTED, |
| 151 | VOLUME_DISABLED |
| 152 | }; |
| 153 | |
| 154 | struct ocfs2_alloc_stats |
| 155 | { |
| 156 | atomic_t moves; |
| 157 | atomic_t local_data; |
| 158 | atomic_t bitmap_data; |
| 159 | atomic_t bg_allocs; |
| 160 | atomic_t bg_extends; |
| 161 | }; |
| 162 | |
| 163 | enum ocfs2_local_alloc_state |
| 164 | { |
| 165 | OCFS2_LA_UNUSED = 0, |
| 166 | OCFS2_LA_ENABLED, |
| 167 | OCFS2_LA_DISABLED |
| 168 | }; |
| 169 | |
| 170 | enum ocfs2_mount_options |
| 171 | { |
| 172 | OCFS2_MOUNT_HB_LOCAL = 1 << 0, /* Heartbeat started in local mode */ |
| 173 | OCFS2_MOUNT_BARRIER = 1 << 1, /* Use block barriers */ |
| 174 | OCFS2_MOUNT_NOINTR = 1 << 2, /* Don't catch signals */ |
| 175 | OCFS2_MOUNT_ERRORS_PANIC = 1 << 3, /* Panic on errors */ |
| 176 | OCFS2_MOUNT_DATA_WRITEBACK = 1 << 4, /* No data ordering */ |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 177 | }; |
| 178 | |
| 179 | #define OCFS2_OSB_SOFT_RO 0x0001 |
| 180 | #define OCFS2_OSB_HARD_RO 0x0002 |
| 181 | #define OCFS2_OSB_ERROR_FS 0x0004 |
| 182 | |
| 183 | struct ocfs2_journal; |
| 184 | struct ocfs2_journal_handle; |
| 185 | struct ocfs2_super |
| 186 | { |
| 187 | u32 osb_id; /* id used by the proc interface */ |
| 188 | struct task_struct *commit_task; |
| 189 | struct super_block *sb; |
| 190 | struct inode *root_inode; |
| 191 | struct inode *sys_root_inode; |
| 192 | struct inode *system_inodes[NUM_SYSTEM_INODES]; |
| 193 | |
| 194 | struct ocfs2_slot_info *slot_info; |
| 195 | |
| 196 | spinlock_t node_map_lock; |
| 197 | struct ocfs2_node_map mounted_map; |
| 198 | struct ocfs2_node_map recovery_map; |
| 199 | struct ocfs2_node_map umount_map; |
| 200 | |
| 201 | u32 num_clusters; |
| 202 | u64 root_blkno; |
| 203 | u64 system_dir_blkno; |
| 204 | u64 bitmap_blkno; |
| 205 | u32 bitmap_cpg; |
| 206 | u8 *uuid; |
| 207 | char *uuid_str; |
| 208 | u8 *vol_label; |
| 209 | u64 first_cluster_group_blkno; |
| 210 | u32 fs_generation; |
| 211 | |
| 212 | u32 s_feature_compat; |
| 213 | u32 s_feature_incompat; |
| 214 | u32 s_feature_ro_compat; |
| 215 | |
| 216 | /* Protects s_next_generaion, osb_flags. Could protect more on |
| 217 | * osb as it's very short lived. */ |
| 218 | spinlock_t osb_lock; |
| 219 | u32 s_next_generation; |
| 220 | unsigned long osb_flags; |
| 221 | |
| 222 | unsigned long s_mount_opt; |
| 223 | |
| 224 | u16 max_slots; |
| 225 | u16 num_nodes; |
| 226 | s16 node_num; |
| 227 | s16 slot_num; |
| 228 | int s_sectsize_bits; |
| 229 | int s_clustersize; |
| 230 | int s_clustersize_bits; |
| 231 | struct proc_dir_entry *proc_sub_dir; /* points to /proc/fs/ocfs2/<maj_min> */ |
| 232 | |
| 233 | atomic_t vol_state; |
Arjan van de Ven | c74ec2f | 2006-01-13 21:54:23 -0800 | [diff] [blame] | 234 | struct mutex recovery_lock; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 235 | struct task_struct *recovery_thread_task; |
| 236 | int disable_recovery; |
| 237 | wait_queue_head_t checkpoint_event; |
| 238 | atomic_t needs_checkpoint; |
| 239 | struct ocfs2_journal *journal; |
| 240 | |
| 241 | enum ocfs2_local_alloc_state local_alloc_state; |
| 242 | struct buffer_head *local_alloc_bh; |
| 243 | |
| 244 | /* Next two fields are for local node slot recovery during |
| 245 | * mount. */ |
| 246 | int dirty; |
| 247 | struct ocfs2_dinode *local_alloc_copy; |
| 248 | |
| 249 | struct ocfs2_alloc_stats alloc_stats; |
| 250 | char dev_str[20]; /* "major,minor" of the device */ |
| 251 | |
| 252 | struct dlm_ctxt *dlm; |
| 253 | struct ocfs2_lock_res osb_super_lockres; |
| 254 | struct ocfs2_lock_res osb_rename_lockres; |
| 255 | struct dlm_eviction_cb osb_eviction_cb; |
| 256 | struct ocfs2_dlm_debug *osb_dlm_debug; |
| 257 | |
| 258 | struct dentry *osb_debug_root; |
| 259 | |
| 260 | wait_queue_head_t recovery_event; |
| 261 | |
| 262 | spinlock_t vote_task_lock; |
| 263 | struct task_struct *vote_task; |
| 264 | wait_queue_head_t vote_event; |
| 265 | unsigned long vote_wake_sequence; |
| 266 | unsigned long vote_work_sequence; |
| 267 | |
| 268 | struct list_head blocked_lock_list; |
| 269 | unsigned long blocked_lock_count; |
| 270 | |
| 271 | struct list_head vote_list; |
| 272 | int vote_count; |
| 273 | |
| 274 | u32 net_key; |
| 275 | spinlock_t net_response_lock; |
| 276 | unsigned int net_response_ids; |
| 277 | struct list_head net_response_list; |
| 278 | |
| 279 | struct o2hb_callback_func osb_hb_up; |
| 280 | struct o2hb_callback_func osb_hb_down; |
| 281 | |
| 282 | struct list_head osb_net_handlers; |
| 283 | |
| 284 | wait_queue_head_t osb_mount_event; |
| 285 | |
| 286 | /* Truncate log info */ |
| 287 | struct inode *osb_tl_inode; |
| 288 | struct buffer_head *osb_tl_bh; |
| 289 | struct work_struct osb_truncate_log_wq; |
Mark Fasheh | b4df6ed | 2006-02-22 17:35:08 -0800 | [diff] [blame] | 290 | |
| 291 | struct ocfs2_node_map osb_recovering_orphan_dirs; |
| 292 | unsigned int *osb_orphan_wipes; |
| 293 | wait_queue_head_t osb_wipe_event; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 294 | }; |
| 295 | |
| 296 | #define OCFS2_SB(sb) ((struct ocfs2_super *)(sb)->s_fs_info) |
| 297 | #define OCFS2_MAX_OSB_ID 65536 |
| 298 | |
| 299 | static inline int ocfs2_should_order_data(struct inode *inode) |
| 300 | { |
| 301 | if (!S_ISREG(inode->i_mode)) |
| 302 | return 0; |
| 303 | if (OCFS2_SB(inode->i_sb)->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) |
| 304 | return 0; |
| 305 | return 1; |
| 306 | } |
| 307 | |
| 308 | /* set / clear functions because cluster events can make these happen |
| 309 | * in parallel so we want the transitions to be atomic. this also |
| 310 | * means that any future flags osb_flags must be protected by spinlock |
| 311 | * too! */ |
| 312 | static inline void ocfs2_set_osb_flag(struct ocfs2_super *osb, |
| 313 | unsigned long flag) |
| 314 | { |
| 315 | spin_lock(&osb->osb_lock); |
| 316 | osb->osb_flags |= flag; |
| 317 | spin_unlock(&osb->osb_lock); |
| 318 | } |
| 319 | |
| 320 | static inline void ocfs2_set_ro_flag(struct ocfs2_super *osb, |
| 321 | int hard) |
| 322 | { |
| 323 | spin_lock(&osb->osb_lock); |
| 324 | osb->osb_flags &= ~(OCFS2_OSB_SOFT_RO|OCFS2_OSB_HARD_RO); |
| 325 | if (hard) |
| 326 | osb->osb_flags |= OCFS2_OSB_HARD_RO; |
| 327 | else |
| 328 | osb->osb_flags |= OCFS2_OSB_SOFT_RO; |
| 329 | spin_unlock(&osb->osb_lock); |
| 330 | } |
| 331 | |
| 332 | static inline int ocfs2_is_hard_readonly(struct ocfs2_super *osb) |
| 333 | { |
| 334 | int ret; |
| 335 | |
| 336 | spin_lock(&osb->osb_lock); |
| 337 | ret = osb->osb_flags & OCFS2_OSB_HARD_RO; |
| 338 | spin_unlock(&osb->osb_lock); |
| 339 | |
| 340 | return ret; |
| 341 | } |
| 342 | |
| 343 | static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb) |
| 344 | { |
| 345 | int ret; |
| 346 | |
| 347 | spin_lock(&osb->osb_lock); |
| 348 | ret = osb->osb_flags & OCFS2_OSB_SOFT_RO; |
| 349 | spin_unlock(&osb->osb_lock); |
| 350 | |
| 351 | return ret; |
| 352 | } |
| 353 | |
| 354 | #define OCFS2_IS_VALID_DINODE(ptr) \ |
| 355 | (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE)) |
| 356 | |
| 357 | #define OCFS2_RO_ON_INVALID_DINODE(__sb, __di) do { \ |
| 358 | typeof(__di) ____di = (__di); \ |
| 359 | ocfs2_error((__sb), \ |
Mark Fasheh | b0697053 | 2006-03-03 10:24:33 -0800 | [diff] [blame] | 360 | "Dinode # %llu has bad signature %.*s", \ |
| 361 | (unsigned long long)(____di)->i_blkno, 7, \ |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 362 | (____di)->i_signature); \ |
| 363 | } while (0); |
| 364 | |
| 365 | #define OCFS2_IS_VALID_EXTENT_BLOCK(ptr) \ |
| 366 | (!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE)) |
| 367 | |
| 368 | #define OCFS2_RO_ON_INVALID_EXTENT_BLOCK(__sb, __eb) do { \ |
| 369 | typeof(__eb) ____eb = (__eb); \ |
| 370 | ocfs2_error((__sb), \ |
Mark Fasheh | b0697053 | 2006-03-03 10:24:33 -0800 | [diff] [blame] | 371 | "Extent Block # %llu has bad signature %.*s", \ |
| 372 | (unsigned long long)(____eb)->h_blkno, 7, \ |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 373 | (____eb)->h_signature); \ |
| 374 | } while (0); |
| 375 | |
| 376 | #define OCFS2_IS_VALID_GROUP_DESC(ptr) \ |
| 377 | (!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE)) |
| 378 | |
| 379 | #define OCFS2_RO_ON_INVALID_GROUP_DESC(__sb, __gd) do { \ |
| 380 | typeof(__gd) ____gd = (__gd); \ |
| 381 | ocfs2_error((__sb), \ |
Mark Fasheh | b0697053 | 2006-03-03 10:24:33 -0800 | [diff] [blame] | 382 | "Group Descriptor # %llu has bad signature %.*s", \ |
| 383 | (unsigned long long)(____gd)->bg_blkno, 7, \ |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 384 | (____gd)->bg_signature); \ |
| 385 | } while (0); |
| 386 | |
| 387 | static inline unsigned long ino_from_blkno(struct super_block *sb, |
| 388 | u64 blkno) |
| 389 | { |
| 390 | return (unsigned long)(blkno & (u64)ULONG_MAX); |
| 391 | } |
| 392 | |
| 393 | static inline u64 ocfs2_clusters_to_blocks(struct super_block *sb, |
| 394 | u32 clusters) |
| 395 | { |
| 396 | int c_to_b_bits = OCFS2_SB(sb)->s_clustersize_bits - |
| 397 | sb->s_blocksize_bits; |
| 398 | |
| 399 | return (u64)clusters << c_to_b_bits; |
| 400 | } |
| 401 | |
| 402 | static inline u32 ocfs2_blocks_to_clusters(struct super_block *sb, |
| 403 | u64 blocks) |
| 404 | { |
| 405 | int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits - |
| 406 | sb->s_blocksize_bits; |
| 407 | |
| 408 | return (u32)(blocks >> b_to_c_bits); |
| 409 | } |
| 410 | |
| 411 | static inline unsigned int ocfs2_clusters_for_bytes(struct super_block *sb, |
| 412 | u64 bytes) |
| 413 | { |
| 414 | int cl_bits = OCFS2_SB(sb)->s_clustersize_bits; |
| 415 | unsigned int clusters; |
| 416 | |
| 417 | bytes += OCFS2_SB(sb)->s_clustersize - 1; |
| 418 | /* OCFS2 just cannot have enough clusters to overflow this */ |
| 419 | clusters = (unsigned int)(bytes >> cl_bits); |
| 420 | |
| 421 | return clusters; |
| 422 | } |
| 423 | |
| 424 | static inline u64 ocfs2_blocks_for_bytes(struct super_block *sb, |
| 425 | u64 bytes) |
| 426 | { |
| 427 | bytes += sb->s_blocksize - 1; |
| 428 | return bytes >> sb->s_blocksize_bits; |
| 429 | } |
| 430 | |
| 431 | static inline u64 ocfs2_clusters_to_bytes(struct super_block *sb, |
| 432 | u32 clusters) |
| 433 | { |
| 434 | return (u64)clusters << OCFS2_SB(sb)->s_clustersize_bits; |
| 435 | } |
| 436 | |
| 437 | static inline u64 ocfs2_align_bytes_to_clusters(struct super_block *sb, |
| 438 | u64 bytes) |
| 439 | { |
| 440 | int cl_bits = OCFS2_SB(sb)->s_clustersize_bits; |
| 441 | unsigned int clusters; |
| 442 | |
| 443 | clusters = ocfs2_clusters_for_bytes(sb, bytes); |
| 444 | return (u64)clusters << cl_bits; |
| 445 | } |
| 446 | |
| 447 | static inline u64 ocfs2_align_bytes_to_blocks(struct super_block *sb, |
| 448 | u64 bytes) |
| 449 | { |
| 450 | u64 blocks; |
| 451 | |
| 452 | blocks = ocfs2_blocks_for_bytes(sb, bytes); |
| 453 | return blocks << sb->s_blocksize_bits; |
| 454 | } |
| 455 | |
| 456 | static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes) |
| 457 | { |
| 458 | return (unsigned long)((bytes + 511) >> 9); |
| 459 | } |
| 460 | |
| 461 | #define ocfs2_set_bit ext2_set_bit |
| 462 | #define ocfs2_clear_bit ext2_clear_bit |
| 463 | #define ocfs2_test_bit ext2_test_bit |
| 464 | #define ocfs2_find_next_zero_bit ext2_find_next_zero_bit |
| 465 | #endif /* OCFS2_H */ |
| 466 | |