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> 36 #include <linux/mutex.h> 37 #include <linux/jbd.h> 38 39 #include "cluster/nodemanager.h" 40 #include "cluster/heartbeat.h" 41 #include "cluster/tcp.h" 42 43 #include "dlm/dlmapi.h" 44 45 #include "ocfs2_fs.h" 46 #include "endian.h" 47 #include "ocfs2_lockid.h" 48 49 /* Most user visible OCFS2 inodes will have very few pieces of 50 * metadata, but larger files (including bitmaps, etc) must be taken 51 * into account when designing an access scheme. We allow a small 52 * amount of inlined blocks to be stored on an array and grow the 53 * structure into a rb tree when necessary. */ 54 #define OCFS2_INODE_MAX_CACHE_ARRAY 2 55 56 struct ocfs2_caching_info { 57 unsigned int ci_num_cached; 58 union { 59 sector_t ci_array[OCFS2_INODE_MAX_CACHE_ARRAY]; 60 struct rb_root ci_tree; 61 } ci_cache; 62 }; 63 64 /* this limits us to 256 nodes 65 * if we need more, we can do a kmalloc for the map */ 66 #define OCFS2_NODE_MAP_MAX_NODES 256 67 struct ocfs2_node_map { 68 u16 num_nodes; 69 unsigned long map[BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES)]; 70 }; 71 72 enum ocfs2_ast_action { 73 OCFS2_AST_INVALID = 0, 74 OCFS2_AST_ATTACH, 75 OCFS2_AST_CONVERT, 76 OCFS2_AST_DOWNCONVERT, 77 }; 78 79 /* actions for an unlockast function to take. */ 80 enum ocfs2_unlock_action { 81 OCFS2_UNLOCK_INVALID = 0, 82 OCFS2_UNLOCK_CANCEL_CONVERT, 83 OCFS2_UNLOCK_DROP_LOCK, 84 }; 85 86 /* ocfs2_lock_res->l_flags flags. */ 87 #define OCFS2_LOCK_ATTACHED (0x00000001) /* have we initialized 88 * the lvb */ 89 #define OCFS2_LOCK_BUSY (0x00000002) /* we are currently in 90 * dlm_lock */ 91 #define OCFS2_LOCK_BLOCKED (0x00000004) /* blocked waiting to 92 * downconvert*/ 93 #define OCFS2_LOCK_LOCAL (0x00000008) /* newly created inode */ 94 #define OCFS2_LOCK_NEEDS_REFRESH (0x00000010) 95 #define OCFS2_LOCK_REFRESHING (0x00000020) 96 #define OCFS2_LOCK_INITIALIZED (0x00000040) /* track initialization 97 * for shutdown paths */ 98 #define OCFS2_LOCK_FREEING (0x00000080) /* help dlmglue track 99 * when to skip queueing 100 * a lock because it's 101 * about to be 102 * dropped. */ 103 #define OCFS2_LOCK_QUEUED (0x00000100) /* queued for downconvert */ 104 #define OCFS2_LOCK_NOCACHE (0x00000200) /* don't use a holder count */ 105 106 struct ocfs2_lock_res_ops; 107 108 typedef void (*ocfs2_lock_callback)(int status, unsigned long data); 109 110 struct ocfs2_lock_res { 111 void *l_priv; 112 struct ocfs2_lock_res_ops *l_ops; 113 spinlock_t l_lock; 114 115 struct list_head l_blocked_list; 116 struct list_head l_mask_waiters; 117 118 enum ocfs2_lock_type l_type; 119 unsigned long l_flags; 120 char l_name[OCFS2_LOCK_ID_MAX_LEN]; 121 int l_level; 122 unsigned int l_ro_holders; 123 unsigned int l_ex_holders; 124 struct dlm_lockstatus l_lksb; 125 126 /* used from AST/BAST funcs. */ 127 enum ocfs2_ast_action l_action; 128 enum ocfs2_unlock_action l_unlock_action; 129 int l_requested; 130 int l_blocking; 131 132 wait_queue_head_t l_event; 133 134 struct list_head l_debug_list; 135 }; 136 137 struct ocfs2_dlm_debug { 138 struct kref d_refcnt; 139 struct dentry *d_locking_state; 140 struct list_head d_lockres_tracking; 141 }; 142 143 enum ocfs2_vol_state 144 { 145 VOLUME_INIT = 0, 146 VOLUME_MOUNTED, 147 VOLUME_DISMOUNTED, 148 VOLUME_DISABLED 149 }; 150 151 struct ocfs2_alloc_stats 152 { 153 atomic_t moves; 154 atomic_t local_data; 155 atomic_t bitmap_data; 156 atomic_t bg_allocs; 157 atomic_t bg_extends; 158 }; 159 160 enum ocfs2_local_alloc_state 161 { 162 OCFS2_LA_UNUSED = 0, 163 OCFS2_LA_ENABLED, 164 OCFS2_LA_DISABLED 165 }; 166 167 enum ocfs2_mount_options 168 { 169 OCFS2_MOUNT_HB_LOCAL = 1 << 0, /* Heartbeat started in local mode */ 170 OCFS2_MOUNT_BARRIER = 1 << 1, /* Use block barriers */ 171 OCFS2_MOUNT_NOINTR = 1 << 2, /* Don't catch signals */ 172 OCFS2_MOUNT_ERRORS_PANIC = 1 << 3, /* Panic on errors */ 173 OCFS2_MOUNT_DATA_WRITEBACK = 1 << 4, /* No data ordering */ 174 }; 175 176 #define OCFS2_OSB_SOFT_RO 0x0001 177 #define OCFS2_OSB_HARD_RO 0x0002 178 #define OCFS2_OSB_ERROR_FS 0x0004 179 #define OCFS2_DEFAULT_ATIME_QUANTUM 60 180 181 struct ocfs2_journal; 182 struct ocfs2_super 183 { 184 struct task_struct *commit_task; 185 struct super_block *sb; 186 struct inode *root_inode; 187 struct inode *sys_root_inode; 188 struct inode *system_inodes[NUM_SYSTEM_INODES]; 189 190 struct ocfs2_slot_info *slot_info; 191 192 spinlock_t node_map_lock; 193 struct ocfs2_node_map recovery_map; 194 195 u64 root_blkno; 196 u64 system_dir_blkno; 197 u64 bitmap_blkno; 198 u32 bitmap_cpg; 199 u8 *uuid; 200 char *uuid_str; 201 u8 *vol_label; 202 u64 first_cluster_group_blkno; 203 u32 fs_generation; 204 205 u32 s_feature_compat; 206 u32 s_feature_incompat; 207 u32 s_feature_ro_compat; 208 209 /* Protects s_next_generaion, osb_flags. Could protect more on 210 * osb as it's very short lived. */ 211 spinlock_t osb_lock; 212 u32 s_next_generation; 213 unsigned long osb_flags; 214 215 unsigned long s_mount_opt; 216 unsigned int s_atime_quantum; 217 218 u16 max_slots; 219 s16 node_num; 220 s16 slot_num; 221 s16 preferred_slot; 222 int s_sectsize_bits; 223 int s_clustersize; 224 int s_clustersize_bits; 225 226 atomic_t vol_state; 227 struct mutex recovery_lock; 228 struct task_struct *recovery_thread_task; 229 int disable_recovery; 230 wait_queue_head_t checkpoint_event; 231 atomic_t needs_checkpoint; 232 struct ocfs2_journal *journal; 233 unsigned long osb_commit_interval; 234 235 int local_alloc_size; 236 enum ocfs2_local_alloc_state local_alloc_state; 237 struct buffer_head *local_alloc_bh; 238 u64 la_last_gd; 239 240 /* Next two fields are for local node slot recovery during 241 * mount. */ 242 int dirty; 243 struct ocfs2_dinode *local_alloc_copy; 244 245 struct ocfs2_alloc_stats alloc_stats; 246 char dev_str[20]; /* "major,minor" of the device */ 247 248 struct dlm_ctxt *dlm; 249 struct ocfs2_lock_res osb_super_lockres; 250 struct ocfs2_lock_res osb_rename_lockres; 251 struct dlm_eviction_cb osb_eviction_cb; 252 struct ocfs2_dlm_debug *osb_dlm_debug; 253 254 struct dentry *osb_debug_root; 255 256 wait_queue_head_t recovery_event; 257 258 spinlock_t dc_task_lock; 259 struct task_struct *dc_task; 260 wait_queue_head_t dc_event; 261 unsigned long dc_wake_sequence; 262 unsigned long dc_work_sequence; 263 264 struct list_head blocked_lock_list; 265 unsigned long blocked_lock_count; 266 267 wait_queue_head_t osb_mount_event; 268 269 /* Truncate log info */ 270 struct inode *osb_tl_inode; 271 struct buffer_head *osb_tl_bh; 272 struct delayed_work osb_truncate_log_wq; 273 274 struct ocfs2_node_map osb_recovering_orphan_dirs; 275 unsigned int *osb_orphan_wipes; 276 wait_queue_head_t osb_wipe_event; 277 }; 278 279 #define OCFS2_SB(sb) ((struct ocfs2_super *)(sb)->s_fs_info) 280 281 static inline int ocfs2_should_order_data(struct inode *inode) 282 { 283 if (!S_ISREG(inode->i_mode)) 284 return 0; 285 if (OCFS2_SB(inode->i_sb)->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) 286 return 0; 287 return 1; 288 } 289 290 static inline int ocfs2_sparse_alloc(struct ocfs2_super *osb) 291 { 292 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC) 293 return 1; 294 return 0; 295 } 296 297 static inline int ocfs2_writes_unwritten_extents(struct ocfs2_super *osb) 298 { 299 /* 300 * Support for sparse files is a pre-requisite 301 */ 302 if (!ocfs2_sparse_alloc(osb)) 303 return 0; 304 305 if (osb->s_feature_ro_compat & OCFS2_FEATURE_RO_COMPAT_UNWRITTEN) 306 return 1; 307 return 0; 308 } 309 310 static inline int ocfs2_supports_inline_data(struct ocfs2_super *osb) 311 { 312 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_INLINE_DATA) 313 return 1; 314 return 0; 315 } 316 317 /* set / clear functions because cluster events can make these happen 318 * in parallel so we want the transitions to be atomic. this also 319 * means that any future flags osb_flags must be protected by spinlock 320 * too! */ 321 static inline void ocfs2_set_osb_flag(struct ocfs2_super *osb, 322 unsigned long flag) 323 { 324 spin_lock(&osb->osb_lock); 325 osb->osb_flags |= flag; 326 spin_unlock(&osb->osb_lock); 327 } 328 329 static inline void ocfs2_set_ro_flag(struct ocfs2_super *osb, 330 int hard) 331 { 332 spin_lock(&osb->osb_lock); 333 osb->osb_flags &= ~(OCFS2_OSB_SOFT_RO|OCFS2_OSB_HARD_RO); 334 if (hard) 335 osb->osb_flags |= OCFS2_OSB_HARD_RO; 336 else 337 osb->osb_flags |= OCFS2_OSB_SOFT_RO; 338 spin_unlock(&osb->osb_lock); 339 } 340 341 static inline int ocfs2_is_hard_readonly(struct ocfs2_super *osb) 342 { 343 int ret; 344 345 spin_lock(&osb->osb_lock); 346 ret = osb->osb_flags & OCFS2_OSB_HARD_RO; 347 spin_unlock(&osb->osb_lock); 348 349 return ret; 350 } 351 352 static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb) 353 { 354 int ret; 355 356 spin_lock(&osb->osb_lock); 357 ret = osb->osb_flags & OCFS2_OSB_SOFT_RO; 358 spin_unlock(&osb->osb_lock); 359 360 return ret; 361 } 362 363 static inline int ocfs2_mount_local(struct ocfs2_super *osb) 364 { 365 return (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT); 366 } 367 368 #define OCFS2_IS_VALID_DINODE(ptr) \ 369 (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE)) 370 371 #define OCFS2_RO_ON_INVALID_DINODE(__sb, __di) do { \ 372 typeof(__di) ____di = (__di); \ 373 ocfs2_error((__sb), \ 374 "Dinode # %llu has bad signature %.*s", \ 375 (unsigned long long)le64_to_cpu((____di)->i_blkno), 7, \ 376 (____di)->i_signature); \ 377 } while (0) 378 379 #define OCFS2_IS_VALID_EXTENT_BLOCK(ptr) \ 380 (!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE)) 381 382 #define OCFS2_RO_ON_INVALID_EXTENT_BLOCK(__sb, __eb) do { \ 383 typeof(__eb) ____eb = (__eb); \ 384 ocfs2_error((__sb), \ 385 "Extent Block # %llu has bad signature %.*s", \ 386 (unsigned long long)le64_to_cpu((____eb)->h_blkno), 7, \ 387 (____eb)->h_signature); \ 388 } while (0) 389 390 #define OCFS2_IS_VALID_GROUP_DESC(ptr) \ 391 (!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE)) 392 393 #define OCFS2_RO_ON_INVALID_GROUP_DESC(__sb, __gd) do { \ 394 typeof(__gd) ____gd = (__gd); \ 395 ocfs2_error((__sb), \ 396 "Group Descriptor # %llu has bad signature %.*s", \ 397 (unsigned long long)le64_to_cpu((____gd)->bg_blkno), 7, \ 398 (____gd)->bg_signature); \ 399 } while (0) 400 401 static inline unsigned long ino_from_blkno(struct super_block *sb, 402 u64 blkno) 403 { 404 return (unsigned long)(blkno & (u64)ULONG_MAX); 405 } 406 407 static inline u64 ocfs2_clusters_to_blocks(struct super_block *sb, 408 u32 clusters) 409 { 410 int c_to_b_bits = OCFS2_SB(sb)->s_clustersize_bits - 411 sb->s_blocksize_bits; 412 413 return (u64)clusters << c_to_b_bits; 414 } 415 416 static inline u32 ocfs2_blocks_to_clusters(struct super_block *sb, 417 u64 blocks) 418 { 419 int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits - 420 sb->s_blocksize_bits; 421 422 return (u32)(blocks >> b_to_c_bits); 423 } 424 425 static inline unsigned int ocfs2_clusters_for_bytes(struct super_block *sb, 426 u64 bytes) 427 { 428 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits; 429 unsigned int clusters; 430 431 bytes += OCFS2_SB(sb)->s_clustersize - 1; 432 /* OCFS2 just cannot have enough clusters to overflow this */ 433 clusters = (unsigned int)(bytes >> cl_bits); 434 435 return clusters; 436 } 437 438 static inline u64 ocfs2_blocks_for_bytes(struct super_block *sb, 439 u64 bytes) 440 { 441 bytes += sb->s_blocksize - 1; 442 return bytes >> sb->s_blocksize_bits; 443 } 444 445 static inline u64 ocfs2_clusters_to_bytes(struct super_block *sb, 446 u32 clusters) 447 { 448 return (u64)clusters << OCFS2_SB(sb)->s_clustersize_bits; 449 } 450 451 static inline u64 ocfs2_align_bytes_to_clusters(struct super_block *sb, 452 u64 bytes) 453 { 454 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits; 455 unsigned int clusters; 456 457 clusters = ocfs2_clusters_for_bytes(sb, bytes); 458 return (u64)clusters << cl_bits; 459 } 460 461 static inline u64 ocfs2_align_bytes_to_blocks(struct super_block *sb, 462 u64 bytes) 463 { 464 u64 blocks; 465 466 blocks = ocfs2_blocks_for_bytes(sb, bytes); 467 return blocks << sb->s_blocksize_bits; 468 } 469 470 static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes) 471 { 472 return (unsigned long)((bytes + 511) >> 9); 473 } 474 475 static inline unsigned int ocfs2_page_index_to_clusters(struct super_block *sb, 476 unsigned long pg_index) 477 { 478 u32 clusters = pg_index; 479 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits; 480 481 if (unlikely(PAGE_CACHE_SHIFT > cbits)) 482 clusters = pg_index << (PAGE_CACHE_SHIFT - cbits); 483 else if (PAGE_CACHE_SHIFT < cbits) 484 clusters = pg_index >> (cbits - PAGE_CACHE_SHIFT); 485 486 return clusters; 487 } 488 489 /* 490 * Find the 1st page index which covers the given clusters. 491 */ 492 static inline pgoff_t ocfs2_align_clusters_to_page_index(struct super_block *sb, 493 u32 clusters) 494 { 495 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits; 496 pgoff_t index = clusters; 497 498 if (PAGE_CACHE_SHIFT > cbits) { 499 index = (pgoff_t)clusters >> (PAGE_CACHE_SHIFT - cbits); 500 } else if (PAGE_CACHE_SHIFT < cbits) { 501 index = (pgoff_t)clusters << (cbits - PAGE_CACHE_SHIFT); 502 } 503 504 return index; 505 } 506 507 static inline unsigned int ocfs2_pages_per_cluster(struct super_block *sb) 508 { 509 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits; 510 unsigned int pages_per_cluster = 1; 511 512 if (PAGE_CACHE_SHIFT < cbits) 513 pages_per_cluster = 1 << (cbits - PAGE_CACHE_SHIFT); 514 515 return pages_per_cluster; 516 } 517 518 #define ocfs2_set_bit ext2_set_bit 519 #define ocfs2_clear_bit ext2_clear_bit 520 #define ocfs2_test_bit ext2_test_bit 521 #define ocfs2_find_next_zero_bit ext2_find_next_zero_bit 522 #endif /* OCFS2_H */ 523 524