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 /* For union ocfs2_dlm_lksb */ 40 #include "stackglue.h" 41 42 #include "ocfs2_fs.h" 43 #include "ocfs2_lockid.h" 44 45 /* Most user visible OCFS2 inodes will have very few pieces of 46 * metadata, but larger files (including bitmaps, etc) must be taken 47 * into account when designing an access scheme. We allow a small 48 * amount of inlined blocks to be stored on an array and grow the 49 * structure into a rb tree when necessary. */ 50 #define OCFS2_INODE_MAX_CACHE_ARRAY 2 51 52 struct ocfs2_caching_info { 53 unsigned int ci_num_cached; 54 union { 55 sector_t ci_array[OCFS2_INODE_MAX_CACHE_ARRAY]; 56 struct rb_root ci_tree; 57 } ci_cache; 58 }; 59 60 /* this limits us to 256 nodes 61 * if we need more, we can do a kmalloc for the map */ 62 #define OCFS2_NODE_MAP_MAX_NODES 256 63 struct ocfs2_node_map { 64 u16 num_nodes; 65 unsigned long map[BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES)]; 66 }; 67 68 enum ocfs2_ast_action { 69 OCFS2_AST_INVALID = 0, 70 OCFS2_AST_ATTACH, 71 OCFS2_AST_CONVERT, 72 OCFS2_AST_DOWNCONVERT, 73 }; 74 75 /* actions for an unlockast function to take. */ 76 enum ocfs2_unlock_action { 77 OCFS2_UNLOCK_INVALID = 0, 78 OCFS2_UNLOCK_CANCEL_CONVERT, 79 OCFS2_UNLOCK_DROP_LOCK, 80 }; 81 82 /* ocfs2_lock_res->l_flags flags. */ 83 #define OCFS2_LOCK_ATTACHED (0x00000001) /* have we initialized 84 * the lvb */ 85 #define OCFS2_LOCK_BUSY (0x00000002) /* we are currently in 86 * dlm_lock */ 87 #define OCFS2_LOCK_BLOCKED (0x00000004) /* blocked waiting to 88 * downconvert*/ 89 #define OCFS2_LOCK_LOCAL (0x00000008) /* newly created inode */ 90 #define OCFS2_LOCK_NEEDS_REFRESH (0x00000010) 91 #define OCFS2_LOCK_REFRESHING (0x00000020) 92 #define OCFS2_LOCK_INITIALIZED (0x00000040) /* track initialization 93 * for shutdown paths */ 94 #define OCFS2_LOCK_FREEING (0x00000080) /* help dlmglue track 95 * when to skip queueing 96 * a lock because it's 97 * about to be 98 * dropped. */ 99 #define OCFS2_LOCK_QUEUED (0x00000100) /* queued for downconvert */ 100 #define OCFS2_LOCK_NOCACHE (0x00000200) /* don't use a holder count */ 101 #define OCFS2_LOCK_PENDING (0x00000400) /* This lockres is pending a 102 call to dlm_lock. Only 103 exists with BUSY set. */ 104 105 struct ocfs2_lock_res_ops; 106 107 typedef void (*ocfs2_lock_callback)(int status, unsigned long data); 108 109 struct ocfs2_lock_res { 110 void *l_priv; 111 struct ocfs2_lock_res_ops *l_ops; 112 spinlock_t l_lock; 113 114 struct list_head l_blocked_list; 115 struct list_head l_mask_waiters; 116 117 enum ocfs2_lock_type l_type; 118 unsigned long l_flags; 119 char l_name[OCFS2_LOCK_ID_MAX_LEN]; 120 int l_level; 121 unsigned int l_ro_holders; 122 unsigned int l_ex_holders; 123 union ocfs2_dlm_lksb l_lksb; 124 125 /* used from AST/BAST funcs. */ 126 enum ocfs2_ast_action l_action; 127 enum ocfs2_unlock_action l_unlock_action; 128 int l_requested; 129 int l_blocking; 130 unsigned int l_pending_gen; 131 132 wait_queue_head_t l_event; 133 134 struct list_head l_debug_list; 135 136 #ifdef CONFIG_OCFS2_FS_STATS 137 unsigned long long l_lock_num_prmode; /* PR acquires */ 138 unsigned long long l_lock_num_exmode; /* EX acquires */ 139 unsigned int l_lock_num_prmode_failed; /* Failed PR gets */ 140 unsigned int l_lock_num_exmode_failed; /* Failed EX gets */ 141 unsigned long long l_lock_total_prmode; /* Tot wait for PR */ 142 unsigned long long l_lock_total_exmode; /* Tot wait for EX */ 143 unsigned int l_lock_max_prmode; /* Max wait for PR */ 144 unsigned int l_lock_max_exmode; /* Max wait for EX */ 145 unsigned int l_lock_refresh; /* Disk refreshes */ 146 #endif 147 }; 148 149 struct ocfs2_dlm_debug { 150 struct kref d_refcnt; 151 struct dentry *d_locking_state; 152 struct list_head d_lockres_tracking; 153 }; 154 155 enum ocfs2_vol_state 156 { 157 VOLUME_INIT = 0, 158 VOLUME_MOUNTED, 159 VOLUME_DISMOUNTED, 160 VOLUME_DISABLED 161 }; 162 163 struct ocfs2_alloc_stats 164 { 165 atomic_t moves; 166 atomic_t local_data; 167 atomic_t bitmap_data; 168 atomic_t bg_allocs; 169 atomic_t bg_extends; 170 }; 171 172 enum ocfs2_local_alloc_state 173 { 174 OCFS2_LA_UNUSED = 0, 175 OCFS2_LA_ENABLED, 176 OCFS2_LA_DISABLED 177 }; 178 179 enum ocfs2_mount_options 180 { 181 OCFS2_MOUNT_HB_LOCAL = 1 << 0, /* Heartbeat started in local mode */ 182 OCFS2_MOUNT_BARRIER = 1 << 1, /* Use block barriers */ 183 OCFS2_MOUNT_NOINTR = 1 << 2, /* Don't catch signals */ 184 OCFS2_MOUNT_ERRORS_PANIC = 1 << 3, /* Panic on errors */ 185 OCFS2_MOUNT_DATA_WRITEBACK = 1 << 4, /* No data ordering */ 186 OCFS2_MOUNT_LOCALFLOCKS = 1 << 5, /* No cluster aware user file locks */ 187 }; 188 189 #define OCFS2_OSB_SOFT_RO 0x0001 190 #define OCFS2_OSB_HARD_RO 0x0002 191 #define OCFS2_OSB_ERROR_FS 0x0004 192 #define OCFS2_DEFAULT_ATIME_QUANTUM 60 193 194 struct ocfs2_journal; 195 struct ocfs2_slot_info; 196 struct ocfs2_recovery_map; 197 struct ocfs2_super 198 { 199 struct task_struct *commit_task; 200 struct super_block *sb; 201 struct inode *root_inode; 202 struct inode *sys_root_inode; 203 struct inode *system_inodes[NUM_SYSTEM_INODES]; 204 205 struct ocfs2_slot_info *slot_info; 206 207 u32 *slot_recovery_generations; 208 209 spinlock_t node_map_lock; 210 211 u64 root_blkno; 212 u64 system_dir_blkno; 213 u64 bitmap_blkno; 214 u32 bitmap_cpg; 215 u8 *uuid; 216 char *uuid_str; 217 u8 *vol_label; 218 u64 first_cluster_group_blkno; 219 u32 fs_generation; 220 221 u32 s_feature_compat; 222 u32 s_feature_incompat; 223 u32 s_feature_ro_compat; 224 225 /* Protects s_next_generation, osb_flags and s_inode_steal_slot. 226 * Could protect more on osb as it's very short lived. 227 */ 228 spinlock_t osb_lock; 229 u32 s_next_generation; 230 unsigned long osb_flags; 231 s16 s_inode_steal_slot; 232 atomic_t s_num_inodes_stolen; 233 234 unsigned long s_mount_opt; 235 unsigned int s_atime_quantum; 236 237 unsigned int max_slots; 238 unsigned int node_num; 239 int slot_num; 240 int preferred_slot; 241 int s_sectsize_bits; 242 int s_clustersize; 243 int s_clustersize_bits; 244 245 atomic_t vol_state; 246 struct mutex recovery_lock; 247 struct ocfs2_recovery_map *recovery_map; 248 struct task_struct *recovery_thread_task; 249 int disable_recovery; 250 wait_queue_head_t checkpoint_event; 251 atomic_t needs_checkpoint; 252 struct ocfs2_journal *journal; 253 unsigned long osb_commit_interval; 254 255 int local_alloc_size; 256 enum ocfs2_local_alloc_state local_alloc_state; 257 struct buffer_head *local_alloc_bh; 258 u64 la_last_gd; 259 260 /* Next two fields are for local node slot recovery during 261 * mount. */ 262 int dirty; 263 struct ocfs2_dinode *local_alloc_copy; 264 265 struct ocfs2_alloc_stats alloc_stats; 266 char dev_str[20]; /* "major,minor" of the device */ 267 268 char osb_cluster_stack[OCFS2_STACK_LABEL_LEN + 1]; 269 struct ocfs2_cluster_connection *cconn; 270 struct ocfs2_lock_res osb_super_lockres; 271 struct ocfs2_lock_res osb_rename_lockres; 272 struct ocfs2_dlm_debug *osb_dlm_debug; 273 274 struct dentry *osb_debug_root; 275 276 wait_queue_head_t recovery_event; 277 278 spinlock_t dc_task_lock; 279 struct task_struct *dc_task; 280 wait_queue_head_t dc_event; 281 unsigned long dc_wake_sequence; 282 unsigned long dc_work_sequence; 283 284 /* 285 * Any thread can add locks to the list, but the downconvert 286 * thread is the only one allowed to remove locks. Any change 287 * to this rule requires updating 288 * ocfs2_downconvert_thread_do_work(). 289 */ 290 struct list_head blocked_lock_list; 291 unsigned long blocked_lock_count; 292 293 wait_queue_head_t osb_mount_event; 294 295 /* Truncate log info */ 296 struct inode *osb_tl_inode; 297 struct buffer_head *osb_tl_bh; 298 struct delayed_work osb_truncate_log_wq; 299 300 struct ocfs2_node_map osb_recovering_orphan_dirs; 301 unsigned int *osb_orphan_wipes; 302 wait_queue_head_t osb_wipe_event; 303 }; 304 305 #define OCFS2_SB(sb) ((struct ocfs2_super *)(sb)->s_fs_info) 306 307 static inline int ocfs2_should_order_data(struct inode *inode) 308 { 309 if (!S_ISREG(inode->i_mode)) 310 return 0; 311 if (OCFS2_SB(inode->i_sb)->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) 312 return 0; 313 return 1; 314 } 315 316 static inline int ocfs2_sparse_alloc(struct ocfs2_super *osb) 317 { 318 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC) 319 return 1; 320 return 0; 321 } 322 323 static inline int ocfs2_writes_unwritten_extents(struct ocfs2_super *osb) 324 { 325 /* 326 * Support for sparse files is a pre-requisite 327 */ 328 if (!ocfs2_sparse_alloc(osb)) 329 return 0; 330 331 if (osb->s_feature_ro_compat & OCFS2_FEATURE_RO_COMPAT_UNWRITTEN) 332 return 1; 333 return 0; 334 } 335 336 static inline int ocfs2_supports_inline_data(struct ocfs2_super *osb) 337 { 338 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_INLINE_DATA) 339 return 1; 340 return 0; 341 } 342 343 /* set / clear functions because cluster events can make these happen 344 * in parallel so we want the transitions to be atomic. this also 345 * means that any future flags osb_flags must be protected by spinlock 346 * too! */ 347 static inline void ocfs2_set_osb_flag(struct ocfs2_super *osb, 348 unsigned long flag) 349 { 350 spin_lock(&osb->osb_lock); 351 osb->osb_flags |= flag; 352 spin_unlock(&osb->osb_lock); 353 } 354 355 static inline void ocfs2_set_ro_flag(struct ocfs2_super *osb, 356 int hard) 357 { 358 spin_lock(&osb->osb_lock); 359 osb->osb_flags &= ~(OCFS2_OSB_SOFT_RO|OCFS2_OSB_HARD_RO); 360 if (hard) 361 osb->osb_flags |= OCFS2_OSB_HARD_RO; 362 else 363 osb->osb_flags |= OCFS2_OSB_SOFT_RO; 364 spin_unlock(&osb->osb_lock); 365 } 366 367 static inline int ocfs2_is_hard_readonly(struct ocfs2_super *osb) 368 { 369 int ret; 370 371 spin_lock(&osb->osb_lock); 372 ret = osb->osb_flags & OCFS2_OSB_HARD_RO; 373 spin_unlock(&osb->osb_lock); 374 375 return ret; 376 } 377 378 static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb) 379 { 380 int ret; 381 382 spin_lock(&osb->osb_lock); 383 ret = osb->osb_flags & OCFS2_OSB_SOFT_RO; 384 spin_unlock(&osb->osb_lock); 385 386 return ret; 387 } 388 389 static inline int ocfs2_userspace_stack(struct ocfs2_super *osb) 390 { 391 return (osb->s_feature_incompat & 392 OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK); 393 } 394 395 static inline int ocfs2_mount_local(struct ocfs2_super *osb) 396 { 397 return (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT); 398 } 399 400 static inline int ocfs2_uses_extended_slot_map(struct ocfs2_super *osb) 401 { 402 return (osb->s_feature_incompat & 403 OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP); 404 } 405 406 407 #define OCFS2_IS_VALID_DINODE(ptr) \ 408 (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE)) 409 410 #define OCFS2_RO_ON_INVALID_DINODE(__sb, __di) do { \ 411 typeof(__di) ____di = (__di); \ 412 ocfs2_error((__sb), \ 413 "Dinode # %llu has bad signature %.*s", \ 414 (unsigned long long)le64_to_cpu((____di)->i_blkno), 7, \ 415 (____di)->i_signature); \ 416 } while (0) 417 418 #define OCFS2_IS_VALID_EXTENT_BLOCK(ptr) \ 419 (!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE)) 420 421 #define OCFS2_RO_ON_INVALID_EXTENT_BLOCK(__sb, __eb) do { \ 422 typeof(__eb) ____eb = (__eb); \ 423 ocfs2_error((__sb), \ 424 "Extent Block # %llu has bad signature %.*s", \ 425 (unsigned long long)le64_to_cpu((____eb)->h_blkno), 7, \ 426 (____eb)->h_signature); \ 427 } while (0) 428 429 #define OCFS2_IS_VALID_GROUP_DESC(ptr) \ 430 (!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE)) 431 432 #define OCFS2_RO_ON_INVALID_GROUP_DESC(__sb, __gd) do { \ 433 typeof(__gd) ____gd = (__gd); \ 434 ocfs2_error((__sb), \ 435 "Group Descriptor # %llu has bad signature %.*s", \ 436 (unsigned long long)le64_to_cpu((____gd)->bg_blkno), 7, \ 437 (____gd)->bg_signature); \ 438 } while (0) 439 440 static inline unsigned long ino_from_blkno(struct super_block *sb, 441 u64 blkno) 442 { 443 return (unsigned long)(blkno & (u64)ULONG_MAX); 444 } 445 446 static inline u64 ocfs2_clusters_to_blocks(struct super_block *sb, 447 u32 clusters) 448 { 449 int c_to_b_bits = OCFS2_SB(sb)->s_clustersize_bits - 450 sb->s_blocksize_bits; 451 452 return (u64)clusters << c_to_b_bits; 453 } 454 455 static inline u32 ocfs2_blocks_to_clusters(struct super_block *sb, 456 u64 blocks) 457 { 458 int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits - 459 sb->s_blocksize_bits; 460 461 return (u32)(blocks >> b_to_c_bits); 462 } 463 464 static inline unsigned int ocfs2_clusters_for_bytes(struct super_block *sb, 465 u64 bytes) 466 { 467 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits; 468 unsigned int clusters; 469 470 bytes += OCFS2_SB(sb)->s_clustersize - 1; 471 /* OCFS2 just cannot have enough clusters to overflow this */ 472 clusters = (unsigned int)(bytes >> cl_bits); 473 474 return clusters; 475 } 476 477 static inline u64 ocfs2_blocks_for_bytes(struct super_block *sb, 478 u64 bytes) 479 { 480 bytes += sb->s_blocksize - 1; 481 return bytes >> sb->s_blocksize_bits; 482 } 483 484 static inline u64 ocfs2_clusters_to_bytes(struct super_block *sb, 485 u32 clusters) 486 { 487 return (u64)clusters << OCFS2_SB(sb)->s_clustersize_bits; 488 } 489 490 static inline u64 ocfs2_align_bytes_to_clusters(struct super_block *sb, 491 u64 bytes) 492 { 493 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits; 494 unsigned int clusters; 495 496 clusters = ocfs2_clusters_for_bytes(sb, bytes); 497 return (u64)clusters << cl_bits; 498 } 499 500 static inline u64 ocfs2_align_bytes_to_blocks(struct super_block *sb, 501 u64 bytes) 502 { 503 u64 blocks; 504 505 blocks = ocfs2_blocks_for_bytes(sb, bytes); 506 return blocks << sb->s_blocksize_bits; 507 } 508 509 static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes) 510 { 511 return (unsigned long)((bytes + 511) >> 9); 512 } 513 514 static inline unsigned int ocfs2_page_index_to_clusters(struct super_block *sb, 515 unsigned long pg_index) 516 { 517 u32 clusters = pg_index; 518 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits; 519 520 if (unlikely(PAGE_CACHE_SHIFT > cbits)) 521 clusters = pg_index << (PAGE_CACHE_SHIFT - cbits); 522 else if (PAGE_CACHE_SHIFT < cbits) 523 clusters = pg_index >> (cbits - PAGE_CACHE_SHIFT); 524 525 return clusters; 526 } 527 528 /* 529 * Find the 1st page index which covers the given clusters. 530 */ 531 static inline pgoff_t ocfs2_align_clusters_to_page_index(struct super_block *sb, 532 u32 clusters) 533 { 534 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits; 535 pgoff_t index = clusters; 536 537 if (PAGE_CACHE_SHIFT > cbits) { 538 index = (pgoff_t)clusters >> (PAGE_CACHE_SHIFT - cbits); 539 } else if (PAGE_CACHE_SHIFT < cbits) { 540 index = (pgoff_t)clusters << (cbits - PAGE_CACHE_SHIFT); 541 } 542 543 return index; 544 } 545 546 static inline unsigned int ocfs2_pages_per_cluster(struct super_block *sb) 547 { 548 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits; 549 unsigned int pages_per_cluster = 1; 550 551 if (PAGE_CACHE_SHIFT < cbits) 552 pages_per_cluster = 1 << (cbits - PAGE_CACHE_SHIFT); 553 554 return pages_per_cluster; 555 } 556 557 static inline void ocfs2_init_inode_steal_slot(struct ocfs2_super *osb) 558 { 559 spin_lock(&osb->osb_lock); 560 osb->s_inode_steal_slot = OCFS2_INVALID_SLOT; 561 spin_unlock(&osb->osb_lock); 562 atomic_set(&osb->s_num_inodes_stolen, 0); 563 } 564 565 static inline void ocfs2_set_inode_steal_slot(struct ocfs2_super *osb, 566 s16 slot) 567 { 568 spin_lock(&osb->osb_lock); 569 osb->s_inode_steal_slot = slot; 570 spin_unlock(&osb->osb_lock); 571 } 572 573 static inline s16 ocfs2_get_inode_steal_slot(struct ocfs2_super *osb) 574 { 575 s16 slot; 576 577 spin_lock(&osb->osb_lock); 578 slot = osb->s_inode_steal_slot; 579 spin_unlock(&osb->osb_lock); 580 581 return slot; 582 } 583 584 #define ocfs2_set_bit ext2_set_bit 585 #define ocfs2_clear_bit ext2_clear_bit 586 #define ocfs2_test_bit ext2_test_bit 587 #define ocfs2_find_next_zero_bit ext2_find_next_zero_bit 588 #endif /* OCFS2_H */ 589 590