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 #ifndef CONFIG_OCFS2_COMPAT_JBD 38 # include <linux/jbd2.h> 39 #else 40 # include <linux/jbd.h> 41 # include "ocfs2_jbd_compat.h" 42 #endif 43 44 /* For union ocfs2_dlm_lksb */ 45 #include "stackglue.h" 46 47 #include "ocfs2_fs.h" 48 #include "ocfs2_lockid.h" 49 50 /* Most user visible OCFS2 inodes will have very few pieces of 51 * metadata, but larger files (including bitmaps, etc) must be taken 52 * into account when designing an access scheme. We allow a small 53 * amount of inlined blocks to be stored on an array and grow the 54 * structure into a rb tree when necessary. */ 55 #define OCFS2_INODE_MAX_CACHE_ARRAY 2 56 57 struct ocfs2_caching_info { 58 unsigned int ci_num_cached; 59 union { 60 sector_t ci_array[OCFS2_INODE_MAX_CACHE_ARRAY]; 61 struct rb_root ci_tree; 62 } ci_cache; 63 }; 64 65 /* this limits us to 256 nodes 66 * if we need more, we can do a kmalloc for the map */ 67 #define OCFS2_NODE_MAP_MAX_NODES 256 68 struct ocfs2_node_map { 69 u16 num_nodes; 70 unsigned long map[BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES)]; 71 }; 72 73 enum ocfs2_ast_action { 74 OCFS2_AST_INVALID = 0, 75 OCFS2_AST_ATTACH, 76 OCFS2_AST_CONVERT, 77 OCFS2_AST_DOWNCONVERT, 78 }; 79 80 /* actions for an unlockast function to take. */ 81 enum ocfs2_unlock_action { 82 OCFS2_UNLOCK_INVALID = 0, 83 OCFS2_UNLOCK_CANCEL_CONVERT, 84 OCFS2_UNLOCK_DROP_LOCK, 85 }; 86 87 /* ocfs2_lock_res->l_flags flags. */ 88 #define OCFS2_LOCK_ATTACHED (0x00000001) /* we have initialized 89 * the lvb */ 90 #define OCFS2_LOCK_BUSY (0x00000002) /* we are currently in 91 * dlm_lock */ 92 #define OCFS2_LOCK_BLOCKED (0x00000004) /* blocked waiting to 93 * downconvert*/ 94 #define OCFS2_LOCK_LOCAL (0x00000008) /* newly created inode */ 95 #define OCFS2_LOCK_NEEDS_REFRESH (0x00000010) 96 #define OCFS2_LOCK_REFRESHING (0x00000020) 97 #define OCFS2_LOCK_INITIALIZED (0x00000040) /* track initialization 98 * for shutdown paths */ 99 #define OCFS2_LOCK_FREEING (0x00000080) /* help dlmglue track 100 * when to skip queueing 101 * a lock because it's 102 * about to be 103 * dropped. */ 104 #define OCFS2_LOCK_QUEUED (0x00000100) /* queued for downconvert */ 105 #define OCFS2_LOCK_NOCACHE (0x00000200) /* don't use a holder count */ 106 #define OCFS2_LOCK_PENDING (0x00000400) /* This lockres is pending a 107 call to dlm_lock. Only 108 exists with BUSY set. */ 109 110 struct ocfs2_lock_res_ops; 111 112 typedef void (*ocfs2_lock_callback)(int status, unsigned long data); 113 114 struct ocfs2_lock_res { 115 void *l_priv; 116 struct ocfs2_lock_res_ops *l_ops; 117 spinlock_t l_lock; 118 119 struct list_head l_blocked_list; 120 struct list_head l_mask_waiters; 121 122 enum ocfs2_lock_type l_type; 123 unsigned long l_flags; 124 char l_name[OCFS2_LOCK_ID_MAX_LEN]; 125 int l_level; 126 unsigned int l_ro_holders; 127 unsigned int l_ex_holders; 128 union ocfs2_dlm_lksb l_lksb; 129 130 /* used from AST/BAST funcs. */ 131 enum ocfs2_ast_action l_action; 132 enum ocfs2_unlock_action l_unlock_action; 133 int l_requested; 134 int l_blocking; 135 unsigned int l_pending_gen; 136 137 wait_queue_head_t l_event; 138 139 struct list_head l_debug_list; 140 141 #ifdef CONFIG_OCFS2_FS_STATS 142 unsigned long long l_lock_num_prmode; /* PR acquires */ 143 unsigned long long l_lock_num_exmode; /* EX acquires */ 144 unsigned int l_lock_num_prmode_failed; /* Failed PR gets */ 145 unsigned int l_lock_num_exmode_failed; /* Failed EX gets */ 146 unsigned long long l_lock_total_prmode; /* Tot wait for PR */ 147 unsigned long long l_lock_total_exmode; /* Tot wait for EX */ 148 unsigned int l_lock_max_prmode; /* Max wait for PR */ 149 unsigned int l_lock_max_exmode; /* Max wait for EX */ 150 unsigned int l_lock_refresh; /* Disk refreshes */ 151 #endif 152 }; 153 154 struct ocfs2_dlm_debug { 155 struct kref d_refcnt; 156 struct dentry *d_locking_state; 157 struct list_head d_lockres_tracking; 158 }; 159 160 enum ocfs2_vol_state 161 { 162 VOLUME_INIT = 0, 163 VOLUME_MOUNTED, 164 VOLUME_MOUNTED_QUOTAS, 165 VOLUME_DISMOUNTED, 166 VOLUME_DISABLED 167 }; 168 169 struct ocfs2_alloc_stats 170 { 171 atomic_t moves; 172 atomic_t local_data; 173 atomic_t bitmap_data; 174 atomic_t bg_allocs; 175 atomic_t bg_extends; 176 }; 177 178 enum ocfs2_local_alloc_state 179 { 180 OCFS2_LA_UNUSED = 0, /* Local alloc will never be used for 181 * this mountpoint. */ 182 OCFS2_LA_ENABLED, /* Local alloc is in use. */ 183 OCFS2_LA_THROTTLED, /* Local alloc is in use, but number 184 * of bits has been reduced. */ 185 OCFS2_LA_DISABLED /* Local alloc has temporarily been 186 * disabled. */ 187 }; 188 189 enum ocfs2_mount_options 190 { 191 OCFS2_MOUNT_HB_LOCAL = 1 << 0, /* Heartbeat started in local mode */ 192 OCFS2_MOUNT_BARRIER = 1 << 1, /* Use block barriers */ 193 OCFS2_MOUNT_NOINTR = 1 << 2, /* Don't catch signals */ 194 OCFS2_MOUNT_ERRORS_PANIC = 1 << 3, /* Panic on errors */ 195 OCFS2_MOUNT_DATA_WRITEBACK = 1 << 4, /* No data ordering */ 196 OCFS2_MOUNT_LOCALFLOCKS = 1 << 5, /* No cluster aware user file locks */ 197 OCFS2_MOUNT_NOUSERXATTR = 1 << 6, /* No user xattr */ 198 OCFS2_MOUNT_INODE64 = 1 << 7, /* Allow inode numbers > 2^32 */ 199 OCFS2_MOUNT_POSIX_ACL = 1 << 8, /* POSIX access control lists */ 200 OCFS2_MOUNT_USRQUOTA = 1 << 9, /* We support user quotas */ 201 OCFS2_MOUNT_GRPQUOTA = 1 << 10, /* We support group quotas */ 202 }; 203 204 #define OCFS2_OSB_SOFT_RO 0x0001 205 #define OCFS2_OSB_HARD_RO 0x0002 206 #define OCFS2_OSB_ERROR_FS 0x0004 207 #define OCFS2_DEFAULT_ATIME_QUANTUM 60 208 209 struct ocfs2_journal; 210 struct ocfs2_slot_info; 211 struct ocfs2_recovery_map; 212 struct ocfs2_quota_recovery; 213 struct ocfs2_super 214 { 215 struct task_struct *commit_task; 216 struct super_block *sb; 217 struct inode *root_inode; 218 struct inode *sys_root_inode; 219 struct inode *system_inodes[NUM_SYSTEM_INODES]; 220 221 struct ocfs2_slot_info *slot_info; 222 223 u32 *slot_recovery_generations; 224 225 spinlock_t node_map_lock; 226 227 u64 root_blkno; 228 u64 system_dir_blkno; 229 u64 bitmap_blkno; 230 u32 bitmap_cpg; 231 u8 *uuid; 232 char *uuid_str; 233 u32 uuid_hash; 234 u8 *vol_label; 235 u64 first_cluster_group_blkno; 236 u32 fs_generation; 237 238 u32 s_feature_compat; 239 u32 s_feature_incompat; 240 u32 s_feature_ro_compat; 241 242 /* Protects s_next_generation, osb_flags and s_inode_steal_slot. 243 * Could protect more on osb as it's very short lived. 244 */ 245 spinlock_t osb_lock; 246 u32 s_next_generation; 247 unsigned long osb_flags; 248 s16 s_inode_steal_slot; 249 atomic_t s_num_inodes_stolen; 250 251 unsigned long s_mount_opt; 252 unsigned int s_atime_quantum; 253 254 unsigned int max_slots; 255 unsigned int node_num; 256 int slot_num; 257 int preferred_slot; 258 int s_sectsize_bits; 259 int s_clustersize; 260 int s_clustersize_bits; 261 unsigned int s_xattr_inline_size; 262 263 atomic_t vol_state; 264 struct mutex recovery_lock; 265 struct ocfs2_recovery_map *recovery_map; 266 struct task_struct *recovery_thread_task; 267 int disable_recovery; 268 wait_queue_head_t checkpoint_event; 269 atomic_t needs_checkpoint; 270 struct ocfs2_journal *journal; 271 unsigned long osb_commit_interval; 272 273 struct delayed_work la_enable_wq; 274 275 /* 276 * Must hold local alloc i_mutex and osb->osb_lock to change 277 * local_alloc_bits. Reads can be done under either lock. 278 */ 279 unsigned int local_alloc_bits; 280 unsigned int local_alloc_default_bits; 281 282 enum ocfs2_local_alloc_state local_alloc_state; /* protected 283 * by osb_lock */ 284 285 struct buffer_head *local_alloc_bh; 286 287 u64 la_last_gd; 288 289 #ifdef CONFIG_OCFS2_FS_STATS 290 struct dentry *local_alloc_debug; 291 char *local_alloc_debug_buf; 292 #endif 293 294 /* Next three fields are for local node slot recovery during 295 * mount. */ 296 int dirty; 297 struct ocfs2_dinode *local_alloc_copy; 298 struct ocfs2_quota_recovery *quota_rec; 299 300 struct ocfs2_alloc_stats alloc_stats; 301 char dev_str[20]; /* "major,minor" of the device */ 302 303 char osb_cluster_stack[OCFS2_STACK_LABEL_LEN + 1]; 304 struct ocfs2_cluster_connection *cconn; 305 struct ocfs2_lock_res osb_super_lockres; 306 struct ocfs2_lock_res osb_rename_lockres; 307 struct ocfs2_dlm_debug *osb_dlm_debug; 308 309 struct dentry *osb_debug_root; 310 311 wait_queue_head_t recovery_event; 312 313 spinlock_t dc_task_lock; 314 struct task_struct *dc_task; 315 wait_queue_head_t dc_event; 316 unsigned long dc_wake_sequence; 317 unsigned long dc_work_sequence; 318 319 /* 320 * Any thread can add locks to the list, but the downconvert 321 * thread is the only one allowed to remove locks. Any change 322 * to this rule requires updating 323 * ocfs2_downconvert_thread_do_work(). 324 */ 325 struct list_head blocked_lock_list; 326 unsigned long blocked_lock_count; 327 328 wait_queue_head_t osb_mount_event; 329 330 /* Truncate log info */ 331 struct inode *osb_tl_inode; 332 struct buffer_head *osb_tl_bh; 333 struct delayed_work osb_truncate_log_wq; 334 335 struct ocfs2_node_map osb_recovering_orphan_dirs; 336 unsigned int *osb_orphan_wipes; 337 wait_queue_head_t osb_wipe_event; 338 }; 339 340 #define OCFS2_SB(sb) ((struct ocfs2_super *)(sb)->s_fs_info) 341 342 static inline int ocfs2_should_order_data(struct inode *inode) 343 { 344 if (!S_ISREG(inode->i_mode)) 345 return 0; 346 if (OCFS2_SB(inode->i_sb)->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) 347 return 0; 348 return 1; 349 } 350 351 static inline int ocfs2_sparse_alloc(struct ocfs2_super *osb) 352 { 353 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC) 354 return 1; 355 return 0; 356 } 357 358 static inline int ocfs2_writes_unwritten_extents(struct ocfs2_super *osb) 359 { 360 /* 361 * Support for sparse files is a pre-requisite 362 */ 363 if (!ocfs2_sparse_alloc(osb)) 364 return 0; 365 366 if (osb->s_feature_ro_compat & OCFS2_FEATURE_RO_COMPAT_UNWRITTEN) 367 return 1; 368 return 0; 369 } 370 371 static inline int ocfs2_supports_inline_data(struct ocfs2_super *osb) 372 { 373 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_INLINE_DATA) 374 return 1; 375 return 0; 376 } 377 378 static inline int ocfs2_supports_xattr(struct ocfs2_super *osb) 379 { 380 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_XATTR) 381 return 1; 382 return 0; 383 } 384 385 static inline int ocfs2_meta_ecc(struct ocfs2_super *osb) 386 { 387 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_META_ECC) 388 return 1; 389 return 0; 390 } 391 392 /* set / clear functions because cluster events can make these happen 393 * in parallel so we want the transitions to be atomic. this also 394 * means that any future flags osb_flags must be protected by spinlock 395 * too! */ 396 static inline void ocfs2_set_osb_flag(struct ocfs2_super *osb, 397 unsigned long flag) 398 { 399 spin_lock(&osb->osb_lock); 400 osb->osb_flags |= flag; 401 spin_unlock(&osb->osb_lock); 402 } 403 404 static inline void ocfs2_set_ro_flag(struct ocfs2_super *osb, 405 int hard) 406 { 407 spin_lock(&osb->osb_lock); 408 osb->osb_flags &= ~(OCFS2_OSB_SOFT_RO|OCFS2_OSB_HARD_RO); 409 if (hard) 410 osb->osb_flags |= OCFS2_OSB_HARD_RO; 411 else 412 osb->osb_flags |= OCFS2_OSB_SOFT_RO; 413 spin_unlock(&osb->osb_lock); 414 } 415 416 static inline int ocfs2_is_hard_readonly(struct ocfs2_super *osb) 417 { 418 int ret; 419 420 spin_lock(&osb->osb_lock); 421 ret = osb->osb_flags & OCFS2_OSB_HARD_RO; 422 spin_unlock(&osb->osb_lock); 423 424 return ret; 425 } 426 427 static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb) 428 { 429 int ret; 430 431 spin_lock(&osb->osb_lock); 432 ret = osb->osb_flags & OCFS2_OSB_SOFT_RO; 433 spin_unlock(&osb->osb_lock); 434 435 return ret; 436 } 437 438 static inline int ocfs2_userspace_stack(struct ocfs2_super *osb) 439 { 440 return (osb->s_feature_incompat & 441 OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK); 442 } 443 444 static inline int ocfs2_mount_local(struct ocfs2_super *osb) 445 { 446 return (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT); 447 } 448 449 static inline int ocfs2_uses_extended_slot_map(struct ocfs2_super *osb) 450 { 451 return (osb->s_feature_incompat & 452 OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP); 453 } 454 455 456 #define OCFS2_IS_VALID_DINODE(ptr) \ 457 (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE)) 458 459 #define OCFS2_IS_VALID_EXTENT_BLOCK(ptr) \ 460 (!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE)) 461 462 #define OCFS2_IS_VALID_GROUP_DESC(ptr) \ 463 (!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE)) 464 465 466 #define OCFS2_IS_VALID_XATTR_BLOCK(ptr) \ 467 (!strcmp((ptr)->xb_signature, OCFS2_XATTR_BLOCK_SIGNATURE)) 468 469 static inline unsigned long ino_from_blkno(struct super_block *sb, 470 u64 blkno) 471 { 472 return (unsigned long)(blkno & (u64)ULONG_MAX); 473 } 474 475 static inline u64 ocfs2_clusters_to_blocks(struct super_block *sb, 476 u32 clusters) 477 { 478 int c_to_b_bits = OCFS2_SB(sb)->s_clustersize_bits - 479 sb->s_blocksize_bits; 480 481 return (u64)clusters << c_to_b_bits; 482 } 483 484 static inline u32 ocfs2_blocks_to_clusters(struct super_block *sb, 485 u64 blocks) 486 { 487 int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits - 488 sb->s_blocksize_bits; 489 490 return (u32)(blocks >> b_to_c_bits); 491 } 492 493 static inline unsigned int ocfs2_clusters_for_bytes(struct super_block *sb, 494 u64 bytes) 495 { 496 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits; 497 unsigned int clusters; 498 499 bytes += OCFS2_SB(sb)->s_clustersize - 1; 500 /* OCFS2 just cannot have enough clusters to overflow this */ 501 clusters = (unsigned int)(bytes >> cl_bits); 502 503 return clusters; 504 } 505 506 static inline u64 ocfs2_blocks_for_bytes(struct super_block *sb, 507 u64 bytes) 508 { 509 bytes += sb->s_blocksize - 1; 510 return bytes >> sb->s_blocksize_bits; 511 } 512 513 static inline u64 ocfs2_clusters_to_bytes(struct super_block *sb, 514 u32 clusters) 515 { 516 return (u64)clusters << OCFS2_SB(sb)->s_clustersize_bits; 517 } 518 519 static inline u64 ocfs2_align_bytes_to_clusters(struct super_block *sb, 520 u64 bytes) 521 { 522 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits; 523 unsigned int clusters; 524 525 clusters = ocfs2_clusters_for_bytes(sb, bytes); 526 return (u64)clusters << cl_bits; 527 } 528 529 static inline u64 ocfs2_align_bytes_to_blocks(struct super_block *sb, 530 u64 bytes) 531 { 532 u64 blocks; 533 534 blocks = ocfs2_blocks_for_bytes(sb, bytes); 535 return blocks << sb->s_blocksize_bits; 536 } 537 538 static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes) 539 { 540 return (unsigned long)((bytes + 511) >> 9); 541 } 542 543 static inline unsigned int ocfs2_page_index_to_clusters(struct super_block *sb, 544 unsigned long pg_index) 545 { 546 u32 clusters = pg_index; 547 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits; 548 549 if (unlikely(PAGE_CACHE_SHIFT > cbits)) 550 clusters = pg_index << (PAGE_CACHE_SHIFT - cbits); 551 else if (PAGE_CACHE_SHIFT < cbits) 552 clusters = pg_index >> (cbits - PAGE_CACHE_SHIFT); 553 554 return clusters; 555 } 556 557 /* 558 * Find the 1st page index which covers the given clusters. 559 */ 560 static inline pgoff_t ocfs2_align_clusters_to_page_index(struct super_block *sb, 561 u32 clusters) 562 { 563 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits; 564 pgoff_t index = clusters; 565 566 if (PAGE_CACHE_SHIFT > cbits) { 567 index = (pgoff_t)clusters >> (PAGE_CACHE_SHIFT - cbits); 568 } else if (PAGE_CACHE_SHIFT < cbits) { 569 index = (pgoff_t)clusters << (cbits - PAGE_CACHE_SHIFT); 570 } 571 572 return index; 573 } 574 575 static inline unsigned int ocfs2_pages_per_cluster(struct super_block *sb) 576 { 577 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits; 578 unsigned int pages_per_cluster = 1; 579 580 if (PAGE_CACHE_SHIFT < cbits) 581 pages_per_cluster = 1 << (cbits - PAGE_CACHE_SHIFT); 582 583 return pages_per_cluster; 584 } 585 586 static inline unsigned int ocfs2_megabytes_to_clusters(struct super_block *sb, 587 unsigned int megs) 588 { 589 BUILD_BUG_ON(OCFS2_MAX_CLUSTERSIZE > 1048576); 590 591 return megs << (20 - OCFS2_SB(sb)->s_clustersize_bits); 592 } 593 594 static inline void ocfs2_init_inode_steal_slot(struct ocfs2_super *osb) 595 { 596 spin_lock(&osb->osb_lock); 597 osb->s_inode_steal_slot = OCFS2_INVALID_SLOT; 598 spin_unlock(&osb->osb_lock); 599 atomic_set(&osb->s_num_inodes_stolen, 0); 600 } 601 602 static inline void ocfs2_set_inode_steal_slot(struct ocfs2_super *osb, 603 s16 slot) 604 { 605 spin_lock(&osb->osb_lock); 606 osb->s_inode_steal_slot = slot; 607 spin_unlock(&osb->osb_lock); 608 } 609 610 static inline s16 ocfs2_get_inode_steal_slot(struct ocfs2_super *osb) 611 { 612 s16 slot; 613 614 spin_lock(&osb->osb_lock); 615 slot = osb->s_inode_steal_slot; 616 spin_unlock(&osb->osb_lock); 617 618 return slot; 619 } 620 621 #define ocfs2_set_bit ext2_set_bit 622 #define ocfs2_clear_bit ext2_clear_bit 623 #define ocfs2_test_bit ext2_test_bit 624 #define ocfs2_find_next_zero_bit ext2_find_next_zero_bit 625 #define ocfs2_find_next_bit ext2_find_next_bit 626 #endif /* OCFS2_H */ 627 628