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) /* have we 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_DISMOUNTED, 165 VOLUME_DISABLED 166 }; 167 168 struct ocfs2_alloc_stats 169 { 170 atomic_t moves; 171 atomic_t local_data; 172 atomic_t bitmap_data; 173 atomic_t bg_allocs; 174 atomic_t bg_extends; 175 }; 176 177 enum ocfs2_local_alloc_state 178 { 179 OCFS2_LA_UNUSED = 0, /* Local alloc will never be used for 180 * this mountpoint. */ 181 OCFS2_LA_ENABLED, /* Local alloc is in use. */ 182 OCFS2_LA_THROTTLED, /* Local alloc is in use, but number 183 * of bits has been reduced. */ 184 OCFS2_LA_DISABLED /* Local alloc has temporarily been 185 * disabled. */ 186 }; 187 188 enum ocfs2_mount_options 189 { 190 OCFS2_MOUNT_HB_LOCAL = 1 << 0, /* Heartbeat started in local mode */ 191 OCFS2_MOUNT_BARRIER = 1 << 1, /* Use block barriers */ 192 OCFS2_MOUNT_NOINTR = 1 << 2, /* Don't catch signals */ 193 OCFS2_MOUNT_ERRORS_PANIC = 1 << 3, /* Panic on errors */ 194 OCFS2_MOUNT_DATA_WRITEBACK = 1 << 4, /* No data ordering */ 195 OCFS2_MOUNT_LOCALFLOCKS = 1 << 5, /* No cluster aware user file locks */ 196 OCFS2_MOUNT_NOUSERXATTR = 1 << 6, /* No user xattr */ 197 OCFS2_MOUNT_INODE64 = 1 << 7, /* Allow inode numbers > 2^32 */ 198 }; 199 200 #define OCFS2_OSB_SOFT_RO 0x0001 201 #define OCFS2_OSB_HARD_RO 0x0002 202 #define OCFS2_OSB_ERROR_FS 0x0004 203 #define OCFS2_DEFAULT_ATIME_QUANTUM 60 204 205 struct ocfs2_journal; 206 struct ocfs2_slot_info; 207 struct ocfs2_recovery_map; 208 struct ocfs2_super 209 { 210 struct task_struct *commit_task; 211 struct super_block *sb; 212 struct inode *root_inode; 213 struct inode *sys_root_inode; 214 struct inode *system_inodes[NUM_SYSTEM_INODES]; 215 216 struct ocfs2_slot_info *slot_info; 217 218 u32 *slot_recovery_generations; 219 220 spinlock_t node_map_lock; 221 222 u64 root_blkno; 223 u64 system_dir_blkno; 224 u64 bitmap_blkno; 225 u32 bitmap_cpg; 226 u8 *uuid; 227 char *uuid_str; 228 u32 uuid_hash; 229 u8 *vol_label; 230 u64 first_cluster_group_blkno; 231 u32 fs_generation; 232 233 u32 s_feature_compat; 234 u32 s_feature_incompat; 235 u32 s_feature_ro_compat; 236 237 /* Protects s_next_generation, osb_flags and s_inode_steal_slot. 238 * Could protect more on osb as it's very short lived. 239 */ 240 spinlock_t osb_lock; 241 u32 s_next_generation; 242 unsigned long osb_flags; 243 s16 s_inode_steal_slot; 244 atomic_t s_num_inodes_stolen; 245 246 unsigned long s_mount_opt; 247 unsigned int s_atime_quantum; 248 249 unsigned int max_slots; 250 unsigned int node_num; 251 int slot_num; 252 int preferred_slot; 253 int s_sectsize_bits; 254 int s_clustersize; 255 int s_clustersize_bits; 256 unsigned int s_xattr_inline_size; 257 258 atomic_t vol_state; 259 struct mutex recovery_lock; 260 struct ocfs2_recovery_map *recovery_map; 261 struct task_struct *recovery_thread_task; 262 int disable_recovery; 263 wait_queue_head_t checkpoint_event; 264 atomic_t needs_checkpoint; 265 struct ocfs2_journal *journal; 266 unsigned long osb_commit_interval; 267 268 struct delayed_work la_enable_wq; 269 270 /* 271 * Must hold local alloc i_mutex and osb->osb_lock to change 272 * local_alloc_bits. Reads can be done under either lock. 273 */ 274 unsigned int local_alloc_bits; 275 unsigned int local_alloc_default_bits; 276 277 enum ocfs2_local_alloc_state local_alloc_state; /* protected 278 * by osb_lock */ 279 280 struct buffer_head *local_alloc_bh; 281 282 u64 la_last_gd; 283 284 #ifdef CONFIG_OCFS2_FS_STATS 285 struct dentry *local_alloc_debug; 286 char *local_alloc_debug_buf; 287 #endif 288 289 /* Next two fields are for local node slot recovery during 290 * mount. */ 291 int dirty; 292 struct ocfs2_dinode *local_alloc_copy; 293 294 struct ocfs2_alloc_stats alloc_stats; 295 char dev_str[20]; /* "major,minor" of the device */ 296 297 char osb_cluster_stack[OCFS2_STACK_LABEL_LEN + 1]; 298 struct ocfs2_cluster_connection *cconn; 299 struct ocfs2_lock_res osb_super_lockres; 300 struct ocfs2_lock_res osb_rename_lockres; 301 struct ocfs2_dlm_debug *osb_dlm_debug; 302 303 struct dentry *osb_debug_root; 304 305 wait_queue_head_t recovery_event; 306 307 spinlock_t dc_task_lock; 308 struct task_struct *dc_task; 309 wait_queue_head_t dc_event; 310 unsigned long dc_wake_sequence; 311 unsigned long dc_work_sequence; 312 313 /* 314 * Any thread can add locks to the list, but the downconvert 315 * thread is the only one allowed to remove locks. Any change 316 * to this rule requires updating 317 * ocfs2_downconvert_thread_do_work(). 318 */ 319 struct list_head blocked_lock_list; 320 unsigned long blocked_lock_count; 321 322 wait_queue_head_t osb_mount_event; 323 324 /* Truncate log info */ 325 struct inode *osb_tl_inode; 326 struct buffer_head *osb_tl_bh; 327 struct delayed_work osb_truncate_log_wq; 328 329 struct ocfs2_node_map osb_recovering_orphan_dirs; 330 unsigned int *osb_orphan_wipes; 331 wait_queue_head_t osb_wipe_event; 332 }; 333 334 #define OCFS2_SB(sb) ((struct ocfs2_super *)(sb)->s_fs_info) 335 336 static inline int ocfs2_should_order_data(struct inode *inode) 337 { 338 if (!S_ISREG(inode->i_mode)) 339 return 0; 340 if (OCFS2_SB(inode->i_sb)->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) 341 return 0; 342 return 1; 343 } 344 345 static inline int ocfs2_sparse_alloc(struct ocfs2_super *osb) 346 { 347 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC) 348 return 1; 349 return 0; 350 } 351 352 static inline int ocfs2_writes_unwritten_extents(struct ocfs2_super *osb) 353 { 354 /* 355 * Support for sparse files is a pre-requisite 356 */ 357 if (!ocfs2_sparse_alloc(osb)) 358 return 0; 359 360 if (osb->s_feature_ro_compat & OCFS2_FEATURE_RO_COMPAT_UNWRITTEN) 361 return 1; 362 return 0; 363 } 364 365 static inline int ocfs2_supports_inline_data(struct ocfs2_super *osb) 366 { 367 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_INLINE_DATA) 368 return 1; 369 return 0; 370 } 371 372 static inline int ocfs2_supports_xattr(struct ocfs2_super *osb) 373 { 374 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_XATTR) 375 return 1; 376 return 0; 377 } 378 379 /* set / clear functions because cluster events can make these happen 380 * in parallel so we want the transitions to be atomic. this also 381 * means that any future flags osb_flags must be protected by spinlock 382 * too! */ 383 static inline void ocfs2_set_osb_flag(struct ocfs2_super *osb, 384 unsigned long flag) 385 { 386 spin_lock(&osb->osb_lock); 387 osb->osb_flags |= flag; 388 spin_unlock(&osb->osb_lock); 389 } 390 391 static inline void ocfs2_set_ro_flag(struct ocfs2_super *osb, 392 int hard) 393 { 394 spin_lock(&osb->osb_lock); 395 osb->osb_flags &= ~(OCFS2_OSB_SOFT_RO|OCFS2_OSB_HARD_RO); 396 if (hard) 397 osb->osb_flags |= OCFS2_OSB_HARD_RO; 398 else 399 osb->osb_flags |= OCFS2_OSB_SOFT_RO; 400 spin_unlock(&osb->osb_lock); 401 } 402 403 static inline int ocfs2_is_hard_readonly(struct ocfs2_super *osb) 404 { 405 int ret; 406 407 spin_lock(&osb->osb_lock); 408 ret = osb->osb_flags & OCFS2_OSB_HARD_RO; 409 spin_unlock(&osb->osb_lock); 410 411 return ret; 412 } 413 414 static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb) 415 { 416 int ret; 417 418 spin_lock(&osb->osb_lock); 419 ret = osb->osb_flags & OCFS2_OSB_SOFT_RO; 420 spin_unlock(&osb->osb_lock); 421 422 return ret; 423 } 424 425 static inline int ocfs2_userspace_stack(struct ocfs2_super *osb) 426 { 427 return (osb->s_feature_incompat & 428 OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK); 429 } 430 431 static inline int ocfs2_mount_local(struct ocfs2_super *osb) 432 { 433 return (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT); 434 } 435 436 static inline int ocfs2_uses_extended_slot_map(struct ocfs2_super *osb) 437 { 438 return (osb->s_feature_incompat & 439 OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP); 440 } 441 442 443 #define OCFS2_IS_VALID_DINODE(ptr) \ 444 (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE)) 445 446 #define OCFS2_RO_ON_INVALID_DINODE(__sb, __di) do { \ 447 typeof(__di) ____di = (__di); \ 448 ocfs2_error((__sb), \ 449 "Dinode # %llu has bad signature %.*s", \ 450 (unsigned long long)le64_to_cpu((____di)->i_blkno), 7, \ 451 (____di)->i_signature); \ 452 } while (0) 453 454 #define OCFS2_IS_VALID_EXTENT_BLOCK(ptr) \ 455 (!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE)) 456 457 #define OCFS2_RO_ON_INVALID_EXTENT_BLOCK(__sb, __eb) do { \ 458 typeof(__eb) ____eb = (__eb); \ 459 ocfs2_error((__sb), \ 460 "Extent Block # %llu has bad signature %.*s", \ 461 (unsigned long long)le64_to_cpu((____eb)->h_blkno), 7, \ 462 (____eb)->h_signature); \ 463 } while (0) 464 465 #define OCFS2_IS_VALID_GROUP_DESC(ptr) \ 466 (!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE)) 467 468 #define OCFS2_RO_ON_INVALID_GROUP_DESC(__sb, __gd) do { \ 469 typeof(__gd) ____gd = (__gd); \ 470 ocfs2_error((__sb), \ 471 "Group Descriptor # %llu has bad signature %.*s", \ 472 (unsigned long long)le64_to_cpu((____gd)->bg_blkno), 7, \ 473 (____gd)->bg_signature); \ 474 } while (0) 475 476 static inline unsigned long ino_from_blkno(struct super_block *sb, 477 u64 blkno) 478 { 479 return (unsigned long)(blkno & (u64)ULONG_MAX); 480 } 481 482 static inline u64 ocfs2_clusters_to_blocks(struct super_block *sb, 483 u32 clusters) 484 { 485 int c_to_b_bits = OCFS2_SB(sb)->s_clustersize_bits - 486 sb->s_blocksize_bits; 487 488 return (u64)clusters << c_to_b_bits; 489 } 490 491 static inline u32 ocfs2_blocks_to_clusters(struct super_block *sb, 492 u64 blocks) 493 { 494 int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits - 495 sb->s_blocksize_bits; 496 497 return (u32)(blocks >> b_to_c_bits); 498 } 499 500 static inline unsigned int ocfs2_clusters_for_bytes(struct super_block *sb, 501 u64 bytes) 502 { 503 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits; 504 unsigned int clusters; 505 506 bytes += OCFS2_SB(sb)->s_clustersize - 1; 507 /* OCFS2 just cannot have enough clusters to overflow this */ 508 clusters = (unsigned int)(bytes >> cl_bits); 509 510 return clusters; 511 } 512 513 static inline u64 ocfs2_blocks_for_bytes(struct super_block *sb, 514 u64 bytes) 515 { 516 bytes += sb->s_blocksize - 1; 517 return bytes >> sb->s_blocksize_bits; 518 } 519 520 static inline u64 ocfs2_clusters_to_bytes(struct super_block *sb, 521 u32 clusters) 522 { 523 return (u64)clusters << OCFS2_SB(sb)->s_clustersize_bits; 524 } 525 526 static inline u64 ocfs2_align_bytes_to_clusters(struct super_block *sb, 527 u64 bytes) 528 { 529 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits; 530 unsigned int clusters; 531 532 clusters = ocfs2_clusters_for_bytes(sb, bytes); 533 return (u64)clusters << cl_bits; 534 } 535 536 static inline u64 ocfs2_align_bytes_to_blocks(struct super_block *sb, 537 u64 bytes) 538 { 539 u64 blocks; 540 541 blocks = ocfs2_blocks_for_bytes(sb, bytes); 542 return blocks << sb->s_blocksize_bits; 543 } 544 545 static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes) 546 { 547 return (unsigned long)((bytes + 511) >> 9); 548 } 549 550 static inline unsigned int ocfs2_page_index_to_clusters(struct super_block *sb, 551 unsigned long pg_index) 552 { 553 u32 clusters = pg_index; 554 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits; 555 556 if (unlikely(PAGE_CACHE_SHIFT > cbits)) 557 clusters = pg_index << (PAGE_CACHE_SHIFT - cbits); 558 else if (PAGE_CACHE_SHIFT < cbits) 559 clusters = pg_index >> (cbits - PAGE_CACHE_SHIFT); 560 561 return clusters; 562 } 563 564 /* 565 * Find the 1st page index which covers the given clusters. 566 */ 567 static inline pgoff_t ocfs2_align_clusters_to_page_index(struct super_block *sb, 568 u32 clusters) 569 { 570 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits; 571 pgoff_t index = clusters; 572 573 if (PAGE_CACHE_SHIFT > cbits) { 574 index = (pgoff_t)clusters >> (PAGE_CACHE_SHIFT - cbits); 575 } else if (PAGE_CACHE_SHIFT < cbits) { 576 index = (pgoff_t)clusters << (cbits - PAGE_CACHE_SHIFT); 577 } 578 579 return index; 580 } 581 582 static inline unsigned int ocfs2_pages_per_cluster(struct super_block *sb) 583 { 584 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits; 585 unsigned int pages_per_cluster = 1; 586 587 if (PAGE_CACHE_SHIFT < cbits) 588 pages_per_cluster = 1 << (cbits - PAGE_CACHE_SHIFT); 589 590 return pages_per_cluster; 591 } 592 593 static inline unsigned int ocfs2_megabytes_to_clusters(struct super_block *sb, 594 unsigned int megs) 595 { 596 BUILD_BUG_ON(OCFS2_MAX_CLUSTERSIZE > 1048576); 597 598 return megs << (20 - OCFS2_SB(sb)->s_clustersize_bits); 599 } 600 601 static inline void ocfs2_init_inode_steal_slot(struct ocfs2_super *osb) 602 { 603 spin_lock(&osb->osb_lock); 604 osb->s_inode_steal_slot = OCFS2_INVALID_SLOT; 605 spin_unlock(&osb->osb_lock); 606 atomic_set(&osb->s_num_inodes_stolen, 0); 607 } 608 609 static inline void ocfs2_set_inode_steal_slot(struct ocfs2_super *osb, 610 s16 slot) 611 { 612 spin_lock(&osb->osb_lock); 613 osb->s_inode_steal_slot = slot; 614 spin_unlock(&osb->osb_lock); 615 } 616 617 static inline s16 ocfs2_get_inode_steal_slot(struct ocfs2_super *osb) 618 { 619 s16 slot; 620 621 spin_lock(&osb->osb_lock); 622 slot = osb->s_inode_steal_slot; 623 spin_unlock(&osb->osb_lock); 624 625 return slot; 626 } 627 628 #define ocfs2_set_bit ext2_set_bit 629 #define ocfs2_clear_bit ext2_clear_bit 630 #define ocfs2_test_bit ext2_test_bit 631 #define ocfs2_find_next_zero_bit ext2_find_next_zero_bit 632 #endif /* OCFS2_H */ 633 634