| 1 | /* SPDX-License-Identifier: GPL-2.0-or-later */ |
| 2 | /* |
| 3 | * ocfs2.h |
| 4 | * |
| 5 | * Defines macros and structures used in OCFS2 |
| 6 | * |
| 7 | * Copyright (C) 2002, 2004 Oracle. All rights reserved. |
| 8 | */ |
| 9 | |
| 10 | #ifndef OCFS2_H |
| 11 | #define OCFS2_H |
| 12 | |
| 13 | #include <linux/spinlock.h> |
| 14 | #include <linux/sched.h> |
| 15 | #include <linux/wait.h> |
| 16 | #include <linux/list.h> |
| 17 | #include <linux/llist.h> |
| 18 | #include <linux/rbtree.h> |
| 19 | #include <linux/workqueue.h> |
| 20 | #include <linux/kref.h> |
| 21 | #include <linux/mutex.h> |
| 22 | #include <linux/lockdep.h> |
| 23 | #include <linux/jbd2.h> |
| 24 | |
| 25 | /* For union ocfs2_dlm_lksb */ |
| 26 | #include "stackglue.h" |
| 27 | |
| 28 | #include "ocfs2_fs.h" |
| 29 | #include "ocfs2_lockid.h" |
| 30 | #include "ocfs2_ioctl.h" |
| 31 | |
| 32 | /* For struct ocfs2_blockcheck_stats */ |
| 33 | #include "blockcheck.h" |
| 34 | |
| 35 | #include "reservations.h" |
| 36 | |
| 37 | #include "filecheck.h" |
| 38 | |
| 39 | /* Caching of metadata buffers */ |
| 40 | |
| 41 | /* Most user visible OCFS2 inodes will have very few pieces of |
| 42 | * metadata, but larger files (including bitmaps, etc) must be taken |
| 43 | * into account when designing an access scheme. We allow a small |
| 44 | * amount of inlined blocks to be stored on an array and grow the |
| 45 | * structure into a rb tree when necessary. */ |
| 46 | #define OCFS2_CACHE_INFO_MAX_ARRAY 2 |
| 47 | |
| 48 | /* Flags for ocfs2_caching_info */ |
| 49 | |
| 50 | enum ocfs2_caching_info_flags { |
| 51 | /* Indicates that the metadata cache is using the inline array */ |
| 52 | OCFS2_CACHE_FL_INLINE = 1<<1, |
| 53 | }; |
| 54 | |
| 55 | struct ocfs2_caching_operations; |
| 56 | struct ocfs2_caching_info { |
| 57 | /* |
| 58 | * The parent structure provides the locks, but because the |
| 59 | * parent structure can differ, it provides locking operations |
| 60 | * to struct ocfs2_caching_info. |
| 61 | */ |
| 62 | const struct ocfs2_caching_operations *ci_ops; |
| 63 | |
| 64 | /* next two are protected by trans_inc_lock */ |
| 65 | /* which transaction were we created on? Zero if none. */ |
| 66 | unsigned long ci_created_trans; |
| 67 | /* last transaction we were a part of. */ |
| 68 | unsigned long ci_last_trans; |
| 69 | |
| 70 | /* Cache structures */ |
| 71 | unsigned int ci_flags; |
| 72 | unsigned int ci_num_cached; |
| 73 | union { |
| 74 | sector_t ci_array[OCFS2_CACHE_INFO_MAX_ARRAY]; |
| 75 | struct rb_root ci_tree; |
| 76 | } ci_cache; |
| 77 | }; |
| 78 | /* |
| 79 | * Need this prototype here instead of in uptodate.h because journal.h |
| 80 | * uses it. |
| 81 | */ |
| 82 | struct super_block *ocfs2_metadata_cache_get_super(struct ocfs2_caching_info *ci); |
| 83 | |
| 84 | /* this limits us to 256 nodes |
| 85 | * if we need more, we can do a kmalloc for the map */ |
| 86 | #define OCFS2_NODE_MAP_MAX_NODES 256 |
| 87 | struct ocfs2_node_map { |
| 88 | u16 num_nodes; |
| 89 | unsigned long map[BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES)]; |
| 90 | }; |
| 91 | |
| 92 | enum ocfs2_ast_action { |
| 93 | OCFS2_AST_INVALID = 0, |
| 94 | OCFS2_AST_ATTACH, |
| 95 | OCFS2_AST_CONVERT, |
| 96 | OCFS2_AST_DOWNCONVERT, |
| 97 | }; |
| 98 | |
| 99 | /* actions for an unlockast function to take. */ |
| 100 | enum ocfs2_unlock_action { |
| 101 | OCFS2_UNLOCK_INVALID = 0, |
| 102 | OCFS2_UNLOCK_CANCEL_CONVERT, |
| 103 | OCFS2_UNLOCK_DROP_LOCK, |
| 104 | }; |
| 105 | |
| 106 | /* ocfs2_lock_res->l_flags flags. */ |
| 107 | #define OCFS2_LOCK_ATTACHED (0x00000001) /* we have initialized |
| 108 | * the lvb */ |
| 109 | #define OCFS2_LOCK_BUSY (0x00000002) /* we are currently in |
| 110 | * dlm_lock */ |
| 111 | #define OCFS2_LOCK_BLOCKED (0x00000004) /* blocked waiting to |
| 112 | * downconvert*/ |
| 113 | #define OCFS2_LOCK_LOCAL (0x00000008) /* newly created inode */ |
| 114 | #define OCFS2_LOCK_NEEDS_REFRESH (0x00000010) |
| 115 | #define OCFS2_LOCK_REFRESHING (0x00000020) |
| 116 | #define OCFS2_LOCK_INITIALIZED (0x00000040) /* track initialization |
| 117 | * for shutdown paths */ |
| 118 | #define OCFS2_LOCK_FREEING (0x00000080) /* help dlmglue track |
| 119 | * when to skip queueing |
| 120 | * a lock because it's |
| 121 | * about to be |
| 122 | * dropped. */ |
| 123 | #define OCFS2_LOCK_QUEUED (0x00000100) /* queued for downconvert */ |
| 124 | #define OCFS2_LOCK_NOCACHE (0x00000200) /* don't use a holder count */ |
| 125 | #define OCFS2_LOCK_PENDING (0x00000400) /* This lockres is pending a |
| 126 | call to dlm_lock. Only |
| 127 | exists with BUSY set. */ |
| 128 | #define OCFS2_LOCK_UPCONVERT_FINISHING (0x00000800) /* blocks the dc thread |
| 129 | * from downconverting |
| 130 | * before the upconvert |
| 131 | * has completed */ |
| 132 | |
| 133 | #define OCFS2_LOCK_NONBLOCK_FINISHED (0x00001000) /* NONBLOCK cluster |
| 134 | * lock has already |
| 135 | * returned, do not block |
| 136 | * dc thread from |
| 137 | * downconverting */ |
| 138 | |
| 139 | struct ocfs2_lock_res_ops; |
| 140 | |
| 141 | typedef void (*ocfs2_lock_callback)(int status, unsigned long data); |
| 142 | |
| 143 | #ifdef CONFIG_OCFS2_FS_STATS |
| 144 | struct ocfs2_lock_stats { |
| 145 | u64 ls_total; /* Total wait in NSEC */ |
| 146 | u32 ls_gets; /* Num acquires */ |
| 147 | u32 ls_fail; /* Num failed acquires */ |
| 148 | |
| 149 | /* Storing max wait in usecs saves 24 bytes per inode */ |
| 150 | u32 ls_max; /* Max wait in USEC */ |
| 151 | u64 ls_last; /* Last unlock time in USEC */ |
| 152 | }; |
| 153 | #endif |
| 154 | |
| 155 | struct ocfs2_lock_res { |
| 156 | void *l_priv; |
| 157 | const struct ocfs2_lock_res_ops *l_ops; |
| 158 | |
| 159 | |
| 160 | struct list_head l_blocked_list; |
| 161 | struct list_head l_mask_waiters; |
| 162 | struct list_head l_holders; |
| 163 | |
| 164 | unsigned long l_flags; |
| 165 | char l_name[OCFS2_LOCK_ID_MAX_LEN]; |
| 166 | unsigned int l_ro_holders; |
| 167 | unsigned int l_ex_holders; |
| 168 | signed char l_level; |
| 169 | signed char l_requested; |
| 170 | signed char l_blocking; |
| 171 | |
| 172 | /* Data packed - type enum ocfs2_lock_type */ |
| 173 | unsigned char l_type; |
| 174 | |
| 175 | /* used from AST/BAST funcs. */ |
| 176 | /* Data packed - enum type ocfs2_ast_action */ |
| 177 | unsigned char l_action; |
| 178 | /* Data packed - enum type ocfs2_unlock_action */ |
| 179 | unsigned char l_unlock_action; |
| 180 | unsigned int l_pending_gen; |
| 181 | |
| 182 | spinlock_t l_lock; |
| 183 | |
| 184 | struct ocfs2_dlm_lksb l_lksb; |
| 185 | |
| 186 | wait_queue_head_t l_event; |
| 187 | |
| 188 | struct list_head l_debug_list; |
| 189 | |
| 190 | #ifdef CONFIG_OCFS2_FS_STATS |
| 191 | struct ocfs2_lock_stats l_lock_prmode; /* PR mode stats */ |
| 192 | u32 l_lock_refresh; /* Disk refreshes */ |
| 193 | u64 l_lock_wait; /* First lock wait time */ |
| 194 | struct ocfs2_lock_stats l_lock_exmode; /* EX mode stats */ |
| 195 | #endif |
| 196 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
| 197 | struct lockdep_map l_lockdep_map; |
| 198 | #endif |
| 199 | }; |
| 200 | |
| 201 | enum ocfs2_orphan_reco_type { |
| 202 | ORPHAN_NO_NEED_TRUNCATE = 0, |
| 203 | ORPHAN_NEED_TRUNCATE, |
| 204 | }; |
| 205 | |
| 206 | enum ocfs2_orphan_scan_state { |
| 207 | ORPHAN_SCAN_ACTIVE, |
| 208 | ORPHAN_SCAN_INACTIVE |
| 209 | }; |
| 210 | |
| 211 | struct ocfs2_orphan_scan { |
| 212 | struct mutex os_lock; |
| 213 | struct ocfs2_super *os_osb; |
| 214 | struct ocfs2_lock_res os_lockres; /* lock to synchronize scans */ |
| 215 | struct delayed_work os_orphan_scan_work; |
| 216 | time64_t os_scantime; /* time this node ran the scan */ |
| 217 | u32 os_count; /* tracks node specific scans */ |
| 218 | u32 os_seqno; /* tracks cluster wide scans */ |
| 219 | atomic_t os_state; /* ACTIVE or INACTIVE */ |
| 220 | }; |
| 221 | |
| 222 | struct ocfs2_dlm_debug { |
| 223 | struct kref d_refcnt; |
| 224 | u32 d_filter_secs; |
| 225 | struct list_head d_lockres_tracking; |
| 226 | }; |
| 227 | |
| 228 | enum ocfs2_vol_state |
| 229 | { |
| 230 | VOLUME_INIT = 0, |
| 231 | VOLUME_MOUNTED, |
| 232 | VOLUME_MOUNTED_QUOTAS, |
| 233 | VOLUME_DISMOUNTED, |
| 234 | VOLUME_DISABLED |
| 235 | }; |
| 236 | |
| 237 | struct ocfs2_alloc_stats |
| 238 | { |
| 239 | atomic_t moves; |
| 240 | atomic_t local_data; |
| 241 | atomic_t bitmap_data; |
| 242 | atomic_t bg_allocs; |
| 243 | atomic_t bg_extends; |
| 244 | }; |
| 245 | |
| 246 | enum ocfs2_local_alloc_state |
| 247 | { |
| 248 | OCFS2_LA_UNUSED = 0, /* Local alloc will never be used for |
| 249 | * this mountpoint. */ |
| 250 | OCFS2_LA_ENABLED, /* Local alloc is in use. */ |
| 251 | OCFS2_LA_THROTTLED, /* Local alloc is in use, but number |
| 252 | * of bits has been reduced. */ |
| 253 | OCFS2_LA_DISABLED /* Local alloc has temporarily been |
| 254 | * disabled. */ |
| 255 | }; |
| 256 | |
| 257 | enum ocfs2_mount_options |
| 258 | { |
| 259 | OCFS2_MOUNT_HB_LOCAL = 1 << 0, /* Local heartbeat */ |
| 260 | OCFS2_MOUNT_BARRIER = 1 << 1, /* Use block barriers */ |
| 261 | OCFS2_MOUNT_NOINTR = 1 << 2, /* Don't catch signals */ |
| 262 | OCFS2_MOUNT_ERRORS_PANIC = 1 << 3, /* Panic on errors */ |
| 263 | OCFS2_MOUNT_DATA_WRITEBACK = 1 << 4, /* No data ordering */ |
| 264 | OCFS2_MOUNT_LOCALFLOCKS = 1 << 5, /* No cluster aware user file locks */ |
| 265 | OCFS2_MOUNT_NOUSERXATTR = 1 << 6, /* No user xattr */ |
| 266 | OCFS2_MOUNT_INODE64 = 1 << 7, /* Allow inode numbers > 2^32 */ |
| 267 | OCFS2_MOUNT_POSIX_ACL = 1 << 8, /* Force POSIX access control lists */ |
| 268 | OCFS2_MOUNT_NO_POSIX_ACL = 1 << 9, /* Disable POSIX access |
| 269 | control lists */ |
| 270 | OCFS2_MOUNT_USRQUOTA = 1 << 10, /* We support user quotas */ |
| 271 | OCFS2_MOUNT_GRPQUOTA = 1 << 11, /* We support group quotas */ |
| 272 | OCFS2_MOUNT_COHERENCY_BUFFERED = 1 << 12, /* Allow concurrent O_DIRECT |
| 273 | writes */ |
| 274 | OCFS2_MOUNT_HB_NONE = 1 << 13, /* No heartbeat */ |
| 275 | OCFS2_MOUNT_HB_GLOBAL = 1 << 14, /* Global heartbeat */ |
| 276 | |
| 277 | OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT = 1 << 15, /* Journal Async Commit */ |
| 278 | OCFS2_MOUNT_ERRORS_CONT = 1 << 16, /* Return EIO to the calling process on error */ |
| 279 | OCFS2_MOUNT_ERRORS_ROFS = 1 << 17, /* Change filesystem to read-only on error */ |
| 280 | }; |
| 281 | |
| 282 | #define OCFS2_OSB_SOFT_RO 0x0001 |
| 283 | #define OCFS2_OSB_HARD_RO 0x0002 |
| 284 | #define OCFS2_OSB_ERROR_FS 0x0004 |
| 285 | #define OCFS2_DEFAULT_ATIME_QUANTUM 60 |
| 286 | |
| 287 | struct ocfs2_triggers { |
| 288 | struct jbd2_buffer_trigger_type ot_triggers; |
| 289 | int ot_offset; |
| 290 | struct super_block *sb; |
| 291 | }; |
| 292 | |
| 293 | enum ocfs2_journal_trigger_type { |
| 294 | OCFS2_JTR_DI, |
| 295 | OCFS2_JTR_EB, |
| 296 | OCFS2_JTR_RB, |
| 297 | OCFS2_JTR_GD, |
| 298 | OCFS2_JTR_DB, |
| 299 | OCFS2_JTR_XB, |
| 300 | OCFS2_JTR_DQ, |
| 301 | OCFS2_JTR_DR, |
| 302 | OCFS2_JTR_DL, |
| 303 | OCFS2_JTR_NONE /* This must be the last entry */ |
| 304 | }; |
| 305 | |
| 306 | #define OCFS2_JOURNAL_TRIGGER_COUNT OCFS2_JTR_NONE |
| 307 | |
| 308 | void ocfs2_initialize_journal_triggers(struct super_block *sb, |
| 309 | struct ocfs2_triggers triggers[]); |
| 310 | |
| 311 | enum ocfs2_recovery_state { |
| 312 | OCFS2_REC_ENABLED = 0, |
| 313 | OCFS2_REC_QUOTA_WANT_DISABLE, |
| 314 | /* |
| 315 | * Must be OCFS2_REC_QUOTA_WANT_DISABLE + 1 for |
| 316 | * ocfs2_recovery_disable_quota() to work. |
| 317 | */ |
| 318 | OCFS2_REC_QUOTA_DISABLED, |
| 319 | OCFS2_REC_WANT_DISABLE, |
| 320 | /* |
| 321 | * Must be OCFS2_REC_WANT_DISABLE + 1 for ocfs2_recovery_exit() to work |
| 322 | */ |
| 323 | OCFS2_REC_DISABLED, |
| 324 | }; |
| 325 | |
| 326 | struct ocfs2_journal; |
| 327 | struct ocfs2_slot_info; |
| 328 | struct ocfs2_recovery_map; |
| 329 | struct ocfs2_replay_map; |
| 330 | struct ocfs2_quota_recovery; |
| 331 | struct ocfs2_super |
| 332 | { |
| 333 | struct task_struct *commit_task; |
| 334 | struct super_block *sb; |
| 335 | struct inode *root_inode; |
| 336 | struct inode *sys_root_inode; |
| 337 | struct inode *global_system_inodes[NUM_GLOBAL_SYSTEM_INODES]; |
| 338 | struct inode **local_system_inodes; |
| 339 | |
| 340 | struct ocfs2_slot_info *slot_info; |
| 341 | |
| 342 | u32 *slot_recovery_generations; |
| 343 | |
| 344 | spinlock_t node_map_lock; |
| 345 | |
| 346 | u64 root_blkno; |
| 347 | u64 system_dir_blkno; |
| 348 | u64 bitmap_blkno; |
| 349 | u32 bitmap_cpg; |
| 350 | char *uuid_str; |
| 351 | u32 uuid_hash; |
| 352 | u8 *vol_label; |
| 353 | u64 first_cluster_group_blkno; |
| 354 | u32 fs_generation; |
| 355 | |
| 356 | u32 s_feature_compat; |
| 357 | u32 s_feature_incompat; |
| 358 | u32 s_feature_ro_compat; |
| 359 | |
| 360 | /* Protects s_next_generation, osb_flags and s_inode_steal_slot. |
| 361 | * Could protect more on osb as it's very short lived. |
| 362 | */ |
| 363 | spinlock_t osb_lock; |
| 364 | u32 s_next_generation; |
| 365 | unsigned long osb_flags; |
| 366 | u16 s_inode_steal_slot; |
| 367 | u16 s_meta_steal_slot; |
| 368 | atomic_t s_num_inodes_stolen; |
| 369 | atomic_t s_num_meta_stolen; |
| 370 | |
| 371 | unsigned long s_mount_opt; |
| 372 | unsigned int s_atime_quantum; |
| 373 | |
| 374 | unsigned int max_slots; |
| 375 | unsigned int node_num; |
| 376 | int slot_num; |
| 377 | int preferred_slot; |
| 378 | int s_sectsize_bits; |
| 379 | int s_clustersize; |
| 380 | int s_clustersize_bits; |
| 381 | unsigned int s_xattr_inline_size; |
| 382 | |
| 383 | atomic_t vol_state; |
| 384 | struct mutex recovery_lock; |
| 385 | struct ocfs2_recovery_map *recovery_map; |
| 386 | struct ocfs2_replay_map *replay_map; |
| 387 | struct task_struct *recovery_thread_task; |
| 388 | enum ocfs2_recovery_state recovery_state; |
| 389 | wait_queue_head_t checkpoint_event; |
| 390 | struct ocfs2_journal *journal; |
| 391 | unsigned long osb_commit_interval; |
| 392 | |
| 393 | /* Journal triggers for checksum */ |
| 394 | struct ocfs2_triggers s_journal_triggers[OCFS2_JOURNAL_TRIGGER_COUNT]; |
| 395 | |
| 396 | struct delayed_work la_enable_wq; |
| 397 | |
| 398 | /* |
| 399 | * Must hold local alloc i_rwsem and osb->osb_lock to change |
| 400 | * local_alloc_bits. Reads can be done under either lock. |
| 401 | */ |
| 402 | unsigned int local_alloc_bits; |
| 403 | unsigned int local_alloc_default_bits; |
| 404 | /* osb_clusters_at_boot can become stale! Do not trust it to |
| 405 | * be up to date. */ |
| 406 | unsigned int osb_clusters_at_boot; |
| 407 | |
| 408 | enum ocfs2_local_alloc_state local_alloc_state; /* protected |
| 409 | * by osb_lock */ |
| 410 | |
| 411 | struct buffer_head *local_alloc_bh; |
| 412 | |
| 413 | u64 la_last_gd; |
| 414 | |
| 415 | struct ocfs2_reservation_map osb_la_resmap; |
| 416 | |
| 417 | unsigned int osb_resv_level; |
| 418 | unsigned int osb_dir_resv_level; |
| 419 | |
| 420 | /* Next two fields are for local node slot recovery during |
| 421 | * mount. */ |
| 422 | struct ocfs2_dinode *local_alloc_copy; |
| 423 | struct ocfs2_quota_recovery *quota_rec; |
| 424 | |
| 425 | struct ocfs2_blockcheck_stats osb_ecc_stats; |
| 426 | struct ocfs2_alloc_stats alloc_stats; |
| 427 | char dev_str[20]; /* "major,minor" of the device */ |
| 428 | |
| 429 | u8 osb_stackflags; |
| 430 | |
| 431 | char osb_cluster_stack[OCFS2_STACK_LABEL_LEN + 1]; |
| 432 | char osb_cluster_name[OCFS2_CLUSTER_NAME_LEN + 1]; |
| 433 | struct ocfs2_cluster_connection *cconn; |
| 434 | struct ocfs2_lock_res osb_super_lockres; |
| 435 | struct ocfs2_lock_res osb_rename_lockres; |
| 436 | struct ocfs2_lock_res osb_nfs_sync_lockres; |
| 437 | struct rw_semaphore nfs_sync_rwlock; |
| 438 | struct ocfs2_lock_res osb_trim_fs_lockres; |
| 439 | struct mutex obs_trim_fs_mutex; |
| 440 | struct ocfs2_dlm_debug *osb_dlm_debug; |
| 441 | |
| 442 | struct dentry *osb_debug_root; |
| 443 | |
| 444 | wait_queue_head_t recovery_event; |
| 445 | |
| 446 | spinlock_t dc_task_lock; |
| 447 | struct task_struct *dc_task; |
| 448 | wait_queue_head_t dc_event; |
| 449 | unsigned long dc_wake_sequence; |
| 450 | unsigned long dc_work_sequence; |
| 451 | |
| 452 | /* |
| 453 | * Any thread can add locks to the list, but the downconvert |
| 454 | * thread is the only one allowed to remove locks. Any change |
| 455 | * to this rule requires updating |
| 456 | * ocfs2_downconvert_thread_do_work(). |
| 457 | */ |
| 458 | struct list_head blocked_lock_list; |
| 459 | unsigned long blocked_lock_count; |
| 460 | |
| 461 | /* List of dquot structures to drop last reference to */ |
| 462 | struct llist_head dquot_drop_list; |
| 463 | struct work_struct dquot_drop_work; |
| 464 | |
| 465 | wait_queue_head_t osb_mount_event; |
| 466 | |
| 467 | /* Truncate log info */ |
| 468 | struct inode *osb_tl_inode; |
| 469 | struct buffer_head *osb_tl_bh; |
| 470 | struct delayed_work osb_truncate_log_wq; |
| 471 | atomic_t osb_tl_disable; |
| 472 | /* |
| 473 | * How many clusters in our truncate log. |
| 474 | * It must be protected by osb_tl_inode->i_rwsem. |
| 475 | */ |
| 476 | unsigned int truncated_clusters; |
| 477 | |
| 478 | struct ocfs2_node_map osb_recovering_orphan_dirs; |
| 479 | unsigned int *osb_orphan_wipes; |
| 480 | wait_queue_head_t osb_wipe_event; |
| 481 | |
| 482 | struct ocfs2_orphan_scan osb_orphan_scan; |
| 483 | |
| 484 | /* used to protect metaecc calculation check of xattr. */ |
| 485 | spinlock_t osb_xattr_lock; |
| 486 | |
| 487 | unsigned int osb_dx_mask; |
| 488 | u32 osb_dx_seed[4]; |
| 489 | |
| 490 | /* the group we used to allocate inodes. */ |
| 491 | u64 osb_inode_alloc_group; |
| 492 | |
| 493 | /* rb tree root for refcount lock. */ |
| 494 | struct rb_root osb_rf_lock_tree; |
| 495 | struct ocfs2_refcount_tree *osb_ref_tree_lru; |
| 496 | |
| 497 | struct mutex system_file_mutex; |
| 498 | |
| 499 | /* |
| 500 | * OCFS2 needs to schedule several different types of work which |
| 501 | * require cluster locking, disk I/O, recovery waits, etc. Since these |
| 502 | * types of work tend to be heavy we avoid using the kernel events |
| 503 | * workqueue and schedule on our own. |
| 504 | */ |
| 505 | struct workqueue_struct *ocfs2_wq; |
| 506 | |
| 507 | /* sysfs directory per partition */ |
| 508 | struct kset *osb_dev_kset; |
| 509 | |
| 510 | /* file check related stuff */ |
| 511 | struct ocfs2_filecheck_sysfs_entry osb_fc_ent; |
| 512 | }; |
| 513 | |
| 514 | #define OCFS2_SB(sb) ((struct ocfs2_super *)(sb)->s_fs_info) |
| 515 | |
| 516 | /* Useful typedef for passing around journal access functions */ |
| 517 | typedef int (*ocfs2_journal_access_func)(handle_t *handle, |
| 518 | struct ocfs2_caching_info *ci, |
| 519 | struct buffer_head *bh, int type); |
| 520 | |
| 521 | static inline int ocfs2_should_order_data(struct inode *inode) |
| 522 | { |
| 523 | if (!S_ISREG(inode->i_mode)) |
| 524 | return 0; |
| 525 | if (OCFS2_SB(inode->i_sb)->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) |
| 526 | return 0; |
| 527 | return 1; |
| 528 | } |
| 529 | |
| 530 | static inline int ocfs2_sparse_alloc(struct ocfs2_super *osb) |
| 531 | { |
| 532 | if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC) |
| 533 | return 1; |
| 534 | return 0; |
| 535 | } |
| 536 | |
| 537 | static inline int ocfs2_writes_unwritten_extents(struct ocfs2_super *osb) |
| 538 | { |
| 539 | /* |
| 540 | * Support for sparse files is a pre-requisite |
| 541 | */ |
| 542 | if (!ocfs2_sparse_alloc(osb)) |
| 543 | return 0; |
| 544 | |
| 545 | if (osb->s_feature_ro_compat & OCFS2_FEATURE_RO_COMPAT_UNWRITTEN) |
| 546 | return 1; |
| 547 | return 0; |
| 548 | } |
| 549 | |
| 550 | static inline int ocfs2_supports_append_dio(struct ocfs2_super *osb) |
| 551 | { |
| 552 | if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_APPEND_DIO) |
| 553 | return 1; |
| 554 | return 0; |
| 555 | } |
| 556 | |
| 557 | |
| 558 | static inline int ocfs2_supports_inline_data(struct ocfs2_super *osb) |
| 559 | { |
| 560 | if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_INLINE_DATA) |
| 561 | return 1; |
| 562 | return 0; |
| 563 | } |
| 564 | |
| 565 | static inline int ocfs2_supports_xattr(struct ocfs2_super *osb) |
| 566 | { |
| 567 | if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_XATTR) |
| 568 | return 1; |
| 569 | return 0; |
| 570 | } |
| 571 | |
| 572 | static inline int ocfs2_meta_ecc(struct ocfs2_super *osb) |
| 573 | { |
| 574 | if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_META_ECC) |
| 575 | return 1; |
| 576 | return 0; |
| 577 | } |
| 578 | |
| 579 | static inline int ocfs2_supports_indexed_dirs(struct ocfs2_super *osb) |
| 580 | { |
| 581 | if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_INDEXED_DIRS) |
| 582 | return 1; |
| 583 | return 0; |
| 584 | } |
| 585 | |
| 586 | static inline int ocfs2_supports_discontig_bg(struct ocfs2_super *osb) |
| 587 | { |
| 588 | if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG) |
| 589 | return 1; |
| 590 | return 0; |
| 591 | } |
| 592 | |
| 593 | static inline unsigned int ocfs2_link_max(struct ocfs2_super *osb) |
| 594 | { |
| 595 | if (ocfs2_supports_indexed_dirs(osb)) |
| 596 | return OCFS2_DX_LINK_MAX; |
| 597 | return OCFS2_LINK_MAX; |
| 598 | } |
| 599 | |
| 600 | static inline unsigned int ocfs2_read_links_count(struct ocfs2_dinode *di) |
| 601 | { |
| 602 | u32 nlink = le16_to_cpu(di->i_links_count); |
| 603 | u32 hi = le16_to_cpu(di->i_links_count_hi); |
| 604 | |
| 605 | nlink |= (hi << OCFS2_LINKS_HI_SHIFT); |
| 606 | |
| 607 | return nlink; |
| 608 | } |
| 609 | |
| 610 | static inline void ocfs2_set_links_count(struct ocfs2_dinode *di, u32 nlink) |
| 611 | { |
| 612 | u16 lo, hi; |
| 613 | |
| 614 | lo = nlink; |
| 615 | hi = nlink >> OCFS2_LINKS_HI_SHIFT; |
| 616 | |
| 617 | di->i_links_count = cpu_to_le16(lo); |
| 618 | di->i_links_count_hi = cpu_to_le16(hi); |
| 619 | } |
| 620 | |
| 621 | static inline void ocfs2_add_links_count(struct ocfs2_dinode *di, int n) |
| 622 | { |
| 623 | u32 links = ocfs2_read_links_count(di); |
| 624 | |
| 625 | links += n; |
| 626 | |
| 627 | ocfs2_set_links_count(di, nlink: links); |
| 628 | } |
| 629 | |
| 630 | static inline int ocfs2_refcount_tree(struct ocfs2_super *osb) |
| 631 | { |
| 632 | if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_REFCOUNT_TREE) |
| 633 | return 1; |
| 634 | return 0; |
| 635 | } |
| 636 | |
| 637 | /* set / clear functions because cluster events can make these happen |
| 638 | * in parallel so we want the transitions to be atomic. this also |
| 639 | * means that any future flags osb_flags must be protected by spinlock |
| 640 | * too! */ |
| 641 | static inline void ocfs2_set_osb_flag(struct ocfs2_super *osb, |
| 642 | unsigned long flag) |
| 643 | { |
| 644 | spin_lock(lock: &osb->osb_lock); |
| 645 | osb->osb_flags |= flag; |
| 646 | spin_unlock(lock: &osb->osb_lock); |
| 647 | } |
| 648 | |
| 649 | static inline void ocfs2_set_ro_flag(struct ocfs2_super *osb, |
| 650 | int hard) |
| 651 | { |
| 652 | spin_lock(lock: &osb->osb_lock); |
| 653 | osb->osb_flags &= ~(OCFS2_OSB_SOFT_RO|OCFS2_OSB_HARD_RO); |
| 654 | if (hard) |
| 655 | osb->osb_flags |= OCFS2_OSB_HARD_RO; |
| 656 | else |
| 657 | osb->osb_flags |= OCFS2_OSB_SOFT_RO; |
| 658 | spin_unlock(lock: &osb->osb_lock); |
| 659 | } |
| 660 | |
| 661 | static inline int ocfs2_is_hard_readonly(struct ocfs2_super *osb) |
| 662 | { |
| 663 | int ret; |
| 664 | |
| 665 | spin_lock(lock: &osb->osb_lock); |
| 666 | ret = osb->osb_flags & OCFS2_OSB_HARD_RO; |
| 667 | spin_unlock(lock: &osb->osb_lock); |
| 668 | |
| 669 | return ret; |
| 670 | } |
| 671 | |
| 672 | static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb) |
| 673 | { |
| 674 | int ret; |
| 675 | |
| 676 | spin_lock(lock: &osb->osb_lock); |
| 677 | ret = osb->osb_flags & OCFS2_OSB_SOFT_RO; |
| 678 | spin_unlock(lock: &osb->osb_lock); |
| 679 | |
| 680 | return ret; |
| 681 | } |
| 682 | |
| 683 | static inline int ocfs2_is_readonly(struct ocfs2_super *osb) |
| 684 | { |
| 685 | int ret; |
| 686 | spin_lock(lock: &osb->osb_lock); |
| 687 | ret = osb->osb_flags & (OCFS2_OSB_SOFT_RO | OCFS2_OSB_HARD_RO); |
| 688 | spin_unlock(lock: &osb->osb_lock); |
| 689 | |
| 690 | return ret; |
| 691 | } |
| 692 | |
| 693 | static inline int ocfs2_emergency_state(struct ocfs2_super *osb) |
| 694 | { |
| 695 | if (ocfs2_is_readonly(osb)) |
| 696 | return -EROFS; |
| 697 | |
| 698 | return 0; |
| 699 | } |
| 700 | |
| 701 | static inline int ocfs2_clusterinfo_valid(struct ocfs2_super *osb) |
| 702 | { |
| 703 | return (osb->s_feature_incompat & |
| 704 | (OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK | |
| 705 | OCFS2_FEATURE_INCOMPAT_CLUSTERINFO)); |
| 706 | } |
| 707 | |
| 708 | static inline int ocfs2_userspace_stack(struct ocfs2_super *osb) |
| 709 | { |
| 710 | if (ocfs2_clusterinfo_valid(osb) && |
| 711 | memcmp(p: osb->osb_cluster_stack, OCFS2_CLASSIC_CLUSTER_STACK, |
| 712 | OCFS2_STACK_LABEL_LEN)) |
| 713 | return 1; |
| 714 | return 0; |
| 715 | } |
| 716 | |
| 717 | static inline int ocfs2_o2cb_stack(struct ocfs2_super *osb) |
| 718 | { |
| 719 | if (ocfs2_clusterinfo_valid(osb) && |
| 720 | !memcmp(p: osb->osb_cluster_stack, OCFS2_CLASSIC_CLUSTER_STACK, |
| 721 | OCFS2_STACK_LABEL_LEN)) |
| 722 | return 1; |
| 723 | return 0; |
| 724 | } |
| 725 | |
| 726 | static inline int ocfs2_cluster_o2cb_global_heartbeat(struct ocfs2_super *osb) |
| 727 | { |
| 728 | return ocfs2_o2cb_stack(osb) && |
| 729 | (osb->osb_stackflags & OCFS2_CLUSTER_O2CB_GLOBAL_HEARTBEAT); |
| 730 | } |
| 731 | |
| 732 | static inline int ocfs2_mount_local(struct ocfs2_super *osb) |
| 733 | { |
| 734 | return (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT); |
| 735 | } |
| 736 | |
| 737 | static inline int ocfs2_uses_extended_slot_map(struct ocfs2_super *osb) |
| 738 | { |
| 739 | return (osb->s_feature_incompat & |
| 740 | OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP); |
| 741 | } |
| 742 | |
| 743 | |
| 744 | #define OCFS2_IS_VALID_DINODE(ptr) \ |
| 745 | (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE)) |
| 746 | |
| 747 | #define OCFS2_IS_VALID_EXTENT_BLOCK(ptr) \ |
| 748 | (!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE)) |
| 749 | |
| 750 | #define OCFS2_IS_VALID_GROUP_DESC(ptr) \ |
| 751 | (!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE)) |
| 752 | |
| 753 | |
| 754 | #define OCFS2_IS_VALID_XATTR_BLOCK(ptr) \ |
| 755 | (!strcmp((ptr)->xb_signature, OCFS2_XATTR_BLOCK_SIGNATURE)) |
| 756 | |
| 757 | #define OCFS2_IS_VALID_DIR_TRAILER(ptr) \ |
| 758 | (!strcmp((ptr)->db_signature, OCFS2_DIR_TRAILER_SIGNATURE)) |
| 759 | |
| 760 | #define OCFS2_IS_VALID_DX_ROOT(ptr) \ |
| 761 | (!strcmp((ptr)->dr_signature, OCFS2_DX_ROOT_SIGNATURE)) |
| 762 | |
| 763 | #define OCFS2_IS_VALID_DX_LEAF(ptr) \ |
| 764 | (!strcmp((ptr)->dl_signature, OCFS2_DX_LEAF_SIGNATURE)) |
| 765 | |
| 766 | #define OCFS2_IS_VALID_REFCOUNT_BLOCK(ptr) \ |
| 767 | (!strcmp((ptr)->rf_signature, OCFS2_REFCOUNT_BLOCK_SIGNATURE)) |
| 768 | |
| 769 | static inline unsigned long ino_from_blkno(struct super_block *sb, |
| 770 | u64 blkno) |
| 771 | { |
| 772 | return (unsigned long)(blkno & (u64)ULONG_MAX); |
| 773 | } |
| 774 | |
| 775 | static inline u64 ocfs2_clusters_to_blocks(struct super_block *sb, |
| 776 | u32 clusters) |
| 777 | { |
| 778 | int c_to_b_bits = OCFS2_SB(sb)->s_clustersize_bits - |
| 779 | sb->s_blocksize_bits; |
| 780 | |
| 781 | return (u64)clusters << c_to_b_bits; |
| 782 | } |
| 783 | |
| 784 | static inline u32 ocfs2_clusters_for_blocks(struct super_block *sb, |
| 785 | u64 blocks) |
| 786 | { |
| 787 | int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits - |
| 788 | sb->s_blocksize_bits; |
| 789 | |
| 790 | blocks += (1 << b_to_c_bits) - 1; |
| 791 | return (u32)(blocks >> b_to_c_bits); |
| 792 | } |
| 793 | |
| 794 | static inline u32 ocfs2_blocks_to_clusters(struct super_block *sb, |
| 795 | u64 blocks) |
| 796 | { |
| 797 | int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits - |
| 798 | sb->s_blocksize_bits; |
| 799 | |
| 800 | return (u32)(blocks >> b_to_c_bits); |
| 801 | } |
| 802 | |
| 803 | static inline unsigned int ocfs2_clusters_for_bytes(struct super_block *sb, |
| 804 | u64 bytes) |
| 805 | { |
| 806 | int cl_bits = OCFS2_SB(sb)->s_clustersize_bits; |
| 807 | unsigned int clusters; |
| 808 | |
| 809 | bytes += OCFS2_SB(sb)->s_clustersize - 1; |
| 810 | /* OCFS2 just cannot have enough clusters to overflow this */ |
| 811 | clusters = (unsigned int)(bytes >> cl_bits); |
| 812 | |
| 813 | return clusters; |
| 814 | } |
| 815 | |
| 816 | static inline unsigned int ocfs2_bytes_to_clusters(struct super_block *sb, |
| 817 | u64 bytes) |
| 818 | { |
| 819 | int cl_bits = OCFS2_SB(sb)->s_clustersize_bits; |
| 820 | unsigned int clusters; |
| 821 | |
| 822 | clusters = (unsigned int)(bytes >> cl_bits); |
| 823 | return clusters; |
| 824 | } |
| 825 | |
| 826 | static inline u64 ocfs2_blocks_for_bytes(struct super_block *sb, |
| 827 | u64 bytes) |
| 828 | { |
| 829 | bytes += sb->s_blocksize - 1; |
| 830 | return bytes >> sb->s_blocksize_bits; |
| 831 | } |
| 832 | |
| 833 | static inline u64 ocfs2_clusters_to_bytes(struct super_block *sb, |
| 834 | u32 clusters) |
| 835 | { |
| 836 | return (u64)clusters << OCFS2_SB(sb)->s_clustersize_bits; |
| 837 | } |
| 838 | |
| 839 | static inline u64 ocfs2_block_to_cluster_start(struct super_block *sb, |
| 840 | u64 blocks) |
| 841 | { |
| 842 | int bits = OCFS2_SB(sb)->s_clustersize_bits - sb->s_blocksize_bits; |
| 843 | unsigned int clusters; |
| 844 | |
| 845 | clusters = ocfs2_blocks_to_clusters(sb, blocks); |
| 846 | return (u64)clusters << bits; |
| 847 | } |
| 848 | |
| 849 | static inline u64 ocfs2_align_bytes_to_clusters(struct super_block *sb, |
| 850 | u64 bytes) |
| 851 | { |
| 852 | int cl_bits = OCFS2_SB(sb)->s_clustersize_bits; |
| 853 | unsigned int clusters; |
| 854 | |
| 855 | clusters = ocfs2_clusters_for_bytes(sb, bytes); |
| 856 | return (u64)clusters << cl_bits; |
| 857 | } |
| 858 | |
| 859 | static inline u64 ocfs2_align_bytes_to_blocks(struct super_block *sb, |
| 860 | u64 bytes) |
| 861 | { |
| 862 | u64 blocks; |
| 863 | |
| 864 | blocks = ocfs2_blocks_for_bytes(sb, bytes); |
| 865 | return blocks << sb->s_blocksize_bits; |
| 866 | } |
| 867 | |
| 868 | static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes) |
| 869 | { |
| 870 | return (unsigned long)((bytes + 511) >> 9); |
| 871 | } |
| 872 | |
| 873 | static inline unsigned int ocfs2_page_index_to_clusters(struct super_block *sb, |
| 874 | unsigned long pg_index) |
| 875 | { |
| 876 | u32 clusters = pg_index; |
| 877 | unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits; |
| 878 | |
| 879 | if (unlikely(PAGE_SHIFT > cbits)) |
| 880 | clusters = pg_index << (PAGE_SHIFT - cbits); |
| 881 | else if (PAGE_SHIFT < cbits) |
| 882 | clusters = pg_index >> (cbits - PAGE_SHIFT); |
| 883 | |
| 884 | return clusters; |
| 885 | } |
| 886 | |
| 887 | /* |
| 888 | * Find the 1st page index which covers the given clusters. |
| 889 | */ |
| 890 | static inline pgoff_t ocfs2_align_clusters_to_page_index(struct super_block *sb, |
| 891 | u32 clusters) |
| 892 | { |
| 893 | unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits; |
| 894 | pgoff_t index = clusters; |
| 895 | |
| 896 | if (PAGE_SHIFT > cbits) { |
| 897 | index = (pgoff_t)clusters >> (PAGE_SHIFT - cbits); |
| 898 | } else if (PAGE_SHIFT < cbits) { |
| 899 | index = (pgoff_t)clusters << (cbits - PAGE_SHIFT); |
| 900 | } |
| 901 | |
| 902 | return index; |
| 903 | } |
| 904 | |
| 905 | static inline unsigned int ocfs2_pages_per_cluster(struct super_block *sb) |
| 906 | { |
| 907 | unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits; |
| 908 | unsigned int pages_per_cluster = 1; |
| 909 | |
| 910 | if (PAGE_SHIFT < cbits) |
| 911 | pages_per_cluster = 1 << (cbits - PAGE_SHIFT); |
| 912 | |
| 913 | return pages_per_cluster; |
| 914 | } |
| 915 | |
| 916 | static inline unsigned int ocfs2_megabytes_to_clusters(struct super_block *sb, |
| 917 | unsigned int megs) |
| 918 | { |
| 919 | BUILD_BUG_ON(OCFS2_MAX_CLUSTERSIZE > 1048576); |
| 920 | |
| 921 | return megs << (20 - OCFS2_SB(sb)->s_clustersize_bits); |
| 922 | } |
| 923 | |
| 924 | static inline unsigned int ocfs2_clusters_to_megabytes(struct super_block *sb, |
| 925 | unsigned int clusters) |
| 926 | { |
| 927 | return clusters >> (20 - OCFS2_SB(sb)->s_clustersize_bits); |
| 928 | } |
| 929 | |
| 930 | static inline void _ocfs2_set_bit(unsigned int bit, unsigned long *bitmap) |
| 931 | { |
| 932 | __set_bit_le(nr: bit, addr: bitmap); |
| 933 | } |
| 934 | #define ocfs2_set_bit(bit, addr) _ocfs2_set_bit((bit), (unsigned long *)(addr)) |
| 935 | |
| 936 | static inline void _ocfs2_clear_bit(unsigned int bit, unsigned long *bitmap) |
| 937 | { |
| 938 | __clear_bit_le(nr: bit, addr: bitmap); |
| 939 | } |
| 940 | #define ocfs2_clear_bit(bit, addr) _ocfs2_clear_bit((bit), (unsigned long *)(addr)) |
| 941 | |
| 942 | #define ocfs2_test_bit test_bit_le |
| 943 | #define ocfs2_find_next_zero_bit find_next_zero_bit_le |
| 944 | #define ocfs2_find_next_bit find_next_bit_le |
| 945 | |
| 946 | static inline void *correct_addr_and_bit_unaligned(int *bit, void *addr) |
| 947 | { |
| 948 | #if BITS_PER_LONG == 64 |
| 949 | *bit += ((unsigned long) addr & 7UL) << 3; |
| 950 | addr = (void *) ((unsigned long) addr & ~7UL); |
| 951 | #elif BITS_PER_LONG == 32 |
| 952 | *bit += ((unsigned long) addr & 3UL) << 3; |
| 953 | addr = (void *) ((unsigned long) addr & ~3UL); |
| 954 | #else |
| 955 | #error "how many bits you are?!" |
| 956 | #endif |
| 957 | return addr; |
| 958 | } |
| 959 | |
| 960 | static inline void ocfs2_set_bit_unaligned(int bit, void *bitmap) |
| 961 | { |
| 962 | bitmap = correct_addr_and_bit_unaligned(bit: &bit, addr: bitmap); |
| 963 | ocfs2_set_bit(bit, bitmap); |
| 964 | } |
| 965 | |
| 966 | static inline void ocfs2_clear_bit_unaligned(int bit, void *bitmap) |
| 967 | { |
| 968 | bitmap = correct_addr_and_bit_unaligned(bit: &bit, addr: bitmap); |
| 969 | ocfs2_clear_bit(bit, bitmap); |
| 970 | } |
| 971 | |
| 972 | static inline int ocfs2_test_bit_unaligned(int bit, void *bitmap) |
| 973 | { |
| 974 | bitmap = correct_addr_and_bit_unaligned(bit: &bit, addr: bitmap); |
| 975 | return ocfs2_test_bit(nr: bit, addr: bitmap); |
| 976 | } |
| 977 | |
| 978 | static inline int ocfs2_find_next_zero_bit_unaligned(void *bitmap, int max, |
| 979 | int start) |
| 980 | { |
| 981 | int fix = 0, ret, tmpmax; |
| 982 | bitmap = correct_addr_and_bit_unaligned(bit: &fix, addr: bitmap); |
| 983 | tmpmax = max + fix; |
| 984 | start += fix; |
| 985 | |
| 986 | ret = ocfs2_find_next_zero_bit(addr: bitmap, size: tmpmax, offset: start) - fix; |
| 987 | if (ret > max) |
| 988 | return max; |
| 989 | return ret; |
| 990 | } |
| 991 | |
| 992 | #endif /* OCFS2_H */ |
| 993 | |
| 994 | |